Comparing alternating pressure mattresses and high-specification foam mattresses to prevent pressure ulcers in high-risk patients: the PRESSURE 2 RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 11/36/33. The contractual start date was in March 2013. The draft report began editorial review in December 2017 and was accepted for publication in July 2018. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Christopher McCabe has received grant funding from the University of Alberta and Alberta Innovates Health Solutions (Edmonton, AB, Canada). Julia Brown is Deputy Director of the National Institute for Health Research (NIHR) Health Technology Assessment (HTA) Clinical Evaluation and Trials Board and has received grant funding for the following studies: the NIHR eRAPID Programme Grants for Applied Research (PGfAR) trial, the NIHR Efficacy and Mechanism Evaluation (EME) GLiSten trial, the NIHR HTA LAVA trial (liver resection surgery versus thermal ablation for colorectal liver metastases), the NIHR EME IntAct (Intraoperative flourescence angiography to prevent Anastomotic leak in rectal cancer surgery), NIHR Research for Patient Benefit LACE (Life After Cancer Epidemiology) trial, the NIHR EME ROLARR (RObotic vs. LAparoscopic Resection for Rectal cancer) trial, the NIHR HTA SaFarI (Sacral nerve stimulation versus the FENIX TM magnetic sphincter augmentation for faecal incontinence: a Randomised Investigation and the NIHR StamINA Programme Development Grant trial. Claire Hulme and E Andrea Nelson have been members of the NIHR HTA Commissioning Board and E Andrea Nelson has received funding for the NIHR HTA CODIFI (Concordance in Diabetic Foot Ulcer Infection) study. Elizabeth McGinnis has received funding for NIHR Health Service and Delivery Research (HSDR) Information Systems, Monitoring and Managing from Ward to Board and the NIHR PGfAR SWHSI (Surgical Wounds Healing by Secondary Intention) trial. Isabelle L Smith and Susanne Coleman have received a NIHR personal fellowship. Benjamin Thorpe has received a NIHR research methods fellowship. Delia Muir has received a Wellcome Trust Engagement Fellowship. Rachael Gilberts has received funding for the NIHR HTA ALPHA (ALitretinoin versus PUVA in severe chronic HAnd eczema) and MIDFUT (Multiple Interventions for Diabetic Foot Ulcer Treatment) trials. Nikki Stubbs has received funding for NIHR PGfAR SWHSI and NIHR HTA AMBER (Abdominal Massage for Bowel Dysfunction Effectiveness Research) trial. Jane Nixon has received funding for NIHR HTA MIDFUT, CODIFI and ALPHA. Sarah Brown has received funding for NIHR HTA ALPHA, MIDFUT, FORVAD (Clinical and cost-effectiveness of posterior cervical FORaminotomy Versus Anterior cervical Discectomy in the treatment of cervical brachialgia: a multicentre, Phase III, randomised controlled trial), the NIHR PGfAR PROMPT (early detection to improve outcome in patients with undiagnosed psoriatic arthritis), ARUK (Arthritis Research UK) SALRISE (SALivary electro-stimulation for the treatment of dry mouth in patients with Sjögren’s syndrome: a multicentRe randomISEd sham-controlled double-blind study).

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Nixon et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Classification

Pressure ulcers range in severity from intact skin with non-blanchable erythema to severe ulcers involving fat, muscle and bone. Various classification systems have been developed8,11,12 using terms such as ‘grade’, ‘stage’ and ‘category’. The most widely used international classifications use a numerical scale of 1–4 to indicate increasing severity of skin and underlying soft tissue/bone involvement, with additional descriptors including ‘unstageable’ and ‘deep tissue injury’ for which a 1–4 classification cannot be determined from clinical examination. 8,12

More recently, Australian and US groups have advocated that the term ‘pressure injury’ be adopted to replace the term ‘pressure ulcer’. 13,14 The European position15 and a NHS Improvement Definitions Task and Finish Group16 both recommend that the term ‘pressure ulcer’ is retained so that the UK is aligned with the most up-to-date International Classification of Diseases, 11th Revision. 17

For the purposes of this report, the term ‘pressure ulcer’ and its abbreviation ‘PU’ will be used.

Risk factors

A systematic review on PU risk factors identified the risk factor domains emerging most frequently in multivariable modelling of independent predictors of PU development as mobility/activity, perfusion (including diabetes) and skin/PU status. 18 Moisture, age, haematological measures, nutrition and general health status are also important, but emerge inconsistently in multivariable modelling. The importance of body temperature and immunity require further confirmatory research and there is no clear evidence that race and gender are important to PU development. 18 The risk factor systematic review informed the development of a PU conceptual framework that maps the potential relationships between the key ‘direct causal factors’ (immobility/inactivity, perfusion and skin status) and the large number of other ‘indirect causal factors’ (e.g. moisture, age, nutrition, acute illness). 9

Importance

Over the past three decades, government health departments, national guideline agencies, commissioners and funders of health care have promoted policies, incentives and guidelines that focus on the prevention of PUs. 5,12,19–22 This reflects the high personal costs incurred by patients1,23 and the high financial costs incurred by health-care funders and providers in the treatment of PUs due to increased length of hospital stay, hospital admission, community nursing, treatments (reconstruction surgery/mattresses/dressings/technical therapies) and complications (serious infection). 24,25 Litigation is also a burden on health-care providers and is predicted to increase because of general societal trends and, in some cases (e.g. the UK), changes in the national policies that have led to investigation by government agencies of severe PUs to detect institutional and professional neglect of vulnerable adults. 21,22

Interventions for pressure ulcer prevention

In all national and international guidelines, there are three main components of PU prevention practice, including:

1. formal risk assessment to identify ‘at risk’ patients and target preventative interventions

2. minimising both the intensity and the duration of exposure to mechanical loads on vulnerable skin sites not adapted to loading8,12 through –

   • minimising exposure to localised pressure through use of pressure-redistributing mattresses/cushions

   • intermittent pressure relief26 through major or minor repositioning of patients and/or mattresses and cushions that alternate temporarily to relieve pressure

   • complete pressure relief through continuous offloading of a specific skin site12 achieved by positioning patients or providing a heel offloading device so that a specific skin site (i.e. a skin site with an existing PU or adverse skin status) is completely offloaded at all times.

3. optimising skin condition through cleansing, drying and minimising exposure to moisture. 12

In this randomised controlled trial (RCT), the focus was on the effectiveness of two interventions: one minimising the intensity of pressure and one minimising both the intensity and the duration of pressure.

Pressure ulcer prevention mattresses

Pressure-relieving mattresses either distribute a patient’s weight over a larger contact area, providing ‘constant low pressure’ to reduce pressure intensity, or mechanically distribute a patient’s weight over a large contact area AND vary the pressure beneath the patient in order to reduce both the intensity and the duration of pressure [alternating pressure mattresses (APMs)]. 27 There are a range of ‘constant low pressure’ mattresses that are classified as ‘low technology’ [e.g. high-specification foam, gel, water] and ‘high technology’ [e.g. electrically powered air and air-fluidised (bead) beds]. All the alternating pressure mattresses are electrically powered and are classified as high technology. 27 Unit costs vary considerably; mattresses used frequently in the UK can be as little as £80–35228 for a low-technology high-specification foam mattress or as much as £500–4000 for a high-technology alternating pressure mattress. 28 In this study, these two main mattress types, utilised within the NHS, were compared, namely (1) high-technology APMs and (2) HSFMs, which are low-technology support surfaces. 27

Table 1 illustrates different types of mattresses and their classifications as ‘low technology’ and ‘high technology’, as adapted from the National Institute for Health and Care Excellence (NICE) guidelines12 and Cochrane systematic review. 27 Only HSFMs and APMs were used in this study.

Mattress intervention effectiveness

Overall in this field, the quality of trials is poor [small underpowered studies without allocation concealment, intention-to-treat (ITT) analysis or a priori sample size estimate]. 27,29 NICE guidelines12 and systematic review evidence27 highlight the fact that NHS resource availability is not based on robust health economic evaluations.

The fifth version of the Cochrane systematic review of support surfaces for PU prevention27 was published in 2015. The review included 59 RCTs, of which 11 compared the effectiveness of APMs with a range of low-technology (including high-specification foam, gel, silicone, water, static air) and high-technology constant low-pressure mattresses (low air loss). Most trials showed no evidence of a difference between treatment groups, although some were too small and underpowered to detect clinically important differences. Pooling of data in a meta-analysis of 10 trials of APMs versus various constant low-pressure mattresses showed no evidence of a difference in effectiveness for PU prevention.

Of the 11 RCTs, only one study compared APMs with a HSFM (viscoelastic)30 as part of a 2 × 2 factorial design incorporating two methods of risk assessment (Braden Scale31 score vs. observation of category 1) and two mattress/repositioning interventions (APM overlay vs. HSFM with turning every 4 hours). The study took place in a single centre and recruited patients from surgery, internal medicine and elderly care. Patients aged ≥ 18 years, with an expected length of stay of > 3 days were recruited and randomised to be ‘risk assessed’ using either the Braden Scale31 or observation of a category 1 skin area. Those identified as being at risk of developing a PU then had a second-level randomisation, with allocation to an APM or HSFM with turning every 4 hours. A total of 447 patients had the second-level randomisation and the incidence of PUs (of category ≥ 2) was 15.3% (34/222) in the APM arm and 15.6% (35/225) in the HSFM plus repositioning arm. Outcome data were recorded by ward staff. There were more heel ulcers reported in the HSFM arm, but more severe ulcers developed in the APM arm. An adjusted analysis incorporating risk assessment was not reported. Results are confounded by the inclusion in the HSFM arm of turning every 4 hours, and the potential for contamination between methods of risk assessment was not explored (including impact on outcome assessment, i.e. staff in the category 1 risk assessment arm were already alert to pressure damage); a major limitation is that outcome data were recorded by ward staff.

The review concluded that the relative merits of APMs and constant low-pressure mattresses are unclear. The review recommended the evaluation of APMs compared with constant low-pressure mattresses (such as HSFMs) because of their widespread use. 27 Similarly, NICE guidelines12 research recommendations acknowledge the limited evidence of effectiveness of pressure-redistributing devices and recommend further research. They particularly note the variation in cost of APMs compared with HSFMs.

The patient perspective

Although NICE guidelines12 are underpinned by the principle that all health-care professionals should deliver patient-centred care, it is exceptional to find a RCT with a patient-centred outcome such as acceptability, comfort or quality of life (QoL). Only four out of 59 RCTs included in the Cochrane review27 report patient comfort or acceptability of the mattress as an outcome and none report a patient’s QoL.

There is evidence from qualitative studies exploring the lived experience of patients with PUs23,32 and secondary end-point data in Pressure RElieving Support SUrfaces: a Randomised Evaluation (PRESSURE) 133 that some patients do not like APMs; a network meta-analysis34 reported that APMs had the lowest probability of being the most comfortable compared with other mattress types. Alternating pressure mattresses comprise large air-filled pockets that inflate and deflate in cycles. The alternating sensation is disliked by some patients and can cause feelings of nausea and affect sleep. In addition, on patient movement, the air-filled pockets are compressed and patients find it difficult to mobilise in bed and also report feeling unstable at the mattress edge, when they are getting in and out of bed, or feeling like they will be ‘rolled out of bed’, creating an unsafe feeling. 23,32,33 Other issues include noise from the pump, technical failure and attendant alarms. This was further supported by members of the Pressure Ulcer Research Service User Network (PURSUN),35 which supports research prioritisation and informs grant development, project delivery and data interpretation. Some members of the group feel strongly that APMs can be uncomfortable and debilitating, restrict movement and independence, exacerbate existing balance/mobility problems and leave patients in need of extra care (i.e. help in repositioning).

Practice guidelines

Previous NICE guidance26 stated that:

. . . decisions about which pressure-relieving device to use should be based on cost considerations and an overall assessment of the individual. Holistic assessment should include all of the following: identified levels of risk, skin assessment, comfort, general health state, lifestyle and abilities, critical care needs and acceptability of the proposed pressure-relieving equipment to the patient and/or carer.

Reproduced with permission from NICE. © NICE 2003 Pressure Ulcer Prevention: Pressure Ulcer Risk Assessment and Prevention, Including the use of Pressure-Relieving Devices (Beds, Mattresses and Overlays) for the Prevention of Pressure Ulcers in Primary and Secondary Care. 26 Available from www.nice.org.uk/guidance/cg7 All rights reserved. Subject to Notice of Rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication

It is not clear what ‘cost considerations’ means, but, in practice, decisions are generally made on unit costs and not cost-effectiveness, this being challenged following publication of PRESSURE 1,33 in which it was demonstrated that, despite no clinical difference between mattresses, there was a 64% probability that the more expensive APM replacement (unit cost ≈£4000) was more cost-effective than the cheaper APM overlay (unit cost ≈£1000).

The current NICE guideline12 for PU prevention utilised the evidence found in the 2011 version of the McInnes et al. 36 Cochrane review and recommended that a HSFM should be used for adults in secondary care settings or those who are assessed as being at a high risk of PUs in community and primary care environments. As this recommendation is based on trials deemed to be of low quality, NICE highlighted the following research priority:

Do pressure redistributing devices reduce the development of pressure ulcers for those who are at risk of developing a pressure ulcer?

Reproduced with permission from NICE. © NICE 2014 Pressure Ulcers: Prevention and Management. 12 Available from www.nice.org.uk/guidance/cg179 All rights reserved. Subject to Notice of Rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication

The NICE guidelines section for PU treatment, however, do include the following caveat:

Use high-specification foam mattresses for adults with a pressure ulcer. If this is not sufficient to redistribute pressure, consider the use of a dynamic support surface.

Reproduced with permission from NICE. © NICE 2014 Pressure Ulcers: Prevention and Management. 12 Available from www.nice.org.uk/guidance/cg179 All rights reserved. Subject to Notice of Rights. NICE guidance is prepared for the National Health Service in England. All NICE guidance is subject to regular review and may be updated or withdrawn. NICE accepts no responsibility for the use of its content in this product/publication

The quality of evidence to inform this recommendation was considered to be low and none of the studies followed up patients for a sufficient time. There is no reference to the category of PU in the recommendation; however, the RCTs used to inform this included PUs of all categories.

NHS practice

Traditional hospital foam mattresses (with a marbled cover) have been superseded by HSFMs with both a ‘high-performance’ foam core and a cover designed to minimise ‘hammocking’ and build-up of moisture. There is good evidence of the benefit of HSFMs compared with traditional hospital mattresses in reducing the incidence of PUs in high-risk patient populations27 and HSFMs are in widespread use in the NHS, with many hospitals providing all patients with a HSFM as standard. Following NICE guidance,26 a HSFM is the recommended ‘minimum’ standard care for those assessed as being ‘vulnerable to PUs’.

Despite the lack of evidence of benefit, APMs are also in widespread use in the UK for at-risk patients. In two studies conducted as part of the Pressure UlceR Programme Of reSEarch (PURPOSE),37 their common use in the NHS was identified. First, the multicentre PU prevalence survey,6 involving nine hospitals across three NHS trusts and ≈3000 patients, found that ≈20% of mattresses in the adult care setting were APMs. Second, in a multicentre cohort study of 634 patients, involving high-risk patients with mobility/activity impairment and/or category 1 PUs,38 mattress allocation by ward staff to the study population was 48% HSFM and 52% APM, reflecting a lack of standardised practice.

History of this research

A Health Technology Assessment (HTA) programme commissioning brief in 1998 included multicentre RCTs to compare the clinical effectiveness and cost-effectiveness of APMs with:

• less costly alternating pressure overlays

• low-technology constant low-pressure alternatives (such as different types of HSFM).

At that time, there was a reluctance by clinicians to randomise high-risk patients to HSFM, so the trial funded by the HTA programme, PRESSURE 1,33,37,39 dealt with only the first of the two research priorities and compared overlay and replacement APMs.

However, since then, many UK hospitals have replaced traditional hospital mattresses with HSFMs as standard for some or all clinical specialties. In addition, NICE guidance12 and the widespread use of profiling electric beds have increased clinical confidence in the use of HSFMs for high-risk patients. Furthermore, qualitative and quantitative evidence suggests that some patients do not like APMs,23,32,33 and results of the PRESSURE 1 trial33 showed a lack of difference in clinical outcomes between expensive alternating pressure replacement mattresses and cheaper alternating pressure overlay mattresses. These developments in the knowledge base and clinical experiences have challenged previously held views of clinical effectiveness based on non-randomised evaluations, which inferred the superiority of APMs and enabled trial design and delivery to address a key clinical question comparing high-technology APMs and low-technology HSFMs in the prevention of new PUs.

Summary

In the light of the priority being given to PU prevention, the high cost and lack of evidence relating to the effectiveness of mattresses in common use in the NHS, ad hoc practice in mattress allocation and the disadvantages and difficulties reported by patients in the use of APMs, we undertook a RCT to compare APMs with HSFMs in a high-risk inpatient population.

Aims and objectives

The primary aim was to determine the clinical effectiveness and cost-effectiveness of APMs and HSFMs when used in conjunction with an electric profiling bed frame in secondary and community inpatients with evidence of acute illness, for the prevention of PUs of category ≥ 2.

Secondary aims were to assess the feasibility of using photography to quantify potential bias in the reporting of the PRESSURE 2 trial primary end point (see Chapter 5) and modify and evaluate the Pressure Ulcer Quality-of-Life (PU-QoL) instrument for use in patients at risk of PUs receiving preventative interventions (see Chapter 6).

Primary trial objective

The primary objective was to compare the time taken to develop a new PU of category ≥ 2 in participants using APM with the time taken to develop a new PU of category ≥ 2 in those using HSFM.

Overview of methods

This chapter outlines the main methods for the trial, including all data collection for the trial and substudy work and the analysis methods for the primary and secondary clinical results, which are detailed in Chapter 3. Analysis methods and results for the main trial health economics and QoL comparisons, the photography substudy and the PU-QoL substudy are detailed in Chapters 4, 5 and 6, respectively. The discussion in Chapter 7 draws the work together.

Trial design

The trial was a multicentre, Phase III, open, prospective, randomised, planned as a double-triangular sequential, parallel-group trial, with two planned interim analyses.

High-risk patients from secondary and community inpatient facilities with evidence of acute illness were randomised on a 1 : 1 basis to receive either APM or HSFM in conjunction with an electric profiling bed.

The intervention phase follow-up period was for a maximum of 60 days (referred to as the ‘treatment phase’) and there was a 30-day post-treatment phase follow-up (referred to as the ‘30-day final follow-up’).

The study protocol for this trial has already been published. 40 A study flow diagram/study summary can be found on page 6 of the trial protocol [see Methods on the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/113633/#/ (accessed 20 November 2018)]. Summary details of the methods are given in the following sections.

Ethics approval

Ethics approval for the study was given by Leeds West Research Ethics Committee [Research Ethics Committee (REC) reference number 13/YH/0066; Integrated Research Application System (IRAS) project identification: 122769]. All trial activity took place in accordance with the ethics-approved protocol, principles of Good Clinical Practice41 and the Declaration of Helsinki. 42

Eligibility and informed consent

Acute secondary and community NHS trust inpatient admissions were screened for eligibility by a clinical research nurse/registered health-care professional (CRN/P) in consultation with ward staff. Patients were eligible at any point during their inpatient stay (and irrespective of trust provider) if they fulfilled the following criteria:

• Had evidence of acute illness through –

  • acute admission to secondary care hospital, community hospital or NHS-funded intermediate care/rehabilitation facility

  • inpatient secondary care, community hospital or NHS-funded intermediate care/rehabilitation facility with an onset of acute illness secondary to elective admission

  • recent secondary care hospital discharge to community hospital or NHS-funded intermediate care/rehabilitation facility.

• Was aged ≥ 18 years.

• Had an expected total length of stay of ≥ 5 days.

• Was at a high risk of PU development as a result of one or more of the following:

  • bedfast/chairfast AND completely immobile/very limited mobility (Braden Scale31 activity score of 1 or 2 and mobility score of 1 or 2)31

  • category 1 PU on any pressure area skin site

  • localised skin pain on a healthy, altered or category 1 pressure area skin site.

• Consented to participate [written, informed consent/witnessed verbal consent/consultee agreement or nearest relative/guardian/welfare attorney (in Scotland)].

• Was expected to comply with the follow-up schedule.

• Was on an electric profiling bed frame.

Patients were excluded if:

• they had previously participated in the PRESSURE 2 trial

• they had a current or previous PU of category ≥ 3

• they had a planned admission to an intensive care unit where standard care was APM provision

• they were unable to receive the intervention (e.g. slept at night in a chair or was unable to transfer to randomised mattress)

• they weighed less or more than mattress weight limits (< 45 kg or > 180 kg)

• it was ethically inappropriate to approach them.

A log was completed of all patients who were screened but not randomised, either because they were ineligible or because they declined participation. The following anonymised information was included:

• age

• gender

• ethnicity

• current mattress type

• date screened

• the reason why they were not eligible or declined participation

• other reason for non-randomisation.

Potentially eligible patients were provided with a verbal explanation of the study and a patient information leaflet [see the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/113633/#/ (accessed 20 November 2018)]. Assenting patients were formally assessed for eligibility and invited to provide informed consent. Witnessed consent process was used [see Witnessed Consent Form on the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/113633/#/ (accessed 20 November 2018)] for those who gave consent but were physically unable to complete the written form.

A large proportion of patients suffering from or at risk of PUs have cognitive impairment affecting their understanding and/or dementia. Cognition affects compliance with repositioning and self-care. To ensure that the study population was representative of the normal NHS clinical population assessed in the course of usual care, recruitment procedures also facilitated consultee or nearest relative/guardian/welfare attorney (in Scotland) agreement [see Consultee Agreement Form on the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/113633/#/ (accessed 20 November 2018)].

Interventions

Participants were randomised to either APM or HSFM products used by the participating centre for a maximum of 60 days. All patients were also allocated an electric profiling bed, as per standard care in the participating centres.

As this was a pragmatic trial, operational specifications for both APMs and HSFMs were defined in the PRESSURE 2 mattress specification guide (MSG) (see Appendix 1). Individual products allocated to trial participants were part of the usual-care mattress stock for each participating hospital, maximising generalisability of the trial findings.

The PRESSURE 2 MSG (see Appendix 1) provided details of foam density and cover details for HSFM and cell height (8.5 cm minimum, 29.4 cm maximum) and cycle time and frequency for APMs. Excluded mattresses, for example hybrid foam/air mattresses, were also specified. During feasibility assessment, mattresses in each centre were reviewed against the PRESSURE 2 MSG and only trusts with sufficient access to trial-eligible mattresses and electric profiling beds were able to take part in the trial. As new mattresses were marketed during the trial period, the PRESSURE 2 MSG was updated.

All mattresses had to comply with The Medical Devices Regulations 2002 SI (Système International d’Unités). 43 The allocated randomised mattress was expected to be provided to the trial patient within 24 hours of randomisation.

Randomisation

Participants were randomised once eligibility was confirmed and the baseline assessments and questionnaires were completed. Participants were randomised in a 1 : 1 allocation ratio, to receive either an APM or a HSFM in conjunction with an electric profiling bed. A computer-generated minimisation program, which incorporated a random element of 0.8, was used to ensure that intervention groups were well balanced for the following factors:

• centre

• PU status (no PU, category 1 or 2 PU)

• secondary care hospital, community hospital/intermediate care or rehabilitation facility

• consent [written, witnessed verbal, or consultee or nearest relative/guardian/welfare attorney (in Scotland) agreement].

The randomisation system included an automated internal check using a patient’s NHS number to confirm that the participant had not been recruited to the trial previously.

Blinding

It was not possible to blind participants, the clinical team or the CRN/P because of the nature of the mattresses (presence/sound of a pump on the APM and different appearance of bed and sheeting on each type of mattress).

A validation substudy, using photography with blinded central review, was therefore included to assess any bias in the reporting of PUs of category ≥ 2, details of which are given in Chapter 5.

Assessments and instruments

Data collection schedule

All baseline and outcome assessments were undertaken by trained CRN/Ps. Treatment phase follow-up assessments were undertaken from randomisation to the end of the treatment phase (a maximum of 60 days). A final 30-day final follow-up visit was undertaken.

Trial completion

Trial completion was defined as the end of the 30-day final follow-up (i.e. 30 days after the end-of-treatment phase), withdrawal or death.

The treatment phase was defined as the time from randomisation to (whichever happened soonest):

• discharge from an eligible inpatient facility

• 60 days from randomisation

• no longer at a high risk.

No longer at a high risk was defined as no PU of category ≥ 1 on any skin site AND no localised skin pain on any pressure area skin site AND improved mobility and activity (Braden Scale31 activity score of 3 or 4 AND mobility score of 3 or 4). 50

End points

Trial organisational structure

The trial sponsor was the University of Leeds and responsibilities were delegated to the CTRU, as detailed in the trial contract.

Trial oversight and management was conducted by the Trial Management Group (TMG), the TSC and the DMEC. All groups met regularly throughout the trial.

Centre-level case report form (CRF) returns, data quality, screening logs, recruitment, patient disposition, protocol deviations and safety were monitored by the CTRU, TMG, TSC and DMEC. In addition, overall mattress compliance was monitored by the TSC and DMEC (with no cause for concern to prompt investigation at a centre level). Compliance with the photography and skin verification substudies and the overall event rate were monitored by the DMEC.

Public involvement methods

The CTRU hosts a network of service users, PURSUN, with personal experience of PUs and/or PU risk. The group consists of patients, carers and family members. PURSUN works to improve public involvement in PU research and raise awareness of the topic. It has an established relationship with the study team and has supported this study from conception. Public involvement activities have included the following:

• Study design. Meetings were held with PURSUN at both the grant application and the protocol development stages. In particular, PURSUN made an important contribution to the development of the eligibility criteria (at its request, patients with existing PUs of category ≥ 3 were excluded from participation in the trial) and to the consent process for the photography substudy (as PUs commonly develop on sensitive areas of the body, such as buttocks, PURSUN advised whether or not and how the consent process should be conducted).

• Developing participant materials.

• Project oversight. A service user co-applicant (KW) was invited to all TMG meetings and two PURSUN members joined the TSC. Update meetings were also held with the wider network throughout the study.

• Results interpretation. One results interpretation workshop has already been held with a small number of PURSUN members. A larger one is planned for 2019. The aim of the workshops is to discuss what the results mean for service users, and to plan a collaborative approach to dissemination and implementation.

Statistical methods

Statistical analysis

All analyses were performed using SAS^® software version 9.4 (SAS Institute Inc., Cary, NC, USA).

A SAP was approved prior to the relevant analysis being conducted [see Methods on the project web page: www.journalslibrary.nihr.ac.uk/programmes/hta/113633/#/ (accessed 20 November 2018)].

Reflecting the double-triangular sequential design, a maximum of three analyses were originally planned with unequally spaced reviews at event-driven coherent cut-off points, as specified below.

Interim analysis 1 was to be conducted after 300 events, corresponding to the earliest time point for stopping the trial by demonstrating overwhelming evidence of efficacy or futility. This would have also corresponded to the minimum number of events required for conducting the economic evaluation. The futility boundaries were constructed as non-binding in order for the DMEC to over-rule a decision of stopping early for futility in the event that a futility boundary is crossed.

Interim analysis 2 was to be conducted after 445 events, corresponding to the number of expected events required for trial termination under futility (with 434 corresponding to the number of events required for demonstrating superiority or inferiority of APM to HSFM).

The final analysis was to be conducted after 588 events had occurred.

In addition, in the event of an early stopping signal for futility, an assessment of the value of continuing with the trial from the NHS decision-making perspective, via an expected value of sample information (EVSI) analysis, was also planned, to inform the deliberations of the DMEC.

Following the unplanned interim analysis and the extension to recruitment, only the final analysis was conducted.

Summary of main changes to the protocol

Throughout the trial, five substantial amendments to the protocol were submitted and approved by the REC. The important changes from a patient perspective were the collection of the QoL information and questionnaires as described in Data collection schedule. Prior to starting recruitment (August 2013), modifications to the eligibility criteria, collection of the NHS number and collection of PU-QoL-UI at baseline were added. Two additional objectives were added in an amendment approved in September 2013, to improve the analysis. These were to compare the:

• incidence of mattress change between patients using APM and those using HSFM

• safety between patients using APM and those using HSFM.

When and who performs the photography substudy data collection were also clarified. In February 2014, the amendment provided details of the EQ-5D-5L, PU-QoL-UI, SF-12, health resource use survey (1-week recall) and PU-QoL-P questionnaires and specified when they would be administered. In April 2014, an urgent safety measure led to the changes in questionnaire administration, Scottish centres were added, which required additional documents to comply with legislation for consenting participants who lacked capacity. The last substantial amendment, implemented in November 2015, made further changes to the frequency of questionnaire administration and incorporated other health-care professionals as well as nurses to collect data.

Participant flow

In total, 15,277 patients were assessed for study eligibility and 2030 randomisations took place between August 2013 and 30 November 2016. There were 41 NHS trusts/health boards comprising 47 centres that opened (i.e. met all local regulatory requirements), of which 39 NHS trusts/health boards and 42 centres recruited participants. Of the recruiting trusts/health boards, the numbers of participants randomised by each trust ranged from 1 to 471, with a median of 24 (Figure 1). The mean recruitment per trust was 1.6 participants per month. The recruiting centres/trusts comprised 25 teaching hospitals (21 acute NHS trusts and two health boards), 13 general hospitals (11 acute) and nine community hospitals (in nine community care NHS trusts). Others were combined acute and community trusts.

FIGURE 1.

Recruitment by organisation.

A Consolidated Standards of Reporting Trials (CONSORT) flow diagram of trial progress is presented in Figure 2. Of the 15,277 patients screened, 877 (5.7%) were not assessed for eligibility; the main reasons for this included ethically inappropriate to approach (n = 482, 55.0%), discharged (n = 157, 17.9%) or missed by the research team (n = 68, 7.8%). Of the 14,400 patients assessed for eligibility, 9323 (64.7%) were ineligible, with the main reasons including not at a high risk of PU development (n = 2180, 23.4%), length of stay expected to be < 5 days (n = 1640, 17.6%), patient (n = 938, 10.1%) or staff (n = 1116, 12.0%) unwilling to change mattress and patient too unwell to change mattress (n = 709, 7.6%). Of the 5077 eligible patients, 2068 (40.7%) consented and 2030 (40.0%) were randomised.

FIGURE 2.

The CONSORT flow diagram. ICU, intensive care unit; TVTM, tissue viability team member. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

The screened population and randomised populations were similar in respect of age, gender and ethnicity (see Appendix 4) with the exception of mattress type at screening. In terms of mattress type, in the screened population, 50% (n = 7640) had been allocated an APM or another high-technology device and 48.8% (n = 7462) had been allocated a HSFM or other low-technology mattress by ward staff, whereas in the randomised population a lower proportion of patients (n = 868, 42.2%) were allocated to APM or another high-technology device than to HSFM or another low-technology device by ward staff (n = 1149, 56.6%) (see Appendix 4, Table 68).

In addition, of the 1116 patients for whom staff were unwilling to change mattress, 78.9% (n = 881) were on APMs or other high-technology mattresses and 20.8% (n = 232) were on HSFM or other low-technology mattresses (see Figure 2). This contrasts with the 938 patients who were unwilling to change mattress, of whom 52.0% (n = 488) were on APMs or other high-technology mattress and 47.7% (n = 447) were on HSFM or other low-technology mattress (see Figure 2).

Of those patients randomised, 1017 (50.1%) were allocated to APM and 1013 (49.9%) to HSFM, with 81.5% of participants in each group receiving the allocated mattress within 48 hours. A total of 119 participants were withdrawn [62 (6.1%) of those were allocated to APM and 57 (5.6%) to HSFM] and 166 patients died between randomisation and withdrawal or the 30-day final follow-up period [82 (8.1%) of those were allocated to APM and 84 (8.3%) were allocated to HSFM]. One patient was inadvertently randomised twice and was withdrawn from trial participation as soon as this was identified. Data arising from the second randomisation were excluded from the analysis population; therefore, the analysis population comprises a total of 2029 participants (see Figure 2).

Baseline characteristics

Patient characteristics were balanced across the mattress groups and are detailed in Tables 2–6. In summary, the study population comprised largely elderly patients (median age 81 years, range 21–105 years); 55.2% (n = 1119) were female and 98.2% (n = 1992) were of white ethnicity. Participants were inpatients for a median of 7 days (range 0–388 days) prior to randomisation.

Participants were inpatients on medical (64.6%, n = 1310), orthopaedics and trauma (22.3%, n = 453), surgical (7.6%, n = 155) and other wards (5.2%, n = 106), including oncology, critical care, neurosciences, renal, gastroenterology and spinal injuries unit (Table 2). Participants were inpatients in secondary care hospitals (69.7%, n = 1414), community hospitals (18.7%, n = 379) and intermediate care/rehabilitation facilities (11.5%, n = 234).

In relation to morbidity and risk of PU development, 15.9% (n = 322) of patients lacked capacity, 44.8% (n = 909) had a history of falls in the preceding month (Table 3), a majority of 96.6% (n = 1961) had limitations to independent movement, and 98.7% (n = 2003) and 92.6% (n = 1879) of patients were classified as ‘at risk’ of PU development on the Pressure Ulcer Risk Primary Or Secondary Evaluation Tool (PURPOSE-T) and Braden Scale,31 respectively (see Table 3).

There were high levels of skin morbidity at baseline, which included 53.4% (n = 1084) of patients reporting pressure-related pain on one or more healthy, altered or category 1 PU skin sites (Tables 4 and 5). The ‘worst’ skin classifications recorded were PUs of category 2 in 7.1% (n = 145) of patients, PUs of category 1 in 11.6% (n = 235) of patients and alterations to intact skin in 66.4% (n = 1347) of patients. On a skin site level, of a total number of 28,421 skin sites, there were 177 (0.6%) PUs of category 2, 409 (1.4%) PUs of category 1, 7569 (26.6%) alterations to intact skin, 510 (1.8%) other conditions/wounds (see Box 1), 486 (1.7%) unable to assess (i.e. bandage/cast in situ), 104 (0.4%) amputations and 268 (0.9%) missing.

Pre-randomisation standard pressure ulcer prevention care

Prior to randomisation, preventative care interventions, including mattress, repositioning frequency, sitting, seating and use of adjuvant devices as provided by the attending ward staff, were balanced across the mattress groups and are detailed in Table 6.

Prior to randomisation, 868 (42.8%) participants had been allocated an APM or other high-technology mattress and 1149 (56.6%) had been allocated a HSFM or other low-technology mattress by ward staff. Overall, 89.8% (n = 1823) of patients required complete or partial assistance to reposition themselves; the frequency was observed to be every 2 hours and every 2–3 hours in 14.5% (n = 294) and 47.7% (n = 967) of patients, respectively, with a small number of patients (n = 116, 5.7%) repositioned less frequently than every 6 hours. Fewer than one-sixth of patients (14.0%, n = 284) patients received adjuvant devices or dressings, of whom 56.7% (n = 161) were provided devices for heel off-loading. One-quarter of the patients were confined to bed (26.7%, n = 541), and, of those who did sit out (73.1%, n = 1483), only 65.5% (n = 971) had a specialist cushion.

Outcomes: intention-to-treat population

Primary end point: time to development of new category ≥ 2 pressure ulcers to the 30-day final follow-up

A total of 160 (7.9%) patients developed at least one new PU of category ≥ 2 between randomisation and 30-day final follow-up, death or withdrawal (Table 7). An absolute difference between mattress groups of 2.0% was observed, corresponding to incidence rates of 6.9% (n = 70) in the APM arm and 8.9% (n = 90) in the HSFM arm. Of those patients who developed a PU of category ≥ 2, the median time to development was 18 days (range 2–86 days) in the APM group and 12 days (range 2–94 days) in the HSFM group. There was no evidence of a difference between mattress groups in terms of the primary end point in the unadjusted Gray’s test (p = 0.1148, Figure 3a) or the adjusted analysis (Fine and Gray model HR 0.76, 95% CI 0.56 to 1.04; exact p-value = 0.0890) (Table 8). The only factor statistically significantly associated with the primary end point was skin status (Wald p-value = 0.0057); specifically, patients with a pre-existing category 1 PU were more at risk of developing a PU of category ≥ 2 than patients who did not have a pre-existing PU (HR 1.83, 95% CI 1.17 to 2.87). Similarly, patients with a pre-existing category 2 PU had an increased risk of developing a new PU of category ≥ 2 compared with patients with no PU (HR 1.83, 95% CI 1.09 to 3.09). Overall, 11.4% (27/236) of patients with a pre-existing category 1 PU at baseline and 12.4% (18/145) of patients with a pre-existing PU of category ≥ 2 at baseline developed a new PU of category ≥ 2, compared with 7.0% (115/1648) of patients without a pre-existing PU at baseline.

TABLE 7 - Number (%) of patients developing a new PU of category ≥ 2 by the 30-day final follow-up, by covariate

The participants did not have any skin site assessments for which the end point could be derived.

Parts of this table have been adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Attribute	New PU of category ≥ 2				Not eligiblea	Total
	Yes	No
		No new PU of category ≥ 2	Died	Withdrawn
PU status at baseline
APM
No PU	49 (6.0)	682 (83.1)	59 (7.2)	30 (3.7)	1 (0.1)	821 (100)
Category 1 PU	11 (8.8)	92 (76.0)	12 (9.6)	7 (5.6)	0 (0.0)	125 (100)
Category 2 PU	10 (14.3)	48 (68.6)	6 (8.6)	3 (4.3)	3 (4.3)	70 (100)
Total	70 (6.9)	825 (81.2)	77 (7.6)	40 (3.9)	4 (0.4)	1016 (100)
HSFM
No PU	66 (8.0)	691 (83.6)	49 (5.9)	20 (2.4)	1 (0.1)	827 (100)
Category 1 PU	16 (14.4)	78 (70.3)	11 (9.9)	5 (4.5)	1 (0.9)	111 (100)
Category 2 PU	8 (10.7)	43 (57.3)	12 (16.0)	6 (8.0)	6 (8.0)	75 (100)
Total	90 (8.9)	812 (80.2)	72 (7.1)	31 (3.1)	8 (0.8)	1013 (100)
Consent type
APM
Written consent	44 (6.2)	595 (84.3)	38 (5.4)	26 (3.7)	3 (0.4)	706 (100)
Witnessed verbal consent	16 (10.6)	111 (73.5)	10 (6.6)	13 (8.6)	1 (0.7)	151 (100)
Consultee agreement	10 (6.3)	119 (74.8)	29 (18.2)	1 (0.6)	0 (0.0)	159 (100)
Total	70 (6.9)	825 (81.2)	77 (7.6)	40 (3.9)	4 (0.4)	1016 (100)
HSFM
Written consent	56 (8.0)	579 (83.0)	34 (4.9)	22 (3.2)	7 (1.0)	698 (100)
Witnessed verbal consent	16 (10.5)	123 (80.9)	7 (4.6)	5 (3.3)	1 (0.7)	152 (100)
Consultee agreement	18 (11.0)	110 (67.5)	31 (19.0)	4 (2.5)	0 (0.0)	163 (100)
Total	90 (8.9)	812 (80.2)	72 (7.1)	31 (3.1)	8 (0.8)	1013 (100)
Care setting
APM
Secondary care hospital	43 (6.1)	569 (80.1)	64 (9.0)	30 (4.2)	4 (0.6)	710 (100)
Community hospital	16 (8.4)	161 (84.3)	9 (4.7)	5 (2.6)	0 (0.0)	191 (100)
Intermediate care/rehabilitation	11 (9.6)	95 (82.6)	4 (3.5)	5 (4.3)	0 (0.0)	115 (100)
Total	70 (6.9)	825 (81.2)	77 (7.6)	40 (3.9)	4 (0.4)	1016 (100)
HSFM
Secondary care hospital	59 (8.4)	558 (79.0)	57 (8.1)	24 (3.4)	8 (1.1)	706 (100)
Community hospital	18 (9.6)	155 (82.4)	13 (6.9)	2 (1.1)	0 (0.0)	188 (100)
Intermediate care/rehabilitation	13 (10.9)	99 (83.2)	2 (1.7)	5 (4.2)	0 (0.0)	119 (100)
Total	59 (8.4)	558 (79.0)	57 (8.1)	24 (3.4)	8 (1.1)	1013 (100)
Pain on a healthy/altered/PU of category 1 at baseline
APM
Yes	35 (6.5)	453 (83.7)	30 (5.5)	23 (4.3)	0 (0.0)	541 (100)
No	33 (7.4)	359 (80.3)	38 (8.5)	17 (3.8)	0 (0.0)	447 (100)
Unable to assess	0 (0.0)	7 (46.7)	8 (53.3)	0 (0.0)	0 (0.0)	15 (100)
Missing	2 (15.4)	6 (46.2)	1 (7.7)	0 (0.0)	4 (30.8)	13 (100)
Total	70 (6.9)	825 (81.2)	77 (7.6)	40 (3.9)	4 (0.4)	1016 (100)
HSFM
Yes	55 (10.1)	438 (80.7)	31 (5.7)	18 (3.3)	1 (0.2)	543 (100)
No	34 (7.7)	358 (80.8)	38 (8.6)	13 (2.9)	0 (0.0)	443 (100)
Unable to assess	1 (6.7)	11 (73.3)	3 (20.0)	0 (0.0)	0 (0.0)	15 (100)
Missing	0 (0.0)	5 (41.7)	0 (0.0)	0 (0.0)	7 (58.3)	12 (100)
Total	90 (8.9)	812 (80.2)	72 (7.1)	31 (3.1)	8 (0.8)	1013 (100)
Condition affecting peripheral circulation
APM
Yes	17 (7.5)	183 (81.0)	17 (7.5)	8 (3.5)	1 (0.4)	226 (100)
No	53 (6.7)	641 (81.3)	60 (7.6)	32 (4.1)	2 (0.3)	788 (100)
Missing	0 (0.0)	1 (50.0)	0 (0.0)	0 (0.0)	1 (50.0)	2 (100)
Total	70 (6.9)	825 (81.2)	77 (7.6)	40 (3.9)	4 (0.4)	1016 (100)
HSFM
Yes	22 (9.6)	180 (78.6)	19 (8.3)	7 (3.1)	1 (0.4)	229 (100)
No	67 (8.6)	630 (80.9)	53 (6.8)	24 (3.1)	5 (0.6)	779 (100)
Missing	1 (20.0)	2 (40.0)	0 (0.0)	0 (0.0)	2 (40.0)	5 (100)
Total	90 (8.9)	812 (80.2)	72 (7.1)	31 (3.1)	8 (0.8)	1013 (100)
Overall	160 (7.9)	1637 (80.7)	149 (7.3)	71 (3.5)	12 (0.6)	2029 (100)

FIGURE 3.

Cumulative incidence functions. (a) Cumulative incidence function for development of a PU of category ≥ 2 by the 30-day final follow-up; (b) cumulative incidence function of development of a PU of category ≥ 2 in treatment phase; (c) cumulative incidence function for development of PU of category ≥ 1 by the 30-day final follow-up; and (d) cumulative incidence function for development of a PU of category ≥ 3 by the 30-day final follow-up. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

TABLE 8 - Time to development of a new PU of category ≥ 2 by the 30-day final follow-up

The p-value obtained from corresponding likelihood ratio tests for the effect of treatment is 0.0890.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Covariate	Incidence rate, n/N (%)	Reference level	HR		Wald p-value
			Point estimate	95% Wald confidence limits
Treatment
HSFM	90/1013 (8.9)	–	–	–	0.0890a
APM	70/1016 (6.9)	Vs. HSFM	0.76	0.56 to 1.04
Skin status
No PU	115/1648 (7.0)	–	–	–	0.0057
Category 1 PU	27/236 (11.4)	Vs. no PU	1.83	1.16 to 2.87
Category 2 PU	18/145 (12.4)	Vs. no PU	1.83	1.09 to 3.09
Consent type
Written	100/1404 (7.1)	–	–	–	0.3025
Witnessed verbal	32/303 (10.6)	Vs. written	1.34	0.90 to 1.99
Consultee agreement	28/322 (8.7)	Vs. written	1.23	0.79 to 1.91
Setting
Secondary care hospital	102/1416 (7.2)	–	–	–	0.6182
Community hospital	34/379 (9.0)	Vs. secondary care hospital	1.06	0.71 to 1.58
NHS intermediate care/rehabilitation facility	24/234 (10.3)	Vs. secondary care hospital	1.26	0.79 to 1.99
Pain on a healthy, altered or category 1 PU skin site
No	67/890 (7.5)	–	–	–	0.5070
Yes	90/1084 (8.3)	Vs. no	1.15	0.82 to 1.61
Unable to assess	1/30 (3.3)	Vs. no	0.38	0.05 to 2.94
Missing	2/25 (8.0)	Vs. no	2.02	0.43 to 9.45
Presence of condition affecting peripheral circulation
No	120/1567 (7.7)	–	–	–	0.5688
Yes	39/455 (8.6)	Vs. no	1.09	0.75 to 1.57
Missing	1/7 (14.3)	Vs. no	2.91	0.35 to 24.51

A total of 213 new PUs of category ≥ 2 were observed on 160 patients. Fewer PUs developed on patients allocated to APM (n = 89) than on patients allocated to HSFM (n = 124). The most common skin sites to develop new PUs were the sacrum (n = 44, 20.7%), left buttock (n = 42, 19.7%), right buttock (n = 38, 17.8%), left heel (n = 27, 12.7%) and right heel (n = 22, 10.3%). The frequency distribution of skin sites within mattress types were similar (see Appendix 4, Table 70).

Secondary end points

Time to development of pressure ulcers of category ≥ 1 to 30-day final follow-up

A total of 350 (17.2%) patients developed at least one new PU of category ≥ 1 between randomisation and 30-day final follow-up, death or withdrawal (see Appendix 4, Table 72). An absolute difference between mattress groups of 3.1% was observed, corresponding to incidence rates of 15.7% (n = 160) in the APM arm and 18.8% (n = 190) in the HSFM arm. Of those patients who developed a PU of category ≥ 1, the median time to development was 12 days (range 1–86 days) in the APM arm and 9 days (range 2–94 days) in the HSFM arm. There was no evidence of a difference between mattress groups in terms of the development of a PU of category ≥ 1 from randomisation to 30-day final follow-up in the unadjusted Gray’s test (p = 0.0908; see Figure 3c) or in the adjusted analysis (Fine and Gray model HR 0.83, 95% CI 0.67 to 1.02; exact p-value = 0.0733). Factors related to the development of a PU of category ≥ 1 included skin status (Wald p-value = 0.0301), consent type (Wald p-value = 0.0140) and pain on a healthy or altered skin site (Wald p-value = 0.0063); HRs and corresponding 95% CIs are provided in Table 10. The incidence rate in patients with no PU at baseline was 16.5% (272/1648), compared with 21.2% (50/236) in patients with a PU of category 1 and 19.3% (28/145) in patients with a pre-existing PU of category 2. In terms of type of consent, 21.4% (69/322) of patients with consultee agreement developed a new PU of category ≥ 1, compared with 19.5% (59/303) of patients with witnessed verbal consent and 15.8% (222/1404) of patients with written consent. The incidence of new PUs of category ≥ 1 was 16.0% (226/1416) in the secondary care setting compared with 17.7% (67/379) in community hospitals and 24.4% (57/234) in NHS intermediate care/rehabilitation facilities. The incidence rate in patients who did not have pain on a healthy or altered skin site at baseline was observed to be 15.6% (147/943), compared with 19.2% (198/1029) in patients who did have pain, 6.7% (2/30) in patients for whom pain could not be assessed and 11.1% (3/27) in patients for whom the pain status was missing.

TABLE 10 - Time to development of new PUs of category ≥ 1 by the 30-day final follow-up

The p-value obtained from corresponding likelihood ratio tests for the effect of treatment is 0.0733.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Covariate	Incidence rate, n/N (%)	Reference level	HR		Wald p-value
			Point estimate	95% Wald confidence limits
Treatment
HSFM	190/1013 (18.8)	–	–	–	0.0741a
APM	160/1016 (15.7)	Vs. HSFM	0.83	0.67 to 1.02
Skin status
No PU	272/1648 (16.5)	–	–	–	0.0301
Category 1 PU	50/236 (21.2)	Vs. no PU	1.52	1.11 to 2.09
Category 2 PU	28/145 (19.3)	Vs. no PU	1.18	0.79 to 1.75
Consent type
Written	222/1404 (15.8)	–	–	–	0.0140
Witnessed verbal	59/303 (19.5)	Vs. written	1.15	0.86 to 1.53
Consultee agreement	69/322 (21.4)	Vs. written	1.52	1.15 to 2.01
Setting
Secondary care hospital	226/1416 (16.0)	–	–	–	0.0970
Community hospital	67/379 (17.7)	Vs. secondary care hospital	0.95	0.72 to 1.26
NHS intermediate care/rehabilitation facility	57/234 (24.4)	Vs. secondary care hospital	1.35	1.01 to 1.82
Pain on a healthy or altered skin site
No	147/943 (15.6)	–	–	–	0.0063
Yes	198/1029 (19.2)	Vs. no	1.38	1.11 to 1.71
Unable to assess	2/30 (6.7)	Vs. no	0.28	0.07 to 1.15
Missing	3/27 (11.1)	Vs. no	1.31	0.40 to 4.36
Presence of condition affecting peripheral circulation
No	259/1567 (16.5)	–	–	–	0.3258
Yes	90/455 (19.8)	Vs. no	1.19	0.93 to 1.51
Missing	1/7 (14.3)	Vs. no	1.85	0.26 to 13.12

Time to development of pressure ulcers of category ≥ 3 to 30-day final follow-up

A total of 32 (1.6%) patients developed at least one new PU of category ≥ 3 between randomisation and 30-day final follow-up, death or withdrawal (see Appendix 4, Table 73). An absolute difference between mattress groups of 0.4% was observed, corresponding to incidence rates of 1.4% (n = 14) in the APM arm and 1.8% (n = 18) in the HSFM arm. Among those patients who developed a PU of category ≥ 3, the median time to development was 16 days (range 2–58 days) in the APM arm and 17 days (range 4–66 days) in the HSFM arm. There was no evidence of a difference between mattress groups in terms of the development of PUs of category ≥ 3 to 30-day final follow-up in the unadjusted Gray’s test (p = 0.5151; see Figure 3d); the adjusted analysis (Fine and Gray model) also showed no evidence of a difference between the APM and HSFM groups (HR 0.81, 95% CI 0.40 to 1.62; exact p-value = 0.5530). Factors related to development of PUs of category ≥ 3 included skin status (Wald p-value = 0.0288), consent type (Wald p-value = 0.0335), pain on a healthy, altered or category 1 PU skin site (Wald p-value < 0.0001) and presence of a condition affecting the peripheral circulation (p < 0.0001); HRs and corresponding 95% CIs are provided in Table 11. Incidence rates within all levels of covariates were low; the incidence rate was 1.3% for patients with no PU (22/1648) or a category 1 PU (3/236) at baseline compared with 4.8% (7/145) of patients with a pre-existing category 2 PU. In terms of consent, the incidence rate was observed to be 1.1% (16/1404) for patients who provided written consent, compared with 2.0% (6/303) for patients who provided witnessed verbal consent and 3.1% (10/322) for patients who were consented via consultee agreement. The presence of pain on a healthy, altered or category 1 PU skin site was statistically significant in the model, with an incidence rate of 0.0% for patients unable to report pain (0/30) and 8.0% for patients for whom the presence of pain was missing (2/25). The incidence rate for patients with pain was marginally higher, at 1.8% (19/1084), than for patients who did not have pain [1.2% (11/890)]. The presence of a condition affecting peripheral circulation was also statistically significant in the model, with an incidence rate of 0.0% (0/7) in patients for whom this factor could not be determined. The incidence rate in patients who had a condition affecting peripheral circulation was 2.2% (10/455), whereas it was 1.4% (22/1567) in patients who did not have such a condition.

TABLE 11 - Time to development of new PUs of category ≥ 3 by the 30-day final follow-up

The p-value obtained from corresponding likelihood ratio tests for the effect of treatment is 0.5530.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Covariate	Incidence rate, n/N (%)	Reference level	HR		Wald p-value
			Point estimate	95% Wald confidence limits
Treatment
HSFM	18/1013 (1.8)	–	–	–	0.5498a
APM	14/1016 (1.4)	Vs. HSFM	0.81	0.40 to 1.62
Skin status
No PU	22/1648 (1.3)	–	–	–	0.0288
Category 1 PU	3/236 (1.3)	Vs. no PU	0.85	0.24 to 2.98
Category 2 PU	7/145 (4.8)	Vs. no PU	3.20	1.33 to 7.71
Consent type
Written	16/1404 (1.1)	–	–	–	0.0335
Witnessed verbal	6/303 (2.0)	Vs. written	1.68	0.66 to 4.28
Consultee agreement	10/322 (3.1)	Vs. written	2.97	1.31 to 6.74
Setting
Secondary care hospital	26/1416 (1.8)	–	–	–	0.3045
Community hospital	3/379 (0.8)	Vs. secondary care hospital	0.43	0.13 to 1.41
NHS intermediate care/rehabilitation facility	3/234 (1.3)	Vs. secondary care hospital	0.61	0.18 to 2.10
Pain on a healthy, altered or category 1 PU skin site
No	11/890 (1.2)	–	–	–	< 0.0001
Yes	19/1084 (1.8)	Vs. no	2.00	0.93 to 4.32
Unable to assess	0/30 (0.0)	Vs. no	0.00	0.00 to 0.00
Missing	2/25 (8.0)	Vs. no	5.90	1.19 to 29.32
Presence of condition affecting peripheral circulation
No	22/1567 (1.4)	–	–	–	< 0.0001
Yes	10/455 (2.2)	Vs. no	1.49	0.70 to 3.15
Missing	0/7 (0.0)	Vs. no	0.00	0.00 to 0.00

A total of 40 new PUs of category ≥ 3 were observed on 32 patients. The number of category 3 PUs was comparable by trial arm (APM, n = 19, vs. HSFM, n = 21). The most common skin sites to develop the first new PU of category ≥ 3 were the sacrum (n = 8, 20.0%), left heel (n = 10, 25.0%) and right heel (n = 12, 30.0%). In terms of the location of new PUs of category ≥ 3, similar proportions were noted on participants in both the APM and the HSFM groups on the major torso skin sites (sacrum and buttocks) and heels (see Appendix 4, Table 74).

Healing of pre-existing pressure ulcers of category 2 to 30-day final follow-up

Of 145 patients who had a pre-existing PU of category 2, 70 (48.3%) were allocated to APM and 75 (51.7%) were allocated to HSFM. The healing rate was observed to be 62.9% (n = 44) for patients allocated to APM and 60.0% (n = 45) for patients allocated to HSFM, an absolute difference of 2.9%. Of the patients who healed, the median time to healing was 20 days (range 2–85 days) in the APM arm and 14 days (range 2–84 days) in the HSFM arm. There was no evidence of a difference between mattress groups in terms of healing by the 30-day final follow-up in the unadjusted Gray’s test (p = 0.7362; Figure 4), and this was reflected in the adjusted analysis (Fine and Gray model HR 1.12, 95% CI 0.74 to 1.68; exact p-value = 0.6122) for the comparison of APM with HSFM. The only factor related to the time to healing of a pre-existing PU of category 2 was the presence of a condition affecting the peripheral circulation (p = 0.0469); HRs and corresponding 95% CIs for all covariates entered into the model are presented in Table 12. The healing rate for patients who had a condition affecting peripheral circulation was 64.2% (68/106) and was 52.6% (20/38) for patients who did not have such a condition.

FIGURE 4.

Cumulative incidence function for healing end point.

TABLE 12 - Analysis model for time to healing of pre-existing PUs

The p-value from the corresponding likelihood ratio test was equal to 0.6122.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Covariate	Healing rate, n/N (%)	Reference level	HR		Wald p-value
			Point estimate	95% Wald confidence limits
Treatment
HSFM	45/75 (60.0)	–	–	–	0.5990a
APM	44/70 (62.9)	Vs. HSFM	1.12	0.74 to 1.68
Consent type
Written	63/102 (61.8)	–	–	–	0.9193
Witnessed verbal	14/23 (60.9)	Vs. written	1.08	0.65 to 1.81
Consultee agreement	12/20 (60.0)	Vs. written	1.12	0.57 to 2.19
Setting
Secondary care hospital	71/111 (64.0)	–	–	–	0.3093
Community hospital	8/20 (40.0)	Vs. secondary care hospital	0.55	0.26 to 1.18
NHS intermediate care/rehabilitation facility	10/14 (71.4)	Vs. secondary care hospital	0.91	0.44 to 1.86
Presence of condition affecting peripheral circulation
No	20/38 (52.6)	–	–	–	0.0469
Yes	68/106 (64.2)	Vs. no	0.59	0.36 to 0.97
Missing	1/1 (100.0)	Vs. no	0.56	0.31 to 1.04

Moderator analysis (exploratory analysis)

The treatment effects for each level of each covariate (risk factor) for the end points of time to development of a PU of category ≥ 2 to 30-day final follow-up (primary end point) and time to development of a PU of category ≥ 2 during the treatment phase are presented in Figures 5–8. The forest plots present the point estimate of the treatment effect, ln(HR), together with the corresponding 95% CI and, alongside the plot, the corresponding HR and 95% CIs. As this is an exploratory analysis, it is noted where there are observed differences in the estimate of the ln(HR) across the levels of each risk factor together with corresponding 95% CI, ln(HR) and corresponding upper 95% CI to the left of zero indicate a benefit in favour of APM rather than HSFM, whereas ln(HR) and lower 95% CI to the right of zero indicate the reverse.

FIGURE 5.

Forest plot of effect sizes within subgroups for the primary end point, by the following subgroups: skin status, mobility, nutrition, consent type. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

FIGURE 6.

Forest plot of effect sizes within subgroups for the primary end point, by the following subgroups: diabetic status, age group, sensory perception, macro- and microcirculation and moisture.

FIGURE 7.

Forest plot effect sizes within subgroups for the treatment phase end point, by the following subgroups: skin status, mobility, nutrition, consent type.

FIGURE 8.

Forest plot of effect sizes within subgroups for the treatment phase end point, by the following subgroups: diabetic status, age group, sensory perception, macro- and microcirculation and moisture.

Incidence rates observed within each level of risk factor are aligned with the conceptual framework. Although there is no evidence of differential treatment effects within risk factors at the 5% significance level, treatment effects from these exploratory analyses suggest that there may be a potential benefit of APM compared with HSFM in patients whose worst skin status was assessed as altered or category 1 PU, those with mobility limitations, those with a nutritional problem and patients who participated via consultee agreement. There is a suggestion that patients aged 75–84 years may benefit more from APM, and those aged ≥ 85 years are not observed to benefit from mattress type. The reason for this is not clear; however, these patients have a higher incidence rate and their treatment effect may be confounded by other correlated risk factors. Further work would be required to investigate this.

Mediator analysis (exploratory analysis)

Figure 9 presents repositioning frequencies by mattress allocation during the treatment phase for all patients in the ITT population for whom repositioning data are available. Throughout the treatment phase, ≥ 50% of patients with complete data in both the APM and the HSFM arms at each visit were repositioned every 2–3 hours or more frequently. However, the proportion of patients repositioned every 2–3 hours or more frequently appears to reduce during the treatment phase in both arms. Of those patients repositioned fewer times than every 2–3 hours, similar proportions in the APM and HSFM arms were positioned every 4–5 hours and every 6–7 hours or fewer during the period between visits 1 and 5, but during the period between visits 6 and 10, patients allocated to APM were repositioned less frequently than those allocated to HSFM (i.e. fewer APM patients were positioned every 4–5 hours and more were positioned every 6–7 hours or fewer times than patients in the HSFM arm).

FIGURE 9.

Repositioning during treatment phase by mattress allocation. a, Proportion calculated on a complete-case basis. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Per-protocol population

The PPP consisted of 1352 (66.6%) patients after removing 545 (26.9%) patients who did not achieve at least 60% compliance with their allocated mattress prior to developing a PU of category ≥ 2, or the end of the treatment phase, whichever occurred sooner. Patients were also excluded (reasons not mutually exclusive) if they were not at a high risk of PU development (n = 42, 2.1%), had a current or previous PU of category ≥ 3 (n = 8, 0.4%), were outside the permitted weight limits (n = 6, 0.3%), did not receive the mattress within 2 days of randomisation (n = 369, 18.2%) and/or the consent form was not received or the consent date was after randomisation (n = 13, 0.6%). Of the 1352 patients in the PPP, 663 (49.0%) were allocated to APM and 689 (51.0%) were allocated to HSFM. There was no evidence of a difference between mattress groups in the incidents of new PUs of category ≥ 2 (adjusted Fine and Gray model HR 0.79, 95% CI 0.54 to 1.16; p = 0.2249). Incidence rates of new PUs of category ≥ 2 by 30-day final follow-up were observed to be 7.2% (n = 42) in the APM arm and 8.4% (n = 58) in the HSFM arm. Incidents of a new PU of category ≥ 2 in the treatment phase sensitivity analysis were observed to be 5.7% (n = 38) in the APM arm and 7.4% (n = 51) in the HSFM arm, a difference of 1.7%, which was marginally significant, indicating some evidence of a difference between mattress groups (adjusted Fine and Gray model HR 0.76, 95% CI 0.32 to 1.00; p = 0.0508).

Introduction

Within-trial economic evaluations were carried out, together with a model-based analysis, to assess the cost-effectiveness of APM compared with HSFM in the prevention of PUs of category ≥ 2 in high-risk patients. The model was used to extrapolate the results over the expected lifetime of the trial participants. The methods used are in line with NICE guidance. 61

Objectives

• The primary objective was to assess the incremental cost-effectiveness of APM compared with HSFM in the prevention of PUs in high-risk patients at 30-day final follow-up (maximum 90 days post randomisation), from the perspectives of the NHS and PSS.

• The secondary objective was to assess the long-term incremental cost-effectiveness of APM compared with HSFM in the prevention of PUs of category ≥ 2 in high-risk patients, from the perspectives of the NHS and PSS.

• A tertiary objective was, in the event of an early stopping signal for futility, we would assess the value of continuing with the trial from the NHS decision-making perspective, via an EVSI analysis, to inform the deliberations of the DMEC. Although a futility boundary was not crossed, this analysis was undertaken twice at the request of the funder. The methods and results are presented in Appendix 3.

Methods

Lifetime decision-analytic model

A decision-analytic model was constructed to estimate the long-term incremental cost-effectiveness of APM compared with HSFM in the prevention of PUs of category ≥ 2 in high-risk patients. Originally, a patient-level simulation model was planned using individual patients defined in terms of age, gender and underlying health condition. However, the economic modelling undertaken as part of the interim trial analysis (see Appendix 3) showed a mean and median age of > 80 years. Given the age of the cohort and the associated likelihood of comorbidities, there was far greater homogeneity within the patient population than anticipated. Following discussion with the trial DMEC, a cohort model was used.

A Markov decision model was constructed using R software version 1.0.136 (The R Foundation for Statistical Computing, Vienna, Austria) (Figure 10). The model structure mirrors that used in the interim VOI analyses (see Appendix 3), but with some differences as the data retrieved from the trial allowed a more detailed patient pathway. The model begins at the point of randomisation and extends for the lifetime of the patients. The model describes patient progression over time through a pathway of health states, with movement between the health states being triggered by development of one or more PUs, healing of PU(s) or death. Compared with the VOI model, the long-term model had two starting points: in hospital PU free or in hospital with a PU. The model also added a stage to account for one PU developed in hospital. The long-term model also used final trial data compared with expectations or partial trial data. Figure 10 shows how patients can move between health states (indicated by the arrows). The model structure was developed in consultations with the clinical expert members of the trial team. It aimed to capture clinical practice according to the retrieved data. Resource use and costs are associated with each health state and patients accumulate costs and health benefits in each state over 3-day cycles. The outcome measure for the model was the QALY. The analysis adopts a lifetime horizon, truncated at 110 years.

FIGURE 10.

Lifetime decision-analytic model to compare APM with HSFM.

The model analysis was conducted from the perspectives of the NHS and PSS. Costs and outcomes were discounted at 3.5%. 61

Results

Within-trial analyses

Sample

The total sample size of the ITT analysis was of 2029 participants (APM arm, n = 1016; HSFM arm, n = 1013).

For the complete-case analysis, 267 participants (APM arm, n = 118; HSFM arm, n = 149) had completed the EQ-5D-5L at all four time points, and 233 had completed the PU-QoL-UI at all four time points (APM arm, n = 107; HSFM arm, n = 126). Figure 11 shows a flow diagram showing the number of complete cases by measurement for resource use and EQ-5D-5L.

FIGURE 11.

Flow chart of completed questionnaires by measurement. a, Patients who responded to all five dimensions of the EQ-5D-5L.

Costs

Tables 80–82 in Appendix 5 contain the average per-patient costs from baseline to 30-day final follow-up for those patients who completed the resource use questionnaires at all four time points. (It should be noted that all these patients were included in the ITT analyses, but only those who had also completed the EQ-5D-5L at all four time points were included in the complete-case analysis.)

The cost driver is, as might be anticipated, inpatient care, although the difference between the total mean cost of inpatient stays between the two arms of the trial is not statistically significant (£2810.08 for the APM arm and £2888.68 for the HSFM arm; p = 0.54). Overall, there are few statistically significant differences in costs across the two arms; however, the majority of costs are lower in the APM arm than in the HSFM arm.

Utility values

The utility values (see Appendix 5, Table 83) of both the EQ-5D-5L and the PU-QoL-UI show no statistically significant differences between the arms of the trial across each of the four time periods (baseline, week 1, week 3 and 30-day final follow-up).

Cost-effectiveness

The results of the within-trial analyses using QALYs derived from the EQ-5D-5L are shown in Table 15.

TABLE 15 - Results of within-trial and lifetime analyses

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Strategy	Total		Incremental		ICER (£)	NMB (£)	Probability of being cost-effective (£20,000 per QALY threshold)	Result
	Cost (£)	QALY	Cost (£)	QALY
Deterministic within-trial analysis (adjusted for baseline costs and QALYs)
APM	4482	0.128				–1918	–	Cost-effective
HSFM	4621	0.127	138	–0.0010	–136,171	–2077	–	Dominated
Probabilistic within-trial analysis (adjusted for baseline costs and QALYs)
APM	4533	0.128				–1979	0.99	Cost-effective
HSFM	4646	0.127	113	–0.0011	–101,699	–2114	0.01	Dominated
Deterministic lifetime analyses
APM	17,736	7.04				123,134	–	Cost-effective
HSFM	18,778	6.98	1042	0.055	–18,979	120,994	–	Dominated
Probabilistic lifetime analyses
APM	17,708	6.91				120,600	0.60	Cost-effective
HSFM	18,661	6.87	953	–0.046	–20,465	118,715	0.40	Dominated

Once baseline adjustments were made, the deterministic analysis shows that the mean total costs of APM and HSFM are £4533 and £4646, respectively, with mean QALYs of 0.128 and 0.127, respectively. Thus, APMs dominate HSFM, as APM has lower costs and higher QALY values. Similar results are seen in the probabilistic analysis in which APMs dominate HSFM (mean total costs of APM and HSFM are £4533 and £4646, respectively, and mean QALYs are 0.128 and 0.127, respectively). Within this analysis, APM has a 99% probability of being cost-effective at a threshold of £20,000. Figures 12 and 13 show the cost-effectiveness plane and CEAC.

FIGURE 12.

Cost-effectiveness plane. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

FIGURE 13.

Cost-effectiveness acceptability curve. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

The sensitivity analyses (see Appendix 5, Tables 84–88) show mixed results. The complete-case analysis showed, unlike the previous analyses, that HSFM was cost-effective. An additional analysis was undertaken using only complete cases for the baseline and 30-day final follow-up. This analysis increased the number of complete cases to 934 (APM arm, n = 460; HSFM arm, n = 474). When baseline adjustment was made, the results here showed that HSFM was the most cost-effective; conversely, when no baseline adjustment was made, APM was the cost-effective strategy (see Appendix 5, Tables 89 and 90). ITT analysis without baseline adjustment found that APM was the cost-effective alternative. The results of the sensitivity analyses using QALYs derived from the PU-QoL-UI similarly found differing results (see Appendix 5, Tables 87 and 88).

Decision-analytic cost-effectiveness model

Within the lifetime cost-effectiveness model, APM dominated HSFM with lower costs and higher QALY gains in both the deterministic and the probabilistic analyses (see Table 15). The probabilistic analysis showed a probability of cost-effectiveness of 60%. The cost-effectiveness plane and CEAC are shown in Figures 14 and 15. The sensitivity analyses (see Table 15) show that the results are robust to changes in the main costs and utility variables.

FIGURE 14.

Cost-effectiveness scatterplot (lifetime analyses).

FIGURE 15.

Cost-effectiveness acceptability curve (lifetime analyses).

The sensitivity analyses (see Appendix 5, Table 91) all show APM to be cost-effective with lower costs and higher QALY values.

Summary

The value for money of the two mattresses showed APM to be cost-effective over both the short and the long term. However, the results were not robust; differences in costs between the two were modest and the differences in the measure of effectiveness were very small.

Background

Currently, there is no objective laboratory measure for the diagnosis of a PU. The accepted ‘gold standard’ is expert clinical examination of the skin areas at risk of PU development. 8,33 Owing to the visible differences between the two interventions (APM and HSFM) in this study, using the ‘gold standard’ for the end-point assessment meant that blinded assessment of the skin was not achievable. It is possible that the trial primary end point could be misreported, either intentionally or not, by clinical research staff. There is also inherent subjectivity with clinical assessments and misclassification of the skin may also occur. 52,79,80 If there is a risk that estimates of the treatment effect will be biased, then the findings of the trial may not be reliable.

To assess and quantify any potential bias through estimates of over- and under-reporting of PUs, one approach would be to take photographs of all skin sites and assess them blinded. The problem with this would be that taking so many photographs would be a burden to patients and staff and raise concerns about patient consent.

Other issues, such as image quality in ambient lighting conditions and the reliability of the identification of a category 2 PU using photographic evidence,79,80 were of concern.

These issues have been only partially explored previously. At the trial design stage, only two studies were found that report the inter-rater reliability of photography assessed by different clinical experts and photography versus clinical assessment. One study compared experts’ photographic assessments, reporting levels of agreement between two groups of experts assessing photographs to classify PUs. 80 The other study compared the ‘gold standard’ expert clinical nurse assessment against assessment using photographs. 79 Although inter-rater agreement was high in both, the former study reported a high proportion of poor-quality images that could not be assessed80 and, in the latter study, half of the PU photographs were of category ≥ 379 and, therefore, not comparable with the primary end point of ‘early’ category 2 PUs.

Other technological solutions were also considered including laser Doppler, light spectroscopy and multispectral imaging, but these detect erythema and the intensity of skin blood flow and are unable to assess the presence of a category 2 PU. 81

Therefore, recognising the need to establish a method for blinded outcome assessment and respond to the funding body’s request, a validation substudy was undertaken in order to address both scientific and practical questions.

Aims and objectives

The main aim of the photographic substudy was to assess the feasibility of using blinded expert central photographic review to quantify potential bias in the reporting of the PRESSURE 2 trial primary end point.

The primary objectives of the substudy were, therefore, to assess:

• over-reporting of PUs of category ≥ 2

• under-reporting of PUs of category ≥ 2.

The secondary objectives of the sub study were to assess:

• rates of consent/potential impact on trial recruitment

• acceptability to patients

• compliance with photographs (i.e. whether or not the intended number of photographs were actually taken)

• compliance with secure transfer of photographs between the research site and the CTRU

• the quality of photographs and confidence of photographic review.

Methods

During the recruitment process, participants were asked to consent to photography of their skin sites; this was an optional consent and did not preclude participation in the trial. Participants were able to opt out of photographs at any time.

Results

A total of 2029 patients were recruited into the PRESSURE 2 trial. A total of 177 PUs of category 2 were observed among 145 patients at baseline, and 213 PUs of category ≥ 2 from 160 patients were observed to develop during the trial; therefore, a total of 390 photographs of PUs of category ≥ 2 were expected. A total of 248 photographs were received.

A total of 264 patients were selected by the 24-hour system to be part of the sample selected for assessment by the independent clinical assessor and a maximum of 528 photographs were expected (two photographs per patient). Only 167 (63.3%) patients actually had a substudy assessment completed, of whom 142 (85.0%) were reported to have had photographs taken. In total, 284 (53.8%) photographs were returned as part of the substudy from 137 (51.9%) patients.

Conclusion

Overall, the reliability of photography as a tool for blinded expert central assessment in PU trials is ‘very good’ (according to the PABAK statistics). There is potential imbalance in terms of the over-reporting of PUs between the arms, with an indication that PUs may have been more likely to be over-reported in the HSFM arm; however, the CIs for the level of agreement for each group overlap and this needs to be considered alongside a potential imbalance in terms of the return of photographs of PUs of category ≥ 2 with a lower return rate for the APM arm and the diagnostic uncertainty associated with blinded expert central photographic review.

Overall, there was ‘very good agreement’ (according to the PABAK statistics) between the independent clinical assessor and CRN/P in their clinical assessment.

Similarly, there may have been under-reporting of PUs of category ≥ 2 by the independent clinical assessor compared with the blinded expert central photographic review, but in both of these cases, the sample size was too small to assess under-reporting any further.

There were varying levels of confidence from the blinded expert reviewers of the photographs, with more confidence demonstrated in skin sites assessed as healthy, altered or category 1. In only a very small number of photographs could no assessment be made.

Background

Patient-centred care requires that interventions be evaluated in terms of patient experience, with patient-reported outcomes (PROs), including HRQoL. A widely accepted definition of HRQoL, useful for clinical trials and health services research, states that:

HRQoL is a multidimensional construct encompassing perceptions of both positive and negative aspects of dimensions, such as physical, emotional, social, and cognitive functions, as well as the negative aspects of somatic discomfort and other symptoms produced by a disease or its treatment.

Reproduced from Osoba89

Integral to this definition is that HRQoL is a subjective phenomenon, so the patient’s assessment is preferred to that of a proxy, and it is multidimensional, including core domains plus symptoms that will differ across diseases, treatments and health interventions. 89,90

Another term closely related to HRQoL is PRO. A PRO is ‘any report of the status of a patient’s health condition that comes directly from the patient, without interpretation of the patient’s response by a clinician or anyone else’. 91 The term PRO emerged to solve the difficulty of finding a universal definition for HRQoL and related concepts. It does not tell what is being measured, only that the patient is providing the data. PROs can be symptoms (e.g. pain, fatigue), aspects of functioning (e.g. physical, psychological, social) and/or multidimensional constructs, such as HRQoL.

Pressure ulcers and interventions aimed at preventing or treating PUs can affect HRQoL in many ways, including pain,92 exudate and odour, and can compromise all areas of patient functioning. 1,93 The presence of symptoms and impaired functioning can have a distal effect on HRQoL outcomes. 94 Intensive interventions for preventing and treating PUs pose added patient burden and further affect HRQoL. 12 This additional impact on patients is related to increased care burden, prolonged rehabilitation, requirement for bed rest, and hospitalisation. 1,47 In this clinical context, evaluating PROs such as symptoms, functioning and HRQoL is particularly important and relevant, and there is enormous potential for PROs to be integral to patient management and recommendations for PUs.

The need for evaluating health-related quality of life in patients at risk of developing pressure ulcers

The possible burden of symptoms and impacts on HRQoL associated with interventions for preventing and treating PUs provide compelling arguments for incorporating the quality of patients’ lives into decisions about PU prevention and management. The PU field is reliant on health-outcome assessment to provide a strong evidence base, incorporating the patient perspective.

Methods of assessing health-related quality of life in patients with pressure ulcers

A standardised approach to assessment of HRQoL is required to reduce variation in both the way questions are asked and how patients respond. For this reason, questionnaires with standard questions about relevant issues incorporating a standard set of response options are used. These are referred to as PRO instruments or measures.

Patient-reported outcome instruments are increasingly used in clinical studies for measuring outcome variables. In this role, PRO instruments are the central dependent variables on which prevention and treatment decisions are made. They can be useful tools for evaluating health changes following interventions if they are fit for purpose and accord with international standards for rigorous measurement. 12,95 PRO instruments specific to PUs could help improve the evidence base through use in research assessing the clinical effectiveness of PU therapies, facilitate clinician–patient communication and shared decision-making, prioritise patient problems and preferences, monitor changes or outcomes of treatment and measure the performance of health-care providers and services, and could be used in clinical audit. 91,95,96

Previous work by this research team has identified PROs that are important to people with PUs,1,48 established the need for patient-reported measures of outcomes specific to PUs,97 and developed and evaluated a PRO instrument to assess PU-specific symptoms and functioning impacts (the PU-QoL instrument). 98 The PU-QoL instrument is researcher administered and comprises three symptom scales, pain (eight items), exudate (eight items) and odour (six items), and seven function scales: four physical functioning – sleep (six items), movement and mobility (nine items), daily activities (eight items) and vitality (five items); two psychological well-being: emotional well-being (15 items) and self-consciousness and appearance (seven items); and one social participation (nine items), plus a single item for itchiness and a single item for global QoL. It is intended for patients who have any category of PU. Patients rate the amount of ‘bother’ attributed ‘during the past week’ on a three-point response scale (0 = not at all to 2 = a lot). Scale scores are generated by summing items and then transforming to a 0–100 scale. High scores indicate greater patient bother. The PU-QoL instrument has been validated for use with patients with PUs and is most appropriate for people with severe PUs, as demonstrated by a lack of items to represent people with little or no bother due to PUs. Given the heterogeneity of the PU population, further work was required to ensure that the PU-QoL instrument scales met the needs of all people with PUs, including patients with superficial PUs, as well as those at risk of PU development and to test whether the PU-QoL instrument scales were responsive to change. The usefulness of a new PRO instrument is demonstrated by multiple applications in different studies and strengthened by an accumulative body of evidence to support scale measurement properties. 97

Aims

In order to enable assessment of PROs in patients receiving preventative interventions who are at risk of developing PUs, the aims of this study were to:

1. modify the PU-QoL instrument for use with patients receiving preventative interventions who are at a high risk of PU development – the PU-QoL-P

2. undertake a comprehensive evaluation of the psychometric measurement properties of the PU-QoL-P instrument to ensure that it is acceptable, reliable, valid and responsive, and suitable for use in the UK health setting.

This would provide a standardised method for evaluating patients’ self-reports of the impact of PU preventative interventions on HRQoL outcomes.

Methods

Results

The final Pressure Ulcer Quality of Life – Prevention instrument

The final PU-QoL-P instrument is a researcher-administered PRO instrument, comprising three symptom scales [pain (12 items), exudate (8 items) and odour (6 items)]; six function scales: four physical functioning [sleep (7 items), movement and mobility (9 items), daily activities (5 items) and malaise (4 items)] and two psychological well-being [emotional well-being (15 items) and self-consciousness and appearance (7 items)]; and three single items for itchiness, appetite and global QoL. Patients rate the amount of ‘bother’ attributed ‘during the past week’ on a three-point response scale (e.g. 0 = not at all to 2 = a lot). Scale scores are generated by summing items and then transforming to a 0–100 scale. High scores indicate greater patient bother. The PU-QoL-P instrument is intended for interview administration, following a user manual, but could be self-completed by patients, depending on their preference and ability. It is suitable for use with any adults at a high risk of PU development receiving preventative interventions in the UK acute and community health-care settings. Scales can be selected depending on the nature of the research. For example, the exudate and odour scales are not intended for people at risk of PU development or with superficial category 1 PUs. Electronically defined ‘skip’ questions have been added to assist in selecting scales relevant to each individual’s circumstance or the exudate and odour scales could be excluded in future prevention trials.

Discussion

The PU field requires a strong evidence base that incorporates assessment of PROs. To fully capture and quantify patients’ perspectives, appropriately developed and validated PRO instruments are required. The PU-QoL-P instrument mostly satisfies criteria for reliability, validity and responsiveness in line with recommended Food and Drug Administration (FDA) guidelines for PRO instruments for use in clinical trials. 91 The ITCs, alpha coefficient and homogeneity coefficient (IIC mean and range) provide evidence towards the reliability and internal construct validity of the PU-QoL-P scales. The results of the factor analysis mostly supported the use of the items, as hypothesised, into six function scales, as was suggested by the results from the original analysis. Some modifications were made to four scales, which were supported by the retrospective analyses, and the changes made are considered conceptually reasonable. However, as changes were made to the scales, some might argue that a further set of data should be collected for further validation purposes. No one test confirms validity; rather, validation of a PRO instrument is an ongoing process, with the accumulation of clinical validation data building a case for a particular instrument functioning effectively in a particular population for a specific purpose. 103

These findings contribute evidence towards support that people with category 2 PUs experience worse symptoms and functioning outcomes than those without PUs. People who are also physically limited and have PUs experience worse mobility outcomes and ability to participate in daily activities than those who have no or only slight physical impairment. PUs are often a secondary comorbidity and a consequence of the primary condition that a patient may be experiencing (e.g. immobility due to prolonged bed rest following extensive surgery). PUs contribute additional impairment in physical functioning outcomes beyond those experienced from other comorbidities. Furthermore, the exploratory hypothesis testing suggests that patients with torso PUs experience worse symptoms and more mobility problems, whereas patients with limb PUs have more problems with sleep quality, daily activities and malaise, have lower emotional well-being and feel more self-consciousness.

Owing to the small sample sizes in the hypothesised known groups, definitive conclusions cannot be made. However, trends were observed in scores in the right direction and, in some scales, moderate to high effect sizes, even though some were not statistically significant. A small sample size affects the standard error, so large CIs might be expected; however, sample size does not affect means, SDs or effect size. Therefore, effect sizes are still relevant and informative, even if non-significant correlations are observed, which, in this case, may be attributed to sample sizes being too small to detect significant differences.

A limitation of the study was scale-to-sample targeting: mean scores were below scale mid-points and all scales exceeded the 10% criterion for floor effects. However, given that it was intended to recruit an at-risk population (i.e. few people had category 1 or 2 PUs at baseline, and none had a category ≥ 3 PU), this finding is expected. The floor effects indicate more homogeneity in the sample than is representative of the general PU population. However, this study sample is representative of a high-risk PU population, who may be experiencing pressure area-related pain but do not experience symptoms associated with severe PUs such as exudate and odour, and the PU-QoL-P version is intended for prevention trials.

The above limitations do not preclude use of the PU-QoL-P instrument. PU-QoL-P scales can be included as one outcome measure, among others, for group comparisons in future PU prevention research (e.g. clinical trials). Work is under way to develop a short form and to test its clinical utility for use in clinical practice. As the PU-QoL-P was developed and evaluated in the UK, the validity and reliability are characteristics of the instrument for a specific population (i.e. UK nationals). A language translation or cross-cultural adaption may be required to ensure that the PU-QoL-P is appropriate for cultures, languages and ethnic groups outside the UK (see the PU-QoL-P instrument website104 for guidance on language translation and cross-cultural adaptation processes). Further research is also needed to investigate self-complete administration mode, and the development of proxy measures and language translations given the prevalence of cognitively impaired patients with PUs. 33,52 Finally, given the small sample sizes for the known-groups and responsiveness analyses, these analyses could be repeated in larger samples. The process of modifying a newly developed PRO instrument is part of an evolving, ongoing measurement process intended to strengthen the hypothesised conceptual relationships with empiric evidence. 97 The usefulness of new measures is therefore demonstrated by multiple applications in different studies (accumulative body of evidence to support scale measurement properties).

Conclusions

The PU-QoL-P instrument provides a means of comprehensively assessing of PU-specific PROs and for quantifying the benefits and harms of PU preventative interventions from a patient’s perspective, thus far lacking in the area. PRO assessment needs to become more commonplace in the PU field so that the goal of PU prevention and management can be to enhance and maintain the HRQoL of people at risk of or with PUs. The PU-QoL-P is a tool with which to evaluate whether or not PU preventative interventions and the health care given achieve this, outcomes that are ultimately best judged by patients themselves.

Summary of findings

Clinical interpretation

Interpretation of the results is complex, in terms of implications for clinical practice, for a number of reasons. First, the event rate was much lower than expected and all differences observed were small. A key question for practitioners is whether or not the overall group differences of 2% in the primary end point and the 2.6% group difference in the treatment phase sensitivity analysis is clinically important when PU incidence rates are low (7.9% and 6.5%, respectively). The difference equates to a number needed to treat of 50 patients for the primary end point and 38 for the treatment phase end point. That means that, in the treatment phase, for every 38 patients allocated an APM, one patient will benefit.

Second, > 10% of participants allocated to APM either refused to change to the APM after randomisation or requested a mattress change because of comfort (118/1017, 11.6%) compared with 3.9% (40/1013) of participants allocated to HSFM. After allocation of mattress, 22.1% (49/222) of participants changed from APM for the first time to aid movement/rehabilitation (patient or ward led) compared with 2.3% (5/220) of participants who changed from HSFM for the first time for the same reason.

Third, data suggest that frequency of repositioning reduced over time and, as the treatment phase progressed, participants allocated to APMs were repositioned less frequently than participants allocated to HSFM (see Figure 9).

A key question by members of PURSUN during a results interpretation event was ‘given the low incidence and the disadvantages of APMs in terms of impact on independent movement and comfort, who will benefit most from APMs?’. The moderator analysis (see Figures 5–8) was included in order to explore the potential benefit of each mattress on patients with known PU risk factors. The analysis suggests that patients who benefit most from APM versus HSFM appear to be those who are completely immobile, have altered skin or PU of category 1 at baseline, have a nutritional problem and those who lack capacity and participate through consultee agreement. In a PU conceptual framework, developed on the basis of epidemiological evidence, immobility, skin condition and tissue perfusion are categorised as direct causal factors, whereas nutritional deficit is classified as an indirect causal factor that impacts on skin condition and tissue perfusion. 9 Lack of capacity is not specified in the conceptual framework but was discussed during its development45 because it was hypothesised that it is a surrogate marker for both direct (immobility) and indirect (sensory perception and response) causal factors affecting repositioning self-care, stimulus to move and ‘compliance’ with repositioning. 45 Direct application of this analysis to practice must be undertaken with caution as it was exploratory, but the results suggest that the impact of altered and category 1 skin status, complete immobility, nutritional deficits and the vulnerability afforded by lack of capacity may be modifiable as risk factors through the use of the APMs, providing marginal gains over HSFMs.

Low event rate

The overall event rate of category ≥ 2 PUs of 7.9% for this participant population was considerably lower than the 2011 grant application sample size estimate, based on a PU incidence rate of 20.5% (APM 18% vs. HSFM 23%), determined by consideration of a range of studies27,105 and two relatively contemporary research studies involving this patient population. 33,38 PRESSURE 133 recruited inpatients between 2001 and 2004 and was a RCT comparing APM overlays with APM replacements (i.e. all participants received an APM) and the incidence of new category 2 PUs was 10.5% overall and 17.7% in acute patients. The pain cohort study38 recruited acutely ill hospital and community patients between 2009 and 2011 and reported that 25.2% of patients developed a new category 2 PU.

A key question regarding the lower than anticipated PU incidence rate is whether this was because the patient population was ‘low risk’ owing to issues around selection bias and equipoise and/or whether it reflects general improvements in clinical practice resulting from national-level PU improvement targets.

In terms of a low-risk population, the screened and randomised populations were similar in respect of age, setting, gender and ethnicity (see Appendix 4, Tables 68 and 69). Of the 15,277 patients screened, the most unwell 1623 (10.6%) were excluded (i.e. it was ethically inappropriate or they felt poorly or unwell); however, the patient population were all acute inpatient admissions, characterised by old age (mean 78.0 years, median 81 years), high levels of pre- and post-randomisation falls (44.8% and 15.3%, respectively) and being at risk on the Braden Scale31 (92.6%) and the PURPOSE-T (98.7%). In addition, adverse skin status at baseline included 11.6% of participants with an existing category 1 PU, 7.1% with a category 2 PU and 53.4% with pain on a healthy, altered or category 1 skin site.

In comparison with the pain cohort population, who had a PU event rate of 25.2%, the study populations were similar in age (pain cohort mean age 77.3 years, median age 80 years). The PRESSURE 2 trial had a higher proportion of participants assessed as being at risk on the Braden Scale31 (pain cohort 72.9%), but a considerably lower proportion of participants with an adverse skin status at baseline (pain cohort 35.5% with an existing category 1 PU, 23.3% with a category 2 PU, 4.1% with a category ≥ 3 PU and 77.1% with pain on a healthy, altered or category 1 skin site). 38

In comparison with the PRESSURE 1 trial combined acute and elective participant population (who had an overall PU incidence of 10.5%), the PRESSURE 2 trial recruited a slightly older population (PRESSURE 1 trial: mean age 75.2 years, median age 76 years), a slightly lower proportion of participants with a category 1 PU at baseline (PRESSURE 1 trial: 15.6%) but a higher proportion of participants with a category 2 PU (PRESSURE 1 trial: 5.7%). 33 A key difference between the two trials is that the PRESSURE 1 trial recruited participants within 24 hours of admission, whereas an unexpected population characteristic in the PRESSURE 2 trial was that the pre-trial duration of hospital stay was a mean of 13 days (median 7 days, range 0–388 days). As the incidence of PU development is highest in the first 2 weeks following admission, the stage at which participants were recruited may have affected PU event rates. However, it is also noteworthy that, in contrast to other prevention trials,27,30 in the PRESSURE 2 trial a small but clinically important minority of 32 (1.6%) participants developed 40 new category 3 PUs.

In terms of mental capacity, screening data from the PRESSURE 1 trial and the pain cohort suggests that the screened population in this study may not be representative of patients at high risk of developing PUs. The PRESSURE 1 trial reported that 39.7% of participants were unable to provide informed consent and in the pain cohort, 24.1% of the screened population were reported to lack capacity, compared with 18.7% of those eligible for the PRESSURE 2 trial. 33,38 However, the actual proportion of participants recruited through consultee agreement was higher in the PRESSURE 2 trial (15.9%) than in the previous study in which consultee agreement was obtained (PRESSURE 1 trial: 4.4%33,52). The inclusion of participants who lack capacity is important, and it was argued that, as there is an element of self-care in repositioning, it was important to include participants who lacked capacity to ensure generalisability. This was evidenced in the results that highlight the poorer outcomes for participants who lack capacity than for participants with capacity in terms of development of PUs of category ≥ 1 (see Table 10) and category ≥ 3 (see Table 11) and for the development of PUs of category ≥ 2 in those allocated to HSFM (see Table 8 and Figures 5–8). It is also noteworthy that participants who provided witnessed verbal consent had a higher incidence of PUs of categories 1, 2 and 3 than those who provided written, informed consent (see Tables 8, 10 and 11); it is suggested that this is also a surrogate indicator of reduced mobility.

From an equipoise perspective, although there were similar numbers of screened patients allocated to APMs and HSFM by ward staff (n = 7640, 50.0%, and n = 7462, 48.8%, respectively), more patients were excluded as the patient or ward staff did not want to change from their existing APM. There was a greater imbalance in the randomised population, with pre-randomisation allocations by ward staff of 42.8% to APM and 56.6% to HSFM (see Appendix 4, Table 68). However, this is only marginally lower than the 48.5% APM allocation by ward staff in the pain cohort participant population.

In terms of impact of general improvements in practice resulting from national targets, these are difficult to elicit from national monitoring such as the Safety Thermometer,106 because of problems of data accuracy4 and difficulties in interpretation of adverse event data. 107,108 However, a key difference between the PRESSURE 2 trial event rate and the pain cohort is that, not only are there fewer participants with an adverse skin status at baseline, but the proportion of participants with a category 1 PU at baseline who subsequently developed a new category 2 PU was much lower in the PRESSURE 2 trial (11.4%) than in the pain cohort study (36.3%).

Overall, the conclusion drawn is that the patient population was indeed characterised by high risk and that the low incidence rate was a result of the prevailing improvements in PU prevention care in the inpatient settings that participated in the trial.

Sample size

A maximum of 588 events (patients developing a new category ≥ 2 PU), corresponding to 2954 participants, were required for the study to have 90% power to detect a difference of 5% in the incidence of category ≥ 2 PUs between APM and HSFM, assuming an incidence rate of 18% on APM and 23% on HSFM, two-sided significance level of 5% and accounting for 6% loss to follow-up. The 5% absolute difference translated to a 24.4% relative difference (absolute difference of 5%/overall event rate of 20.5%) and a hazard ratio of 0.759.

The trial recruited participants more slowly than anticipated originally, leading to a smaller sample size than the planned maximum sample size (2030 compared with 2954), and was therefore underpowered for detecting an absolute difference of 5% at an overall event rate of 20.5%. In addition, the overall event rate was also far lower than originally anticipated (7.9% compared with 20.5%).

Two extension request scenarios were submitted. The first was a fully costed extension request of 14 months to continue recruitment to demonstrate superiority of either mattress for the revised clinically relevant difference of 3.3%, centred on an event rate of 10% with at least 80% power, and to estimate the treatment effect under the assumption of futility with a clinically meaningful improvement in precision. The second was a 6-month extension (requiring no additional HTA programme funding) with the expectation that at least 1996 patients would be recruited and that, although underpowered, it would improve the precision of the estimated treatment effect by at least 3% compared with the trial stopping at the originally planned timescale. The HTA programme approved the 6-month (no additional funding) extension rather than the fully costed extension on the basis that the latter was not seen to be value for money, taking into account the low improvement in precision.

After the interim analysis, the TMG was informed that the event rate was lower than expected and a range of scenarios were considered to inform further discussions with the oversight committees. The TMG noted that the difference relative to the overall event rate was also important when considering the event rates rather than just the absolute difference. Scenarios of the overall event rate centred at 5%, 10% and 15% were considered. The TMG came to a consensus that a relative difference of 33.3% centred on an overall event rate of 10%, corresponding to an absolute difference of 3.3%, was considered to be clinically meaningful and the preferred minimum clinically relevant difference. If the overall event rate was 5%, a relative difference of 50%, corresponding to an absolute difference of 2.5%, was considered to be clinically meaningful and the preferred minimum clinically relevant difference. The TMG also concluded that, with an overall event rate of 5%, a relative difference as small as 40%, corresponding to an absolute difference of 2%, was not considered to be of clinical relevance and that, at these event rates, the decision over which type of mattress to manage a patient on would be based on patient choice, cost and clinical judgement, including factors such as rehabilitation need and self-care considerations.

A HR of 0.76 was observed in the primary analysis of the primary end point, which is in line with the expected HR in the original sample size. However, as the corresponding relative difference was 25.3%, centred on the overall event rate of 7.9%, the a priori discussions with the TMG suggest that this is not considered to be a clinically relevant difference.

In the treatment phase sensitivity analysis, the estimate of the HR was 0.66; the corresponding relative difference of 40%, centred on an overall event rate of 6.5%, may be considered clinically relevant. However in light of the a priori discussion with the TMG, as the overall event rate is low, mattress provision should also be guided by patient choice, clinical opinion and consideration of risk factors for which APMs may have a potential benefit, for example patients with altered or category 1 PU skin, patients who are completely immobile or lack capacity to consent or those with nutritional problems.

Risk factors

The fact that risk factors were found to be predictive of PU development is in line with previous work and adds to the growing body of evidence that a key risk factor in immobile patients is skin status. 8,18 In all adjusted analyses, the presence of a category 1 or 2 PU at baseline was statistically significant in the model (see Tables 8, 10 and 11).

Other factors important in determining outcome in the adjusted analysis for both the category 1 and 3 end points included type of consent and the presence of pain on a healthy, altered (or category 1) skin site. In addition, setting was found to be important in the category 1 model and presence of condition affecting the peripheral circulation in the category 3 model (see Tables 10 and 11).

Usual care

A strength of the study is that data were collected in order that standard care (in terms of efforts made to minimise exposure to mechanical load) in the study population could be characterised. This is the first mattress RCT to provide a detailed description of standard care provision for the patient population27 and will support interpretation and wider application of study findings.

It is noteworthy that 99.6% of participants had an electric profiling bed, as this was a requirement of the trial, and in a population at baseline for which 42.8% of participants were allocated an APM, approximately one in six participants were repositioning more frequently than every 2 hours, whereas 1 in 20 participants were repositioned less frequently than every 6 hours (which is not in line with good practice recommendations). 8,12 Moreover, although one in four participants were confined to bed, one in five of the study population sat out in a chair for > 8 hours a day and 27.6% of those sitting out had a standard chair with no specialist cushion. Adjuvant devices (such as heel off-loading devices and dressings) were used in one in seven participants. A limitation of the data is that the time durations for repositioning and sitting were estimated from a combination of information including patient self-report, staff report, health-care records and observation, and the reliability has not been assessed.

Standard care interventions were also recorded during the treatment phase of the study and a mediator analysis was planned, but could not be undertaken because of the lack of available methods when there are competing risks (see Chapter 2, Final analysis; Mediator analysis). It is noteworthy that participants in the APM arm appeared to be repositioned more frequently than those in the HSFM arm; these data will be explored in more detail in planned work for the future and inform the development of a realist evaluation109 to further inform trial interpretation.

A limitation of the trial is that the discharge date for all trial participants was not collected, as this was recorded on the health resource utilisation questionnaire, and, in line with other inpatient trials,27 post-discharge mattress provision was not recorded. The large proportion of mattress changes during ward transfers and the incidence of PUs from the end of treatment to 30-day final follow-up does indicate, however, that work is required to improve continuity of care as patients transition through services110 and further supports the appropriateness of the treatment phase sensitivity analysis.

Compliance

Overall, mattress compliance was good, with 81.5% of participants on their allocated mattress within 2 days of randomisation for each treatment group. The median time spent on the allocated mattress was high and only 9.3% (n = 94) of those randomised to APM and 10.9% (n = 110) of those randomised to HSFM did not receive their allocated mattress at any point during the treatment phase. As a result, compliance was better than expected compared with the PRESSURE 1 trial. 33 Delays in randomised mattress provision were mainly logistical and, when participants were not allocated their mattress on the day of randomisation, they were, typically, on the alternative mattress pre randomisation.

In total, 158 participants decided, after randomisation, that they did not want to move onto their allocated mattress or requested a mattress change from their allocated mattress. There was an imbalance in this between the two groups: 118 (5.9%) participants randomised to APM and 40 (2.0%) participants randomised to HSFM did not move or requested a change.

Conversely, there were 238 participants whom ward staff decided after randomisation that they did not want to move onto their allocated mattress or who requested a mattress change from their allocated mattress for clinical reasons. This was imbalanced between the two groups: 52 (21.8%) participants allocated to APM and 186 (78.2%) participants allocated to HSFM were not moved onto their allocated mattress or had a change of mattress requested for clinical reasons.

In general, more participants were moved from the APMs to HSFMs for comfort and to aid rehabilitation (reflecting the wider literature)23,32,34,52 and more participants were moved from HSFMs to APMs because of a deteriorating clinical condition.

From a clinical practice perspective, a large number of mattress changes resulted from ward moves. Of the first mattress change after the randomised mattress was received, 40 (18%) changes from APM and 20 (9.1%) changes from HSFM were due to a ward move. Overall, there were 442 (21.8%) participants who had 1320 mattress changes during the course of the treatment phase, although 606 (45.9%) were instigated by the CRN/P to maximise trial compliance. This reflects a lack of continuity during ward transfers, reported in previous research. 110

Safety

Overall, no safety concerns were indicated for either APMs or HSFMs. There were no RU SAEs and only three mattress-related events, none of which was classified as serious.

The proportion of deaths was similar in both trial arms (APM 8.1% vs. HSFM 8.3%) and consistent with a previous trial. 33 There was a difference between re-admission rates (APM 8.1% vs. HSFM 6.1%), but no re-admissions were reported as being mattress related.

The participant population was characterised by high levels of falling in the month preceding randomisation; these levels are in line with those reported in the 2015 national falls audit. 111 Concerns expressed by patients in previous research about feeling unsafe on APMs32,33 were not reflected in falls, with 14.9% (n = 152) of participants in the APM arm and 15.7% (n = 159) of participants in the HSFM arm reported to have fallen, with similar numbers on the allocated mattress at the time of the fall and a high proportion of total falls (n = 486) occurring after the treatment phase (62.3%, n = 303). The fall rate for the number of participants who had a fall is consistent with those reported in acutely ill hospital populations;112 however, the proportion of total falls that resulted in serious injury (5.6%, 27/486) is similar to the 5% rate reported in older people in community-dwelling settings and much lower than the reported 10–25% for institutional falls resulting in fracture, laceration or need for hospital care. 112

Health economic methodological considerations

Small differences in utility values derived from the EuroQol-5 Dimensions, three-level version (EQ-5D-3L), have often been attributed to a lack of sensitivity in the instrument. 113 In an attempt to address this, the EQ-5D-5L was used. The EQ-5D-5L was developed to improve sensitivity. 64 However, recent preliminary evidence suggests that use of the EQ-5D-5L rather than the EQ-5D-3L ‘causes a decrease in the incremental QALY gain from effective health technologies and therefore technologies appear less cost-effective’. 113 Despite the current uncertainty in the choice of EQ-5D-5L measure, a strength of this analysis is the use of data collected using the PU-QoL-UI, a preference-based measure developed to assess the impact of PUs on HRQoL. 37 The sensitivity analyses using this measure produced similar results to the primary analyses, giving the research team confidence in the results and the conclusions drawn.

Although the APM dominates the HSFM in all base-case scenarios (i.e. within-trial and long-term analysis), the difference in terms of QoL between the two types of mattresses is negligible (less than half a day for the within-trial model and 11 days of full health for the long-term model); it could be argued that this could be seen as a zero random effect. Despite the probabilistic sensitivity analysis of both the within-trial and the long-term models suggesting a > 50% probability of APM being cost-effective, the slight difference in the QALYs suggest that the results should be viewed with caution. In terms of costs alone, however, both analyses suggest that using APM would be the less expensive option.

The analyses are limited by a number of factors, not least the under-recruitment, discussed in more detail later in this chapter, and the high proportion of missing data. The base-case analyses mirrored the methods used in the clinical analyses, using ITT. However, primarily because of missing EQ-5D-5L data, the number of complete cases was relatively small (n = 267) compared with the overall sample of the study (n = 2029), which, despite adjustment at baseline and imputation, is a limitation of the study. To address this, the analysis was conducted under the assumption that missing data were MAR, based on a descriptive analysis performed on the missing data. Visual and logistic regression analyses indicated that the data were unlikely to be MCAR. Although MAR was a plausible assumption, it is recognised that MNAR cannot be completely ruled out as some variables relevant to the economic data are unknown, which may affect the results.

When modelling the long-term cost-effectiveness, there is a risk of structural uncertainty – that an excluded clinical pathway could change the results. 114 This is important here as, over the time of the trial, there was a shift from HSFM provision as the minimum standard of provision for all inpatients to substitution of HSFM for hybrid foam and self-adjusting air-cell mattresses to reduce the use of APMs. As noted in Research delivery, Identifying centres prepared to randomise to high-specification foam mattresses, technology shift is an ongoing issue in medical device trials52 in the current regulatory environment, which does not require evidence of effectiveness prior to product marketing. 43

No previous cost-effectiveness analyses were identified that had modelled long-term cost-effectiveness in this area (e.g. Padula et al.,115 model for 1 year from admission to hospital). For both the VOI for the interim analyses and the final, long-term cost-effectiveness modelling, the model structures were validated and discussed with the clinical trial team. The structure of the model was adapted between the two analyses to incorporate additional information from the trial, which allowed a more detailed patient pathway. Specifically, the model also added a stage to account for one PU developed in hospital that was not included in the VOI. In addition, the trial included a 30-day post-treatment phase follow-up assessment. It is well documented that a PU can take up to 22 weeks to heal, with implications for the amount of resources required to treat it and the impact on the QoL of the patient. This additional measurement gave further information regarding the potential long-term implications of the two analysed mattresses.

Trial methodological considerations

The trial was planned as a double-triangular group sequential trial, which is considered an efficient design as it allows for the possibility of early stopping if superiority of either mattress has been demonstrated or if the trial proves to be futile. Therefore, the trial design optimised the potential for producing clinical evidence on the effectiveness of the mattress earlier than in a conventional fixed design. A conclusion of futility was also considered an important outcome for the trial as both mattresses are currently used in clinical practice. Stopping boundaries on a group sequential trial provide statistical guidance to the DMEC on the recommendations for stopping the trial early; however, other factors, including mattress compliance, participant safety, cost-effectiveness and other information external to the trial, were considered in order to make a fully informed decision on the future progression of the trial.

Moreover, the trial utilised an early primary end point, time to developing a PU of category ≥ 2 to a maximum of 90 days post randomisation, thereby allowing stopping of the trial in a timely manner.

The primary end point required close monitoring in order to plan when a prespecified interim analysis needed to be conducted. A benefit of the close monitoring allowed the overall event rate, centred across both mattress arms, to be reviewed against the original trial assumptions. It was considered important to report the estimate of the overall event rate to the DMEC at the scheduled meetings to allow recommendations to be made to the TSC and, subsequently, the funder on future design modifications. The uncertainty of the overall event rate and the impact this has on the accuracy of the power for detecting the prespecified treatment effect led to the funder requesting an unplanned early interim analysis. The recommendations following the unplanned interim analysis were to continue recruiting to the trial; however, the overall event rate observed led to discussions with the TMG on revised minimum clinically relevant differences between mattresses for varying centred event rates. These discussions acknowledged that the event rate was lower than originally assumed but the actual overall event rate observed at the interim analysis was not disclosed, thereby mitigating the risk of any operational bias in the future conduct of the trial. The output from these discussions informed the final target sample size of 1996 participants to be recruited with a 6-month recruitment extension, which allowed a target clinically relevant difference of 4% to be detected with at least 80% power, assuming an overall event rate of 10%.

Observing a lower event rate than originally assumed resulted in not reaching the required number of events to conduct the first planned interim analysis; therefore, the trial was modified to a fixed design with one final analysis.

The unplanned interim analysis did not account for competing risks and also deviated from the assumption of proportional hazards. However, the estimate of the treatment effect observed in the piecewise Cox model and also in the sensitivity analysis conducted on the treatment phase supported the final primary end-point analysis results. The difference was not statistically significant at the point of conducting the interim analysis, suggesting that it was worthwhile to continue the trial following the interim analysis.

A large number of data were reported for 14 skin sites (plus other skin sites identified post hoc) with repeated assessments over a maximum of 90 days post randomisation. All of these data were then reduced into one record per participant, reporting whether or not the participant developed a new PU of category ≥ 2 (i.e. the end point) and the corresponding time to development (or competing risk/censoring). Therefore, although this was a clinically meaningful end point, the multilevel nature of the data arising from skin sites nested within participants and assessed at repeated time points is ignored. This means that potentially important information, such as intermediate changes in skin status, multiple events and other risk factors, fails to be taken into account in the end-point derivation.

The primary end point was defined as time to developing a PU of category ≥ 2 to a maximum of 90 days post randomisation, with a sensitivity analysis conducted on the time to developing a PU of category ≥ 2 during the treatment phase. Interpretation of the trial results is first and foremost based on the analysis of the primary end point, with the sensitivity analysis used to support these findings. Therefore, the results of the sensitivity analysis would always be reported together with those from the primary end-point analysis. The sensitivity analysis was driven by discussions with clinical research staff, who thought that the PU development during the treatment phase was clinically important because factors such as discharge plans could affect PU development at 30 days post treatment.

This is the first study to include a longer-term post-treatment phase follow-up perspective. 27 It could be argued that the primary end point should have been the shorter-term end point of time to developing a PU of category ≥ 2 by the end of the treatment phase, with the 30-day final follow-up as the primary time point for the cost-effectiveness evaluation. This could be considered a more clinically meaningful time frame owing to the majority of participants (132/160, 83%) developing their new category ≥ 2 PUs during the treatment phase, the relevance of the outcome to the institution providing the mattress intervention and the different patient pathways following the treatment phase and associated discharge. In addition, the long interval between the end of the treatment phase and follow-up at 30 days will have resulted in imprecise reporting of the actual date a PU developed during this time period. However, the longer-term outcome provides a realistic estimate of effectiveness within current NHS inpatient and community services. The pros and cons of the primary end-point definition options require further consideration for future trials.

Research delivery

Overall, trial conduct within the NHS research teams was good, with NHS centres returning high-quality data, better than expected intervention compliance and lower than expected losses to follow-up. The centre variation in recruitment and, in particular, recruitment to the top six centres was a reflection of both local CRN/P provision and ‘top-up’ trial-funded CRN/Ps at centres with evidence of available patients and capacity/ability to expand the research team.

There were five main challenges for the delivery of the PRESSURE 2 trial within the NHS clinical service and research infrastructure. This are outlined in the following sections.

Patient and public involvement

Implications for practice

• Alternating pressure mattresses confer a small treatment-phase benefit on acutely ill inpatients who are bedfast/chairfast and/or have a category 1 PU, which is diminished over time.

• APM patient compliance, the very low PU incidence rate observed and small group differences indicate the need for improved indicators for targeting of APMs.

• Individualised decision-making should take into account skin status, patient preferences (e.g. movement ability and rehabilitation needs) and the presence of factors that may be modifiable through APM allocation, including being completely immobile, having nutritional deficits, lacking capacity and/or having altered skin/category 1 PU.

• Patients with existing category 1 and 2 PUs are most at risk of subsequent PUs of category ≥ 2 and require targeted secondary prevention.

• Improved communication is required prior to ward transfers to improve continuity of PU prevention care.

• Improvements are required to ensure continuity of PU prevention post discharge.

Implications for research

• This study should be incorporated into a meta-analysis of APM versus HSFM RCTs.

• Objective measurement instruments of key risk factors is required to better inform risk stratification and preventative interventions in practice.

• Further analysis is required to explore the relationship between mental capacity, levels of independent movement and repositioning, nutritional status and PU development.

• Further research is required to explore ‘what works for whom and in what circumstances’ to better inform mattress provision for high-risk patients.

• The health economic analysis was limited by missing data; however, the difference in QoL outcomes between the trial arms was negligible and the difference in cost was small, suggesting no need for further research.

• Central blinded expert photographic review is a reliable method for assessing PU outcomes in research. A robust method to enable repeated photographic assessments, which minimises patient burden while enabling sensitivity analyses, requires development.

• Clinical end points should be considered for PU research during the treatment phase because skin changes can occur very quickly and may be influenced by factors such as discharge plans.

• Skin site-level data collected in PU research should be detailed in order to understand how skin changes over time. Further methodological work is required to be able to fully utilise these data in the analysis of trial outcomes.

• The PU-QoL-P tool is suitable for capturing patient-reported functioning (core domains of HRQoL) and PU-area pain in patients at risk of PU development and for quantifying the benefits of associated preventative interventions from the patient’s perspective, thus far lacking. It can be used in research with adults at risk of PU development in all UK health-care settings. Further research is needed to investigate self-complete administration mode, and the development of proxy measures and language translations, given the prevalence of cognitively impaired patients with PUs.

Participants

We wish to express our gratitude to all patients who participated in this research, for giving up their time and participating in this research study.

Trial Steering Committee members

Data Monitoring and Ethics Committee members

Pressure Ulcer Service User Network UK

We would like to sincerely thank all members of PURSUN UK for their continued hard work and dedication, including those on the TSC and other members involved through participation in general PURSUN UK meetings where the PRESSURE 2 trial was discussed.

Research team members at participating sites

We are grateful to all the clinical research team members and co-authors from the 41 NHS Trusts who have participated in this research and made it a part of their busy daily schedules. Without the commitment and support of the PIs, Tissue Viability Nurse Specialists, CRN/Ps and support staff in gaining local permissions, recruiting patients and collecting data for this trial would not have been possible.

Contributions of authors

Jane Nixon (Professor of Tissue Viability and Clinical Trials Research) was the Chief Investigator and is responsible for the conception of the PRESSURE 2 trial. She contributed to the design of the trial, protocol development, protocol implementation, interpretation of data, drafting and revising the report for intellectual content and approval of the final version of the report. She also acts as guarantor.

Sarah Brown (Principal Medical Statistician) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation, statistical design and oversight, and interpretation of data. She also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

Isabelle L Smith (Medical Statistician) contributed to the design of the trial, protocol development, protocol implementation, statistical design, statistical analysis of the clinical results and photography analysis and interpretation of data. She also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

Elizabeth McGinnis (Tissue Viability Nurse Consultant and PRESSURE 2 trial Clinical Co-ordinator) was a co-applicant and contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was responsible for protocol implementation and local co-ordination of approvals and data acquisition for Leeds Teaching Hospitals NHS Trust. She was on the blinded expert photographic assessment panel. She also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

Armando Vargas-Palacios (Research Fellow in Health Economics) contributed to protocol implementation, the health economic analysis plan, data analysis and data interpretation. He also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

E Andrea Nelson (Professor of Wound Healing) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Julia Brown (Professor of Clinical Trials Research) was a co-applicant. She contributed to the conception of the trial, design of the PU-QoL-P substudy, protocol development, scientific oversight and interpretation of the PU-QoL-P substudy. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Susanne Coleman (NIHR post-doctoral Research Fellow) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was on the blinded expert photographic assessment panel. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Howard Collier (Senior Data Manager) contributed to the design of the trial, protocol development, protocol implementation and interpretation of data. He also contributed to revising the report for intellectual content and approval of the final version of the report.

Catherine Fernandez (Head of Trial Management) contributed to the design of the trial, protocol development, protocol implementation and interpretation of data. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Rachael Gilberts (Senior Trial Co-ordinator) contributed to the design of the trial, protocol development, protocol implementation and interpretation of data. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Valerie Henderson (Tissue Viability Nurse Specialist) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was responsible for protocol implementation and local co-ordination of approvals and data acquisition for Northumbria Healthcare NHS Foundation Trust. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Christopher McCabe (Professor of Health Economics) was a co-applicant. He was responsible for the conception of the trial and was the study lead for the health economic VOI analysis, including design, data interpretation and drafting the VOI analyses. He also contributed to revising the report for intellectual content and approval of the final version of the report.

Delia Muir (PPI Officer) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was responsible for the co-ordination of PPI. She also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

Claudia Rutherford (Senior Research Fellow) was a co-applicant. She led the conception and design of the PU-QoL-P substudy, including protocol development, analysis plan and data analysis and interpretation. She drafted the PU-QoL-P substudy and contributed to revising the report for intellectual content and approval of the final version of the report.

Nikki Stubbs (Tissue Viability Nurse Specialist) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was responsible for protocol implementation and local co-ordination of approvals and data acquisition for Leeds Community Healthcare NHS Trust. She was on the blinded expert photography assessment panel. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Benjamin Thorpe (Research Methods Fellow) contributed to the design of the mediator and moderator analyses, data analysis and data interpretation. He also contributed to revising the report for intellectual content and approval of the final version of the report.

Klemens Wallner (Research Assistant) was responsible for the health economic VOI analysis, including design, data analysis, data interpretation and drafting the VOI analysis. He also contributed to revising the report for intellectual content and approval of the final version of the report.

Kay Walker (PURSUN representative) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She also contributed to drafting the report and approval of the final version of the report.

Lyn Wilson (Research Manager and PRESSURE 2 trial Clinical Co-ordinator) was a co-applicant. She contributed to the conception and design of the trial, protocol development, protocol implementation and interpretation of data. She was responsible for protocol implementation and local co-ordination of approvals and data acquisition for Mid Yorkshire Hospitals NHS Trust. She was on the blinded expert photography assessment panel. She also contributed to revising the report for intellectual content and approval of the final version of the report.

Claire Hulme (Professor of Health Economics) was a co-applicant. She contributed to the conception and design of the trial, protocol development, oversight of the health economic data analysis and conduct, and data interpretation. She also contributed to drafting and revising the report for intellectual content and approval of the final version of the report.

Publication

Nixon J, Smith IL, Brown S, McGinnis E, Vargas-Palacios A, Nelson EA, et al. Pressure relieving support surfaces for pressure ulcer prevention (PRESSURE 2): clinical and health economic results of a randomised controlled trial [published online ahead of print September 3 2019]. EClinicalMedicine 2019.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.

Mattress specification guideline: PRESSURE 2 trial

All centres are requested to provide a list of mattresses that are in use in each hospital and to be used in the trial.

1. All participants will have an electric profiling bed frame as an adjunct to the trial mattress.

2. All participants will be randomised to either a HSFM or an APM. Overlay or replacement mattresses may be used.

3. After randomisation, an eligible mattress will be sourced and allocated by the clinical research nurse.

4. All mattresses and their use will comply with the Medical Devices Regulations SI2002/618.

End points to be analysed

Primary end point

The primary end point was the time taken to develop a new category ≥ 2 PU from randomisation to 30 days post treatment phase or withdrawal/death.

Derivation of primary end point

Each participant had a minimum of 14 prespecified skin sites [spine/back, sacrum, left and right buttocks, ischial tuberosities, trochanters (hips), heels, ankles and elbows] assessed at baseline and at every follow-up assessment thereafter. There was the option at each assessment to add other additional skin sites that were not prespecified on the CRF. The data collection process outlined above led to repeated measures for each skin site. Therefore, before deriving whether or not a participant developed a new category ≥ 2 PU (and the time to development), the derivation of whether or not a new category ≥ 2 PU has developed since baseline needed to be defined on a skin site basis (Figure 16). This led to a data set with one record per skin site. These data were then used to derive the primary end point data set with one record per participant that included a variable denoting whether or not the participant developed a category ≥ 2 PU at any skin site and the time to develop the first new PU or censoring time. A summary of the derivation of the primary end point (time to first new category ≥ 2 PU on a patient level) is provided in Figure 17 and Table 27.

FIGURE 16.

Derivation of PU development on a skin site level (one record per skin site per participant). a, Patients with a category 3/4/U PU should have been excluded from the trial in accordance with the eligibility criteria; however they have been included in this derivation to account for participants who may have been incorrectly recruited to the trial. N/A, not applicable; U, unstageable. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

FIGURE 17.

Derivation of PU development at a patient level. a, Evaluable skin sites are such that skin site level outcome can be determined as ‘develops PU’ or ‘does not develop PU’ (see Figure 12). b, The maximum follow-up period was 95 days, to allow for time windows around visits during the treatment phase and around the final visit after the treatment phase. Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

TABLE 27 - App.2 Derivation of the event/censor variable and the number of days to developing a new category ≥ 2 PU (on a patient basis)

Death and some withdrawals were considered as competing risks as they prevent a PU occurring or being observed.

Data collected on the same day that participants were withdrawn from data collection due to clinical condition were excluded from the derivation of end points. This was owing to ethics considerations around a patient’s data being used if they had lost capacity or, for example, were receiving palliative care. The decision over whether or not to censor these patients earlier was made in conjunction with the chief investigator (or delegate) and is based solely on the reason for withdrawal, blind to mattress provision, skin status and time since entering the study.

Adapted from Nixon et al. 44 © The Authors, 2019. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.

Scenario	Number of days	Event/censor variable
Patient is observed to develop a new category ≥ 2 PU before end of follow-up	Number of days between date of randomisation and date first new category ≥ 2 PU was observed (i.e. date first recorded on the CRF)	Event
Patient dies before end of follow-up not having developed a category ≥ 2 PU	Number of days between date of randomisation and date of last evaluable skin assessment	Censoreda at date of last evaluable skin assessment
Patient withdraws from the trial before end of follow-up not having developed a category ≥ 2 PU	Number of days between date of randomisation and date of last follow-up assessment when an assessment on an evaluable skin site was made	Patient is censoreda at date of last evaluable skin assessmentb
Patient is not observed to develop a new category ≥ 2 PU before the end of follow-up (including patients who were lost to follow-up)	Number of days between date of randomisation and date of last follow up assessment where an assessment on an evaluable skin site was made	Patient is censored at date of last evaluable skin site assessment

The time to development of a new category ≥ 2 PU and the value of the censor variable (a variable denoting whether or not the participant was observed to develop a category ≥ 2 PU) will be derived during the analysis process, as detailed in Table 27.