Pressure UlceR Programme Of reSEarch (PURPOSE): using mixed methods (systematic reviews, prospective cohort, case study, consensus and psychometrics) to identify patient and organisational risk, develop a risk assessment tool and patient-reported outcome Quality of Life and Health Utility measures

Pain prevalence in the hospital population

Objectives were to:

1. estimate the unattributed pressure area-related pain prevalence in a hospital population of patients with and without pressure ulcers

2. estimate the pressure area-related pain prevalence in patients with no observable pressure ulcers

3. estimate the pressure ulcer pain prevalence in patients with pressure ulcers

4. describe the intensity and type of pressure area-related and pressure ulcer pain in a hospital population of patients with and without pressure ulcers

5. explore the association between pain intensity, type of pain and pressure ulcer classification in a hospital population of patients with and without pressure ulcers.

Pain prevalence in the community population

Objectives were to:

1. estimate the prevalence of unattributed pressure area-related pain within a community population of patients with pressure ulcers

2. assess the intensity and type of pressure area-related and pressure ulcer pain within a community population of patients with pressure ulcers

3. describe the intensity and type of pressure area-related and pressure ulcer pain within a community population of patients with pressure ulcers

4. explore the association between pain intensity, type of pain and pressure ulcer classification within a community population of patients with pressure ulcers

5. compare and contrast community pressure ulcer prevalence case-finding methods.

Study design

We undertook nested, multicentre, cross-sectional studies in three acute hospital NHS trusts35 and two large community trusts in England45 embedding the pain prevalence study into routine pressure ulcer audits. To identify patients who had unattributed pressure area-related pain, questions about pain were added to the routine annual pressure ulcer prevalence audits undertaken in the participating NHS trusts. To estimate the prevalence of pressure area-related and pressure ulcer pain and to explore the association between pain and pressure ulcer classification, patients who reported pain were invited and consented to undergo a full pain assessment (see Appendix 5 for the consent form).

Nesting the pain prevalence study within routine pressure ulcer prevalence audits meant that we collected data for the total eligible patient population in each setting. The routine NHS audits collect unlinked anonymous data and patient consent is not required to ensure that accurate pressure ulcer prevalence data are obtained for the total eligible population. Nesting the study within routine pressure ulcer prevalence audits, however, limited the number of data items that could be collected. Furthermore, in the community the two trusts defined their denominator population differently and adopted different pressure ulcer case-finding methods. 48

Setting

Three acute NHS hospital trusts took part. One trust included three district general hospitals. The other two NHS trusts were large teaching hospitals and together included four main and two satellite hospitals. This meant that the patient population consisted of those in general secondary care and regional/supraregional specialist services from a total of nine hospitals.

The community NHS trusts consisted of locality 1, serving an urban population of 292,179, and locality 2, serving a rural population of 311,991. 49

Each NHS community trust provides general and tissue viability specialist nursing care to patients residing in their own homes and residential homes as well as community/rehabilitation/hospice inpatient facilities. In addition, each trust provides tissue viability specialist nursing care to patients residing in nursing home settings.

Routine pressure ulcer audit: hospital setting

Routine pressure ulcer audit: community setting

Pain prevalence eligibility criteria

Assessments

Unlinked anonymised individual patient audit data were recorded by a designated ward or community nurse trained in the use of the data collection form and skin assessment as part of the preparation for the audit. Data recorded included place of assessment (hospital/community and ward specialty, patient’s own home, nursing home, residential home, hospice, community bed), date of birth, gender, height, weight, ethnicity, mobility and risk assessment scale total score (using either the Waterlow score50 or the Braden scale51 as per local policy). Skin was assessed using the 1998 EPUAP24 classification and recorded for a minimum of 13 skin sites (sacrum, left and right buttocks, ischial tuberosities, hips, heels, elbows and ankles). The 1998 EPUAP classification (and not the revised EPUAP/NPUAP 2009 version)1 was used as this was the version in routine use at the participating centres/localities. In addition, the presence of an unstageable pressure ulcer, other type of wound or normal skin were confirmed or skin status was recorded as not applicable (e.g. amputation) or unable to assess.

When patients were assessed as not able to report pain this was recorded along with the reasons for ineligibility. When ward/community staff indicated that the patient was well and able to report pain, a member of the tissue viability team asked the patient two pain screening questions as follows:35,45

1. At any time, do you get pain, soreness or discomfort at a pressure area (prompt: back, bottom, heels, elbows or other as appropriate to the patient)? (Yes or no)

2. Do you think this is related to either your pressure ulcer OR lying in bed for a long time OR sitting for a long time? (Yes or no)

These questions were adapted from the case screening questions used in a large postal survey of pain prevalence in the UK. 35,52 Unlinked anonymous individual patient data were recorded for both questions. The site of the pain, soreness or discomfort was not recorded (i.e. the pressure area/pressure ulcer pain was unattributable to individual body sites).

Staff training and preparation

Ward and community nurses were trained locally as per local trust standard pressure ulcer audit practice. Members of the tissue viability team were trained in study procedures, including pain assessment and skin assessments, by the programme manager (LW). No formal inter-rater reliability assessment was undertaken as previous research has demonstrated high agreement between specialist nurses and clinical research nurses in skin assessment and pressure ulcer classification. 56

Data processing

All data returned to the Clinical Trials Research Unit (CTRU) for data processing were anonymous. Data were entered into a bespoke MACRO (version 3; MACRO, Infermed, London, UK) database and range and consistency data checks were carried out to assess the accuracy of the data.

Sample size

For the priori sample size we planned to use a minimum of two acute NHS trusts with an estimated patient population of 2000 and two community NHS trusts with an estimated community nursing caseload of 6000 community patients; therefore, it was planned to include approximately 8000 patients in the prevalence audit.

The a priori sample size was based on the following assumptions:

• the prevalence of pressure ulcers in hospital patients is 10% and in community patients is 5%2

• 30% of patients have pressure ulcers of grades 2–4/unstageable ulcers, of whom 25–50% would report pressure-area related pain40–43

• 70% of patients have pressure ulcers of grade 1, of whom 10–30% would report pressure area-related pain

• 90% of hospital and 95% of community patients have no pressure ulcers

• 2.5–5% of patients without pressure ulcers would report pressure area-related pain.

Based on these assumptions we estimated that between 259 and 555 patients would report pressure area-related pain (i.e. that 3–7% of patients would report pressure area-related pain; Table 2). A sample of 8000 patients would enable us to estimate a pressure area-related pain prevalence of 3% to within ±0.38% (n = 7742) and a pressure area-related pain prevalence of 7% to within ±0.56% (n = 7975).

Analysis

Analysis included data summaries and no inferential statistical testing was planned or undertaken. The denominator for the acute hospital pressure ulcer prevalence was the total inpatient population. The community pressure ulcer prevalence was calculated per 1000 of the estimated total population of adults aged ≥ 18 years for each site (240,038 locality 1 population aged ≥ 18 years; 251,891 locality 2 population aged ≥ 18 years). 49

‘Percentages were calculated using the total number of patients from the relevant population as the denominator (i.e. including all patients with missing data for that variable).’45 When another skin condition or chronic wound was indicated, the specific skin site was excluded from the analysis. ‘All analyses were carried out using SAS software [version 9.2; SAS Institute Inc., Cary, NC, USA]. All percentages were rounded to 1 decimal place. Means, medians, standard deviations [SDs] and ranges were summarised to one more decimal place than the data collected.’45

Type of pain was determined using the results of the seven-item LANSS scale,5 with the responses to each of the seven items scored and summed to provide a total score. Pain was classified as inflammatory pain when the LANSS total score was < 12 and neuropathic pain if the LANSS total score was ≥ 12.

Ethical approval

The studies were approved by the Leeds Central Research Ethics Committee prior to data collection (reference number 09/H1313/14).

Pressure ulcer prevalence: hospital population

Data collection was undertaken between 15 September 2009 and 3 March 2010. From across nine acute hospitals, a total of 3397 patients (see Appendix 1) were included in the routine pressure ulcer prevalence surveys and this is our target hospital population. 50 Figure 2 details the flow of patients through each stage of the process. The number of patients audited by specialty is presented in Table 3.

FIGURE 2.

Participant flow: hospital population. PU, pressure ulcer; UPAR, unattributed pressure area-related.

The median age of patients was 70 years [mean 65.8 (SD 19.23), range 18–103 years]. The numbers of men and women were similar (48.7% male; 1655/3397) and 7.2% (243/3397) were non-Caucasian. 35 In total, 53.9% of patients (1830/3397) were assessed using the Waterlow scale and of these 1062 (58.0%) were classified as ‘at risk’ (score of ≥ 10); 46.1% of patients (1567/3397) were assessed using the Braden scale and of these 532 (34.0%) were classified as ‘at risk’ (score of ≤ 18) (Table 4).

Of the 3397 patients included, 502 (14.8%) were reported to have 1066 pressure ulcers [median 1.0, mean 2.1 (SD 1.63), range 1–13 per patient]. The majority (70.5%; 752/1066) of reported pressure ulcers were grade 1, approximately one-fifth were grade 2 (22.2%; 237/1066) and a small percentage (7.2%; 77/1066) were grades 3–4/unstageable. 35

Detailed pain assessment: hospital population

Of the 327 who answered ‘yes’ to both pain screening questions, 164 (50.2%) were not able, or declined, to participate in the full pain assessment and 163 (49.8%) consented. Three patients were subsequently withdrawn and therefore the analysis population of eligible patients with unattributed pressure area-related pain who participated in the detailed pain assessment was 160 (Figure 3).

FIGURE 3.

Participant flow: detailed pain and skin assessments: hospital population.

The median age of these 160 patients was 69.0 years [mean 66.2 (SD 18.23), range 18–99 years] and half (80/160; 50.0%) were men. Almost three-quarters (114/160; 71.3%) of patients with unattributed pressure area-related pain were assessed as ‘at risk’ on the Braden scale and four (2.5%) were non-Caucasian. A total of 75 (46.9%) patients were reported to have 139 pressure ulcers [median 1.0, mean 1.9 (SD 1.23), range 1–5 per patient] (see Table 4).

A total of 2090 skin sites were assessed (see Figure 3), with 1933 skin sites assessed as normal, 139 assessed as pressure ulcers and skin status was not able to be assessed for 18 sites. The majority (69.8%) of reported pressure ulcers were grade 1, 23.0% (32/139) were grade 2 and 7.2% (10/139) were grades 3–4/unstageable (see Table 4).

Pain was reported by 157 patients on 298 skin sites (mean 1.9, SD 1.17, range 1–7 per patient). This included pressure area-related pain reported on 9.8% (190/1933) of skin sites assessed as ‘normal’ and pressure ulcer pain for 68.0% (66/97) of grade 1 pressure ulcers, 84.4% (27/32) of grade 2 pressure ulcers and 90.0% (9/10) of grades 3–4/unstageable ulcers (Table 5). The worst pain intensity reported by each patient ranged from 1 to 10, with a mean of 5.4 (SD 2.30) and median of 5.0.

TABLE 5 - Detailed pain assessment: number of times pain reported by skin classification: hospital population

Note: 160 patients completed the detailed pain assessment. Each patient had 13 skin assessments and there were 10 ‘other’ sites assessed. The overall total therefore corresponds to (160 × 13) + 10 = 2090 skin assessments.

Skin classification	Yes, n (%)	No, n (%)	Missing, n (%)	Total, n (%)
Normal skin	190 (9.8)	1730 (89.5)	13 (0.7)	1933 (100.0)
Grade 1	66 (68.0)	30 (30.9)	1 (1.0)	97 (100.0)
Grade 2	27 (84.4)	5 (15.6)	0 (0.0)	32 (100.0)
Grade 3	4 (100.0)	0 (0.0)	0 (0.0)	4 (100.0)
Grade 4	3 (100.0)	0 (0.0)	0 (0.0)	3 (100.0)
Unstageable	2 (66.7)	1 (33.3)	0 (0.0)	3 (100.0)
Unable to assess	2 (22.2)	1 (11.1)	6 (66.7)	9 (100.0)
Classification missing	4 (44.4)	5 (55.6)	0 (0.0)	9 (100.0)
Total	298 (14.3)	1772 (84.8)	20 (1.0)	2090 (100.0)

The distribution of pain intensity is similar for each grade of pressure ulcer (Figure 4). In total, 128 patients identified one skin site for LANSS assessment (89 torso and 39 limb skin sites) and 29 patients identified both a torso and a limb skin site for LANSS assessment, providing 118 torso and 68 limb LANSS assessments. Nociceptive pain was dominant in both torso and limb skin sites, with 70.3% (83/118) of painful torso skin sites and 60.3% (41/68) of painful limb skin sites scoring < 12 on the LANSS assessment (Table 6). Neuropathic pain was observed on skin assessed as normal as well as for all grades of pressure ulcer (see Table 6).

FIGURE 4.

Pain intensity by skin classification: hospital population. (a) Normal skin; (b) grade 1 pressure ulcers; (c) grade 2 pressure ulcers; and (d) grades 3–4/unstageable pressure ulcers.

TABLE 6 - Type of pain by skin classification for the most painful torso and limb areas: hospital population

The denominator here is the number of patients who completed the LANSS assessment for a torso skin.

The denominator here is the number of patients who completed the LANSS assessment for a limb skin site.

Location	Skin classification	Nociceptive, n (%)	Neuropathic, n (%)	Missing, n (%)	Total, n (%)
Torso	Normal skin	56 (76.7)	17 (23.3)	0 (0.0)	73 (100.0)
	Grade 1	12 (54.5)	8 (36.4)	2 (9.1)	22 (100.0)
	Grade 2	12 (70.6)	5 (29.4)	0 (0.0)	17 (100.0)
	Grade 3	2 (66.7)	1 (33.3)	0 (0.0)	3 (100.0)
	Grade 4	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)
	Unstageable	1 (100.0)	0 (0.0)	0 (0.0)	1 (100.0)
	Missing	0 (0.0)	1 (50.0)	1 (50.0)	2 (100.0)
	Totala	83 (70.3)	32 (27.1)	3 (2.5)	118 (100.0)
Limb	Normal skin	27 (61.4)	16 (36.4)	1 (2.3)	44 (100.0)
	Grade 1	11 (73.3)	3 (20.0)	1 (6.7)	15 (100.0)
	Grade 2	1 (20.0)	4 (80.0)	0 (0.0)	5 (100.0)
	Grade 3	0 (0.0)	1 (100.0)	0 (0.0)	1 (100.0)
	Grade 4	1 (100.0)	0 (0.0)	0 (0.0)	1 (100.0)
	Unstageable	1 (100.0)	0 (0.0)	0 (0.0)	1 (100.0)
	Missing	0 (0.0)	0 (0.0)	1 (100.0)	1 (100.0)
	Totalb	41 (60.3)	24 (35.3)	3 (4.4)	68 (100.0)

Pressure ulcer prevalence: community population

Locality 1 collected data between 8 February and 2 April 2010 and locality 2 collected data between 12 April and 7 May 2010. Figure 5 shows the patient flow through the stages of the process. 45

FIGURE 5.

Participant flow: community population. Grade 1-U, grade 1 or above. a, Locality 1 audited all patients on the caseload whereas locality 2 audited all patients with existing pressure damage. Adapted from McGinnis et al. 45 © 2014 McGinnis et al. ; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. 45

‘The two community NHS Trusts identified 287 patients with Grade 1–4/Unstageable pressure damage. The case finding methods resulted in differing prevalence rates. In locality 1, 1680 patients were assessed, and of these 185 patients were assessed as having a pressure ulcer ≥ Grade 1, a prevalence rate of 0.77 per 1000 [(185/240,038) × 1000 adults]. In locality 2, 102 patients were identified from the community nursing caseloads and assessed as having a Grade ≥ 1 pressure ulcer, a prevalence rate of 0.40 per 1000 [(102/251,891) × 1000 adults]’45. A notable difference between the two sites, and one that could also contribute to the difference in reported prevalence, was the patients’ place of residence. In locality 1, 93 out of 185 (50.3%) patients were resident in a nursing home, whereas in locality 2 only five out of 103 (4.9%) patients were resident in a nursing home.

The median age of patients with pressure ulcers was 81 years (mean 77.8, SD 13.44, range 23–106 years), just over one-third of patients were male (100/287; 34.8%),45 89.6% (251/280) were assessed as being ‘at risk’ using either the Waterlow scale or the Braden scale and only 1.4% (4/287) were non-Caucasian (Table 7).

The 287 patients with pressure ulcers were reported to have 440 ulcers [median 1, mean 1.5 (SD 0.83), range 1–5 per patient]. About one-third of pressure ulcers (155/440; 35.2%) were grade 1, 40.2% (177/440) were grade 2 and 24.5% (108/440) were grades 3–4/unstageable. 45

Detailed pain assessment: community population

‘Of the 133 patients with unattributed pressure area-related pain, 96 were not able or declined to participate in the full pain assessment [see Figure 5]. Therefore, the analysis population of eligible patients who consented to the detailed pain assessment was 27.8% (37/133) of the population reporting pain.’45

The median age of these 37 patients was 75.0 years (mean 72.6, SD 15.31, range 23–98 years). Most (70.3%) patients were assessed in their own home, with the remainder assessed in residential or nursing homes, rehabilitation units or palliative care units. Nine patients (24.3%) were male and all were white (see Table 7). 45

‘A total of 481 skin sites were assessed [Figure 6], including 427 skin sites assessed as normal and 54 PUs (mean 1.5 per patient, SD 0.65, range 1–3). Approximately a third of PUs were Grade 1 (37.0%; n = 20/54), Grade 2 (31.5%; n = 17/54) and Grade 3–4/U (31.5%; n = 17/54) [see Table 7], with 29 (53.7%) located on a torso skin site and 25 (46.3%) located on a limb skin site.’45

FIGURE 6.

Participant flow: detailed pain and skin assessments: community population. PU, pressure ulcer. Reproduced from McGinnis et al. 45 © 2014 McGinnis et al. ; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

The 37 patients reported pain on 53 out of 481 (11.0%) skin sites [median 1.0, mean 1.4 (SD 0.65), range 1–3 per patient]. ‘No pressure area related pain was reported on normal skin whilst patients reported PU pain for 98.1% (n = 53/54) of all PUs [Table 8]. Pain intensity ranged from 1–10, with a mean of 6.4 (SD 2.53) and median of 7.0. There is a slightly skewed distribution of pain intensity with very similar pain levels for each grade of PU [Figure 7]. Thirty-one patients identified one skin site for LANSS assessment (n = 19 torso and n = 15 limb) and six patients identified both a torso and a limb skin site for LANSS assessment providing a total of 22 torso and 18 limb LANSS assessments. Neuropathic pain was slightly dominant in both torso and limb skin sites, with 54.5% (n = 12/22) of torso PUs and 61.1% (n = 11/18) of limb PUs scoring ≥ 12 on the LANSS assessment’ (Table 9).

TABLE 8 - Detailed pain assessment: number of times pain reported by skin classification: community population

Reproduced from McGinnis et al. 45 © 2014 McGinnis et al. ; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

	Yes, n (%)	No, n (%)	Total, n (%)
Normal skin	0 (0.0)	427 (100.0)	427(100.0)
Grade 1	19 (95.0)	1 (5.0)	20 (100.0)
Grade 2	17 (100.0)	0 (0.0)	17 (100.0)
Grade 3	8 (100.0)	0 (0.0)	8 (100.0)
Grade 4	5 (100.0)	0 (0.0)	5 (100.0)
Unstageable	4 (100.0)	0 (0.0)	4 (100.0)
Total	53 (11.0)	428 (89.0)	481 (100.0)

FIGURE 7.

Pain intensity by skin classification: community population. (a) grade 1; (b) grade 2; and (c) grades 3–4/unstageable. Reproduced from McGinnis et al. 45 © 2014 McGinnis et al. ; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited.

TABLE 9 - Type of pain by skin classification for the most painful torso and limb areas

The denominator here is the number of patients who completed the LANSS for a Torso skin site.

The denominator here is the number of patients who completed the LANSS for a Limb skin site.

Source: reproduced from McGinnis et al. 45 © 2014 McGinnis et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. 45

Location	Skin classification	Nociceptive, N (%)	Neuropathic, N (%)	Missing, N (%)	Total, N (%)
Torso	Grade 1	3 (42.9%)	4 (57.1%)	0 (0.0%)	7 (100.0%)
	Grade 2	3 (33.3%)	6 (66.7%)	0 (0.0%)	9 (100.0%)
	Grade 3	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
	Grade 4	4 (80.0%)	1 (20.0%)	0 (0.0%)	5 (100.0%)
	Unstageable	0 (0.0%)	1 (100.0%)	0 (0.0%)	1 (100.0%)
	Totala	10 (45.5%)	12 (54.5%)	0 (0.0%)	22 (100.0%)
Limb	Grade 1	2 (50.0%)	2 (50.0%)	0 (0.0%)	4 (100.0%)
	Grade 2	1 (20.0%)	4 (80.0%)	0 (0.0%)	5 (100.0%)
	Grade 3	2 (33.3%)	3 (50.0%)	1 (16.7%)	6 (100.0%)
	Grade 4	0 (0.0%)	0 (0.0%)	0 (0.0%)	0 (0.0%)
	Unstageable	1 (33.3%)	2 (66.7%)	0 (0.0%)	3 (100.0%)
	Totalb	6 (33.3%)	11 (61.1%)	1 (5.6%)	18 (100.0%)

Study design

We undertook a multicentre prospective cohort study. We recorded the presence of key risk factors, skin status and pain at baseline with twice-weekly follow-up for up to 30 days from registration to identify the development of new category 2 or above pressure ulcers (see Appendix 6 for the study protocol).

Setting

Hospital patients were recruited from vascular, trauma, orthopaedic and medical/elderly wards and community NHS patients were recruited from their place of normal residence (own/residential/nursing home) and community inpatient facilities.

Eligibility criteria

Screening

Participating research sites were required to complete a log of all patients screened for eligibility. The following anonymised information was collected: age, gender, ethnicity and reason not eligible for study participation or the reason eligible but declined.

Recruitment and consent

When eligibility was indicated by the attending clinical team, patients were flagged to a member of the trust tissue viability team. A full verbal explanation of the study patient information leaflet (see Appendix 7) was provided for the patient to consider, including detailed information about the rationale, design and personal implications of the study. Assenting patients were then invited to provide written informed consent/witnessed consent (see Appendix 8) and eligibility was confirmed prior to registration using a central 24-hour telephone registration system.

Assessments

At registration, baseline demographics, skin status, pain status, analgesic use, mattress intervention, the Braden subscales51 and risk factors, including diabetic status, presence of other chronic wounds, history of weight loss and body mass index (BMI) were recorded.

Pressure ulcer interventions

All participating trusts had pressure ulcer prevention policies and guidelines, which included risk assessment, mattress and repositioning guidance. The cohort study participants received pressure ulcer prevention interventions as allocated by the attending clinical teams and as per local policies and guidelines. We recorded mattress provision, which was categorised as non-pressure relieving (e.g. domestic mattress or silicone fibre overlay), static pressure relieving (e.g. high-specification foam, viscoelastic foam, air filled, gel filled) and dynamic pressure relieving (e.g. alternating pressure, low air loss, air fluidised).

Staff training and preparation

Consistent with methods used for the pain prevalence study, members of the tissue viability team were trained and no formal assessment of inter-rater reliability was undertaken (see Pain prevalence in hospital and community populations, Methods).

Data processing

Data processing methods were consistent with those described in Pain prevalence in hospital and community populations (see Methods). In addition, data queries and missing data were chased until the data were confirmed as correct or unavailable. Batch validation and consistency data checks were also carried out to check the accuracy of the data.

Sample size

For risk factor studies using logistic regression it is recommended that at least 10 patients with the event of interest are needed for reliable estimation of effects. 59 Our aim was to assess whether or not the presence of localised skin pain is predictive of new category 2 (or above) pressure ulcer development, after adjusting for the effects of other known risk factors.

We prespecified seven risk factors identified from our risk factor systematic review and emerging conceptual framework46 (see Chapter 5): age, diabetes status, nutritional status, Braden mobility subscale score, presence of skin alterations, presence of a category 1 ulcer on any site and patient setting (hospital or community). In addition, because a patient’s perception of pain is likely to be affected by the use of analgesics or other forms of pain relief, this was also included as a prespecified covariate, resulting in a model potentially including nine factors (i.e. pain, the seven prespecified risk factors and analgesic use).

A nine-factor model would therefore require a minimum of 90 patients to develop a new pressure ulcer of category 2 or above. In the absence of prospective data for community-based patient populations,46 the sample size estimate was based on previous research in acute hospital patients,37,58,60 suggesting that approximately 15% of patients would develop a new pressure ulcer of category 2 or above within 30 days of entering the study. Based on this assumption and allowing for potential loss to follow-up of 5%, we estimated that we would require 632 patients to be recruited to this study.

A further consideration in appraising the sample size estimate was the prevalence of pain at study entry. As no previous work in this field had been undertaken we considered a range of prevalence rates. Table 10 shows the largest difference in pressure ulcer incidence that could be detected with a minimum of 80% power if 10% or 20% of patients reported pain at study entry. We estimated pressure ulcer event rates in the patients without pain of 10% and 15% for each case and assumed that patients with pain at study entry are more likely to develop a new ulcer than those without pain at entry.

We estimated that, if we recruited 600 patients (after accounting for 5% loss to follow up) with 60 (10%) of them having pain on study entry, this would allow us to detect a statistically significant difference (p < 0.05) of 13.2% between those with and those without pain using a chi-square test (80% power, 5% significance) if 10% of patients without pain and 23.2% of those with pain developed a new pressure ulcer within the 30-day follow-up period, corresponding to an odds ratio (OR) of 2.72 [95% confidence interval (CI) 1.40 to 5.27].

As this was an exploratory study and there was uncertainty around the assumptions made to estimate the sample size, the proportion of patients with pain at baseline and the incidence of pressure ulcer development was monitored by the statistical team and chief investigator and reported to a subgroup of the programme steering committee.

End point definition

The primary end point was defined as the development of a new category 2 or above pressure ulcer on any skin site after registration and before the end of follow-up. End of follow-up was defined as no longer at high risk (see Inclusion criteria), patient transferred to a non-participating setting, death or the end of study follow-up (30 days), whichever was the earliest event.

The secondary end point was defined as the time in days after registration to development of the first new category 2 or above pressure ulcer or to the end of follow-up for patients who were not observed to develop a new category 2 or above pressure ulcer during follow-up. Patients who were not observed to develop a new category 2 or above pressure ulcer were censored at the end of follow-up.

Analysis population

The defined analysis population was all patients for whom a primary end point could be determined, that is, patients with at least one follow-up skin assessment completed.

Analysis methods

Baseline characteristics

In total, 397 (65.9%) patients were registered from the acute setting and 205 (34.1%) patients were registered from the community setting, half of whom were located in rehabilitation inpatient settings (104/205, 50.7%; Table 11).

Patient characteristics are detailed in Tables 12 and 13. In summary, across both groups there were fewer male patients (38.7%), all but one patient was Caucasian (99.8%) and one-quarter of patients were diabetic (25.4%). The median age of patients was 80 years (range 21–101 years). One-quarter of patients had a history of weight loss (24.4%), with 10.6% assessed as underweight by BMI and 22.3% assessed as obese. In total, 21.1% of patients had a chronic wound. The majority of patients recruited to the study had alterations to intact skin (62.0%), almost half had a category 1 pressure ulcer (48.2%) over one-quarter had a category 2 ulcer (27.2%) and 4.0% had a category 3, 4 or unstageable ulcer.

Pain was reported on healthy, altered and category 1 skin sites by 77.1% of patients (see Table 12). Only 3.5% of patients were completely immobile, with the majority of patients assessed as having either very limited (43.5%) or slightly limited (38.4%) mobility according to the Braden mobility subscale (see Table 13).

The majority of patients were receiving either a static pressure-relieving (45.3%) or a dynamic pressure-relieving (48.5%) mattress. In the community 15.1% of mattresses were described as non-pressure relieving whereas in the hospital setting this was the case for only 1.3% of patients (see Table 12).

Primary end point analysis

The primary end point was whether or not a patient developed a new category 2 or above pressure ulcer on any skin site after registration and before the end of follow-up. Skin sites with a category 2 or above pressure ulcer or recorded as not applicable/unable to assess (e.g. amputated limb or bandage/dressing in situ) at baseline were excluded from the primary end point analysis. The overall incidence of new category 2 or above pressure ulcers in the analysis population was 152 out of 602 (25.2%).

A total of 464 (77.1%) of the study population reported pressure area pain on skin assessed clinically as normal, altered or with a category 1 pressure ulcer and, of these, 130 (28.0%) developed a category 2 (or above) pressure ulcer compared with 22 (15.9%) patients with no pain at baseline (Table 14).

The results of the univariate analysis are presented in Table 15. Factors that had a statistically significant association with the odds of developing a new category 2 or above pressure ulcer included presence of skin alterations (OR 1.79, 95% CI 1.20 to 2.66; p = 0.0045), presence of at least one category 1 pressure ulcer (OR 3.20, 95% CI 2.63 to 4.74; p < 0.0001) and presence of pain on a healthy, altered or category 1 skin site at baseline (OR 2.05, 95% CI 1.25 to 3.38; p = 0.0047). There was marginal evidence that diabetic status was associated with the odds of developing a new category 2 or above pressure ulcer, with the odds higher for patients with diabetes than for patients without diabetes (OR 1.45, 95% CI 0.97 to 2.18; p = 0.0722). The following factors were not statistically significant: age, history of previous weight loss, Braden mobility subscale score, setting and analgesic use. Therefore, there was no evidence of an association between these factors and the odds of developing a new category 2 or above pressure ulcer.

The final multivariable model obtained from the variable selection carried out included three variables: the presence of a category 1 pressure ulcer (OR 3.25, 95% CI 2.17 to 4.86; p < 0.0001), the presence of skin alterations (OR 1.98, 95% CI 1.30 to 3.00; p = 0.0014) and the presence of pain on a healthy, altered or category 1 skin site at baseline (OR 1.56, 95% CI 0.93 to 2.63; p = 0.0931) (Table 16). Therefore, there was significant evidence that the presence of a category 1 pressure ulcer and presence of skin alterations are risk factors for developing a category 2 or above ulcer. After adjusting for these risk factors, there was marginal evidence that the presence of pain is a further risk factor for developing a category 2 or above pressure ulcer.

Overdispersion analysis

The model obtained in the primary end point analysis was further developed by considering the inclusion of other variables for which data were collected using further forwards and backwards stepwise variable selection. The results of the univariate analyses are shown in Table 15.

The final multivariable model obtained from the further variable selection carried out included six variables. These included (1) the presence of a category 1 pressure ulcer (OR 3.20, 95% CI 2.11 to 4.85; p < 0.0001), (2) the presence of skin alterations (OR 1.90, 95% CI 1.24 to 2.91; p = 0.0032), (3) the presence of pain on a healthy, altered or category 1 skin site at baseline (OR 1.85, 95% CI 1.07 to 3.20; p = 0.0271), (4) the presence of a category 2 ulcer (OR 2.09, 95% CI 1.35 to 3.23; p = 0.0009), (5) the presence of a chronic wound (OR 1.66, 95% CI 1.04 to 2.62, p = 0.0277; Table 17). In addition, (6) there was significant evidence that Braden activity was related to the odds of developing a new category 2 or above pressure ulcer (p = 0.0476). As shown in Table 17, the odds of developing a category 2 or above ulcer were greater for patients who were chairfast than for patients who were bedfast (OR 1.86, 95% CI 1.03 to 2.29) whereas there was no evidence that patients who walk occasionally or frequently were more likely to develop a category 2 or above pressure ulcer than patients who were bedfast (the 95% CIs for the ORs straddle 1). After adjusting for these risk factors, there was significant evidence that the presence of pain is a further risk factor for developing a category 2 or above pressure ulcer.

Correlations between explanatory variables were examined and each of the comparisons yielded low associations. Table 18 shows the associations between the explanatory variables. Cross-tabulations of explanatory variables were also produced for further information and are shown in Appendix 9.

Time-to-event analysis

The variables that were statistically significantly (at the 10% level) associated with the time to onset of a new category 2 or above pressure ulcer in the univariate analyses were age [acceleration factor (AF) 1.01, 95% CI 1.00 to 1.03; p = 0.0354], Braden mobility subscale score (AF 1.37, 95% CI 1.00 to 1.87; p = 0.0498), the presence of a category 1 pressure ulcer (AF 2.63, 95% CI 1.93 to 3.58; p < 0.0001) and the presence of pain on a healthy, altered or category 1 skin site (AF 2.68, 95% CI 1.86 to 3.86; p < 0.0001). In addition, there was marginal evidence that the presence of skin alterations was associated with the time to onset of a new category 2 or above pressure ulcer (AF 1.40, 95% CI 0.99 to 1.97; p = 0.0593; Table 19).

The final accelerated failure time multivariable model obtained from forwards and backwards variable selection included the covariates presence of a category 1 pressure ulcer and presence of pain on a healthy, altered or category 1 skin site and so both are risk factors for reducing the time to develop a category 2 or above pressure ulcer (Table 20). Unlike the primary analysis, the presence of skin alterations has not been included in the final model but was shown to have a significant association with time to onset of a new category 2 or above pressure ulcer in the univariate analysis and therefore may explain why skin alterations were not included in the final model. The cumulative incidence function for the presence of skin alterations is very similar to the cumulative incidence function for the presence of pain on a healthy, altered or category 1 pressure ulcer skin site, although the cumulative incidence functions indicate that there is a larger difference between pain categories than between skin alteration categories. The final model shows that patients are likely to develop a new category 2 or above pressure ulcer 2.32 times faster if they have a category 1 pressure ulcer at baseline than if they do not have a category 1 pressure ulcer at baseline (AF 2.32, 95% CI 1.73 to 3.12; p < 0.0001). In addition, patients are likely to develop a new category 2 or above pressure ulcer 2.28 times faster if they have pressure area-related pain at a healthy, altered or category 1 skin site than if they do not have pressure area-related pain at a healthy, altered or category 1 skin site (AF 2.28, 95% CI 1.59 to 3.27; p < 0.0001).

Cumulative incidence functions for the presence of skin alterations, the presence of a category 1 pressure ulcer and the presence of pain on a healthy, altered or category 1 skin site at baseline are presented in Figure 9.

FIGURE 9.

Cumulative incidence plots. (a) Cumulative incidence plot of time to develop a new category 2 pressure ulcer by presence of skin alterations at baseline; (b) cumulative incidence plot of time to develop a new category 2 pressure ulcer by presence of a category 1 PU at baseline; and (c) cumulative incidence plot of time to develop a new category 2 pressure ulcer by presence of pain on a healthy, altered or category 1 pressure ulcer skin site at baseline.

Skin site-level analysis

The analysis population of 602 patients had a combined total of 7863 potential skin sites assessed (Table 21), of which the majority (77.5%) were observed as being healthy skin sites. Pain was reported more frequently with more severe skin status, that is, 63.1% of category 1 skin sites were observed to have pain compared with 40.3% of skin sites with alterations and 6.4% of healthy skin sites (see Table 21).

Of the total 7863 skin sites, 7483 (95.2%) were evaluable in the analysis, that is, all skin sites that were observed as being healthy, altered or category 1 at baseline and which had at least one follow-up assessment (i.e. the end point could be derived for that skin site). Overall, 223 (3.0%) of the evaluable skin sites developed a new category 2 or above pressure ulcer, and the incidence for skin sites with pain at baseline was higher at 10.3% than that for skin sites with no pain at baseline (1.7%) (Table 22). Similarly, the incidence of a new category 2 or above pressure ulcer was observed to increase with the severity of skin status at baseline (i.e. the incidence for skin sites with a category 1 pressure ulcer at baseline was 18.2% compared with 6.4% for sites with skin alterations and 1.1% for healthy skin sites at baseline).

The results of the univariate analysis are presented in Table 23. Factors that had a statistically significant association with the odds of developing a new category 2 or above pressure ulcer were skin status, which consisted of two levels – skin alterations (OR 6.29, 95% CI 4.21 to 9.40; p < 0.0001) and category 1 pressure ulcer (OR 27.34, 95% CI 18.5 to 40.4; p < 0.0001) – and the presence of pain at baseline on a healthy, altered or category 1 skin site (OR 8.68, 95% CI 6.30 to 11.97; p < 0.0001). The following factors were not statistically significant: age, diabetic status, history of previous weight loss, Braden mobility subscale score, setting and analgesic use. Therefore, there was no evidence of an association between these factors and the odds of developing a new category 2 or above pressure ulcer at the skin site level.

The final multivariable model obtained from the variable selection carried out included two variables: skin status, which consisted of two levels – skin alterations (OR 4.65, 95% CI 3.01 to 7.18; p < 0.0001) and category 1 pressure ulcer (OR 17.30, 95% CI 11.09 to 27.00; p < 0.0001) – and the presence of pain on a healthy, altered or category 1 skin site (OR 2.25, 95% CI 1.53 to 3.29; p < 0.0001; Table 24). Therefore, there was significant evidence that skin status is a risk factor for developing a category 2 or above pressure ulcer after adjusting for between-patient variation. After adjusting for skin status and between-patient variation there was strong evidence that the presence of pain is a further risk factor for developing a category 2 or above pressure ulcer.

Design

The research design was substantially influenced by two arguments. The first stemmed from discussions at the start of the study. Although our principal objective concerned the effect of organisational context on the development of severe pressure ulcers, we realised that we could not simply assume that a relationship between the two existed and could be studied empirically. Indeed, the available empirical evidence is limited, but the literature offers three distinct explanations,97,98 namely:

1. pressure ulcers develop following a mistake made by an individual clinician99,100

2. they develop as a result of a sequence of otherwise unconnected mistakes101–103

3. there are systemic weaknesses in the organisation and delivery of care, such that the regime is one in which pressure ulcers are more likely to develop. 104–106

We decided that it would be necessary to discriminate between these candidate explanations to establish whether, and how, the organisational context helped to explain the development of severe pressure ulcers or alternatively played no role. We should also include two other logical possibilities, in order to identify their role in explaining development or eliminate them from consideration, namely (1) the behaviour of clinicians had no effect on the development of a pressure ulcer, which would have developed whatever they had done, and (2) there was an alternative explanation, which had not previously been reported or hypothesised.

The second argument flowed from the nature of the domain that we were studying. Severe pressure ulcers occur relatively rarely and can develop in a wide range of settings over periods of days or weeks. It is not currently possible to predict who will develop them and who will not; it is only possible to identify people who have already developed a severe pressure ulcer. A prospective study was therefore not feasible. We opted to identify patients who had developed severe pressure ulcers and to reconstruct what had happened to them. This led us to adopt a retrospective case study design. A process-tracing case study method was used to capture the experiences of eight individuals who had developed severe pressure ulcers. 107 Accounts of their experiences were developed, which were then compared and contrasted to identify common features and hence common explanations. The two arguments taken together led us to develop a novel research design.

Setting

Eight patients were recruited in six NHS trusts in Yorkshire, England. Four patients’ accounts occurred wholly or mainly in acute hospitals, three mainly in their own homes and one in a combination of a community hospital and an acute hospital rehabilitation ward. The decision to recruit eight patients was pragmatic: each account took approximately 4 months from initial interview to completion of analysis and we were therefore able to complete eight accounts with the resources available to this study.

Eligibility

Recruitment and consent

Participants were sampled partly to maximise the diversity of individuals and the contexts in which they developed severe pressure ulcers and partly purposively (see Appendix 10). Eligible patients were identified by members of the local tissue viability nurse teams at one of the six study sites in Yorkshire, England, who informed them about the study and provided them with a study information leaflet and an ‘agree to be contacted by the researcher’ form (see Appendix 11). Consent to participate was obtained from individuals and, when appropriate, also from their main carers (see Appendix 12).

Data collection

Data were collected by a field researcher (LP) with a non-clinical background from five sources, namely interviews with individuals who had developed a severe pressure ulcer (and, when relevant, their main carers), interviews with clinical and other staff who had been involved in their care, clinical records, other documents relevant to the account such as critical incident reports, and relevant local policy documents (e.g. on the conduct of skin risk assessments) (Figure 10, stage 1). A parallel review of patient notes was undertaken by a tissue viability nurse at each study site.

FIGURE 10.

Analysis and review of individual accounts.

Patients were interviewed first and invited to give their account of the reasons why, in their view, their severe pressure ulcer had developed. Interviews were semistructured and lasted between 30 and 90 minutes. They were digitally recorded and transcribed.

The patient interview was used to direct the next phase of data collection, which involved accessing and reviewing nursing, medical and therapist notes, clinical incident reports and other documents (e.g. staff rotas for key periods of time in the patients’ accounts). An initial analysis of the documents was undertaken and, on the basis of the analysis and the patient account, an initial interview schedule was drawn up. The analysis was also used to identify members of staff who were likely to be able to provide useful information about the development of the severe pressure ulcer. After the initial interviews, the researcher discussed the emerging possible explanations with the site tissue viability nurse specialist and a list of further interviewees was agreed. It is worth noting that, for this and subsequent analyses, the focus was on understanding how severe pressure ulcers developed, but a range of contextual information was provided in interviewees’ responses and in the documentation provided that could be used to discriminate between the explanations identified earlier.

Interviewees for hospital-based accounts included matrons, ward nurses, health-care assistants, ward clerks, ward managers, physiotherapists and consultants. In community settings, interviewees included district nurses, home care assistants and therapists (Table 25). At least seven interviews were conducted for each account except for the eighth account, in which an individual developed a severe pressure ulcer in her own home after a fall and few health professionals had useful information about her circumstances. Few professionals were directly involved in identification of and response to her ulcer, and comprehensive notes were available about her assessment and treatment only once she came into contact with health services. In total, 70 interviews were conducted across the eight accounts. Judgements were made about the time period that each account needed to cover and the extent of the documentation that was needed. In some instances, both were extended when it became clear that the histories were longer or more complicated than at first appeared.

Analysis

Transcripts of patients’ interviews were reviewed and key passages that set out their accounts of events were included verbatim at the start of each account. This was done partly because the patient (and in some cases also the carer) was the only person who had been present throughout and partly to guard against losing sight of the ‘patient’s voice’ in subsequent analysis. A Microsoft Access^® 2010 database (Microsoft Corporation, Redmond, WA, USA) was then created for each account and used to organise key decisions and actions into a chronological sequence. Patient- and carer-derived data were recorded in one column, clinician interview data in a second and clinical record and other documentary sources in a third (see Figure 10, stage 2). The presentation of data in parallel columns made it possible to develop a chronological account of events, identify consistencies and inconsistencies between different data sources and assess the ‘strength’ of evidence available about key events, reflected in the number and quality of sources. These data were used as the basis of a single, provisional timeline of events.

The site principal investigator, who in each case was a nurse with a specialist interest in tissue viability, undertook a parallel review, based solely on available patient records and on other available documentation, including local guidelines and critical incident reports. The review followed the guidance for reviews of critical incidents in the NHS. 94 The investigator wrote a report, identifying key decisions and actions in chronological order, including departures from local guidelines. The field researcher and site principal investigator then met and compared their accounts, identifying consistencies and inconsistencies (e.g. actions that the nurse judged as important that were not included in the researcher’s account). The timeline in each initial account was revised in light of additional facts or insights generated (see Figure 10, stage 3).

The eight accounts

The eight patients were selected, in part, to maximise diversity (Table 26). Unsurprisingly, then, there were marked differences in the details of their treatment and care and different explanations were offered by those interviewed for the development of severe pressure ulcers.

Seven of the eight – the exception being number 8 – exhibited widely recognised risk factors and had complex treatment and care needs. In a number of accounts some staff who were interviewed blamed the patients, on the basis that they had not complied with advice on managing their risks (e.g. shifting position regularly). But patients themselves, in the same accounts, generally pointed to specific actions or omissions, such as the failure to be turned regularly overnight, to be provided with a specialised mattress or to act on patients’ comments about their own risks.

Participants approached interviews in very different ways. Some patients and carers were very clear in their own minds about what had gone wrong whereas others were very reluctant to criticise any aspect of their care. Similarly, some accounts involved individuals moving between wards, or an account developed against a background of a major ward reorganisation. The significance attributed to these moves varied, including, for example, the failure to transfer notes with patients, which placed receiving staff at a disadvantage, and staff feeling harassed because they were working in an unfamiliar environment during a reorganisation.

Elimination of hypotheses

The diverse group of individuals all had the same outcome: a severe pressure ulcer. In one account (number 8), the review teams judged that the development of the pressure ulcer was unavoidable, because the individual concerned developed a severe pressure ulcer in her own home, before any health professional saw her. In another account (number 3) there was a single precipitating event and in three other accounts (numbers 2, 4 and 6) there was a sequence of events. But the clinical subgroup and subsequent reviewers all judged that the organisational context made development more likely in seven of the eight accounts (Table 27). It is possible that the organisational context made the ‘key events’ in accounts 2–4 and 6 more likely, although we cannot make causal inferences with any confidence on the basis of our evidence. The evidence suggests that the term ‘organisational context’ covers two distinct concepts. The first concerns the prevailing cultures in the settings where severe pressure ulcers developed (see Cross-patient themes). The second relates to what one might term the functional characteristics of those settings, particularly nursing staff shortages, staff who (justifiably or not) did not believe that they had enough time for proper treatment and care and wider organisational issues, including contemporaneous reconfigurations of services in four of the accounts.

Cross-patient themes

The process of eliminating hypotheses, and the analysis of common themes across the eight individuals, led to the identification of three broad themes. First, the ‘voices’ of those who developed severe pressure ulcers, and of their carers when they were involved, were not heard by staff. As noted earlier, the individuals themselves behaved differently and had different relationships with clinical staff, but failures to heed information were evident in several accounts. There were examples of patients making repeated appeals for pain and discomfort to be addressed and expressing concerns about their own well-being that were not heeded over periods of hours or even days. In some instances, these appeals seem to have been dismissed by staff, that is, they were heard but were not taken seriously. Patients were also blamed for the development of their pressure ulcers on the basis that they did not comply with instructions that they were given, and were branded as ‘difficult’, even when they had a cognitive impairment.

Second, there were failures to recognise and act on warning signs. Risk assessments were not undertaken when they should have been and, in some cases, they were undertaken only several days after admission to an acute hospital ward. Evidence of pre-existing clinical risks in records was not acted on in six of the seven cases in which the environment was judged to have made development more likely. Action was not taken promptly when overt evidence, including the presence of a category 2 pressure ulcer, was identified. In interviews, there were a number of instances of staff blaming colleagues for these failures. Conversely, there was evidence of poor documentation, so that adherence with patients’ care plans was not recorded and, in some instances, direct evidence of skin redness or a pressure ulcer was not recorded. Some health-care assistants, who provided direct care, observed that they lacked the appropriate training to identify and record risks or were not allowed to record them.

Third, there were co-ordination failures, between patients, carers and staff, between staff in the same setting, between staff in different settings in the same organisation (e.g. two wards) and between staff in different organisations. Sometimes this was manifested as interprofessional communication failure and sometimes there was poor communication between the same professional groups in two locations. One example of the latter came in a postoperative setting where risks were not properly communicated between the anaesthetic recovery unit and the postoperative ward. In other accounts, records were not moved with an individual so that key information was not available in a new setting. Again, in staff interviews these problems were often recognised but in contexts in which interviewees were defensive, tending to blame others rather than taking responsibility for their individual or collective actions.

It would be possible to interpret these co-ordination failures as clear evidence of failure by individuals or teams. But there is a corollary to this point: nurses and health-care assistants, in particular, could find themselves working in conditions in which they had limited information about individuals and their risks (e.g. patients had an unknown diagnosis) or in which records had not travelled with a patient from another location. It is possible, therefore, that individual members of staff behaved reasonably in the contexts in which they found themselves; the problems lay more with the overall co-ordination of treatment and care.

The larger and broader finding – the mid-range theory arising from the findings – is that individuals developed severe pressure ulcers in environments where there were problems with the prevailing culture. This cultural explanation binds together the otherwise separate points made above. In most accounts there was a combination of problems – some staff blaming colleagues or describing patients as ‘difficult’, poor documentation and failures to act on repeated clear warning signs (i.e. to step up care provision when a superficial pressure ulcer was observed), so that a severe pressure ulcer was ‘allowed’ to develop. In these contexts responsibilities were not clear and staff interviews pointed to problems with team working, extending beyond the specific accounts that we were producing.

Design: Pressure Ulcer Research Service User Network UK workshop model

We organised a workshop that was developed and facilitated by the PURPOSE PPI officer, the severe pressure ulcer study field researcher and one member of PURSUN UK. The workshop was designed along the broad lines of a public enquiry – albeit a benign one – in which participants would be invited to act as ‘expert witnesses’ in a case that was presented to them.

Three types of material were prepared before the workshop. First, one patient’s account of her health problem and treatment was used to create a brief for a simulated patient. The PURSUN UK member with specialist expertise in health-related role play took on the role of the patient from the case. The researcher conducting the inquiry then interviewed the simulated patient about their experiences. This was presented live at the workshop. Various simulated patient models have been used in the UK since the late 1970s, typically in communication skills training or assessment for health professionals. 32 The approach was adapted here for use in a research context. Second, professionals’ accounts of events were filmed and edited into short videos. Here, actors were given a brief, prepared by the workshop facilitators, and asked to improvise a piece to camera (they did not read from a transcript, in part to avoid using verbatim quotes, which might have led to the patient being identified). Third, a visual timeline of events was presented using Prezi software (Prezi, Budapest).

The workshop was held on 17 May 2012 in Leeds. It was attended by nine members of PURSUN UK, six members of the research project team and two NHS PPI managers.

Workshop evaluation

The workshop was, as far as we were aware, innovative and we were unsure how successful it would be. We therefore decided to evaluate it. Five participants took part in videoed interviews and one further audio-recorded interview was carried out by the PURPOSE PPI officer. The interviews took place both during the workshop, to provide immediate reactions, and afterwards, giving participants time to reflect before commenting. Three participants also chose to provide written feedback. Themes from the interviews and written feedback were then collated by the PURPOSE PPI officer. Participants were also offered the opportunity to provide anonymous feedback through a neutral third party but in the event this option was not used. Clips from the videoed interviews are available at http://youtu.be/bgg6zkbILrg (accessed 24 February 2015).

The workshop model

Participants found the metaphor of a public enquiry useful as it helped to convey the design of the field study and the rationale for the ‘expert witness’ roles in the workshop. Participants found the simulated patient interview engaging and valued the interactive nature of the session. One member of PURSUN UK commented that she would not have become involved in the project if it had required her to take part in a complex, paper-based exercise. The live interview also provided a snapshot into a part of the research process that few had previously experienced. As the simulated patient stayed ‘in role’ during discussions, workshop participants were able to briefly step into the shoes of the interviewer, asking follow-up questions and checking assumptions.

Impact on workshop participants

Service users reported an increased understanding of research processes in general as a result of the workshop. Some members also said that it had made them think more about their own and their family’s health, particularly in relation to preventing pressure ulcers. Some service users also reported an increased feeling of empathy for the health professionals dealing with such complex cases.

Members of the project team valued the dialogue with service users in a non-clinical context, despite some initial concerns about working in this way. Everyone felt that the collaborative nature of the workshop was important, particularly the fact that academics and nurses got to hear service users’ opinions first hand.

Face validity

The workshop also provided us with feedback – admittedly for just one of the eight accounts – about the validity of the account. Members of PURSUN UK arrived at a similar interpretation of events as the nurses and other experts involved in the formal analysis reported in Retrospective study of the development of severe pressure ulcers.

The research team had already identified a need to involve patients and carers in the critical review of severe pressure ulcers in the NHS (see Implementation project: the review of critical incidents). This was supported by PURSUN UK members and one member has worked with the study team during 2013. PURSUN UK also highlighted the importance of patient/carer engagement in pressure ulcer prevention and the role that professionals can play in facilitating that engagement. Conversely, we recognise that the severe pressure ulcer case study findings were filtered through the materials prepared for the workshop. This was necessary partly to make the findings accessible and partly to protect the anonymity of the individual whose account was used.

Findings and reflections

The substantive findings of the reviews were consistent with those from the retrospective study (Table 28). Both reviews using the new method identified additional contributing decisions and actions, over and above those identified in the parallel root cause analyses. The patients and carers did not identify the key decisions and actions themselves; their descriptions allowed the investigators to do so. In both cases the patients and carers valued the opportunity to ‘tell their own story’.

Feedback from ward staff

A meeting was held with the staff from the ward where the hospital pressure ulcer incident had taken place. This was to feed back the findings of the pilot investigation and to evaluate the process by eliciting staff views. Generally, staff thought that the process was more thorough than the current root cause analysis process and was more informative about the reasons why the pressure ulcer developed. They felt that the patient’s account not only told them a lot about the development of the pressure ulcer but also gave them important feedback about other aspects of care. There were observations in the patient’s account that they could not have obtained from any other source (e.g. an existing sore on the ankle). Some staff felt that it made ‘uncomfortable reading’ but that it was to be expected when a category 3 pressure ulcer had occurred. They agreed that it was no more uncomfortable than the current process. They acknowledged the value of more staff involvement in the process as people may take more note of the outcome of an investigation if they are part of the process. They also noted that nurses are currently carrying out root cause analysis without any training. Finally, the idea that an independent person should lead the investigation (i.e. not someone working on that ward) was supported.

Design

The approach was based on the systematic review methods recommended for questions of effectiveness20,127 and adapted to identify risk factor studies, with consideration of the methodological limitations including bias and confounding associated with observational studies. 83,85

Study eligibility

Methodological quality criteria were integrated into the inclusion and exclusion criteria of the systematic review, developed from principles of good research conduct in observational studies and randomised controlled trials that minimise bias. 82,86,89,128

Data sources

Fourteen electronic databases were searched, from inception until March 2010: Allied and Complementary Medicine Database (AMED), British Nursing Index (BNI), MEDLINE, EMBASE, PsycINFO, Cumulative Index to Nursing and Allied Health Literature (CINAHL), The Cochrane Library, ProQuest, Networked Digital Library of Theses and Dissertations, International Theses in Progress, Theses Canada Portal, Australian Digital Theses Program, Russian Academy of Sciences Bibliographies and Index to Theses. The search strategy (see Appendix 14) sought to identify all published and unpublished research studies investigating risk factors for the development of pressure ulcers. The search strategy was designed with guidance from the collaborative team and included pressure ulcer search terms,130 Ovid maximum sensitivity filters for prognosis and aetiology or harm and an Ovid maximum sensitivity filter for randomised controlled trials. 127

In addition, we hand searched specialist journals and conference proceedings, contacted 13 experts, searched the UK national research websites and performed a citation search on all included studies and systematic reviews identified in the search (see Appendix 14).

Study selection

Abstracts were screened for relevance by one reviewer (CG) and checked by a second (JN). Articles assessed as potentially relevant were obtained in full and reviewed against the eligibility criteria by one reviewer (CG or SC) and checked by another (JN). When the statistical methods were unclear and eligibility could not be determined, statistical review was undertaken (JB). Disagreements were dealt with through consensus.

Data extraction

When studies fulfilled the eligibility criteria data were extracted by a single reviewer (CG or SC) and checked by a second reviewer (JN). When data were missing from the publication, attempts were made to contact the authors. When duplicate publications of patient data sets were identified, the most detailed report was used for data extraction. Experts in the field were asked to review/data extract abstracts and articles not published in English (see Acknowledgements).

Quality assessment

There are no guidelines for the quality assessment of risk factor studies and so we developed an assessment framework based on guidelines for assessing quality and risk of bias in prognostic studies and on methodological considerations in the analysis, meta-analysis and publication of observational studies. 81–89 Each study was appraised by two reviewers (JN, SC) and the following methodological limitations were noted when present: baseline characteristics not adequately described; inadequate measurement of risk factors (e.g. record review); inappropriate cut-points used for continuous data; and time-dependent covariates included in the analysis without appropriate adjustment.

In addition, specific consideration was given to the following criteria:

1. Is there a sufficient number of events (rule of thumb: ≥ 10 events per risk factor)?

2. Are there sufficient data to assess the adequacy of the methods and analysis?

3. Is the strategy for model building (i.e. inclusion of variables) appropriate and based on a conceptual framework?

4. Is the selected model adequate for the design?

Each of the four criteria was assessed to see whether or not they were met (yes/no/partial/unsure), which provided a structured approach for the classification of overall study quality. We classified studies as being of high, moderate, low and very low quality using the following criteria:

• high-quality studies: ‘yes’ for all criteria

• moderate-quality studies: ‘yes’ for criterion 1 and at least two other criteria

• low-quality studies: ‘no’ for criterion 1 and ‘no’ or ‘partial’ for two other criteria

• very low-quality studies: ‘no’ for criterion 1 and ‘no’ or ‘partial’ for all three remaining criteria.

Data synthesis

Meta-analysis of the data was not feasible for this review because of heterogeneity in the study designs, patient populations, risk factor descriptors, interventions used and outcomes reported. As the main aim was to identify risk factors rather than quantify the effect size of the relationship between these factors and pressure ulcer development, a narrative synthesis was carried out. 127

For each study, all factors entered into multivariable modelling and those that emerged as significant (p ≤ 0.05) were identified. For studies using stepwise regression, we included non-significant factors (p ≥ 0.05) if these were reported in the final model as being independently associated with pressure ulcer development.

Risk factors were categorised into domains and subdomains by collating related factors from the source articles into a grouping (domain). An example of a domain and subdomain would be the domain of skin/ulcer status and the subdomains of stage/grade 1, existing pressure ulcer, previous pressure ulcer and general skin status. Evidence tables (see Appendix 15 for an example), were generated for each risk factor subdomain, with a summary narrative synthesis by subdomain and domain. For each subdomain, the total number of studies entering the variable, the total number of studies in which the variable emerged in the multivariable analyses and the quality of the studies are summarised. In the evidence tables (see Appendix 15 for an example), grade and stage of pressure ulcer are recorded as reported in individual studies.

Study quality

The included studies comprised seven high-quality, 10 moderate-quality, 27 low-quality and 10 very low-quality studies (Table 29). The low- and very low-quality studies had inadequate numbers of pressure ulcers and other methodological limitations.

Emerging risk factor domains/subdomains

The review identified 15 risk factor domains and 46 subdomains (Table 30). The number and quality of studies in which associated risk variables emerged in multivariable modelling and the number and quality of studies in which associated risk variables did not emerge are detailed in Table 30 [full evidence tables are available at: http://medhealth.leeds.ac.uk/downloads/download/657/systematic_review_evidence_tables (accessed August 2015)]. The review highlighted three primary risk factor domains:

1. Mobility/activity (immobility), with the subdomains of mobility subscales, mobility/activity, activities of daily living (ADL) and activity (bedfast/chairfast/immobile descriptors) emerging most consistently.

2. Skin/ulcer status, with the subdomain of stage/grade 1 pressure ulcer emerging most consistently. General skin status was also found to be important but the diverse variables (e.g. dry sacral skin, mottled skin, unhealthy skin, skin redness, baseline skin trauma) made interpretation difficult.

3. Perfusion, with the subdomain of diabetes emerging strongly in the high-/moderate-quality studies. Evidence from the large number of other perfusion-related variables suggests that factors that impair circulation increase the probability of pressure ulcer development but the evidence is limited by the large range of variable descriptors and study quality. Further research is needed in this area.