Probiotics for Antibiotic-Associated Diarrhoea (PAAD): a prospective observational study of antibiotic-associated diarrhoea (including Clostridium difficile-associated diarrhoea) in care homes

Database searches

We searched MEDLINE and EMBASE prior to starting this study, and our findings indicated that the evidence base for the use of probiotics is incomplete. In the light of our searches we have made recommendations about what should be done taking into consideration this uncertainty and about future research. 42 More specifically, we were not able to identify evidence for or against using probiotics to prevent AAD in people admitted to LTCF/care homes or in primary care. It is essential that research is carried out in the setting in which the evidence will be applied, the findings of hospital-based studies may not apply to the care home setting, as antibiotics, routes of administration and patient profiles differ between hospitals and care homes.

Probiotics for Antibiotic-Associated Diarrhoea stage 1: primary objectives

1. To conduct prospective systematic ascertainment of the incidence of AAD in care homes.

2. To allow an appraisal of the estimated sample size for a RCT in PAAD stage 2.

Probiotics for Antibiotic-Associated Diarrhoea stage 1: secondary objectives and auxiliary study objectives

1. To conduct prospective systematic ascertainment of antibiotic use in care homes.

2. To estimate the risk of AAD overall and from particular antibiotics in care home settings.

3. To identify barriers and implementation issues in conducting a trial of AAD prevention/amelioration in a care home setting.

4. To determine the prevalence of asymptomatic C. difficile carriage in residents within selected care homes.

The qualitative study

1. To conduct focus groups and individual qualitative interviews with care home residents, their family, care home staff and general practitioners (GPs).

2. To explore the ethical and practical issues of consent and assent, particularly the topic of advanced consent, for elderly residents who may/may not have capacity to consent.

Probiotics for Antibiotic-Associated Diarrhoea stage 2 objectives

The primary objective was to:

1. assess the effectiveness of probiotics taken in conjunction with antibiotic treatment in reducing the incidence of AAD.

The secondary objectives were to:

1. assess the effectiveness of probiotics taken in conjunction with antibiotic treatment in reducing the incidence of CDAD

2. evaluate the impact of probiotics in conjunction with antibiotic treatment on the functional status and QoL

3. evaluate the cost-effectiveness of probiotics in conjunction with antibiotic treatment in reducing AAD.

Study design

The PAAD study stage 1 was a prospective observational cohort study of antibiotic prescribing and associated AAD in care home residents (Figure 1). The study was conducted between November 2010 and March 2012 in care homes in South Wales. We aimed to recruit a total of 270 care home residents and follow up each resident for 12 months. The South East Wales Research Ethics Committee (REC) approved the study (10/WSE03/31). Agreement to conduct the study at the care homes was either given by the care home manager, regional manager or care homeowner (if privately owned) or given by the local authority (LA).

FIGURE 1.

Study design. CRF, case report form.

Study objectives

Participants and recruitment

Study procedures

Data collection

Prior to the recruitment of residents, care home staff were asked to complete a care home information CRF. The purpose of this CRF was to elicit summary information about the care home characteristics and, to that end, it incorporated questions asking about the number of beds, number of residents, staffing levels and whether or not there was an infection control policy in place. A further CRF was completed at the end of the study to provide a measure of change in number of residents and staffing levels over the study period.

Data about residents participating in the observational study were collected in two phases (1) ‘baseline’ data were collected as soon as possible following consent and (2) ‘follow-up’ data were collected when a resident recruited into the study was prescribed antibiotics at any point during the 12 months following recruitment (Figure 2).

FIGURE 2.

Data flow of recruited residents.

Serious adverse event reporting was undertaken for all recruited residents regardless of whether or not they were prescribed antibiotics during the 12-month observational period. The CRFs included standard screening and assessment tools (among other questions devised by the research team) (Table 2).

TABLE 2 - Screening and assessment tools included in CRFs

Tool	Description of tool	When recorded
MUST	MUST is a five-step screening tool used to assess nutritional risk. The algorithm requires information about current body mass index, weight loss and current health status in order to provide the health professional with a score used to guide nutritional management of the patient46	When recruited into the study (CRF 03) Each time the resident was prescribed an antibiotic (CRF 05)
Clinical Frailty Scale	The Clinical Frailty Scale is a visual aid used to assess a person’s perceived frailty. It has been used as a predictor of death or need for entry into care facility. The tool uses nine pictures, each with brief descriptions, on a Likert-type scale in which options vary between very fit and very severely frail or terminally ill47	When recruited into the study (CRF 03)
BSC	The BSC is a visual aid used by health professionals as a diagnostic marker of digestive health. The scale depicts seven pictures of human faeces in various forms, ranging from faeces described as type 1 (separate hard lumps, like nuts; hard to pass) to type 7 (watery, no solid pieces, i.e. entirely liquid)	On the CRFs that accompanied the baseline stool sample (CRF 04) When a stool bowel movement was recorded following an antibiotic prescription (CRF 06) and/or diagnostic stool sample(s) (CRF 07)

Table 3 details, for each PAAD study CRF, time point for data collection, a brief description of the type of data collected and who had overall responsibility for collecting the data. The research team requested that the original CRF should be sent to the PAAD study trial manager and a copy kept in the site file in each care home.

TABLE 3 - Data collection table

CRF	Data collected	Time point	Collected by
CRF 01 – care home information	Type, size, occupancy of home. Staffing levels and infection control training	Prior to commencing recruitment	Care home staff
CRF 02 – recruitment	Eligibility confirmation, age, gender, capacity status	Immediately post consent	Care home staff
CRF 03 – resident characteristics	Medical history, recent hospitalisation, recent antibiotic use	Immediately post consent	Care home staff
CRF 04 – baseline stool sample	Date and time of stool collection. Consistency of stool	Within 1 week from consent	Care home staff
CRF 05 – antibiotic prescription	Name, dose, route, frequency, duration and indication of antibiotic	Day of commencement of antibiotic treatment until duration end date	Care home staff/NISCHR CRC research officers
CRF 06 – weekly stool monitoring chart	Time and consistency of bowel movements. Whether or not a stool sample was taken	Day of commencement of antibiotic treatment, every day for duration of antibiotic + 8 weeks	Care home staff/NISCHR CRC research officers
CRF 07 – diagnostic stool sample	Date and time of stool collection. Consistency of stool	When participant develops diarrhoea within the 8-week follow-up period following antibiotics	Care home staff
CRF 08 – withdrawal	Date and reason for withdrawal	Following a withdrawal	Care home staff/NISCHR CRC research officers
CRF 09 – SAE reporting	Details of adverse event (outcome, description, seriousness, expectedness and causality)	Following any SAE	Care home staff/NISCHR CRC research officers
CRF 10 – antibiotic notes search	Name, dose, route, frequency, duration and indication of antibiotics prescribed in the 3 months prior to study entry	At the end of the recruitment period at the care home	Care home staff/NISCHR CRC research officers

Statistical methods

Substantial changes to study protocol

Over the course of the study a number of changes were made to the PAAD stage 1 protocol. Initially, the intention was to include a sample of care homes of differing sizes from those with 20 beds or fewer to those with 100 beds or more. However, there were very few nursing and dual-registered care homes with 20 beds or fewer in the South-East Wales region. As the most important factor in the care home sample was care home type (nursing, residential and dual registered) it was decided to recruit three care homes of each type regardless of size.

Several changes were made to improve recruitment and data collection. These included allowing verbal consent (with witnesses present) to be provided if the resident or consultee was unable to sign the consent form and enabling the research team to take copies of recruited residents’ MAR sheets at specific time points to ensure the collection of valid concomitant medication data. It also came to light that many of the residents and their consultees were frustrated that they were unable to give consent at the time they were approached (i.e. directly after reading and digesting the information on the participant information sheet) because the protocol stated a mandatory 24-hour ‘consideration’ time before they could join the study. As a result of this the protocol was amended (with REC approval) to allow residents/consultees an opportunity to provide consent at the time of their choosing.

Care homes

Eleven care homes were recruited to the study. However, one withdrew before any residents were recruited. Residents were therefore recruited from 10 care homes: four nursing, four residential and two dual-registered homes. Nine homes were privately managed and one was managed by a LA. The median number of beds was 39.5 (IQR 31.0–50.0) and the median number of residents at the time of recruitment was 33.0 (IQR 28.0–50.0). The median number of staff working in a care home over a typical 24-hour period was 16.0 (IQR 14.0–25.0), with 20.5% (n = 41) categorised as short-term staff members (i.e. employed for less than 12 months) (Table 4 and Figure 3).

FIGURE 3.

Care home flow diagram.

Participants

Three of the 389 residents (or consultees) approached were ineligible and 107 declined participation. A total of 279 residents were therefore recruited (71.7%). Of those recruited, 19 withdrew, 16 (84%) because they moved to a non-participating care home. Five of the 19 residents who withdrew from the study also withdrew permission for data already collected to be used; therefore, our analyses are based on a maximum of 274 residents (Figure 4). There were 81 hospitalisations reported during the study period, with at least one hospitalisation reported for 58 (21.2%) residents (incidence rate of 0.14 hospitalisations per resident-year, 95% CI 0.09 to 0.20 hospitalisations). In total, 64 residents died during the study period. No deaths were deemed study related. Residents were observed for a median of 310 days (IQR 230–364 days).

FIGURE 4.

Resident flow diagram. a, Some prescriptions overlapped in time and so did the stool monitoring periods in these instances.

Descriptive data

Residents had a median age of 86 years (IQR 82–90 years), with 20.4% (n = 56) < 80 years old, 57.7% (n = 158) between 80 and 90 years old and 21.9% (n = 60) older than 90 years. The majority of residents (75.9%) were female. Overall, 28.5% (n = 78) had capacity to provide informed consent for themselves. Few residents had any of the prespecified relevant serious medical conditions. At baseline, 7.7% (n = 21) had faecal incontinence with loose stools, 1.8% (n = 5) had diarrhoea and 66.8% (n = 183) routinely used incontinence pads. A ‘very fit to managing well’ classification was attributed to 13.5% (n = 37) of residents, 51.1% (n = 140) were classed as ‘vulnerable to moderately frail’ and 35.4% (n = 97) as ‘severely frail to terminally ill’. Nursing homes had more frail residents than residential homes. At baseline, 63.0% (n = 172) were classified as having a low nutritional risk status, 14.3% (n = 39) as medium risk and 22.7% (n = 62) as high risk. In the 4 weeks prior to recruitment, 6.9% (n = 19) had been admitted to hospital and 20.8% (n = 57) of residents were prescribed antibiotics (Table 5).

Interim analysis

At the interim time point, we had approached 363 care home residents and recruited 260, giving a recruitment rate of 72% (95% CI 67% to 77%). We recorded at least one antibiotic prescription for 119 residents, giving an antibiotic prescribing rate of 46% (95% CI 40% to 52%). There were 152 antibiotic prescriptions recorded at the interim time point, 51 of which with a corresponding episode of AAD, giving an AAD rate of 34% (95% CI 25% to 42%). As all criteria had been met, we were permitted to begin designing a RCT of probiotics to prevent/ameliorate AAD in care home residents (Table 6).

Antibiotic prescriptions

There were 609 antibiotic prescriptions recorded over the study period, with 73.7% (n = 202) of residents being prescribed at least one antibiotic course. We found an incidence of 2.16 antibiotic prescriptions per care home resident-year (95% CI 1.90 to 2.46 prescriptions).

Antibiotics were prescribed for a median of 7.0 days (IQR 6.0–7.0 days), with 14.7% (n = 88) prescribed for less than 5 days, 74.9% (n = 447) prescribed for between 5 and 7 days, 6.7% (n = 40) prescribed for between 8 and 10 days and 3.7% (n = 22) prescribed for more than 10 days.

Antibiotics were most commonly prescribed for UTIs [29.3% (n = 177) of all antibiotic prescriptions], followed by URTIs [28.8% (n = 174)], skin/connective/soft tissue infections [18.2% (n = 110)] and lower respiratory tract infections (LRTIs) [18.0% (n = 109)]. The proportion of antibiotics prescribed for each indication varied by care home type. For example, prescriptions for UTIs accounted for 25.4% (n = 71) of all prescriptions in nursing homes, 43.5% (n = 70) in residential homes and 21.8% (n = 36) in dual-registered homes (Figure 5). The five most commonly prescribed antibiotics were amoxicillin [30.6% (n = 186) of all prescriptions], trimethoprim [12.7% (n = 77)], flucloxacillin [10.4% (n = 63)], nitrofurantoin [9.5% (n = 58)] and co-amoxiclav [8.7% (n = 53)]. A wide range of antibiotics were prescribed for each indication (Figure 6). The fluoroquinolone ciprofloxacin, previously implicated in CDAD, was prescribed relatively uncommonly, accounting for 3.0% (n = 18) of all antibiotic prescriptions recorded during the study period. Prescriptions of ciprofloxacin were primarily given for UTIs [44.4% (n = 8) of ciprofloxacin prescriptions] and skin/connective or soft tissue infections [27.8% (n = 5)]. Ciprofloxacin was also prescribed for LRTIs in two cases, for one UTI, for one gastrointestinal infection and also in one instance as prophylaxis.

FIGURE 5.

Reason for antibiotic prescriptions by care home type.

FIGURE 6.

Type of antibiotic prescribed by indication.

While the total number of antibiotic prescriptions each month varied considerably (range 0.09–0.29 prescriptions and average 0.21 prescriptions per resident), there was no obvious marked seasonal variation (Figure 7).

FIGURE 7.

Antibiotic prescription rate by study month (stacked by antibiotic type). The height of each bar represents the total antibiotic prescribing rate per resident for the corresponding study month.

Compared with nursing home residents, those residents from dual-registered care homes had a significantly lower chance of being prescribed antibiotics during the study period (OR 0.38, 95% CI 0.18 to 0.79; p = 0.009). The odds of being prescribed an antibiotic during the study period were 2.64 times higher for residents who had been prescribed antibiotics in the 4 weeks before study entry (95% CI 1.17 to 5.99; p = 0.020). Exposure to antibiotics was similar in residents who were vulnerable to moderately frail or severely frail to terminally ill and those who were very fit to managing well (Table 7).

Antibiotic-associated diarrhoea

Three antibiotic prescriptions (from two residents) provided no corresponding stool monitoring data. From the remaining 606 antibiotic prescriptions there were 571 stool monitoring periods, ranging between 1 and 11 weeks. The discrepancy between the number of prescriptions and monitoring periods arose because residents could be prescribed multiple antibiotics in the same week (hence, there was only one ongoing monitoring period). There were 447 unique episodes of AAD reported, with 43.5% (n = 87) of residents who were prescribed antibiotics experiencing at least one episode of AAD during the study period. There were 0.57 episodes of AAD per care home resident-year for those prescribed antibiotics (95% CI 0.41 to 0.81 episodes).

After controlling for length of stool-monitoring period, the odds of developing AAD during a stool monitoring period were more than twice as high in residents who were prescribed co-amoxiclav (OR compared with no co-amoxiclav prescription 2.19, 95% CI 1.06 to 4.52; p = 0.033), with the first AAD episode also occurring sooner (median time to first AAD episode for residents who were and were not prescribed co-amoxiclav was 23 days compared with 28 days, respectively; HR 2.08, 95% CI 1.18 to 3.66; p = 0.011). Time to first AAD episode was also shorter for residents who routinely used incontinence pads than for residents who did not (median 26 vs. 42 days, respectively; HR 2.54, 95% CI 1.26 to 5.13; p = 0.009). Compared with residents in nursing homes, those in residential homes were significantly less likely to develop AAD during the study period (OR 0.12, 95% CI 0.05 to 0.27; p < 0.001) (Table 8) and experienced a slower time to first AAD episode (median for residents in residential homes 56 days, median for residents in nursing homes 21 days; HR 0.14, 95% CI 0.06 to 0.32; p < 0.001) (Table 9 and Figure 8).

FIGURE 8.

Survival curves illustrating the association between care home type and time to first AAD episode.

Clostridium difficile-associated diarrhoea

Of the 447 unique episodes of AAD, corresponding microbiological data from stool samples were available for only 55. C. difficile was cultured from eight of the sample. C. difficile was also detected in a further five stool samples taken after residents experienced loose stools (i.e. BSC type 5–7 stools) although the frequency of stools did not meet our definition of AAD. The 13 samples were obtained from nine residents in the same care home. In total, 12 samples were toxin B positive and there were nine different ribotypes (005, 010, 014, 020, 026, 027, 106, 160 and 193). No ribotype was found in more than one resident (Table 10).

Participant flow and recruitment

Of the 274 residents recruited, 80.7% (n = 221) provided a stool sample at study entry. Samples were collected from all 10 participating care homes, with collection rates between care homes varying from 66.7% to 96.4% (Table 11). Participants who provided samples had a median age of 86.0 years (IQR 82.0–90.0 years) and 78.3% (n = 173) were women. Over one-fifth (n = 49) had been prescribed antibiotics in the 4 weeks prior to study entry and 7.2% (n = 16) of participants had been admitted to hospital in this time frame. There were no statistically significant differences between participants who did and did not provide samples (Table 12).

Baseline data

Outcomes and estimates

Antibiotic resistance

Organisms which were resistant to one or more of the tested antibiotics were found in 51.6% (n = 114) of participants. Enterobacteriaceae resistant to ciprofloxacin were found in 47.1% (n = 104) of participants, to gentamicin in 12.2% (n = 27), to third-generation cephalosporins in 10.9% (n = 24), to meropenem in 1.9% (n = 4) and to vancomycin in 1.4% (n = 3). There was wide variation in the proportion of participants providing stool samples containing bacteria resistant to various antibiotics in different types of care homes, with care home-level intracluster correlations (ICCs) ranging from 0.00 for meropenem and vancomycin to 0.16 for gentamicin (Table 16). Of the four residents who provided samples containing meropenem-resistant isolates, one was prescribed an antibiotic (amoxicillin) and none had been hospitalised in the 4 weeks prior to study entry. Although resistance to ciprofloxacin, gentamicin and third-generation cephalosporins was mainly found in E. coli [95.2% (n = 99), 63.0% (n = 17) and 73.9% (n = 17), respectively], resistance to meropenem was exclusively found in Pseudomonas spp. (Figure 9). Resistant bacterial load, expressed as a percentage of total bacterial load, varied between antibiotic classes (Figure 10). For example, where isolates resistant to ciprofloxacin were found in stool samples, ciprofloxacin-resistant isolates accounted for less than 1% of the total quantity of isolates cultured in samples 12.8% of the time.

FIGURE 9.

Cultured isolates and their resistance to antibiotics.

FIGURE 10.

Percentage of resistant bacterial load by antibiotic class.

Factors associated with carriage of antibiotic-resistant bacteria

The odds of participants carrying antibiotic-resistant bacteria in their stools significantly increased with age (OR for 80 years or older 2.54, 95% CI 1.23 to 5.26; p = 0.012) and previous antibiotic use in the 4 weeks prior to study entry (OR 2.21, 95% CI 1.11 to 4.43; p = 0.025). The odds of carrying antibiotic-resistant bacteria were significantly lower for participants in residential homes than for those in nursing homes (OR 0.44, 95% CI 0.21 to 0.91; p = 0.028). After controlling for the age of the resident, clinical frailty status was not significantly associated with carriage of antibiotic-resistant bacteria (Table 17).

Prevalence of Clostridium difficile

Clostridium difficile was cultured in 7.2% (n = 16) of stool samples. The prevalence varied between care homes, from no C. difficile detected in four care homes to C. difficile being detected in 19.4% (n = 7) of samples from a single home.

One of the 16 samples from which C. difficile was cultured was toxin B negative, with the remaining 15 being toxin B positive. There were 11 different ribotypes (001, 002, 005, 015, 021, 027, 062, 103, 106, 160 and 193). In the majority of cases, ribotypes were unique to each home, with one instance of a ribotype occurring within the same home twice (n = 106). The care home level ICC was 0.29 (Table 18).

Barriers and implementation issues identified and lessons learned

Summary of main findings

This prospective observational cohort study of 274 residents from 10 care homes across South Wales found that antibiotics were prescribed at a rate of 2.16 per resident-year, with almost three-quarters of residents prescribed at least one antibiotic course during the 16-month study period (median length of follow-up 310 days). Residents were over two and a half times more likely to be prescribed antibiotics during the study period if they had been prescribed antibiotics in the 4 weeks prior to study entry. Residents were prescribed a wide range of antibiotics for each indication. The incidence of AAD in those prescribed antibiotics was 0.57 episodes per resident-year, with 43.5% of residents prescribed antibiotics experiencing at least one episode of AAD. Not all episodes of diarrhoea following antibiotic use can be ascribed to antibiotics, and our study does not seek to demonstrate causality, merely association. CDAD occurred in less than 15% of residents who developed AAD and for whom a stool sample was sent for microbiological analysis. Although all CDAD episodes were found in residents from the same care home, the unique ribotypes suggest that the episodes were not associated with an outbreak. Residents in nursing homes were more likely to be prescribed antibiotics and experience AAD. Residents were more likely to experience AAD if they had been prescribed co-amoxiclav or if they routinely used incontinence pads. The majority of residents (80.7%) provided a stool sample at baseline. Analysis of these samples demonstrated that there were high levels of carriage of antibiotic-resistant organisms in care home residents, particularly of ciprofloxacin-resistant organisms. Although recent hospitalisation was not associated with carriage of resistant organisms, recent antibiotic use was. Carriage of resistant organisms was significantly higher in nursing homes than it was in residential homes, even after controlling for clinical frailty. C. difficile was more common in nursing homes, but there was little suggestion of clustering by care home type.

Specific aims of the qualitative exploration of practical and ethical issues

1. To collect data on the various merits and problems associated with a number of models of consent for both PAAD stage 1 and PAAD stage 2 for residents who either have capacity or lack capacity.

2. To establish the views of residents, relatives, care home staff and GPs regarding participation in two study designs (observational vs. CTIMP).

3. To establish the feasibility and acceptability of taking advance consent for research trial procedures (as would be required in PAAD stage 2) in care home residents.

4. To establish views of relatives, residents and care home staff on participating in PAAD stage 1.

5. To establish care home staff views about the study training for staff, the requirements of the study and its impact on the care home including the implications of potentially finding C. difficile in the care home population.

We used qualitative research methods because this allowed us to explore in depth respondents’ views and experiences, including topics that we were unable to predict in advance. We thought it important to understand the views of a range of stakeholders and, so, our respondents included residents, relatives, care home staff and GPs who have a responsibility for the general medical care of residents and who may be asked to assess eligibility for research studies. We recruited through 10 care homes in South Wales, which were participating in PAAD stage 1. Data collection was undertaken through a combination of face-to face interviews with residents, relatives and GPs to facilitate in-depth reflection of respondents’ own involvement in the study and focus groups among care home staff to facilitate discussion about their collective involvement in research.

A shorter version of this exploration of ethical and practical challenges of conducting research in care home settings has previously appeared in the journal Trials. 53

Recruitment of residents

Care home staff approached residents whom they felt had the mental capacity to consent to the qualitative study. Residents were considered eligible to participate in the qualitative study even if they had not given consent to participate in PAAD stage 1. A total of 14 residents consented to the study, all of whom had consented to take part in PAAD stage 1. A research nurse conducted interviews with informants in a private room in the care home, usually the resident’s bedroom. Interviews lasted between 9 and 54 minutes, with an average of 23 minutes. Participating residents were offered £25 in recognition of their time.

Recruitment of relatives

Relatives of residents, four partners and 10 sons or daughters, were also invited by the care home staff to be interviewed (all of whom had given advice as a personal consultee that their relative should be part of PAAD stage 1) and all consented. Once again, interviews were conducted in a private room within the care home by a research nurse. Interviews lasted between 12 and 31 minutes, with an average of 19 minutes. Participating relatives were offered £25 in recognition of their time.

Recruitment of care home staff

Each care home was asked to nominate three members of staff who were most closely involved with the PAAD study. Staff were invited to participate in a focus group that took place at a city-centre hotel. A total of 19 staff from 10 care homes participated in the focus groups. Two focus groups were held, one with senior staff (10 participants) and the other with junior staff (nine participants). Each group discussion lasted approximately 90 minutes. The focus groups were facilitated by a research nurse and a qualitative researcher, both of whom were experienced in running focus groups. Neither had had prior contact with the care home staff regarding the PAAD study. As care home staff were participating ‘off duty’, £25 was offered to each participant. Lunch was provided, and travel expenses were reimbursed.

Recruitment of general practitioners

Senior care home staff were asked to name the main GP who attended residents within the care home. Letters of invitation were sent to all 11 GPs named and two GPs responded. In view of this, the researchers directly contacted 69 other GPs in the health board area to request an interview, which resulted in 10 GPs agreeing to take part in face-to-face interviews. Three of these GPs were aligned to a care home participating in PAAD stage 1, while seven were not but nevertheless attended residents in care homes. GP interviews lasted between 20 and 35 minutes with an average of 27 minutes. GPs were offered £50 for their time.

Background

Results from PAAD stage 1 demonstrated sufficient grounds for progressing to stage 2: a double-blind RCT to assess the clinical effectiveness of probiotics taken in conjunction with antibiotic treatment in reducing the incidence of AAD. The process required to set up the RCT was lengthy and resource intensive, largely because the study involved a CTIMP. The Medicines for Human Use (Clinical Trials): Regulations 200448 requires a favourable opinion from an ethics committee, and authorisation from the competent authority, before a CTIMP can begin. The challenges encountered in obtaining statutory approvals, and in the process of research governance in care homes, both unique to establishing a RCT in this research-naive environment, resulted in significant delay. A number of strategies were employed to minimise the impact of these issues. However, during this period of delay, new evidence emerged from the probiotic lactobacilli and bifidobacteria in AAD and C. difficile diarrhoea in the elderly (PLACIDE) trial55 regarding the effectiveness of probiotics in reducing the incidence of AAD and CDAD in older hospital inpatients. As a result, discussion took place with members of the Trial Management Group (TMG), Trial Steering Committee (TSC) and the National Institute for Health Research (NIHR) Health Technology Assessment (HTA) funding programme (our funders), and the decision was made not to progress to the recruitment phase of PAAD stage 2. Therefore, the planned objectives, to evaluate the clinical effectiveness and cost-effectiveness of probiotics in conjunction with antibiotic treatment in reducing AAD, could not be delivered.

Objectives

The primary objective was to assess the effect of probiotics, taken in conjunction with antibiotics, on the incidence of AAD. This would be ascertained by comparing the proportion of residents in each arm who experienced at least one episode of AAD during the 8 weeks following randomisation.

The PAAD study stage 2 also had a secondary objective: to assess the cost-effectiveness of probiotics for AAD in care home residents. Total costs would be determined by multiplying the recorded data on health-care resource use (including the cost of the probiotic) by relevant unit costs and assessing these against quality-adjusted life-years generated from the European Quality of Life-5 Dimensions (EQ-5D) data. 56 Standard economic evaluation methods57 for dealing with skewed cost data, parameter uncertainty and joint uncertainty in the cost-effectiveness ratio were to be applied.

Design and aims

The PAAD study stage 2 was planned to be a multicentre, double-blinded, placebo-controlled, two-arm, individually randomised trial of the effectiveness of probiotics taken in conjunction with antibiotic treatment in reducing incidence of AAD in care home residents.

The study design involved obtaining consent for the resident to be randomised to receive a probiotic or matched placebo taken in conjunction with an antibiotic, should an oral antibiotic be prescribed by a responsible clinician in the usual course of care at any point during the subsequent 12 months. Residents and their relatives (where applicable) would be fully informed of the study using written materials and posters in care homes, supplemented with verbal explanations. Consent was to be taken from residents who were able and willing to provide this. In the event of cognitive impairment that limited ability to provide fully informed consent, relatives and/or personal legal representatives would be approached to provide consent on their behalf.

Consented, enrolled residents who were prescribed an oral antibiotic in the course of usual care at any time during the 12 months after recruitment would be assessed by the GP to ensure that they remained eligible to participate in the trial. If the resident was eligible, he or she would be individually randomised to receive probiotics (VSL#3), or placebo, in addition to the antibiotic prescription. Residents would be followed up by a research nurse for 8 weeks following randomisation (see Appendix 1).

The aim was to randomise 400 residents (200 per arm) from approximately 24 care homes. We planned to recruit from care homes that had at least 50 residents. Smaller care homes would be considered if the recruitment potential in these care homes was considered adequate. GPs serving the recruited care homes would have played a pivotal role in recruitment and it was therefore essential to secure their commitment to the study at the same time as the commitment of the care home (see Appendix 2).

Participants

Residents would be eligible for the study if they met all of the inclusion criteria and none of the exclusion criteria. Inclusion criteria at consent were resident in a care home for 24 hours or more, with a minimum planned period of residence of 1 month, and able to provide informed consent or have a representative who could provide consent for inclusion. Residents met the exclusion criteria if they were severely immunocompromised, had an artificial heart valve in situ, had a medical history of acute pancreatitis, required nasojejunal feeding/nasogastric feeding, currently had a colostomy or did not have capacity and were already receiving a probiotic.

Residents who had provided advanced consent and remained eligible for study medication would be randomised to receive either the probiotic or the placebo once an antibiotic for an acute infection had been prescribed by selection of the next sequentially numbered study medication pack.

Intervention

There were two planned intervention arms. The active arm was the nutritional supplement VSL#3 probiotic, which contained approximately 450 billion live lactic acid bacteria and bifidobacteria, together with maltose and silicon dioxide. There were eight different strains of potentially beneficial bacteria: Streptococcus thermophilus Orla-Jensen 1919, Bifidobacterium breve Reuter 1963, Bifidobacterium longum Reuter 1963, Bifidobacterium infantis Reuter 1963, Lactobacillus acidophilus (Moro 1900) Hansen and Mocquot 1970, Lactobacillus plantarum (Orla-Jensen 1919) Bergey et al. 1923, Lactobacillus paracasei Collins et al. 1989 and Lactobacillus delbrueckii ssp. bulgaricus (Orla-Jensen 1919) Weiss et al. 1984.

The matching placebo arm of the study consisted of freeze-dried powder (4.4 g), matched for taste, consistency, odour and colour.

The resident was to be given one sachet twice a day for 21 days. The sachet would be opened and the contents stirred into 25–50 ml of cold water or any non-fizzy drink or sprinkled onto cold food and consumed immediately. The study medication would be given in between the antibiotic therapy and not in conjunction with the antibiotic. The study medication would be commenced within 72 hours of the resident being prescribed an antibiotic.

Outcomes

The primary outcome was the occurrence of at least one episode of AAD during the 8 weeks following randomisation. AAD was defined as three or more loose stools (defined as 5–7 on the BSC) in a 24-hour period following a period of normal stool consistency.

Secondary outcomes included the occurrence of CDAD. This was defined as having a diagnostic stool sample (following the occurrence of AAD) containing C. difficile toxin A or B. The duration, frequency and recurrence of AAD were also to be measured, where the duration of AAD was defined as the total number of consecutive 24-hour periods that a resident had AAD. The frequency of AAD was defined as the number of episodes of AAD a resident experienced during the 8-week follow-up period. AAD episodes were considered unique only if they were separated by a period of at least 3 days of ‘normal’ stool consistency (defined as 1–4 on the BSC).

Other secondary outcome measures included:

1. Recovery from the illness that triggered the prescription of antibiotic treatment.

2. Adherence to the study intervention and antibiotic treatment.

3. Health-care resource use and unplanned hospitalisations, including all-cause and AAD related, during the 8-week follow-up period.

Health-related QoL was also to be measured using the EQ-5D56 as a self-reported or proxy measure. Adverse events included reported symptoms such as vomiting, abdominal pain, excessive flatulence, bloating and skin rashes. All-cause mortality during the 8-week follow-up period would also be measured.

Sample size

A total of 400 residents (200 per arm) needed to be randomised in order to achieve 80% power, at the 5% level, to detect a 50% relative reduction in the incidence of AAD in those given probiotic alongside antibiotic treatment, compared with placebo alongside antibiotic treatment. The sample size was calculated using the interim data analysis from PAAD stage 1 assuming an AAD incidence of 25% in the placebo arm. The sample size was adjusted for a 20% rate of withdrawal and loss to follow-up (including death) using data from PAAD stage 1.

Taking into consideration the results from the interim analysis performed in PAAD stage 1 and a more recent interim assessment of antibiotic prescribing and AAD (from first antibiotic prescription), assuming an AAD incidence of 25% in the placebo arm, in order to randomise 400 residents, advanced consent was required from at least 607 residents (assuming that 66% would be prescribed at least one course of antibiotics during the 12-month monitoring period and subsequently randomised). Approximately 1214 residents would be approached in order to achieve this consent rate (assuming that only 50% will provide consent). With an average of 60 residents per home, we needed to recruit from a minimum of 21 care homes. In order to allow for care home drop-out, more than anticipated withdrawals/drop-out/declines and fewer than anticipated care homes, we aimed to recruit from 24 care homes.

Data collection

Data would be collected for 8 weeks following randomisation, to include a daily diary which records symptoms and medications used and side-effects, a stool chart and QoL questionnaire (EQ-5D).

Data collection would be undertaken by care home staff, following training in completion of study-specific data collection tools, supported by research nurses who would visit care homes regularly to assist with recruitment and randomisation and study-specific procedures.

Analysis

For the primary analysis, a logistic regression model would be constructed with ‘AAD during the 8-week follow-up’ (YES/NO) as the dependent variable and ‘study arm’ (probiotic/placebo) as an explanatory variable. The analysis would be adjusted for the potential clustering of residents within care homes via multilevel analysis, and potential, prespecified confounding/risk factors as specified in the statistical analysis plan. The primary analysis was to be based on the intention-to-treat principle. In a similar way to the primary analysis, a logistic regression model would be constructed for the secondary analysis, with ‘returned positive stool sample for C. difficile toxin A and B following AAD’ (YES/NO) as the dependent variable and ‘study arm’ (probiotic /placebo) as an explanatory variable, in order to investigate the effect of probiotics on the development of CDAD during the 8-week follow-up period. In an analogous way to the primary analysis, this analysis was to be adjusted for clustering (if present) and confounding factors.

Ethics approval

The REC provides an independent review of research involving humans, and is required to ensure that the research reaches the required ethical standards. 58 The Medicines for Human Use (Clinical Trials): Regulations 200448 require that those incapable of giving consent should be given special protection. The REC must consider whether or not there is justification provided by the research team for involving adults lacking capacity in the trial. An application for ethics approval was submitted to the South East Wales REC Panel D. Ethics approval for the study was not granted initially, as the REC requested further clarification from the research team on a number of issues including the remuneration to care homes for taking part in the research, the potential recruitment of residents without mental capacity and also the assessment of mental capacity.

Data obtained from PAAD stage 1 showed that the majority of eligible residents (69%) lacked capacity. This evidence, together with the need for a representative sample population without bias, was provided as support for the inclusion of residents without capacity to the REC. The MCA (2005)45 states that there is a presumption that a person has mental capacity, unless there is concern to the contrary. If concerns were raised about a resident’s capacity to consent to participate, an assessment of mental capacity would be conducted initially by senior care home staff familiar with the resident. If the care home staff or researchers were uncertain about a resident’s mental capacity, the research nurses trained and experienced in assessment of mental capacity would conduct an assessment using a standardised template and record the findings.

The REC required further justification for inclusion of residents without capacity from the research team in order to meet the conditions set out in the Medicines for Human Use (Clinical Trials) Regulations 2004 Informed Consent in Clinical Trials Guidance. 59 Preliminary findings available from the PAAD stage 1 showed that there were significant differences between the groups of participants with and without capacity in terms of prescription rates for antibiotics, frailty scores, nutritional risk and rates of AAD. These data were provided to the REC as evidence for the inclusion of residents without capacity as the population most at risk of antibiotic treatment and the associated adverse consequences and, therefore, most likely to benefit from the intervention. Ethics approval was granted following this evidence.

The process for gaining REC approval for this trial took 4 months. In contrast, PAAD stage 1 received full approval from a different REC within 1 month based on the initial application.

Understandably, the process required for ethics approval for a RCT in care home populations has been more complex than in a non-CTIMP study based in the same population. Gaining approval took a longer period of time, and significantly more evidence was required for the justification for including residents without capacity.

Competent authority approval

The intervention that was planned to be used in the trial was a probiotic preparation (VSL#3) that is commercially available to the general public as a food supplement. The research team presented evidence that the manufacturers class the probiotic as a food supplement that is commercially available without prescription and is low risk, and that its use within the trial was within its indication in the investigator brochure. The Medicines and Healthcare products Regulatory Agency (MHRA) subsequently confirmed that the trial was classified as a CTIMP and did require clinical trial authorisation (CTA).

Before granting CTA, the MHRA requires proof that the whole manufacturing process has been carried out in accordance with terms and conditions of a marketing authorisation (MA) for the product to ensure that strict quality assurance is maintained throughout the process. A qualified person (QP) provides certification for each batch of a medicinal product in accordance with good manufacturing practice. The product used in this trial had a complex manufacturing and importation process before reaching the trial site. The request for CTA was not initially accepted as the MHRA required further information regarding the MA; however, the status of the product as an IMP was difficult to reconcile with its licenced manufacture and production as a food supplement, despite being listed as a medicine in the British National Formulary.

The delays in MHRA approval and the contract impacted on labelling of the IMP and placebo and delayed commencement of randomisation. This in turn had a knock-on effect on the start of enrolling residents into the trial and obtaining consent, as it was undesirable to have a large gap between consent and beginning the process of randomising residents should they be prescribed an antibiotic.

The classification as a CTIMP had a significant impact on the time to commencement of the trial. The discrepancy between the statutory authorisations requirement for an IMP, and the absence of any such documentation for the product used in the trial, had a significant impact on the continuation of the trial.

Research governance

Research governance in care homes is complex owing to their status as non-NHS institutions, the ambiguous nature of the environment as both a public and a private space and the unique nature of the setting where the community of residents must be considered as a whole in addition to each individual. 60

The contractual and financial arrangements of the resident as self-funding, LA funded or a mix of both will have an impact on research governance. This is further complicated by whether the care provider is a commercial or non-commercial organisation (independent sector, voluntary sector or LA owned). This means that residents may be covered by different research governance processes, and may change from one process to another if their individual circumstances change.

As non-NHS institutions, there was a lack of clarity regarding where responsibility for research governance lay for the care homes involved in this trial in Wales. This was further clouded by the classification of the trial as a CTIMP and the fact that, although care homes would be the main sites, GP practices would also be sites and are NHS organisations. This resulted in delays with gaining site-specific assessment (SSA), as non-NHS SSAs are carried out by the REC; however, the responsibility for NHS SSAs lies with NHS research and development departments. The trial also covered a number of NHS trusts and LAs’ social service departments. Each of these organisations conducted their own approvals process.

The question of who would be acting as PI for the trial sites also raised difficulties. The structure varies between each individual care home; however, those with the most responsibility on a day-to-day basis are care home managers. Care home managers usually have a nursing background, but some are not registered health-care professionals and have a managerial background.

The regulations governing CTIMPs conducted at more than one site define ‘investigator’ as the ‘authorised health professional’ responsible for the conduct of that trial at a trial site and, if the trial is conducted by a team of health professionals at a site, the investigator is the leader responsible for that team. The regulations define health professional as a doctor, dentist, nurse or pharmacist. The MHRA advised that only care home managers who fulfilled these criteria would be eligible to act as investigator for the trial. It also highlighted that GCP guidelines state that each individual involved in conducting a trial shall be qualified by education, training and experience to perform his or her tasks. Although GCP training and extensive study-specific training was provided to care home managers acting as PI for the site, there was almost no prior research experience among this group. Extensive training and support is required by those with responsibility as PI in research-naive care homes.

Research governance in care homes lacks clarity. The process will differ between care homes, care home residents, LAs and regions. Where care homes are participating in trials, particularly those involving IMP and both NHS and non-NHS sites, the process for identifying responsibility for research governance is problematic and resource intensive.

General practitioner recruitment

There is no standardised model of health-care provision for residents in care homes. Individual care homes have individual policies regarding provision of GP services. This may encourage residents to register with a ‘preferred practice’ that provides health care for a large number of residents at the home, or encourage them to remain with their existing GP when moving into the home. A number of factors may influence this, including geographical location (rural or urban), relationship with local GP practices and financial arrangements.

The trial was designed to minimise the involvement of residents’ GPs because of experience from the observational study and feedback from GPs involved in a qualitative study that investigated the consent process for the RCT in PAAD stage 1. GPs had concerns about the impact on their workload if they were required to check eligibility for care home residents and preferred a streamlined approach which minimised their role. The study protocol stated that a research nurse would assess eligibility for the trial and the responsible clinician at the resident’s GP practice would provide final approval for the resident’s inclusion. However, recruiting GP practices whose patients were resident in participating care homes proved challenging because of difficulties in arranging meetings with the busy clinicians and their reluctance to commit to this responsibility. If care homes had residents registered with a number of GP practices, agreement to participate had to be sought from a number of partners from each practice.

The practical difficulties experienced when seeking agreement to participate from GP practices, and the requirements for each practice to be a trial site in terms of GCP training and study-specific training for each GP, proved to be a major barrier to conducting the trial. Advice was sought from the MHRA in an effort to further minimise the role of the resident’s GP. Clarification was sought as to whether or not a registered nurse could confirm eligibility by accessing care home records for the resident or, where required, their medical notes held by the GP. The MHRA’s response was explicit that the decision whether or not a subject is eligible for entry into a clinical trial is considered to be a medical decision and, therefore, must be made by a medically qualified doctor.

A strategy for communicating with GPs was developed to minimise the burden on them. This included the availability of online GCP training along with drawing up a letter of agreement in place of a more comprehensive practice agreement.

Care home recruitment

The RCT initially involved approaching 32 care homes. Five agreed to participate, with a number of other homes expressing interest awaiting further follow-up of the interest to participate. Twenty care homes declined to participate, four of these at a care home company group meeting. Reasons given for declining were involvement in other studies (n = 2), concerns regarding residents’ suitability for research as a result of their cognitive impairment (n = 1) and workload (n = 1). Twelve care homes gave no reason for declining. It is unclear if there are any differences between homes by category (nursing vs. residential) in willingness to participate. The dependency and care requirements of residents will vary between the two, as will staffing levels and skill mix within the homes. Future studies in this area may need to consider employing a GP clinical fellow to take this responsibility.

All participating care homes had a dual registration for nursing and residential care. There was a wide variation in the ratio of residential to nursing beds, from 1 : 3 to 15 : 3. It is also unknown whether or not the proportion of residents in the care home who have a cognitive disorder has an effect on agreement to participate. All participating care homes had residents with and without mental capacity residing in the home.

Contacting care homes was a very time-consuming process. Consent for the care home to participate is required from the care home manager and also, in some cases, the regional manager for the care home group. A consensus of agreement from the care home staff is also required for the care home manager to agree to participate. Consent in care homes has been described as a two-tiered process, where obtaining consent at institutional/community level is required before progressing to consent from individuals. 61 Care home managers were generally positive about becoming involved in a research study and welcomed the engagement with the researchers. The potential benefits of collaborating with a higher education institute, and the opportunity to gain additional training and skills, were viewed as incentives to participate in the study. Care home staff also welcomed the opportunity to examine current practice and the potential for changing practice, where there was evidence to support this. This was seen in one participating care home, with the adoption of BSC monitoring in a care home where there had previously been unsupported attempts to introduce it by some care home staff. The process for recruiting 24 care homes was originally expected to take 5 months but this was extended to 13 months as a result of the above difficulties.

Indemnity

Other issues that have arisen include the indemnity insurance requirements for care homes participating in research, particularly CTIMP studies, as non-NHS sites. The REC confirmed that care homes were required to provide evidence of adequate insurance indemnity. Indemnity is provided by the sponsor for research activity and, therefore, the sponsor would be liable for any non-negligent harm resulting from activity conducted in accordance with the protocol. However, any negligence on behalf of the site (for example harm caused by protocol deviations such as giving an incorrect dose) would not be covered by the sponsor’s insurance and, although the likelihood of a claim arising out of the trial is small, should be covered by the site insurance. The REC was concerned that harm resulting from activity carried out in contravention of the protocol would not be covered by existing policies as this would not form part of the care home staff’s usual activities. The REC received advice from the Health Research Authority that, if the activities did not form part of their routine professional practice, care homes should be asked to take out an extension to cover research activity under their existing insurance.

Advice from research networks with experience in research in this setting [NIHR ENRICH (Enabling Research in Care Homes) programme] suggested that, where a study involved an investigative product or an intervention, indemnity would need to be checked on a case-by-case basis. Discussion was sought with a participating care home’s broker whether research would be covered under existing insurance or whether additional insurance was required. The broker confirmed that additional conditions would be required to be met, such as evidence of study-specific training, prior to providing additional cover for research activity.

This requirement for each care home acting as a research site to provide indemnity for negligence on its behalf had not previously been encountered by the sponsor and, therefore, was not detailed in the existing contract with the care homes. A requirement for additional indemnity further impacted on the costs and time required to set up a RCT in the care home sector and may impact on care homes agreeing to participate.

Reasons for early termination of the trial

The extended time required for setting up the trial resulted in the anticipated date for first participant recruitment being significantly delayed. Staff in the first care home were trained and ready to begin recruitment 8 months after the initial anticipated date. However, CTA had not yet been granted and, therefore, recruitment could not commence.

During this period of delay, the TMG became aware of emerging evidence that there was no longer sufficient scientific evidence regarding the effectiveness of probiotics in reducing the incidence of AAD and CDAD in older hospital inpatients. 55 The PLACIDE study recommended that no further studies assessing probiotics for AAD should be undertaken until further evidence is generated regarding which strains maybe effective in reducing AAD (in vitro evidence).

The TMG considered that the PLACIDE study population is likely to be very similar in terms of age and frailty to a population recruited from care homes and that, although there were some differences in the strains of probiotic organisms used in the PLACIDE study IMP and in VSL#3, there was also some overlap (Box 2). The sample size of the PLACIDE study and the apparently tight CIs around their estimates suggested that PAAD stage 2 would be unlikely to come to a different finding about the effect of probiotics on the duration and severity of AAD.

Despite the differences between the two studies, the TMG considered that there was no longer equipoise regarding the study question. However, it was agreed that continuing to build a research base within a care home setting was vital.

This was considered to have a significant impact on the justification for commencing recruitment to PAAD stage 2. Following consultation with the appropriate steering committees, and presentation to the HTA, the decision was made that recruitment should not commence and the project should close.

During the close-down period, the study team continued the existing engagement with care homes to establish a research base in the care home sector. This process was intended to optimise the value of resources invested in research in this setting and aimed to support future research in the care homes sector. Part of this process involved care home research priority setting exercises with stakeholders.

During the study development stage

The PAAD research team worked closely with a director of seven care homes in Wales. A number of meetings were held with the primary aim of:

• obtaining more information on the structure of care homes

• gaining GP involvement and model of care delivery

• recruiting care homes

• estimating costs of conducting the research

• determining who could provide approval of research in care homes and finally conducting the study

• determining the additional resources required

• resolving consent issues

• determining prescription and treatment schedules

• assessing the feasibility of administering the probiotic and taking stool samples.

Valuable advice was provided which enhanced the final project design.

During the study

End of the study

Introduction

The PAAD stage 1 prospective observational cohort study of 274 residents from 10 care homes across South Wales found that antibiotics were prescribed at a rate of 2.16 per resident-year, with almost three-quarters of residents prescribed at least one antibiotic course during the 16-month study period (median length of follow-up 310 days). Residents were over 2.5 times more likely to be prescribed antibiotics during the study period if they had been prescribed antibiotics in the 4 weeks prior to study entry. Residents were prescribed a wide range of antibiotics for each indication. The incidence of AAD in those prescribed antibiotics was 0.57 episodes per resident-year, with 43.5% of residents prescribed antibiotics experiencing at least one episode of AAD. CDAD occurred in > 15% of residents who developed AAD and for whom a stool sample was sent for microbiological analysis. Although all CDAD episodes were found in residents from the same care home, the unique ribotypes suggest that the episodes were not associated with an outbreak. Residents in nursing homes were more likely to be prescribed antibiotics and experience AAD. The risk of AAD with prescriptions of co-amoxiclav was double that associated with other antibiotics. Risk of AAD was also increased in those who routinely used incontinence pads.

Analysis of stool samples taken at study entry showed that over half of residents carried organisms resistant to antibiotics, with resistance to ciprofloxacin particularly high. The proportion of participants carrying antibiotic-resistant organisms varied between different types of care homes. We found that, while recent hospitalisation was not associated with carriage of resistant organisms, recent antibiotic use was. We also found that carriage of resistant organisms was significantly more likely for residents who were prescribed antibiotics in the 4 weeks prior to study entry, and for those in nursing homes, even after controlling for clinical frailty. C. difficile was more common in nursing homes, but there was little suggestion of clustering of type by home.

Strengths and limitations of the prospective observational study

We conducted this study in a population and setting difficult to research,76,77 but we were able to estimate the rates of participation, antibiotic prescription and AAD. Despite needing to obtain advice about resident participation from consultees for the majority of participants in care homes, the recruitment target was achieved in PAAD stage 1. This study can be considered to have a low risk of bias because of relatively high inclusion rates in each care home and relatively complete follow-up data. 78 Although every effort was made to maintain the validity of the antibiotic and stool data, it is possible that this was not always achieved. Prescriptions during periods of hospitalisation were not transcribed onto study CRFs, so the number of antibiotic prescriptions may have been under-reported. Antibiotic prescriptions that were not recorded for the purpose of the study may also have resulted in a lack of stool data for the defined follow-up period. It is, therefore, possible that we have underestimated the incidence of AAD. Residents who were recorded as routinely experiencing three or more loose stools in a 24-hour period before observations were started were not classed as experiencing AAD during the study. We assumed that if stool data were missing, AAD was unlikely to have occurred, but there may have been some episodes of AAD that were not recorded. Our approach has, therefore, provided a conservative lower bound of estimated AAD in care homes. Although we found that residents in residential homes are significantly less likely to experience AAD than those in nursing homes, this may be in part because of the increased mobility, with associated less intense monitoring of residents’ stools. We also found that residents were more likely to experience AAD if they routinely used incontinence pads. This may be a real effect, but could reflect easier observation and reporting of loose stool types, or be an artefact as a result of stool and urine mixing in an incontinence pad. Not all episodes of diarrhoea following antibiotic use can be ascribed to antibiotics and our study does not seek to demonstrate causality, merely association.

Strengths of the stool sample study include its prospective nature and that we were also able to obtain a stool sample from residents who were asymptomatic or not acutely unwell.

Comparisons to existing research

Despite most residents lacking capacity to consent to inclusion in the study, a majority in each of the 10 homes were included and the population is representative of those now living in care homes in the UK. 79,80 Residents were predominantly aged 80 years and above, frail and at high nutritional risk. Levels of nutritional risk were similar to levels found in care homes in the 2011 Nutritional Screening Survey. 81 Residential, nursing and dual-registered homes were included; nearly all were privately managed, and some specialised in looking after the elderly mentally infirm. We found little seasonal variation in antibiotic prescribing and AAD, consistent with the mandatory and voluntary surveillance data. 17

Our estimate of antibiotic prescribing is consistent with an estimate obtained from a study conducted in North Wales,82 where 203 antibiotic prescriptions were recorded over a 9-week period in 15 nursing homes (giving an incidence rate of 2.3 antibiotic prescriptions per resident-year). A study conducted in nursing homes in Sweden83 found a rate of 0.51 antibiotic prescriptions per resident-year.

Estimates of AAD range from virtually zero in low-risk groups to over 40% in those at higher risk, but these estimates are derived from control groups in trials of interventions for AAD, and not prospective observational studies such as ours. 42

Carriage of third-generation cephalosporin resistance was higher than in published rates for urinary isolates in Wales (5%), despite the overall reduction in cephalosporin use that has been achieved between 2006–11,33 supporting arguments that the prevalence of extended-spectrum beta-lactamase-producing bacteria is rising in Europe. 34 Resistance to fluoroquinolones (ciprofloxacin) was also higher than previously published rates. Fluoroquinolone resistance in community urinary coliforms in 2011 in Wales among adults aged 80 years and over was 16.4%. Recent antibiotic use has been shown to be a key risk factor for carriage of antibiotic-resistant organisms in previous studies. 20,22–25,32,84,85 Previous antibiotic use has also been shown to be a risk factor for the presence of extended-spectrum beta-lactamase-producing strain of bacteria in long-term care. 86 A dose–response association has been proposed by one study27 and multidrug resistance has been linked to non-hospital antimicrobial consumption. 29 Nursing home residence has been previously shown to increase the risk of carriage of antibiotic-resistant organisms. 30,41,84 Although we found no association between recent hospitalisation and carriage of resistant organisms, an association was found in a recent study investigating risk factors for drug-resistant pathogens in community-acquired and health care-associated pneumonia. 84 The sample of care home residents recruited for this aspect of the research was representative of those recruited to the main cohort study. This group has been shown elsewhere to be representative of residents living in care homes across the UK. 79,80 The rate of C. difficile infection in England was approximately 366.9 per 100,000 in 2011 for people aged over 65 years, while the same rate was approximately 42.5 for those under 65 years of age. ⁷ This indicates that C. difficile is still an issue of concern for the specific age group, even though there was a 53% decrease in overall numbers of reported cases between 2008 and 2011. 8

Introduction

The qualitative study about consent and recruitment of care home residents to research found that residents, relatives, care home staff and GPs are generally supportive of older adults in care homes participating in research studies. However, respondents were concerned about the best way of facilitating this, and some subtle, but important, differences of opinion in how older adults should be recruited into studies were apparent.

Limitations of the qualitative study on consent issues

Although the qualitative study reported here incorporates views from a wide range of stakeholders, we acknowledge that our data may be biased. For example, we were able to interview only relatives who were regular visitors to the care homes, residents who had capacity and care home staff who were interested in participating in research. Although residents and relatives who had declined their consent to participate in PAAD stage 1 were invited to participate in the qualitative study, only those residents and relatives who had agreed to participate in PAAD stage 1 agreed to participate in the qualitative study. This could have resulted in a biased sample that included more people who were more accepting of research. We also do not know how many residents were approached by care home staff to participate and declined to be interviewed, which may be a further source of potential bias. Furthermore, the GPs who agreed to be interviewed may have been more knowledgeable of research ethics. We also acknowledge that our respondents may not have been so supportive of research had the IMP been perceived as more harmful.

Comparisons to existing research

Previous research has shown that an accurate and comprehensive patient information sheet is not enough to ensure comprehension of the main issues. 87 In our study this was evidenced by relatives struggling with long and detailed information sheets which care home staff sometimes supplemented, or replaced, with the short, graphically illustrated, information sheet which had been designed for residents. The data from our qualitative study therefore confirm the importance of researchers and ethics committees making every effort to improve the quality of consent discussions and consent documents, ensuring use of familiar words and ideas, short sentences and paragraphs and routinely computing the reading ease of information sheets. 88

One important issue relates to the comprehension and retention of study information to achieve an optimal level of informed consent. Residents and relatives were generally accepting of studies, which they believed to be low risk, and answered important questions that were in the interests of care home residents and the care home. Advanced consent was generally seen as acceptable in low-risk studies, and there was consensus that if someone lost capacity during a trial that they had consented to participate in, they should not necessarily be withdrawn from the trial unless the trial itself changed meaningfully.

The comprehension and retention of information by residents and their relatives becomes even more of a challenge in a study which lasts a reasonably long period of time (12 months in the case of PAAD stage 2), and in which some relatives, asked to act in the capacity of a personal consultee or a personal legal representative, are themselves frail and at risk of becoming cognitively impaired. However, these specific challenges, and challenges raised by other researchers, are not a reason to exclude older adults from research. Although it is important not to underestimate the difficulties, recruitment strategies need to be developed in order to maximise the involvement of frail older people while at the same time protecting their right to decline participation. As participant information sheets are highly standardised, those who recruit older adults to research studies will need training in methods of communication which promote collaborative decision making and enable them to convey complex principles such as equipoise and randomisation in appropriate language. Recruitment of older people, both with and without capacity, to research studies can be complex but the awareness and integrity of the researcher is fundamental to maintaining the principle of non-exploitation. 68 Particular attention should be given to assessing visual and auditory and mental status, compensating for communication and sensory deficits through the use of innovative consent procedures such as pictorial information sheets and interactive tools that help researchers to assess the potential participants’ understanding and retention of information relevant to the study. 89

The second main issue relates to the role of the personal consultee, or personal legal representative, particularly when a research subject loses mental capacity during a research study or trial. Some relatives expressed the view that the person to whom they were related was too elderly or frail and their participation in research was inappropriate, despite the resident themselves having had capacity to decide that they wanted to take part. If these relatives were to act as personal consultees or legal representatives it is uncertain whether their advice or decision on inclusion would reliably reflect the resident’s wish rather than their own. This was indeed raised as a concern by one resident. It is noteworthy also that 1 of the 10 GPs interviewed did not agree with including adults without capacity into any trial of a medicinal product. It is perhaps worth exploring the extent of this attitude among a larger and broader sample of GPs, health professionals, and care home staff, as if this belief is widely held it is likely that it will be a barrier to improving recruitment of this under-represented group into research trials, despite their heavy use of a wide range of health-care technologies.

In a study that lasts 12 months, it is likely that some care home residents who originally have capacity to give consent may lose mental capacity. Furthermore, it is not unusual in such an age group for capacity to be temporarily lost, for example as a result of delirium associated with a urinary tract or chest infection. 90 One important finding of our study was that many staff, relatives and residents felt that, if a resident loses capacity during a clinical trial, then a legal representative must be consulted about that resident’s continued participation in the study. This is in contrast to the majority of GPs, who believed that this would not need to be the case. The MCA (2005)45 makes provision for a personal consultee to advise on the continued participation of the research subject in studies, but this is in contrast to clinical trials of medicinal products, when consent from an adult to participate in the trial remains valid, even after loss of capacity, provided the trial is not significantly altered.

Although care home environments pose practical problems to researchers, our findings should inform other research teams that care home residents, relatives and staff may be generally very supportive of participating in research. Given that older people are the heaviest consumers of health and social care, challenging ageist practices and attitudes in research is important if they are to gain maximum benefit from advances in understanding and management. Ethics committees are also in a strong position to influence research practice and to reduce unethical discrimination on the grounds of age or social situation, by challenging researchers who continue to exclude these groups from research.

Research teams should consider using models of advanced consent for trials of interventions where there is a high likelihood that participants may not have capacity at the point of randomisation. We would also recommend that researchers who are conducting trials of medicinal products in care home settings ensure that care home staff and residents are made aware that residents who have consented to a study have the legal right to remain in the trial if they lose capacity during the study period, provided the person has not indicated otherwise. 48,49

Main lessons learned from set up of Probiotics for Antibiotic-Associated Diarrhoea study stage 2

The process of setting up a RCT in a care home setting has been more complex than the process of setting up an observational study in the same setting and RCTs in other health-care settings. The additional statutory approvals required for a CTIMP resulted in a longer and more resource-intensive process than expected. Gaining ethics approval took a longer period of time than anticipated, and required significantly more evidence for the justification for including residents without capacity, than in the case of the observational study.

The classification of the study, which involved a widely available food supplement, as a CTIMP had a significant impact on the ability of the trial to proceed. The discrepancy between the statutory authorisations required for a CTIMP and those needed for an IMP, and the absence of any such documentation for a product generally characterised as a food supplement, had a significant impact on the continuation of the trial. The implementation by the MHRA of a risk-adaptive approach is intended to simplify the processes for initiating and conducting trials where a trial is considered to be in a low-risk category. This subsequent change in approach may have reduced the difficulties experienced during the setting up of the trial if it had been implemented prior to the initial application for approval.

There is a lack of clarity in research governance in care homes, particularly those involving IMP and both NHS and non-NHS sites. The process for identifying responsibility for research governance is problematic and resource intensive. Early consideration is required to ensure that sufficient indemnity is provided to cover research activity at non-NHS sites not previously familiar with conducting research.

Recruitment of care homes, and GPs whose patients are resident in participating care homes, is a very time-consuming process. Strategies can be developed to streamline the process and minimise the burden anticipated by those who raised initial concerns about participating. However, the multiple layers of permissions and agreement required at each level of the process before residents can be approached to participate, remains an arduous process.

The requirement for additional support by researchers to care homes that are research-naive settings had been identified during PAAD stage 1. However, the need to train care home staff conducting a CTIMP on issues such as randomisation and administration of IMP further increased the level of training required. The difficulties with care home staff availability for training required researchers to be flexible and imaginative when designing and delivering training.

The need for a more structured and evidence-based approach to delivery of health care to care home residents requires greater participation from this under-represented group. The challenges encountered during this study, and the strategies used to minimise their impact, can be used by researchers to inform the design and development of future research studies in care homes.

Research support

Hayley Prout (research associate) recruited participants to the qualitative study and conducted all interviews and one of the focus groups.

Administrative support

Mrs Judith Evans (research administrator), Mrs Sian Dawes and Mr Rhys Thomas.

Members of the Trial Management Group

Dr Meirion Evans, Dr Anthony Johansen, Dr Neil Wigglesworth, Ms Julia Townson and all those who contributed to the study management.

Members of the Trial Steering Committee

Professor Peter Crome (chairperson), Dr Samuel Coenen, Mr Graham Tanner and Rachel Kemp.

Members of the independent Data Monitoring Committee

Dr Steven George, Professor Matthew Hickman, Dr Andrew Lovering and Professor John McLeod.

The Specialist Antimicrobial Chemotherapy Unit conducted the microbiological analysis.

Institutional support

We acknowledge with thanks the trial funders, the UK NIHR HTA. SEWTU is funded by the Welsh Government through NISCHR and the authors gratefully acknowledge SEWTU’s contribution to the study.

Services and participants

We acknowledge, with thanks, all of the services, the residents, their representatives and the care home staff who participated in the study.

We thank all those not otherwise mentioned above who have contributed to the PAAD study.