Clinical stakeholders’ opinions on the use of selective decontamination of the digestive tract in critically ill patients in intensive care units: an international Delphi study

Sample

A Delphi process gauges opinions from a panel of experts [14]. For this study, we defined experts as those in a position to exert decisional authority and be key stakeholders with regard to ICU policy about SDD or to lead its implementation into clinical practice (if that was to happen). Experts were not defined as experts in SDD. We identified four key stakeholder groups: ICU physicians, ICU pharmacists, ICU clinical leads (medical and nursing), and clinical microbiologists and infectious disease specialists who had ICU responsibility.

Our target sample size was 10 participants per stakeholder group in each of the three geographical zones (40 per zone, 120 in total) [15]. The research team compiled lists of their respective stakeholder groups and assessed individuals according to predetermined diversity factors (geographical location, ICU bed number, current SDD practice and academic affiliation). To achieve a final sample size of 10 in all stakeholder groups in each zone, we initially oversampled up to three participants in each of the four groups per zone. Round 1 was a telephone-delivered semi-structured interviewing using a topic guide developed using the Theoretical Domains Framework. All participants who participated in Round 1 were invited to participate in Round 2. Only those who responded at Round 2 were invited to participate in Round 3. Full analysis of the Round 1 interviews of this Delphi study are reported elsewhere due to the differing nature of the data (qualitative), and the high total volume of data from the study makes concise reporting of all data in one paper impossible [16].

Development of the Delphi materials

Participants

We recruited 141 participants into Round 1 and retained 118 to the end of Round 3 (82% retention) [16]. The breakdown of participation by stakeholder group and retention in the study is displayed in Additional file 2. Median years in practice was 15 (IQR 11 to 21) and median number of ICU beds was 20 (IQR 15 to 28), with 109 participants working in academic/academic-affiliated centres and nine in nonacademic centres.

Participants’ self-reported knowledge of SDD

Participants’ perception of their own knowledge was measured with the item ‘I know the SDD evidence base well enough to have an informed opinion of its use’. Overall, participants rated themselves as knowledgeable about SDD (median score ≥6). However, 37 (31%) rated themselves <5 on the nine-point scale. Participants reporting a low level of perceived knowledge in Round 2 showed greater change in their responses from Round 2 to Round 3 (mean change 0.19, standard deviation 0.19) than participants whose perceived knowledge was >5 (mean change 0.09, standard deviation 0.13; P = 0.002).

Stability of opinions

There was high stability of responses from Round 2 to Round 3, both at the individual level (>69% of change scores in the range −1 to +1) and at the group level (mean differences ranging from 0 to 0.52). Stability across stakeholder groups showed there was a greater level of change from Round 2 to Round 3 in the clinical lead group (mean 0.18, standard deviation 0.2) than in other stakeholder groups (means 0.09 to 0.11, standard deviation 0.16 to 0.2; P = 0.044).

What was important to responders?

Opinions about SDD

Very few participants reported strong opposition to SDD (although scores ranged from 1 to 9; Figure 1). Participants indicated that their hospitals tried to reduce antibiotic use, reduce the widespread use of other strategies for tackling HAI, had low ventilator-associated pneumonia (VAP) rates and the lack of clinical priority for SDD (Figure 1). More items related to this topic are reported in Additional file 1.

Opinions on the validity and adequacy of the evidence base and the consequences of implementing SDD

There was moderately strong agreement (median 7, IQR 6 to 8) that research to date has not adequately addressed concerns about antibiotic resistance, and that the evidence base is not generalisable (Figure 2). There was low-level agreement (median 6, IQR 5 to 7) that SDD increases antibiotic resistance but neutrality about whether SDD increases Clostridium difficile, whether SDD benefits the patients to whom it is delivered and whether the risks of SDD outweigh the benefits. More items related to these topics are reported in Additional file 1. Generally, these responses suggest clinical equipoise relating to the benefits and harms of SDD.

Opinions about the likely barriers to implementation

There was strong agreement that the decision to adopt SDD requires consensus amongst colleagues (median 9, IQR 8 to 9), appraisal of the current evidence base (median 9, IQR 8 to 9), and agreement on which patients will receive SDD (median 8, IQR 8 to 9). There was strong agreement (median 9, IQR 8 to 9), that the skills to administer SDD fall within existing clinical competencies (Figure 3). Participants reported their perceptions of conflicting opinions on antibiotic use among microbiology and intensive care physicians (Additional files 3 and 4).

Opinions on feasibility of further research

There was strong agreement from participants regarding the need for an international RCT of SDD but there was significant variability with regard to the impact of concerns about antibiotic resistance on willingness to participate (Figure 4). Participants felt it was ethically acceptable to conduct further RCTs evaluating the effectiveness of SDD (median 8, IQR 7 to 9). Participants were in favour of mortality as a primary outcome, a cost–benefit analysis, monitoring of antibiotic resistance before, during and after the trial, and a control arm that would include either VAP bundles (the UK and Canada) or usual care (Australia/New Zealand) (Figure 4). More items related to this topic are reported in Additional file 1.

Comparisons between stakeholder groups

We compared responses from the four different stakeholders groups for the items ‘I am opposed to SDD’ and ‘The risks of SDD outweigh the benefits’. For ‘I am opposed to SDD’, ICU physicians had a median 5 (IQR 3 to 5), microbiologists a median 6 (IQR 5 to 7), ICU pharmacists a median 5 (IQR 4 to 5) and clinical leads a median 5 (IQR 3 to 5). For ‘The risks of SDD outweigh the benefits’, ICU physicians had a median 3 (IQR 2 to 5), microbiologists a median 5 (IQR 4 to 7), ICU pharmacists a median 5 (IQR 3 to 6) and clinical leads a median 3 (IQR 2 to 5). Figures are presented in Additional files 3 and 4.

Opinions about SDD

We did not find significant opposition to SDD despite agreement that the evidence base is currently seen as unconvincing and the fact that very few centres have implemented SDD. Participants identified that their hospitals were currently targeting VAP and HAI using other methods, and SDD was not a topic of clinical discussion. This perceived low priority of SDD could act as a major challenge to SDD implementation or to the performance of a future trial. When we analysed differences between key stakeholder groups we found that microbiologists were more likely to be opposed to SDD and, along with ICU pharmacists, were more likely to think that ‘the risks of SDD outweigh the benefits’. This could be explained by the microbiologists and the ICU pharmacists having a more ecological view on antibiotic usage whereas ICU clinicians may be more able to identify with the individual patient benefits.

Opinions about the evidence base

Participants were not persuaded of the internal or external validity of the evidence base and demonstrated clinical equipoise with regard to the benefits and harms of SDD. We conducted a detailed examination of responses in the UK data to check whether equipoise was associated with poor knowledge of the evidence base, and found that the mid-point on the beliefs about consequences items was endorsed by a large proportion of participants (n = 21) who assessed their knowledge of the evidence base as high. Hence, equipoise was unlikely to have arisen from ignorance of the evidence base. Superficially, this equipoise would seem surprising in the presence of such a large evidence base suggesting mortality benefit [9], but since the literature to date does not clearly identify the effects of SDD on key issues such as antibiotic resistance, this seems a reasonable stance and is a key external validity issue. That said, evidence to date does not suggest harmful effects of SDD on antibiotic resistance [20]. The findings about equipoise have clear implications for further research. If there is certainty about the benefits of a clinical intervention, it is regarded as unethical to conduct further evaluation by randomising patients to a usual care control group [21]. Consistent with this principle, there was a high level of agreement that it would be ethically acceptable to conduct further SDD research.

It is interesting to compare the evidence base for SDD (which has not been widely implemented) with the evidence for low-dose steroids [22] or tight glycaemic control [23]. SDD remains not widely implemented despite a large (although clearly incomplete) evidence base, whereas low-dose steroids in septic shock and tight glycaemic control were implemented extraordinarily widely with only one supportive RCT and, in the case of steroids, previous evidence of harm [24]. This possibly suggests that extra-scientific factors are at play in such examples.

Barriers to implementation

This finding of clinical equipoise was confirmed when we identified that issues related to benefits and harms of SDD were also the most important barriers to implementation. The decision process (to adopt) and consensus development were also identified as barriers. The need for appraisal of evidence was also identified as a barrier, which seems surprising considering the high number of meta-analyses in the literature. These issues may explain the very low levels of implementation to date, with no ICUs in Canada, Australia and New Zealand currently undertaking SDD and only 5% of ICUs in the UK undertaking SDD (RR Canter, S Harvey, DA Harrison, et al., Survey and observational analysis of current use of selective decontamination of the digestive tract in UK critical care units, British Journal of Anaesthesia 2013, submitted). Other practice issues such as ease of delivery, skills, patient side-effects and cost were not identified as barriers. Nursing and pharmacy workloads were identified as likely to increase but were not identified as barriers as such. Few participants delivered SDD in their practice and were therefore not ideally placed to make these assessments. Interviews with clinical staff in centres delivering SDD provided a similar perspective [25]. These results indicate the need for new research to address the ongoing uncertainties with the evidence base, followed by high-quality translation research if future studies suggest benefit without harmful effects on antibiotic resistance patterns.

Future research

We found a very high level of agreement that further SDD research was ethical, and most participants would support their centre being involved in a RCT to evaluate the effectiveness of SDD. The potential of SDD to increase antibiotic resistance, despite being identified as a very important issue, would not necessarily limit trial participation. There were high levels of agreement that such a trial should include pre-trial, during-trial and post-trial monitoring of antibiotic resistance, however. Mortality was favoured as the primary outcome, with a cost–benefit analysis seen as desirable. It is clear with the strong weighting participants gave to antibiotic resistance that this outcome needs to be given equal stance in any future studies. Participants identified that the control arm should receive VAP bundles in the UK and Canada, and usual care in Australia and New Zealand. This could be influenced by the lack of national VAP guidelines in Australia or New Zealand. Further data on trial design issues were identified during contemporary interviews of intensive care triallists as another part of our research programme (JJ Francis, EM Duncan, ME Prior, GS MacLennan, SU Dombrowski, G Bellingan, MK Campbell, MP Eccles, L Rose, KM Rowan, R Shulman, APR Wilson, BH Cuthbertson, Selective decontamination of the digestive tract in critically ill patients treated in intensive care units: a mixed-methods feasibility study (the SuDDICU study), Health Technology Assessment 2013, in press). Such data can help inform practitioners and researchers on the need, appropriateness and design of future research.

Strengths and limitations

A strength of this study was its grounding in a theoretical framework that enabled us to distinguish between factors related to the clinical evidence and potential barriers related to more practical issues to do with professional roles, resources and the management of change. Further strengths of this study were inclusion of four key clinical stakeholder groups with influence on SDD policy or delivery, thereby enabling a wide range of professional opinions from those likely to influence local policies. The study was completed by 118 clinicians from the four different groups from three geographical regions with high retention across rounds, making this the largest Delphi study to date in intensive care to our knowledge.

There was evidence of high stability of responses between Rounds 2 and 3, indicating that, when given three opportunities to consider their opinions, and viewing the spread of opinions across the cohort, participants reported consistent opinions about SDD over time. Indices of stability may mask individual instability if different stakeholder groups change their opinions in opposite directions. In this study there was evidence of stability at both the individual and group levels. This enhances our confidence that the identified opinions will be relevant over time, unless the profile of evidence, or the knowledge of the evidence, changes.

It is rare in Delphi studies to identify consensus around uncertainty (for example, in the range 4 to 6 on the nine-point scale), where this exists, and to contrast this with consensus around agreement (range of 7 to 9 on the nine-point scale). Given the overarching objective of the study, consensus around uncertainty was of great importance.

Self-rated knowledge of the field was generally high, although one-third of the participants rated their knowledge of the evidence base as uncertain or low. This variation in knowledge of the evidence base is a potential limitation of the study because it makes stated uncertainty difficult to interpret. There has been longstanding debate regarding the meaning of the neutral response in Likert scales [26]. This point is important because it could lead to an incorrect assessment of the presence of equipoise and the move to conduct a new clinical trial of SDD if neutral scores on key items were interpreted as equipoise or uncertainty when actually they could reflect lack of knowledge. By close scrutiny of the data we have identified that participants who rated their SDD knowledge as high were strongly represented among those who endorsed the neutral response relating to the consequences of SDD, reflecting clinical equipoise. We did not ask any questions about the participants’ views on the play off between mortality benefit and antibiotic resistance rates, but a separate publication reports on this point from the perspective of the ethical issues discussed by the participants.

Other weaknesses include the risk that our sampling framework failed to capture all important opinions on this subject. However, evidence from one randomised study indicates that the Framework can result in the generation of a wider range of opinions than using more open approaches to eliciting opinion [27]. Furthermore, in the Round 2 and Round 3 questionnaires we included all opinions emerging from the Round 1 interviews, including minority opinions, to ensure that all opinions were considered by all participants in the questionnaire rounds. The sample size was appropriate for a Delphi study but was numerically small when considering the scale of clinical practice across the four nations. Finally, the sample was taken from English-speaking countries only. Although these countries are widely geographically situated, we acknowledge that this is a limitation of our study.