A process evaluation of a randomized clinical trial of selective decontamination of the digestive tract in ventilated adults admitted to intensive care: the SuDDICU PE study

Abstract

Background

The SuDDICU trial evaluated the effectiveness of selective digestive decontamination (SDD), a prophylactic antibiotic regimen comprising oral, gastric and intravenous (IV) components in ventilated adults admitted to intensive care. The trial recruited 9289 participants in 26 clusters across Australia and Canada. There was no difference in the primary outcome of in-hospital mortality, but a reduction in new positive blood cultures and antibiotic organisms. We conducted a concomitant process evaluation in Canadian sites. Our objectives were to quantify participant reach, intervention fidelity and understand drivers of implementation.

Methods

Multiple methods process evaluation comprising quantitative data on participant reach and intervention fidelity, and qualitative data using key informant interviews and non-participant observation. Analyses were informed by the Consolidated Framework for Implementation Research (CFIR).

Results

Of 6317 patients screened in Canadian clusters, 577/3020 (19.1%) screened during the intervention phase were eligible but not included; 47/3297 (1.4%) in the control phase. There were 274/1432 (19.1%) participants without oral or gastric SDD components commenced within 6 h of enrolment; 54/1432 (3.8%) for the IV component. Excluding enrolment and extubation days, the median (IQR) number of ventilator days was 5 (2, 11), with 7438/12,299 (60.5%) ventilator days with 100% SDD oral dose fidelity and 7076/12,299 (57.5%) ventilator days with 100% SDD gastric dose fidelity. Considering all 4 days of IV antibiotic administration, there were 4663/5160 (90.4%) days with 100% fidelity. We interviewed 30 clinical/research team members and conducted 40 h of non-participant observation. Considering CFIR domains, individual, intervention, implementation and contextual influences on fidelity included (1) variable understanding of trial importance/objectives; (2) trial complexity, specifically drug preparation and administration resulting in additional and unanticipated workload, exacerbated at some sites due to physical ICU layout and usual drug preparation/administration practices; (3) concerns about the SDD oral aspect and overall antibiotic side effects; (4) lack of sustained implementation education; (5) lack of research staff availability out-of-hours; and (6) the broader context of running a complex trial during the COVID-19 pandemic.

Conclusions

We identified moderate fidelity within the Canadian sites participating in the SuDDICU trial, with important fidelity drivers identified across all CFIR domains.

Trial registration

SuDDICU Trial—ClinicalTrials.gov NCT02389036 (registered 03/17/2015); https://clinicaltrials.gov/ct2/show/NCT02389036.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13063-026-09935-6.

Background

Hospital-acquired infections (HAIs) are a significant public health issue globally, causing more than one million deaths per year [1]. More than 20 million people are treated in an intensive care unit (ICU) annually. Between 20 and 50% of ICU patients develop a HAI [2], with approximately 20% of ICU patients dying in the hospital [3]. Selective digestive decontamination (SDD) is one strategy to prevent HAIs, thereby reducing mortality. It comprises regular application of non-absorbable antibiotics to the mouth, oropharynx and stomach combined with a short course of IV antibiotics [4]. SDD modifies the balance between potentially pathogenic (Gram-negative aerobic bacteria) and other less pathogenic gastrointestinal flora.

Evidence from numerous randomized clinical trials (RCTs) identifies a decrease in HAIs and mortality with SDD [5]. Yet, despite decades of research, SDD adoption remains limited to only a few regions/countries [6]. This is largely due to the unproven, yet justified, concerns about increased antibiotic resistance [7,8 ]. Given low adoption but a promising evidence base, the Selective Decontamination of the Digestive Tract in Intensive Care Unit Patients (SuDDICU) international research collaboration launched an international multi-centre, cluster crossover randomized clinical trial of SDD commencing May 2017 and finishing in 2023 [9]. Twenty-six clusters (ICUs) in Canada (n = 7) and Australia (n = 19) participated, recruiting 9289 patients to the randomised component and a further 10,711 to the accompanying ecological study. The trial found no difference in the primary outcome of in-hospital mortality, but a reduction in new positive blood cultures and antibiotic organisms [10].

SDD is a complex intervention [11] comprised of multiple components with healthcare professionals interacting across several organizational departments. Intervention fidelity, i.e. the extent to which an intervention is delivered as planned, can be challenged in trials of complex interventions [12]. This is because there is a need to disentangle the influence of interrelated and interdependent components on fidelity [13]. Low intervention fidelity can lead to misinterpretation of treatment effect estimates, threatening trial internal validity [14]. Process evaluations conducted in parallel to RCTs explore trial implementation factors influencing intervention fidelity, causal mechanisms, and contextual factors associated with trial outcomes [15,16 ].

The SuDDICU Canadian process evaluation aimed to generate data to quantify SDD intervention fidelity and understand the influence of barriers, facilitators, context, and trial implementation strategies. Our specific objectives were to (1) quantify the ability of clusters (ICUs) to implement the SDD intervention (i.e. participant reach and protocol fidelity) over the 12-month intervention period; (2) understand organizational context and processes influencing implementation; and (3) explore diverse healthcare professional perspectives on the SDD intervention and associated trial processes.

Methods

The SuDDICU intervention

The SuDDICU intervention included (1) topical application of SDD antibiotic and anti-fungal paste to the buccal cavity and oropharynx with concomitant oral care (4 times/day) until tracheal extubation for a maximum of 90 days; (2) SDD suspension to the gut via a naso/orogastric or jejunal tube (4 times/day) until tracheal extubation for a maximum of 90 days; and (3) 4-day course of IV antibiotics (cefotaxime, ceftriaxone or ciprofloxacin). Given the cluster, crossover design, ICUs remained in the intervention arm for twelve months, with all patients meeting trial inclusion criteria and none of the exclusion criteria, enrolled into the trial using a deferred consent model. Although individual informed consent may be waived in cluster trials, deferred consent for the intervention group was required by Health Canada and the Research Ethics Committee approving the study.

Site initiation visits provided education to clinical and research staff to implement the study intervention and collect study data. These visits were in-person with site representatives including medicine, nursing, pharmacy, and the research delivery team. As well as the trial protocol and standard operating procedures, sites were provided with an educational video describing the trial objectives and drug preparation and administration processes. During trial conduct, the chief investigator (BHC) and central coordinating team were available to address site queries; no further in-person ongoing trial education was provided. Midway through the trial, based on site feedback, a trial newsletter was introduced with recruitment statistics and frequently asked questions sent every three months.

Study design

We conducted a multi-method process evaluation within six Canadian ICUs participating in the SuDDICU trial, a cluster crossover trial. Of these, five were in tertiary academic centres, with one ICU located in a regional hospital. All ICUs admitted a mixed medical-surgical population; three are designated a Level I trauma centre and another Level II Regional trauma centre, therefore also admitting mahor trauma. ICU bed numbers ranged from 20 to 40, with an annual volume of mechanically ventilated patients ranging from 500 to 1000. As recommended for process evaluation studies¹⁷, we first conducted iterative key stakeholder interviews to generate our logic model (See supplementary material). We used quantitative methods to assess trial reach and intervention fidelity and qualitative methods to determine implementation enablers and barriers. Quantitative and qualitative data were collected in parallel.

Ethical considerations

We obtained research ethics approval from Clinical Trials Ontario (CTO), a multi-centre research ethics review board [CTO Project #1662]. Individual participation in surveys, observations, and key informant interviews was voluntary, with individual informed consent obtained before data collection. The study was conducted in accordance with the Declaration of Helsinki and every effort made to protect the life, health, dignity, and rights of our research subjects.

Quantitative methods

Fidelity metrics

We collected data on the following fidelity metrics: (1) time of drug commencement: SuDDICU protocol states commencement within 6 h of institution of mechanical ventilation and ICU admission; (2) fidelity of daily administration: SuDDICU protocol states 6 hourly dosing for both oral and gastric preperations (i.e. 4 doses of each/ventilator day); daily (drug dependent) for 4 days for the IV component; and (3) IV antibiotic administered as per SuDDICU protocol: 3rd generation cephalosporin (or ciprofloxacin if cephalosporin allergic.

We defined a protocol deviation as any drug dose omitted without a valid reason for omission. Valid reasons were defined a priori as (1) inability to administer an oral or gastric SDD dose due to inability to access the mouth or gut; (2) clinical contraindication as decided by the prescriber or nurse administering the drug (although there was no requirement to identify the specific contraindication); (3) undergoing procedures outside of the ICU; and (4) patient or family member refusal.

Data collection

To assess participant reach, we examined cluster screening and study enrolment logs to determine the proportion of participants who were eligible but not recruited. To assess intervention fidelity, we accessed data on daily SDD drug administration (oral, gastric, and IV) collected for the trial. Drug administration data, including dose omissions and their reasons, were documented by the nurse administering each dose in the electronic medical record as part of usual drug administration practices. These data were transferred to a study-specific case report form by site research coordinators. Drug reconciliation against drug study kits was done by the site’s pharmacy to maintain drug accountability logs.

Implementation process survey

Participant recruitment

To recruit participants, at each site, a senior clinical team member (primary site investigator or research manager) distributed recruitment emails/fliers to all staff managing (clinically or through research delivery) patients enrolled in SuDDICU.

Data collection

We used the Normalisation MeAsure Development questionnaire (NoMAD) [18] to quantitatively assess the perspectives of healthcare professionals on the implementation of the SuDDICU trial. The NoMAD [20] uestionnaire is a set of 23 survey items based on the Normalization Process Theory [21]. It assesses implementation processes from the perspectives of healthcare professionals directly involved in the work of implementing complex interventions. We administered the NoMAD questionnaire using SurveyMonkey (Ottawa, Canada), sent via a link embedded in an email, with paper copies provided in the ICUs where we conducted in-person observations.

Qualitative methods

Participant recruitment

We recruited participants using the same strategies described above in the implementation survey methods. We used convenience sampling for observation and interviews during a randomly selected week of the intervention period. Our goal was to interview at a minimum the site’s primary investigator, research coordinator and one or two staff members involved in trial processes at each site. We aimed to observe care processes on all enrolled patients during the scheduled observational periods.

Data collection

We collected qualitative data at six of the seven ICUs (the seventh site did not consent to participation in this aspect of the process evaluation) at two timepoints (August 2020 and December 2022) when participating ICUs were in the intervention phase. Qualitative data collection methods included (1) observation of site initiation visits; (2) go-along interviews and bedside observation; and (3) key informant interviews. A trained qualitative researcher accompanied the SuDDICU investigator team to all site initiation visits to make observations and take field notes regarding local implementation planning. Once participant recruitment commenced, we conducted in-person ‘go-along’ interviews [19] with nurses while observing SuDDICU intervention delivery, particularly the oral and gastric components. Go-along interviews also explored SDD-associated nurse workload, ease/difficulty of SDD administration, and observed unintended consequences or complications. Following bedside observation, detailed field notes describing SDD delivery were generated.

We conducted key-informant semi-structured interviews recruiting site investigators, bedside nurses, research staff, and pharmacy staff involved in study drug management. Using an interview guide informed by our logic model (see Supplementary Material), we explored perceptions and experiences of the SDD intervention, trial implementation, and intervention fidelity. To further understand staff experience with SDD, interviews explored how it fit with other care tasks, delivery barriers and facilitators, and any unintended consequences. Interviews were conducted in-person or by telephone by a trained qualitative non-clinician researcher with a background in sociology. Interviews were audio-recorded and professionally transcribed.

Data analysis methods

Quantitative data

We assessed SDD drug administration fidelity, defined as the proportion of drug doses administered by ventilator day and by patient, considering oral, gastric and IV components separately. We excluded the day of study enrolment and the day of extubation, as patients would only be eligible for a proportion of the drug regimen doses. We used counts and proportions to summarize the (1) participants who commenced SDD components within 6 h of ventilation; (2) participants with 100%, 75% and 50% fidelity of drug doses; (3) ventilator days SDD was delivered at the correct frequency, i.e. no missing doses; and (4) ventilator days when doses were omitted and reasons for ommission. For oral and gastric doses, we excluded the day of enrolment and day of extubation; for IV antibiotics, we included the first four days of administration. We considered 80–100% adherence as high fidelity; 51–79% moderate; and < 50% low fidelity based on classifications used in evaluations of intervention fidelity [20,21 ]. We used counts and proportions to summarize responses to the NoMAD questionnaire.

Qualitative data

We analyzed the qualitative dataset using standard thematic analysis techniques and a largely deductive approach [20]. We employed the Consolidated Framework for Implementation Research (CFIR) [22,23 ] to organize emerging themes and develop the final analytic framework. Study team members (KND, MBS and KA) independently read and reread the data, attached descriptive codes to text segments in each transcript, grouped codes into broad topic-oriented categories, and then compared all text segments belonging to the same category. Categories were refined through an iterative process, determining main themes and organization with the CFIR framework. Additionally, we compared codes within and across study sites (constant comparison technique) to monitor emerging themes and seek alternative explanations during CFIR application [23]. The team met regularly during this process to discuss initial coding, theme generation, review themes, and CFIR construct applicability. This analysis was discussed iteratively with the larger evaluation team, including members who are ICU clinicians with expert understanding of the SDD intervention and ICU characteristics, to interrogate interpretation coherence. Agreement on the validity and applicability of the final analytic framework was reached through consensus.

We then leveraged the strength of four methods of triangulated data collection: quantitative reach and fidelity data, the NoMAD questionnaire assessing implementation, direct non-participant observation, and key informant interviews. These methods informed an understanding of the reality of what happened during the trial and the potential impact on trial outcomes.

Results

Quantitative fidelity data

A total of 3020 patients were screened in the seven clusters when randomized to the SDD arm, of whom 577 (19.1%) were eligible but not included. There were 613/3020 (20.2%) patients not included, with no reason provided; 6 participants withdrew on the first study day. When clusters were in the control arm, 3297 patients were screened, of whom 47 (1.4%) were eligible but not included. A further 1145 (34.7%) eligible patients were not included during the control arm, with no reason provided. No participants withdrew on study day one.

Oral SDD doses

There were 274/1432 (19.1%) SDD participants who did not commence oral SDD drugs within 6 h of enrolment. Excluding the day of enrolment and extubation, and therefore 295 patients who were ventilated for < 48 h, we identified 727/1137 (63.9%) patients with 100% oral drug administration fidelity. There were 181/1137 (15.9%) patients with < 75% fidelity and 80/1137 (7.0%) with < 50% fidelity (Fig. 1). Again excluding the day of enrolment and extubation, there were 12,299 ventilation days when oral SDD doses should have been administered (median (interquartile range (IQR) ventilator days for intervention arm participants was 5 (2, 11)). There were 7438/12,299 (60.5%) ventilator days with 100% SDD oral dose fidelity. Of these 12,299 ventilator days, 2496 (20.3%) days had at least one oral SDD dose (i.e. more than one dose may have been omitted that day) omitted for valid reasons; 2091 (17%) days had at least one oral SDD dose omitted for invalid reasons, i.e. protocol violations. The most common valid reason for omitting an oral SDD dose was the inability to administer the oral paste. Ventilator days when patient or family member refusal occurred were minimal, 27 (0.2%) and 9 (0.1%), respectively (Table 1).

Fig. 1.

SuDDICU drug administration fidelity by patient for oral, gastric and IV antibiotics. We excluded the day of enrolment and the day of extubation to establish fidelity for the oral and gastric doses, given that these would be partial days. We considered the first four days of study enrolment for IV antibiotic fidelity. For all drug administration routes, we considered that fidelity was present if less than the prescribed doses were administered, but there was a valid reason provided for not administering the drug

Table 1.

SDD oral paste omitted doses by ventilator days

N = 12,299 ventilator daysa		Ventilator days with doses omittedbn (%)
Valid reasons for oral dose not given (no protocol deviation)
Inability to administer oral paste		1479 (12.0)
Dose omitted as clinically contraindicated		815 (6.6)
Patient undergoing investigations/procedures or not in the ICU		165 (1.3)
Patient refused intervention		27 (0.2)
Substitute decision maker or family member refused intervention		9 (0.1)
No stock available at site		1 (0.01)
Invalid reasons for oral dose not given (protocol deviations)
Dose omitted by omission		1525 (12.4)
Incomplete dose of oral paste administered		313 (2.5)
Unknown reason		253 (2.1)

^aExcludes the day of enrolment and the day of extubation

^bData represent ventilator days where 1 or more doses of gastric SDD were omitted

There were also 274 ventilator days when the gastric dose was not commenced within 6 hours of enrolment

Gastric SDD doses

There were 274/1432 (19.1%) SDD participants who did not commence gastric SDD drugs within 6 h of enrolment. Excluding the day of enrolment and extubation, and therefore 295 patients who were ventilated for < 48 h, we identified 721/1137 (63.4%) patients with 100% gastric drug administration fidelity on all ventilated days. There were 181/1137 (15.9%) patients with < 75% fidelity and 80/1137 (7.0%) with < 50% fidelity (Fig. 1). There were 7076/12,299 (57.5%) ventilator days with 100% SDD gastric dose fidelity. Of these 12,299 ventilator days, 2836 (23.1%) days had at least 1 gastric dose omitted for valid reasons; 2096 (17.1%) days had doses omitted for invalid reasons, i.e. protocol violations. The most common valid reason for omitting a gastric SDD dose was inability to administer due to no naso/oro gastric or jejunal tube in situ (Table 2).

Table 2.

SDD gastric suspension doses omitted doses by ventilator days

N = 12,299 ventilator daysa	Ventilator days with doses omittedb n (%)
Valid reasons for gastric dose not given (no protocol deviation)
Unable to administer gastric suspension	1948 (15.8)
Dose omitted as clinically contraindicated	732 (6.0)
Patient undergoing investigations/procedures or not in the ICU	140 (1.1)
Substitute decision maker or family refused intervention	10 (0.1)
Patient refused intervention	5 (0.04)
No stock available at site	1 (0.01)
Invalid reasons for gastric dose not given (protocol deviations)
Dose omitted by omission	1497 (12.2)
Incomplete dose of SDD gastric administered	337 (2.7)
Unknown reason	262 (2.13)

^aExcludes the day of enrolment and the day of extubation

^bData represent ventilator days where 1 or more doses of gastric SDD were omitted

There were also 274 ventilator days when the gastric dose was not commenced within 6 hours of enrolment

IV antibiotics

There were 54/1432 (3.8%) participants who did not commence the IV component within 6 h of enrolment. Considering all 4 days of IV antibiotic administration, there were 1064/1432 (74.3%) patients with 100% administration fidelity, 187 (13.1%) had < 75% fidelity, and 55 (3.8%) had < 50% fidelity (Fig. 1). There were 5160 ventilation days when IV antibiotics doses should have been administered. Of these, 4663 (90.4%) days had no dose omissions, i.e. 100% fidelity. Of these 5160 days, 5 (0.1%) had a valid reason for omission; 497 (9.6) days had doses omitted for invalid reasons, i.e. protocol violations. The most common protocol violation was that the IV antibiotic was not prescribed (Table 3).

Table 3.

Intravenous antibiotics doses omissions by ventilator days

N = 5,160 ventilator daysa	Ventilator days with doses omittedb n (%)
Valid reasons for IV antibiotic dose not given (no protocol deviation)
Dose omitted as outside ICU at that time	1 (0.02)
Dose omitted as clinically contraindicated/clinician judgment	4 (0.08)
Invalid reasons for gastric dose not given (protocol deviations)
IV antibiotic not prescribed (while patient was ventilated)	201 (3.9)
IV antibiotic stopped (before ventilation discontinued)	112 (2.2)
Dose not given as enrolled late in the evening	54 (1.1)
IV antibiotic dose not administered	19 (0.4)
IV antibiotic additional dose given	1 (0.02)
Unknown reason	110 (2.1)

^aThis includes all ventilator days when patients were enrolled in the SuDDICU intervention arm – i.e., not excluding the day of enrolment and the day of extubation

^bData represent ventilator days where 1 or more doses of IV antibiotics were omitted

Implementation survey

For the NoMAD survey, we received 29 responses across all six units participating in this element of the process evaluation, 67% from nurses. While 68% of respondents reported feeling familiar with SDD, only 47% considered it as part of their normal work, with only 44% seeing the potential value of SDD in their work. Only 43% felt management adequately supported SDD in their unit, with 47% indicating they agreed with the statement ‘My ICU has a shared understanding of the purpose of SDD’. When asked if they would continue to support SDD after the trial, 54% selected either neutral or disagree.

Qualitative

We conducted 40 h of observation with go-along bedside interviews and recruited 31 participants for key informant interviews from all six participating ICUs. Participants included ICU nurses, a dietician, pharmacy staff, research staff (coordinators and managers) and physicians serving as site investigators (Table 4). We report our qualitative findings using the CFIR domains (Fig. 2), moving from the individual level to the broader influence of events external to the ICUs (i.e. the outer setting. Exemplar quotes from the qualitative interview data is provided in the supplementary material (Supplement A).

Table 4.

SuDDICU qualitative key informant participant demographics

Individual interviews	Number of participants (N = 31)
Gender
• Female	22
• Male	9
Occupational group
• MD	6
• RN	14
• Research staff	8
• Allied health	3
Years in Practice
• > 5 years	5
• 5–10 years	5
• 10 + years	21

Fig. 2.

Summary of qualitative results depicted in a visual format reflecting the CFIR Framework

Characteristics of individuals: clinical and research staff perceptions

Despite well-attended site initiation meetings, we identified variability in the understanding of the SuDDICU trial primary objective and importance among staff in key intervention delivery roles, namely ICU nurses. Furthermore, some participants stated that SDD had no relative advantage over current practice (i.e. chlorohexidine rinse for mouth care). Participants perceived the workload required to prepare, particularly the gastric suspension and administer SDD, was underappreciated when designing the trial. Other negative perceptions expressed by ICU nurse participants were that the oral component made the patients’ mouths look dirty and unkempt. Some nurses also perceived SDD increased diarrhea, increasing their workload. These perceptions influenced trial buy-in, potentially affecting SDD intervention fidelity.

Research team participants expressed concerns about trial delivery, with recruitment and drug accountability processes perceived as onerous and more than initially anticipated. These participants felt that such a complex trial and the volume of the data collection required a dedicated and funded research officer (as opposed to incorporation into the existing ICU research portfolio) with appropriate compensation for the required work.

Despite these challenges, characteristics of individuals with influence and/or power over trial implementation included motivated and experienced site investigators and research coordinators, enabling successful trial delivery.

Intervention domain: the SDD intervention

Participants described the intervention as ‘very complicated’, given three drug components, delivered at multiple time points, with challenges in terms of gastric SDD preparation and oral SDD administration. Gastric SDD preparation required reconstitution of the powdered drug with 100 mL of sterile water, providing a five-day supply that then required refrigeration. This process was perceived as a major implementation barrier as the mixing process was onerous and time-consuming. Who prepared the gastric suspension, when, where and how it was prepared, labelled, and stored differed across sites. For example, in some sites, responsibility for preparation was that of the bedside ICU nurse; in others, it was prepared by the pharmacy or a research coordinator. We also observed various approaches to gastric SDD preparation with workarounds, including those that went against drug preparation education. These include preparing large quantities in advance or bringing multiple doses to the bedside.

The oral SDD component was described by nurse participants as challenging due to the endotracheal tube, with the application process considered uncomfortable for patients. Observations confirmed that nurses often struggled to apply the oral paste evenly to the oral cavity, as the endotracheal tube limited access to the mouth. If nurses perceived that low patient acceptability related to oral administration (i.e. patients were agitated or tried to withdraw), or nurses attributed intervention side effects (i.e. diarrhea), participants reported that an oral or gastric SDD dose may not be given. Despite sedation, patients were observed to flinch or become agitated when the oral paste was administered, with nurses trying to soothe them during application. Nurse participants also described how keeping their patients neat, clean and comfortable is an important part of their care, so the messiness of the oral paste went against these important norms.

Implementation process domain: how the trial was implemented

Participants considered the trial initiation education materials and support during the early phase to be very good and easy to access. However, despite sites being provided with an educational video on drug administration processes, participants perceived ongoing training to be insufficiently sustained or not delivered in an accessible manner, e.g. only delivered on day shifts and not repeated for new staff. Participants also perceived underutilization of the nurse educator role. These aspects were viewed as problematic given factors such as high staff turnover, unanticipated trial interruptions due to the COVID-19 pandemic, and questions arising as the trial progressed. Participants perceived the SuDDICU newsletter initiated mid-way through the trial as useful for updating sites on the trial progress. However, this was viewed as one-way communication with insufficient communication channels with the central research team to raise questions or concerns. This was viewed as a barrier to successful implementation, particularly if the SuDDICU intervention was perceived as not easily modified to fit within a participating ICU’s current practices.

Some participants expressed concern about the trial’s deferred consent model, stating they felt this model did not reflect true informed consent. The reasons for deferred consent were not well understood and therefore did not align with participant values. Finally, research staff perceived trial data collection as very administration/paperwork heavy, with requirements for detailed medication reconciliation activities found to be particularly challenging.

Inner setting domain: the intensive care units

Most ICUs were research-intensive units with pre-existing research infrastructure and regular involvement in research delivery. Participants described site investigators as leading clinician-scientists (physicians) who regularly held site investigator roles with research coordinators experienced with implementing clinical trials. Despite these strengths, staff awareness of the SuDDICU trial objectives and processes was perceived to decline during participation in the intervention arm, with a disconnect between study processes that were agreed upon at the site initiation meeting (which did not necessarily include frontline nurses) and day-to-day processes at the bedside.

The physical organization/layout of participating ICUs presented unanticipated challenges arising from medication preparation, storage requirements and the location of the ICU medication storage areas in relation to bedspaces or patient rooms. The reconstitution of the gastric SDD five-day supply took up to 20 min due to poor solubility. Gastric and oral SDD components had to be refrigerated as the drug became inactive at room temperature. Therefore, nurses had to leave the bedside for a significant time to prepare the study drug, causing workload issues as another nurse was required to monitor two patients in their absence. To overcome this issue, one ICU devolved this responsibility to the research coordinator; another site to the pharmacy department. Nurse participants described how the need for refrigeration meant the oral SDD component was not readily available at the bedside to coincide with other mouthcare, again adding to workload or resulting in workarounds with doses moved to the bedside.

Another inner setting element influencing participant reach was the availability of research coordinators and site investigators on the study designation log. Although the study protocol required participant enrollment over 24 h and the 7-day week, this was not feasible in some ICUs due to research staff unavailability and physician coverage models.

Outer setting domain: the world outside the ICU during the trial

The most influential and unanticipated outer setting event impacting the SuDDICU trial was the COVID-19 pandemic, resulting in periods of total trial shutdown as mandated by the Ontario government. The pandemic also caused a clinical and research staffing crisis, forcing research and data collection responsibilities on other non-ICU clinical or research staff. When allowed to recommence, research staff were additionally burdened by COVID-19 prioritised research, making the SuDDICU trial delivery challenging. The pandemic also interrupted SDD drug availability, which was manufactured in Australia and airfreighted to Canada. These factors extended the trial timeline, resulting in delayed completion and the need for funding extensions.

Discussion

This process evaluation of the SuDDICU trial, using a robust, multi-method evaluation of SDD drug administration and trial implementation, found moderate intervention fidelity. Fidelity was lowest, albeit similar, for oral and gastric SDD doses. We also identified differences in participant reach when clusters were in the intervention compared to the control period, possibly reflecting selection bias. Individual, intervention, implementation and contextual influences on fidelity included (1) variable understanding of the trial importance/objectives; (2) trial complexity, specifically drug preparation and administration resulting in additional and unanticipated workload, made worse in some contexts due to the physical ICU layout and usual drug preparation/administration practices; (3) concerns about oral administration and drug side effects; (4) lack of sustained implementation education; (5) research staff unavailability out-of-hours; and (6) the broader context of running this complex trial during the COVID-19 pandemic.

Factors influencing SDD intervention fidelity are not necessarily unique to the SuDDICU trial. For example, complexity and workload of SDD administration, perceived challenges to oral care delivery, and need for more comprehensive education are known barriers to SDD administration outside the context of a trial in ICUs that routinely administer SDD [24]. The need for ongoing trial education and two-way communication have been identified as successful implementation strategies in other studies [25,26 ]. Furthermore, the need for comprehensive, ongoing and repeated education is enhanced in ICU trials given the large volume of staff, particularly ICU nurses, and workforce turnover or medical staff rotations [27]. All trials depend on human behaviours as they require providers (in SuDDICU, ICU doctors and nurses prescribing and administering the SDD drug regimen) to perform actions that they would not do otherwise and can significantly influence fidelity [28]. Trials also require participants to adhere to the intervention; however, this is not pertinent to SuDDICU given all participants were critically ill and invasively ventilated. Therefore, all trials need to consider behavioural strategies designed to optimize buy-in and thus enhance fidelity. The external influence of the pandemic that increased patient acuity, nursing workload and risk to nurses administering drugs to the oral cavity, and created challenges with study pauses and interrupted drug supply. This could not have been accounted for in the study design but highlights the need for a dedicated research infrastructure that can be flexible during times when the healthcare system is unpredictably challenged [29].

Our fidelity findings from the Canadian sites are interesting in the context of the SuDDICU trial that demonstrated no effect on the primary outcome of hospital mortality but reductions in important microbiological secondary outcomes and overall antibiotic exposure [10]. With the caveat that process evaluation data are only available from the Canadian sites, the moderate fidelity we detected may indicate that the full SDD regimen as provided in the trial may not be required to achieve benefit in terms of reducing new positive blood cultures and antibiotic-resistant organisms. Indeed, in a before-and-after study enrolling 1958 participants, a three, as opposed to four daily dose regimen, had similar outcomes in terms of microbiological and clinical outcomes [30].

Inadequate understanding of the realities of implementation, resulting in poor intervention fidelity within trials, can lead to misinterpretation of the estimate of effect and to potentially effective interventions being discarded [30]. While it is unlikely another trial of this rigour will be conducted evaluating SDD effectiveness, findings from our process evaluation should be considered in planning for SDD implementation into standard care. They are also valuable for extrapolation to design considerations in other ICU trials with threats to buy-in, intervention fidelity, and implementation. Evaluation of site trial readiness, understanding of existing work practices and physical ICU design, gauging staff values and promoting understanding of the research underpinning the trial for all professions involved in trial delivery would have enhanced trial fidelity.

Strengths and limitations of this study

As with all pragmatic research, this process evaluation has both strengths and limitations. We conducted a detailed evaluation of all intervention delivery and employed rigorous qualitative research methods conducted longitudinally, in multiple sites and by trained qualitative methodologists. The triangulation of multiple data types is a significant strength and allowed us to look at both the reality of implementation and perceptions of why things happened the way they did in real-time.

In terms of limitations, we were reliant on the nurses administering SDD drugs to provide a rationale for why a dose was omitted, with many citing a clinical contraindication that could not be adjudicated. Due to pandemic research restrictions, we were unable to undertake the volume of bedside observations originally planned. We also experienced challenges in interview recruitment at some sites due to limited access and staff feeling overwhelmed.

Conclusion

In this multiple method process evaluation of the SuDDICU trial, one of the largest trials with a complex intervention conducted in intensive care to-date, we found moderate intervention fidelity influenced by individual (i.e. understanding of trial importance and buy-in), intervention (i.e. workload, oral care and concern for side effects), implementation (i.e. need for ongoing education and site support) and contextual (i.e. unit design, current practices and the COVID-19 pandemic) implementation barriers. These findings are valuable both to inform design considerations in future ICU trials to address threats to buy-in, intervention fidelity, and implementation, as well as for the implementation of SDD into routine practice.

Abbreviations

CFIR

Consolidated Framework for Implementation Research

SuDDICU

Selective Decontamination of the Digestive Tract in the Intensive Care Unit

Authors’ contributions

**KND -** Conceptualization, Methodology, Data curation, Writing- Original draft preparation and editing. **MBS -** Conceptualization, Methodology, Data curation, Writing- Original draft preparation and editing. **KA -** Data collection and analysis, Writing- Original draft preparation and editing. **SM -** Conceptualization, Methodology, Writing- Original draft preparation and editing. **RP -** Methodology, Data curation, Writing- Original draft preparation and editing. **BC -** Conceptualization, Methodology, Data curation, Writing- Original draft preparation and editing. **LR -** Conceptualization, Methodology, Data curation and analysis, Writing- Original draft preparation and editing.

Funding

The SuDDICU process evaluation (PJT-159744) and conduct of the SuDDICU trial in Canadian sites was funded by the Candain Institutes of Health Research (CIHR).

Data availability

The data that support the findings of this study are available upon request and with the permission from primary investigator, but restrictions apply to the availability of the data and so are not publicly available.

Declarations

Ethics approval and consent to participate

The study was approved by Clinical Trials Ontario Research Ethics Board, and all research participants provided informed consent prior to data collection.

Competing interests

The authors declare no competing interests.