Dissemination interventions to improve healthcare workers’ adherence with infection prevention and control guidelines: a systematic review and meta-analysis

We searched Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 9) in the Cochrane Library (searched on 23 September 2020); MEDLINE (via Ovid; 1946 to 23 September 2020); Embase (via Ovid; 1974 to 23 September 2020); and Cochrane COVID-19 Study Register (February 2020 to 23 September 2020; http://​covid-19.​cochrane.​org). We screened the references of related Cochrane systematic reviews and the list of references of the included studies.

An information specialist conducted our search of the literature, which was revised by a content expert. Complete information on the search strategies is available in the protocol. We limited the searches to randomized controlled trials (RCTs) and no other limits were applied. Search outputs were imported into Covidence platform (www.​covidence.​org) to remove duplicates and perform further review steps.

The team of review authors (MTS, TFG, EC, ENS, JOMB) in pairs and independently screened titles and abstracts at Covidence platform. After screening the first 100 studies, the team met to assess disagreements and adjust the selection process. We resolved disagreements by consensus. The same process was applied to select studies in full text that were considered eligible based on title and abstract screening.

We used the Cochrane risk-of-bias tool for RCT version 1 [29], integrated with Covidence [30], to assess the included studies (dual; second reviewer checks all judgements). We judged the risk of bias as “low,” “high,” or “unclear” and provided support for judgement of the following items: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, incomplete outcome data, selective outcome reporting, and other sources of bias. We adopted “unclear risk” only in cases of lack of information about the methods.

All authors extracted data from the studies (MTS, TFG, EC, ENS, JOMB) using a customized form in Covidence, which were cross-checked by a second author (MTS, TFG).

We collected characteristics of the studies (author, year of research, country, setting, study design, inclusion and exclusion criteria, sponsorship source, conflicts of interest), characteristics of the study participants, description of the interventions, and results.

We sought data for adherence to IPC guidelines in each intervention group assessed in the studies according to the nature of the data. We grouped the outcomes of similar enough studies according to the intervention and longest available follow-up. For vaccine uptake, we collected the number of healthcare workers vaccinated and the total number of personnel assessed in each group. Hand hygiene compliance data relied on the number of hand hygiene actions by all hand hygiene opportunities (before patient contact, before aseptic task, after body fluid exposure, after patient contact, after contact with patient surroundings). Knowledge about IPC data was based on the number of individuals assessed and measured for knowledge in each group (mean and standard deviation of the test score or score improvement and interquartile range).

We calculated the mean differences (MD) for knowledge on IPC and risk ratios (RR) of vaccination uptake and hand hygiene compliance outcomes along with 95% confidence intervals (CI). Outcome effect of each intervention was assessed in comparison to usual activities or other strategies. As studies’ interventions relied on multiple dissemination interventions, effects were presented separately into “combined strategies vs. usual activities” and “combined strategies vs. single strategies.” We adopted random-effects meta-analysis for all outcomes [27], considering the outcomes as related but slightly divergent intervention effects. For the cluster RCTs included, we calculated the design effect using the intracluster correlation coefficient, the number of clusters and the average sample size of each cluster. We calculated the RR by entering the sample size and the number of results adjusted by the design effect [29]. We used Stata (version 14.2) to calculate all meta-analyses. When meta-analysis was not feasible, we synthesized the results narratively. We assessed the presence of heterogeneity by inspecting forest plots and calculated the I² statistic and Chi² test. In visually discrepant results in the forest plots distribution, we considered as substantial heterogeneity results with significant Chi² test (p < 010) and I² statistic > 50% [27].

We judged available outcomes (vaccination uptake, hand hygiene compliance, and knowledge) using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to assess the certainty of the evidence in its five domains: limitations, indirectness, imprecision, inconsistency, and other factors [31]). We rated the certainty of the evidence of each outcome as “very low,” “low,” “moderate,” or “high” and prepared evidence profiles and summary of findings tables of the effects of combined strategies in comparison to the controls (usual activities or single strategies).

Out of 6941 retrieved records and 2 additional records identified through other sources, we screened 5698 unique records after duplicate removal based on title and abstracts. We assessed full text of 38 studies and included 14 studies [32‐45] in this systematic review and meta-analysis (Fig. 1).

We excluded 15 full-text studies. Five had an ineligible population [46‐50], four had an ineligible intervention [51‐54], three had an ineligible study design [55‐57], two had ineligible outcomes [58, 59], one occurred in offices and thus had an ineligible setting [60]. Nine studies were ongoing up to the conclusion of this review [61‐69], of which four started in 2020 [64‐67]. Two trials were registered in 2007 and 2009 and remained “ongoing” in their protocols [68, 69]. All of them assessed combined dissemination strategies to improve IPC for healthcare workers, including education, training, audit and feedback, positive deviance, a voice enabled virtual assistant (Amazon Alexa device), gamification, and evidence-based telehealth [61‐67].

We included seven parallel RCTs and seven cluster-RCTs, conducted from 2004 to 2020 and funded mainly with research sponsorship (Table 1). Over 65,370 healthcare workers of all categories were assessed for infection prevention and control adherence outcomes that included influenza vaccination uptake, hand hygiene compliance, and knowledge on infection prevention and control. Two studies did not state the number of healthcare workers assessed, just the number of opportunities for hand hygiene [37, 44].

Figure 2 displays the interventions assessed by included studies. All studies based their dissemination of implementation strategies on educational interventions, including materials, meetings, and games [32‐35, 37‐45, 70]. Six studies that assessed hand hygiene compliance [33, 37, 40, 42, 44, 70] used adapted versions of the WHO multimodal hand hygiene improvement strategy, which includes provision of products and infrastructure for hand hygiene, education, observation and feedback, reminders in the workplace, and creation of a safety culture. Three studies [38, 39, 44] conducted surveys and focus group sessions to tailor their dissemination interventions. Monitoring the performance of the delivery of healthcare was employed in two [33, 44], and audit and feedback in four studies [33, 37, 42, 44]. Patient-mediated interventions were used in one of the experimental groups in one study [44], and public release of performance data was part of the intervention in another [39].

Most studies were held in hospitals [34, 37‐39, 41‐45], three in nursing homes [32, 33, 40], one in primary healthcare center [70], and one in a reference clinic for chronic diseases [35]. Nine studies took place in Europe, three in Asia, and two in America (Table 1).

The main biases of the studies were lack of blinding of participants, personnel, and outcomes assessors (Fig. 3). Sequence generation, allocation concealment, and incomplete outcome data affected over one quarter of studies. No study was free from risk of bias (Fig. 3).

Nine studies used adequate methods for random sequence generation [32, 33, 37, 38, 41‐45] and were at low risk of selection bias. Four did not describe the randomization method and were classified as unclear [34, 39, 40, 70]. One study [35] relied on an alphabetical list of the workers, leading to high risk of bias in randomization.

Nine included studies adequately concealed the allocation [32, 33, 37, 38, 41‐45], two were not clear about this process [34, 70], and three did not conceal the allocation [35, 39, 40], regarded as high risk of selection bias.

Three studies adequately blinded participants and personnel and were at low risk of bias [33, 35, 45]. It was not possible to blind participants and the personnel to their group due to the nature of interventions in 11 studies classified as high risk of performance bias [32, 34, 37‐44, 70]. Nine studies did not blind their outcomes assessors to the intervention and we rated as high risk of detection bias [33, 34, 37‐40, 43, 44, 70]. Five RCTs blinded the outcomes assessors, considered as low risk of detection bias [32, 35, 41, 42, 45].

Four studies were at high risk of attrition bias [32, 39, 45, 70] due to losses of facilities or participants during follow up. The other 10 studies had no problem regarding incomplete data; thus, we considered them to be at low risk of attrition bias [33‐35, 37, 38, 40, 41, 43, 44]. We assessed all studies as having a low risk of reporting bias, since they reported the outcomes as described in their protocol or methods.

Combined dissemination strategies improved the influenza vaccination uptake compared to usual activities (RR 1.59, 95% CI 1.54 to 1.81; 4 studies [32, 35, 38, 39], 53,913 participants; I² = 0%; moderate-certainty evidence; Fig. 4). We downgraded the certainty of evidence by one level for study limitations (Table 2).

Combined dissemination strategies may have little effect or no effect on influenza vaccination uptake, compared to a single dissemination strategy (RR 1.01; 95% CI 0.95 to 1.07; 2 studies [41, 43]; 8340 participants; I² = 0%; low-certainty evidence; Fig. 4). We downgraded the certainty of evidence by two levels for limitations and imprecision (Table 2).

Combined dissemination strategies compared to usual activities improved healthcare workers’ hand hygiene compliance (RR 1.70; 95% CI 1.03 to 2.83; 4 studies [33, 37, 40, 70]; 2134 hand hygiene opportunities; I² = 92.2%; moderate-certainty evidence; Fig. 5). We downgraded the certainty of evidence by one level for study limitations (Table 2). As directions of studies’ effects were similar, serious heterogeneity was disregarded. We did not find any factor (year, design, settings, participants, sample size, intervention, or funding) among these studies that explained the statistical heterogeneity.

Combined dissemination strategies may have little effect or no effect on hand hygiene compliance, compared to a single dissemination strategy (RR 1.16; 95% CI 0.99 to 1.36; 2 studies [42, 71]; 3358 hand hygiene opportunities; I² = 85%; low-certainty evidence; Fig. 5). Homogeneity in the directions of studies’ effects led us to disregard serious inconsistency. We downgraded the certainty of evidence by two levels due to study limitations and imprecision (Table 2).

One study assessed whether educational materials and meetings compared to usual activities would improve knowledge about preventive behaviors against healthcare-associated infections in hospital nurses [34]. The researchers assessed knowledge in the pre-intervention, post-intervention and 4 months later, using a questionnaire with a scale ranging from 0 (insufficient knowledge) to 10 (adequate knowledge). These combined dissemination strategies improved healthcare workers’ knowledge of preventive behaviors on IPC, compared with usual activities (MD 4.10; 95% CI 3.39 to 4.81 in post-intervention; MD 4.1; 95% CI 3.36 to 4.84 in 4 months later; 120 participants; very low-certainty evidence). Due to very low certainty evidence—downgraded by three levels for study limitations, indirectness, and imprecision—, we are uncertain of this effect (Table 2).

One study assessed whether the educational game plus the pre-hospital COVID-19 guidelines compared to the guideline alone would improve knowledge about the use of protective equipment [45]. The researchers measured knowledge using an online survey about the choice of personal protective equipment in short clinical scenarios on a scale of percentage of correct answers (0 to 100%). We are uncertain if combined dissemination strategies impact on healthcare workers’ knowledge of use of protective equipment, compared to a single strategy (17% IQR 8, 33% versus 8% IQR 8, 33%; p = 0.27; 173 participants; very low-certainty evidence). We downgraded the certainty of evidence by three levels for study limitations, indirectness, and imprecision (Table 2).