The principal finding is that while a multimodal intervention based on the WHO’s Hand Hygiene Improvement Strategy led to improvements in hand hygiene compliance, the improvements were small and would not be expected to substantially reduce HCAI rates. To our knowledge, this is the first time a hospital-wide implementation of the WHO hand hygiene strategy has been evaluated in a randomized trial in a LMIC, so it is important to consider why it was unsuccessful at achieving the desired change. We believe three main factors are likely to have been important. First, human resource limitations are likely to have played a role. Because we wanted our findings to be generalizable to other resource-limited hospitals in LMICs, the intervention only made use of resources normally available to the hospital and its infection control team. We therefore deliberately did not employ additional infection control staff for delivering the intervention; nor did we bring in external expertise in behavior change or social marketing, and all promotional and training materials were produced in-house without professional help. We did, however, provide access to high quality training in infection control to members of the hospital infection control team before the intervention started, with members of the ICT attending a course on “Healthcare-associated Infection Prevention and Management” held by ESCMID-SHEA Training Course in Hospital Epidemiology in 2014. We also translated the WHO’s hand hygiene strategy into Thai (https://doi.org/10.21954/ou.ro.0000d7f3), to ensure that infection control staff were able to acquire a good knowledge of the WHO strategy. We provided additional resources for evaluating the intervention to ensure that this did not in any way impact on the ICT’s workload. Nonetheless, the ICT were not always able to deliver the intervention as planned because of competing infection control and organizational priorities including routine surveillance, responding to outbreaks and preparing for a nationally mandated hospital accreditation event. While we did not attempt to assess qualitative aspects of the intervention components which were implemented, clearly some approaches to providing reminders in the workplace, training, system change, promoting a safety climate, and giving feedback are likely to be more effective than others. We can speculate that had more resources, including time and professional expertise in designing effective behavior change interventions, been devoted to these components there would have been more chance of effecting meaningful improvements.

Second, while certain wards did take ownership of the intervention as had been the intention, resulting in some substantial ward-specific improvements in hand hygiene behavior, a more common pattern was that the intervention was not fully implemented or was implemented late. The ICT were not always able to obtain the co-operation of the majority of ward staff despite high-level endorsement of the intervention by the hospital director. This reflects the fact that although the intervention was explicitly endorsed by senior hospital management, organizational factors and other priorities sometimes got in the way. For example, hospital staff changes occurring during the study caused some delays in implementing the intervention. Other studies have reported similar problems, with higher priority work preventing effective implementation of hand hygiene interventions.²⁴

Third, physical resource limitations may have also played a role, particularly for opportunities when hand washing with soap and water was indicated, and insufficient provision of conveniently located sinks and lack of provision of towels at sinks were both reported by hospital staff as likely contributory factors to poor hand hygiene compliance (https://doi.org/10.21954/ou.ro.0000d7f3)

The baseline hand hygiene compliance in this study of about 10% is low, but not atypical. Similar pre-intervention rates have been reported in other studies in both low- and high-income countries.^25-27A recent systematic review found compliance rates of below 20% to be common for all types of HCW.²⁸

The most extensive previous evaluation of the WHO’s multimodal strategy for improving hand hygiene in LMIC settings used a before-after study design and included three countries: Mali, Costa Rica and Pakistan.^27,29 Substantial improvements in hand hygiene in all three countries were reported: respectively, from 8% pre- to 22% post-intervention compliance in Mali, from 40% to 60% in Costa Rica, and from 38% to 59% in Pakistan. Although the before-after design employed in this study is considered to be weak because of its vulnerability to secular trends, the magnitude of the changes reported is consistent with moderate to large beneficial effects. In a similar manner to our study, the interventions were obtained mainly with the use of local resources and expertise. One important difference, however, is the direct involvement of the WHO team who developed the multimodal strategy. Assuming that the changes seen were caused by the intervention, it is possible that the involvement of such a high-profile team helped to galvanize behavior change amongst hospital staff and overcome organizational inertia. More recently, one other study in Sub-Saharan Africa has reported substantial increases in hand hygiene compliance (from 34% at baseline to 69% post-intervention), but again using a weak before-after study design. Assessment of the published literature also needs to take into account the likelihood of publication bias, whereby small (often methodologically weak) study designs are more likely to be reported if they show a positive result.^13,30 A previous systematic review of hand hygiene interventions that included such before-after studies found clear evidence of such publication bias,³⁰ and it is notable that large studies of hand hygiene interventions employing the strongest study designs (and with pre-registered protocols) have consistently shown the smallest effects.^31-33 Indeed, our estimated effect size is consistent with those seen in previous interrupted time series studies.³¹ For these reasons, we believe that our experience of an intervention that achieved a statistically but not clinically significant improvement in hand hygiene may be representative of an underlying problem that has been largely neglected from the published literature. Indeed, the difficulties in implementing the intervention as planned in this study echo those from a previous CRCT of a hand hygiene improvement intervention,³⁴ which found that poor fidelity to the planned intervention may have substantially diluted the potential benefits. Competing priorities with other quality improvement initiatives were identified as one possible cause of these implementation difficulties.

Well-resourced national campaigns modelled on the WHO strategy have helped to bring large, persistent and well-documented improvements in hand hygiene compliance in some high-income countries and these have been shown to be associated with reductions in infection.^35,36There is also some evidence that the WHO strategy can be further improved by adding interventions including goal-setting, reward incentives, and accountability.¹⁰ In our study, goal-setting was used as an additional intervention by two wards who chose to have routine meetings to evaluate the intervention plan, revise the goals and plan to achieve target rates of hand hygiene. It has also been proposed that team–based approaches could help promote hand hygiene with ward-based workshops to discuss the workflow and make ward-specific plans.²⁴ Using this approach Huis et al. found high improvement in certain wards that were successful in making culture changes at the ward to make staff more aware of performing hand hygiene.³⁷ In our study, there were also examples of creative ideas originating from ward staff to promote hand hygiene. For example, some wards set a hand hygiene day in every month. On these days there were activities to promote hand hygiene in the ward in a similar manner to the annual hospital hand hygiene day.

The principal strength of our study is the strong design used to evaluate the impact of the WHO strategy,^16,34 and the hospital-wide intervention. Most previous studies from LMICs have evaluated interventions only in more restricted study populations.

The study’s main limitation was that, as with other hand hygiene studies based on direct observation, Hawthorne effects may have distorted HCW hand hygiene behavior. We had hoped to overcome this problem by independently assessing hand hygiene compliance by measuring volume of alcohol-based hand sanitizer use on each ward, though this proved impossible to do reliably in practice. A second limitation is the possibility of contamination between clusters, as interventions made on one ward could affect staff behavior on other wards. This would be expected to result in overall increases in compliance over time, but might lead to smaller differences between pre- and post-intervention periods in implementing wards.