Community health workers adherence to referral guidelines: evidence from studies introducing RDTs in two malaria transmission settings in Uganda

In many sub-Saharan African countries where malaria is endemic, community health workers (CHWs) have received renewed interest to deliver primary healthcare in areas with poor access to public health services, and CHW programmes to treat common childhood infections of malaria, pneumonia and diarrhoea (known as integrated community case management (iCCM)) have been introduced in over 25 countries with the aim of reducing under-five mortality [1, 2]. Previous studies have shown that adequately trained CHWs can correctly diagnose and treat children with uncomplicated malaria, resulting in mortality reductions [3, 4]. Although CHW training usually includes guidance on when children should be referred to a fully qualified health worker, few studies have examined CHW adherence to these referral guidelines [3, 5‐7]. Guidelines can serve to: (1) support CHWs to make appropriate referral decisions; (2) encourage caregivers to seek further care from health facilities; and, (3) ensure CHWs do not risk managing illnesses they are not trained for, and thus limit adverse outcomes that may arise if children do not receive attention from qualified health workers.

In contrast to the growing literature on CHW adherence to treatment guidelines, there is much less describing adherence to referral guidelines, and findings are mixed, with referral ranging from 9 to 83% [8‐10]. Provision of guidelines to CHWs on which illnesses should be referred to health centres is the start of a complex process involving CHWs, caregivers and health workers in health facilities [11]. First, CHWs need to have the skills to identify and distinguish children with severe signs and symptoms from those who do not, and refer them promptly. Second, when referrals are made, caregivers need to adhere to the advice. Last, health facilities need to be prepared to receive, assess and treat referred cases promptly and effectively. The lack of evidence on referral has been highlighted as a priority for further research to inform the implementation and scale-up of iCCM globally [12‐15].

Previous studies on adherence to referral were small-scale evaluations involving a few CHWs, conducted in the context of a presumptive clinical diagnosis for malaria, before WHO-recommended parasitological testing with malaria rapid diagnostics tests (RDTs) at all levels of the health system, including the community [16]. There is urgent need for contemporary data, based on larger samples, to be representative of the range of referral practices amongst CHWs. In this analysis, data collected during trials conducted in Uganda to evaluate the effect of a CHW intervention using RDTs on malaria treatment, was used to assess CHW adherence to referral guidelines [17]. These two trials provided an opportunity to describe CHW adherence to referral guidelines and explore the factors related to adherence when CHWs were trained to use RDTs and prescribe artemisinin-based combination therapy (ACT).

These results suggest that CHWs failed to refer up to 60% of children who should have been referred according to the referral guideline (high fever temperature ≥38.5 °C). However, this study also found that CHWs using RDTs in the low transmission setting adhered more frequently to referral guidance compared to CHWs using presumptive diagnosis based solely on signs and symptoms; both RDT-negative and -positive children were more likely to be referred compared to CHWs not testing with an RDT. CHW adherence to ACT treatment guidelines was no better: all children with a high fever who were RDT-positive (or not tested with an RDT) should have received rectal artesunate pre-referral treatment, but fewer than 20% of these children received the correct treatment. Failure to prescribe pre-referral treatment (rectal artesunate) and refer is a concern because not referring these children for further management has the potential to worsen their condition.

Children with high fevers not referred were often RDT-positive and prescribed an ACT, suggesting CHWs may have overlooked the need for referral once an ACT was given. In contrast, RDT-negative children not prescribed ACT were more likely to be referred to the nearest health centre for further management. These results also show poor CHW adherence to rectal artesunate prescribing guidelines; the WHO guidance for the management of severe malaria in primary care recommends pre-referral treatment with rectal artesunate and referral [25]. Children with high fevers who visited CHWs using a presumptive diagnosis and those who were RDT positive should all have received rectal artesunate and been referred according to the study guidelines, due to their increased risk of severe malaria (Additional file 1: Figure S1). However, in both settings children with a high fever who were RDT positive or not tested were frequently prescribed an ACT and not referred. Despite the low use of rectal artesunate overall, it is reassuring that when CHWs did prescribe rectal artesunate, they always referred.

In addition to malaria test result and treatment, the CHWs’ decision to refer also differed according to both the day and season of visit, in one study area. This could possibly be due to CHWs knowing that local health centres (HCII and III) would be closed during the weekend and that access would be difficult due to the poor condition of roads during the rainy season, factors which might affect whether the caregiver would take their child to a health centre. In contrast, in the other transmission setting there was no evidence of an association between the day or season of visit and referral. This difference could have been due to the existence of a nearby mission-run hospital open during the weekend, enabling CHWs to refer in the belief that caregivers would be likely to seek further care. There is also the possibility that CHWs did not refer because caregivers influenced the decision to refer. Although this was not captured systematically, ad-hoc evidence from the open-text comments section of the CHW treatment record forms confirm instances where caregivers refused referral and CHWs did not refer them (3% of all comments made). Similar observations were made by Winch et al. who found CHWs were more likely to make a referral when they knew caregivers would comply with their advice [26].

Adherence to the referral guidelines by CHWs was generally higher in the low transmission setting compared to the moderate-to-high transmission setting. This may be due to differences between the two trial sites. First, CHWs in the low transmission setting saw fewer children than CHWs in the moderate-to-high transmission setting, and may have been less trusting of the malaria test and/or their competence in deciding when to refer. CHWs who are less confident in their role may prefer to refer according to the guidelines and not risk further complications in the child. In contrast, CHWs in the moderate-to-high transmission experienced higher number of visits, will have obtained case management experience more rapidly, and may have made their own judgements on when to prescribe rectal artesunate or refer. Similar patterns have been reported in other community-based studies in Uganda, where drug shop vendors became increasingly confident in their skills to manage clients and were reluctant to refer clients [27]. Second, the low transmission setting is historically prone to epidemics, children develop acquired immunity more slowly due to reduced malaria exposure, and may be more likely to develop severe malaria if not treated [28, 29]. CHWs may have been aware of this risk and more inclined to refer. Finally, the presence of a nearby mission-run hospital with an insurance scheme in the low transmission setting may have enabled CHWs to refer knowing care would be available; no such hospital or insurance scheme was present in the moderate-to-high transmission setting.

In this analysis, adherence to referral guidelines was assessed through review of treatment records kept by CHWs; temperature was as an indicator of whether a referral should have been made because it was recorded for all children. There are potential limitations to this approach. First, direct observation with re-examination by a medical professional is an established method to assess health worker adherence to case management guidelines in hospitals [30, 31]. This analysis was limited to data available from the two trials, which did not use this method of evaluation. However, prior studies which have used direct observation to evaluate CHW performance have found evidence of the Hawthorne effect, where CHWs may have followed guidelines more accurately under observation in a clinical setting compared to their community environment [8, 32]. Register review has previously been found to approximate results from a medical professional directly observing CHW performance [31, 32], and can offer a number of advantages: we were able to screen and analyse approximately 22,000 records in less time, with fewer resources compared to a direct observation method, without having to remove CHWs or health centre workers from their normal work. Second, the analysis was limited to data that were routinely recorded, but other unrecorded factors could also have influenced a CHW’s decision to refer or not, including perceptions of the availability and quality of care available at local facilities, caregiver demands, and interpersonal relationships. However, since CHW supplies were regularly monitored as part of the trial, it cannot be certain that non-adherence to the guideline was not due to stock-outs of rectal artesunate or referral forms. Finally, the single criterion (temperature ≥38.5 °C) used in this analysis is not necessarily representative of all criteria for referral, of all children needing referral, or adherence to referral guidelines for other symptoms. Indeed, when children with high fever were referred CHWs often reported additional severe signs and symptoms for referral alongside the temperature criterion, suggesting that CHWs may have thought that high fever on its own did not justify a referral. It is conceivable that other severe signs and symptoms may have had higher referral rates; also, that referral might be lower for non-severe signs and symptoms.

Despite these limitations the findings are comparable with two earlier studies in Ghana where CHWs used a presumptive malaria diagnosis, which found similar patterns of low referral [9, 10]. In one, 5 out of 17 children with signs of severe disease requiring referral were referred with a form; the authors thought this might reflect CHWs’ limited confidence in their diagnosis and a preference to re-assess later the need for referral [9]. It is possible in this study that CHWs may have also followed up and reassessed severely ill children for referral, however this was not systematically recorded.

This is one of the first studies to investigate CHW adherence to referral guidelines when RDTs have been implemented as part of a community case management programme, and although CHWs in these trials were trained only to treat malaria, the findings may also have relevance for referral guidance within iCCM programmes. The use of RDTs in any programme will identify some children that a CHW is not equipped to treat, and guidance for managing RDT-negative children is needed. In both settings CHWs referred eligible children more frequently when they were RDT-negative compared to when they were RDT-positive or not RDT tested. Nonetheless, it is probably impractical that all children who test RDT-negative should be referred without consideration of other signs or symptoms. CHWs in this study were trained to identify more than 20 signs and symptoms for referral. Is this too many? Can inclusion of non-severe signs diminish the perceived importance of adherence to referral guidelines? More worrying, there was strong evidence that children treated by a CHW were not referred. Treatment for one condition can thus increase the risk that CHWs overlook other signs and symptoms requiring referral, underlining the emphasis that needs to be placed on the referral criteria during CHW training. Although children with a high fever are at increased risk of febrile convulsions, it is possible that, in the absence of other symptoms, high fever was not perceived by CHW or caregiver to be serious enough to justify referral. Caregivers may prefer to observe the progression of their child’s fever at home and may not comply with CHW referral advice to visit a health centre. Poor compliance to referral by caregivers has been reported in studies in Uganda and Sierra Leone [33, 34]. There is considerable heterogeneity in national and iCCM guidelines on the referral criteria used in the countries that are currently scaling-up iCCM [35]. Some, but not all, include high fever as a criterion for referral. Although the findings presented here are not necessarily generalizable to other referral criteria, the results highlight the need to understand how CHWs make decisions on when to refer, and the factors that influence these decisions. Failure to refer may result in delayed treatment seeking and poorer child health, which would partly undo the benefits of increasing access to primary healthcare afforded by using CHWs [36, 37].

In this study, CHWs tended not to refer children presenting with high fever (temperature ≥38.5 °C) if they had confirmed malaria diagnosis with RDT and prescribed an ACT. This practice was inconsistent with the CHW training guidelines that recommend referring children presenting with one or more severe symptoms. In other settings where CHW interventions are being implemented, further research is required to fully understand when and how CHWs decide to refer children, and the factors that influence their decision, in order to refine the guidelines and improve management of febrile illness in children.