Study population

We enrolled 67 CHW from the 15 villages. All participants completed the baseline survey, and 63 (94.0%) completed the follow-up survey. Two CHW withdrew from the parent study after completion of the baseline survey, and two others were not available at the time the follow-up survey was administered. A subset of fifteen CHW completed the individual semi-structured interviews.

Sociodemographic characteristics and the experience of the participating CHW are detailed in Table 2. In brief, the median age was 40 years (interquartile range (IQR): 35–47) and 61.2% (41/67) were female. They had a median of 9.5 (IQR: 4.0–14.0) years of experience as a CHW and saw a median of 10 (IQR: 7–15) children per month. The full cohort and interview subgroup were similar in age, sex, and educational level. The interview subgroup had slightly fewer years of experience (8.0 [IQR: 4.0-12.0]), saw slightly more children per month (12.5 [IQR: 8–20]), and had a higher proportion of individuals with primary occupations other than subsistence farmer.

Table 2 Baseline characteristics of participating village health workers (CHW) treating children with acute respiratory illness (ARI) in the 15 study villages in Western UgandaKnowledge of antibiotics

Knowledge of antibiotics was assessed at baseline and follow-up, the results of which are summarized in Table 3. At baseline, the majority of CHW identified amoxicillin as an antibiotic (60/63, 95.2%), associated antibiotics with treatment of bacterial as opposed to viral infections (52/63, 82.5%) and knew that antibiotics should be stopped only after the full prescribed course has been taken (57/63, 90.5%). Over two-thirds were able to identify at least one of the risks of antibiotics (44/63, 69.8%). For all questions, the proportion who answered correctly increased on the follow-up survey, except for the question focused on differentiating antibiotics from anti-inflammatory drugs (51/62, 82.3% v. 41/63, 65.1%; p = 0.03).

Table 3 Knowledge of antibiotics. Proportion of CHW who answered questions regarding knowledge of antibiotics correctly before and after participating in the STAR study, limited to observations where responses were available for both time points (n = 63)Antibiotic use practices

At baseline, all CHW reported giving antibiotics at most or some patient visits. When specifically asked about children presenting with cough, 26.9% (18/67) reported giving antibiotics at most visits and 67.2% (45/67) at some visits. When asked about resources used to make antibiotic treatment decisions, all who responded (n = 66) selected the SJCA, which they often referred to as “the bible” in the semi-structured interviews [ID: 10, 40, 48, 55, 57]. During the interviews, the CHW described that, they mostly relied on “counting the breathing rate of a child depending on their age group” [ID: 15] and prescribing amoxicillin if the rate was high, the technique that is outlined in the iCCM SCJA.

In total, 14 of the 15 CHW interviewed as part of this study expressed challenges or concerns associated with relying on breathing rate as an antibiotic prescription tool. They shared anecdotes of children “push[ing] the timer while you are trying to count”, disturbing the test “by shaking themselves”, or altering their breathing rate upon starting to cry, affecting their ability to properly diagnose [ID: 50, 57, 03]. Some also noted that there are multiple causes of fast breathing other than pneumonia, including malaria, fever, crying, or fear. This complicated the use of this diagnostic tool for identifying pneumonia and may have led to unnecessary antibiotic use. “A high temperature can lead to fast breathing. In such cases, I would give them amoxylin (sp) when am also doubting. I would also think that the amoxylin (sp) is not appropriate because it could be the high temperature that causes fast breathing.” [ID: 03].

Antibiotic perceptions

Overall, CHW perceived antibiotics as harmful when used incorrectly, both at baseline and follow-up. Over 80% recognized prescribing antibiotics if a patient does not need them as potentially harmful to the patient (Table 4 and S1), and > 90% agreed that excessive consumption of antibiotics was bad for health (Table 4) at both time points. Ten of the 15 interviewed CHW mentioned this viewpoint as well. As one CHW stated, “I think, if I give or if a person takes antibiotics when they don’t need them, they might damage their life because, instead of the antibiotics curing the illness, they might fight with the body cells that are actually responsible for protection of the body against yet their body disease. It’s because the person will have taken antibiotics for a wrong illness. So I still think that the antibiotics will cause harm on the person’s health.” [ID: 26] Relatedly, over 90% of CHW thought it was “not ok to obtain” antibiotics without seeing a CHW or clinician in a healthcare facility at both baseline and follow-up (Table 4).

Table 4 Perception of antibiotics. tTop panel shows proportion of CHW participating in both baseline and follow-up surveys (n = 63) who either strongly agreed or agreed with each statement at baseline and follow-up. Frequency of each of the five possible responses is included in Table S1 . Mean likert scores for each statement at baseline and follow-up are reported and compared by Wilcoxon signed rank test in Table S2 . Bottom panel shows frequency of responses to three additional multiple choice questions regarding perceptions of antibiotic use

Discussion of the consequences associated with antimicrobial overuse, including antimicrobial resistance, emerged during semi-structured interviews, as noted by a CHW with 7 years of experience sharing, “if the child takes antibiotics when they are not sick, the child may get used to the medicine, and when they get sick in the future, the medicine may not treat them” [ID: 11]. Health concerns elevated by CHW in association with antibiotic overconsumption included liver damage, harm to the body’s immune response, longer recovery times and increased incidence of subsequent illnesses [ID: 22, 26, 10, 15]. In total, 13 of the 15 CHW interviewed acknowledged appropriate patient health concerns stemming from antibiotic overuse. When asked if they agreed that bacteria developing resistance to antibiotics presented a major problem in Uganda on the surveys, slightly over half of CHW either agreed or strongly agreed with this statement (baseline mean score 2.64; Table 4, S1, and S2).

CHW also noted the potential benefits of antibiotic use with 5 of the 15 interviewed CHW commenting that they can be effective for treatment of respiratory illness. “I find these antibiotics helpful because when children take this medicine, they get well and that becomes helpful to us since it shows that we are doing a great job. So I don’t see any problem with using these antibiotics.” [ID: 50] However, CHW shifted their perspective on prescribing antibiotics as standard practice between before and after the STAR study. The proportion of CHW who believed amoxicillin is a useful treatment for all respiratory infections decreased between baseline and follow-up (Table 4 and S1). As one CHW stated during the interview, “we learn that it’s not everybody that has cough has got a bacterial infection.” [ID: 03].

At baseline, 49.2% of CHW agreed or strongly agreed that prescribing antibiotics “when in doubt” was good practice just in case the patient had pneumonia (mean response score 3.13; Table 4, S1, and S2). This number dropped to 6.4% agreement or strong agreement in the follow-up survey (mean response score 3.98). This shift coincided with a slight increase in confidence in CHW’s ability to prescribe antibiotics correctly, 77.8% of whom felt “very confident” at follow-up compared to 71.4% at baseline (Table 4). Interestingly, despite 75.4% of study participants believing that antibiotics were overprescribed by CHW in Uganda at baseline, just 14.3% held this view during follow-up (Table 4 and S1). Multiple CHW reported in the interviews that the use of the STAR SCJA (including the CRP test) and/or participation in the study and its associated trainings directly led to shifts in their practice – “I knew I was using the knowledge I had acquired during the trainings and not guess my work.” [ID: 30].

Caregiver and community dynamics

Caregiver expectations and the therapeutic relationship between the caregiver and provider can affect antibiotic treatment decisions. When asked to assess caregiver influence, over 80% of CHW disagreed that antibiotics were prescribed solely because caregivers desired them in both baseline and follow up-surveys. The interviews overall suggested that CHW are valued and respected by the community, with a CHW sharing, “the parents appreciate me a lot and thank me for treating their child” [ID: 43]. The majority of CHW (11 of 15) expressed appreciation in return, as noted by a CHW who shared, “I wanted to help my community members so that they do not always struggle moving long distances to go and look for medical services” [ID: 64], and another who described, “The community entrusted me and raised my name and said they want me to work as a VHT. So when they shared that idea with me, I accepted it and agreed to work for the people.” [ID:03] Despite this mutual respect, six CHW described caregivers attempting to influence their care, including coercion and accusations of inadequate or insufficient care, and others did report some initial mistrust of the study treatment algorithm. Furthermore, over 40% agreed or strongly agreed that caregivers get antibiotics from the pharmacy on their own if a CHW or clinician does not prescribe one (Table 4 and S1). However, two CHW reported gaining caregiver trust in the STAR SCJA once they witnessed a child heal without antibiotics.

In Uganda, there is a robust network of mostly unlicensed drug shops that sell antibiotics without a prescription [23]. When asked whether caregivers obtain antibiotics on their own from the drug shops without seeing a VHT or clinician, over 60% in the baseline survey and 70% in the follow up-survey either agreed or strongly agreed. (Table 4 and S1). Five of 15 CHW also described caregivers purchasing antibiotics from drug shops if the CHW does not prescribe them or in place of presenting to the health facility when referred there by the CHW.

STAR algorithm engagement

Semi-structured interviews suggested the STAR SCJA was easy to use, improved treatment accuracy, aided decision-making and facilitated communication with caregivers. One CHW shared, “it has clear instructions; it guides you on how to treat children with fast breathing and also how to handle those that don’t have fast breathing” [ID: 10]. Another CHW added, “it was so easy to understand because some information in the STAR study was also in the ICCM Sick Child Job Aid” [ID: 43].

Decision making was improved through the STAR SCJA by reminding CHW of diagnostic steps they may have skipped, guiding them on how and when to administer medication, and helping them account for concurrent symptoms or complications. When caregivers insisted on receiving antibiotics, CHW reported leveraging the CRP test results as evidence for their decision-making, helping them navigate caregiver dynamics. One CHW shared, “sometimes you find a parent insisting that even if their child has no pneumonia, I should still give him or her medication but now for us we had already undergone training on how you can sit down with a care giver and explain to them on how risky that can be to a child and indeed if you explained to them so well about the risks then they would get convinced” [ID: 30]. As described by another male CHW with more than 10 years of experience, “my tests showed that the child does not require antibiotics at that point. I then explained to the caregiver that, even if the child is coughing, they do not require antibiotics at this time” [ID: 26].

The effects of the STAR algorithm on antimicrobial resistance also emerged during qualitative interviews. A CHW with 11 years of experience reflected on the algorithm as having “widened [his] knowledge that sometimes, [CHW] prescribe antibiotics to people who do not require them” [ID: 26]. Another CHW added, “I got more skills on how to treat children with cough and I even started understanding that even if a child is coughing, it may not necessarily imply that the child has pneumonia” [ID: 43]. Moreover, a CHW recalled how, “the CRP test changed [her] ability to diagnose patients in a way that [she] started to realize that sometimes [CHW] might have been prescribing antibiotics to clients who do not need them, yet the patient could have healed without any antibiotics” [ID: 003]. The CHW with 11 years of experience also acknowledged the value of the CRP test by saying, “the CRP test is good…it guides us to give antibiotics to children who actually need them” [ID: 26].

As such, CHW favored the CRP diagnostic test and the STAR SCJA for its perceived improved accuracy. As summarized by one CHW, “any other illnesses can cause fast breathing among children when using a timer, so I felt that a CRP test is more accurate because you are testing an illness directly”. He continued, “it changed my ability in a way that I was confident in the test” [ID: 35].

Recommendations

Recommendations from CHW included further training and tools for evaluating sick children (both with ARI and other illnesses not currently covered by iCCM), increased availability of CRP test kits, initiation of community discussions on antibiotic use and antimicrobial resistance, and adoption of the STAR SCJA indefinitely through its integration with the iCCM SCJA.

All CHW completing the baseline and follow-up surveys welcomed more training or reference materials on antibiotics (Table 4 and S1). The cohort also all either agreed or strongly agreed that it would be helpful to have more diagnostic tests to help determine what causes a child’s illness (Table 4 and S1). The proportion who strongly agreed about both these statements increased at follow-up compared to baseline (TableS1 ). Similar sentiments were expressed in the semi-structured interviews. As shared by one CHW, “the new job aid would also include diagnosing and treating flu among children since most caregivers bring to us children with high fever, and they turn out negative for both malaria and cough, so we suspect that sometimes the fever might be caused by flu” [ID: 15]. Another stated, “…if they can introduce more tests like the CRP test then we shall be able to overcome making the mistakes we were always doing in the past” [ID: 64]. The desire to keep using the CRP test kits was also captured by a CHW who explained, “we can never be accurate when using the timer…but then the CRP test may instead give me accurate results since I don’t have to worry about a child disturbing me” [ID: 30].

Nine CHW elevated the idea of community discussions to create awareness of antibiotic resistance and explain the purpose of the job aids used by CHW. One CHW commented how the community “should be advised on the effects of taking medicine without being sure of what illness you are sick of” [ID: 57]. This sentiment was echoed by another CHW sharing, “sensitization is needed because now ignorance is causing more diseases to us” [ID: 30]. One even proposed that “if we as CHWs can always be trained with the community members in attendance too then they can also understand the way we do our work as far as treatment of children is concerned” [ID: 43]. Lastly, nine CHW interviewed expressed a desire to adopt the STAR SCJA indefinitely, as outlined by one CHW who shared, “this study should not stop here so that all the other people that have not yet benefited from the study can also benefit so that we can all receive this knowledge.” [ID: 26].