- Open Access
Integrated community case management by drug sellers influences appropriate treatment of paediatric febrile illness in South Western Uganda: a quasi-experimental study
Malaria Journalvolume 16, Article number: 425 (2017)
Fever case management is a major challenge for improved child health globally, despite existence of cheap and effective child survival health technologies. The integrated Community Case Management (iCCM) intervention of paediatric febrile illnesses though adopted by Uganda’s Ministry of Health to be implemented by community health workers, has not addressed the inaccess to life-saving medicines and diagnostics. Therefore, the iCCM intervention was implemented in private drug shops and evaluated for its effect on appropriate treatment of paediatric fever in a low malaria transmission setting in South Western Uganda.
From June 2013 to September 2015, the effect of the iCCM intervention on drug seller paediatric fever management and adherence to iCCM guidelines was assessed in a quasi-experimental study in South Western Uganda. A total of 212 care-seeker exit interviews were done before and 285 after in the intervention arm as compared to 216 before and 268 care-seeker interviews at the end of the study period in the comparison arm. The intervention effect was assessed by difference-in-difference analysis of drug seller treatment practices against national treatment recommendations between the intervention and comparison arms. Observed proportions among care-seeker interviews were compared with corresponding proportions from 5795 child visits recorded in patient registries and 49 direct observations of drug seller–care-seeker encounters in intervention drug shops.
The iCCM intervention increased the appropriate treatment of uncomplicated malaria, pneumonia symptoms and non-bloody diarrhoea by 80.2% (95% CI 53.2–107.2), 65.5% (95% CI 51.6–79.4) and 31.4% (95% CI 1.6–61.2), respectively. Within the intervention arm, drug seller scores on appropriate treatment for pneumonia symptoms and diagnostic test use were the same among care-seeker exit interviews and direct observation. A linear trend (negative slope, − 0.009 p value < 0.001) was observed for proportions of child cases prescribed any antimicrobial medicine in the intervention arm drug shops.
The iCCM intervention improved appropriate treatment for uncomplicated malaria, pneumonia symptoms and diarrhoea. Drug seller adherence to iCCM guidelines was high, without causing excessive prescription of antimicrobial medicines in this study. Further research should assess whether this effect is sustained over time and under routine supervision models.
Uganda is one of only ten countries in sub-Saharan Africa that achieved millennium development goal 4 of a two-third reduction in the 1990 levels of under-five mortality by 2015 . However, the sustainable development goals (SDG) framework has proposed a further reduction from the current 55–25 deaths per 1000 live births by 2030 [2, 3]. Acute febrile illnesses of malaria, pneumonia and diarrhoea account for 45–60% of these deaths [3, 4]. Many Ugandan febrile children either do not receive prompt healthcare or get the wrong treatment or go untreated, despite existence of cheap and effective child medicines and diagnostics [5, 6]. In household surveys, just 60% of the children suspected with uncomplicated malaria receive an anti-malarial medicine and only 23% get the recommended artemisinin-based combination therapy (ACT). Moreover, only 17% get parasitological diagnosis  and a meagre 14% receive ACT within 24 h from onset of fever . Similarly, only 31% of under-five children who present with pneumonia symptoms receive an antibiotic, and 35% of diarrhoea cases receive oral rehydration salts (ORS), and only 2% receive zinc tablets .
The integrated community case management (iCCM) for paediatric febrile illnesses programme is a community level intervention where community health workers (CHWs) are trained on integrated case management of febrile illnesses. They are provided with the point-of-care diagnostics (such as antigen-based malaria rapid diagnostics tests and respiratory rate counters) and medicines (including dose-packed colour-coded ACT and dispersible tablet amoxicillin) to detect illness, classify and treat U5 children accordingly [9, 10]. CHWs are community-based volunteers, also referred to as the village health teams (VHTs) selected by village elders and peers from within the community.
Whereas studies done in rural areas in Uganda, Ethiopia and Zambia [5, 11,12,13,14,15,16] demonstrate that the iCCM intervention is an effective programme with a 40% potential reduction of U5 mortality , its implementation in Uganda has been fragmented and so far rolled-out to only 34 of the 112 districts [17, 18]. Additionally, iCCM implementation has been characterized by challenges of inadequate supervision, unreliable medicine and equipment supply chains, low motivation and retention of CHWs, weak monitoring and evaluation systems and low uptake of CHWs-services in certain areas [19, 20]. Consequently, poor children in remote and underserved areas in Uganda have not fully harnessed the benefits of the iCCM intervention.
To complement the unreliable, sometimes inaccessible or even non-existent facility-based services and community health programmes, people seek fever care from private drug shops and clinics [7, 21]. Drug shops are small-scale medicine outlets that are granted licenses to sell a limited list of medicines by the National Drug Authority (NDA) following successful vetting of personnel, physical premises and payment of prescribed fees . The drug shops and outlets that serve remote areas have minimal health infrastructure  manned by people with health training from zero to 2 years [24,25,26]. In rural areas, up to 53% of care-seekers with U5 children first seek fever care from drug shops [6, 7, 27]. However, these drug outlets occasionally provide treatments that are inconsistent with evidence-based clinical guidelines and are potentially harmful [24, 27]. Yet, these outlets still remain the first port-of-call for poor and disadvantaged children with febrile illnesses .
A pilot study in Eastern Uganda infers that targeted interventions to improve drug seller treatment practice increase coverage and access to high quality care for childhood fevers in underserved high malaria endemic regions. In addition, malaria rapid diagnostic tests (RDTs) results are adhered to in prescribing anti-malarial or alternative fever treatments [29, 30]. It is not clear from this study, however, how a lower malaria transmission rate affects compliance to RDTs results and use of alternative treatments for febrile conditions [31,32,33]. In low malaria transmission areas, the drug seller may have to make a choice between financial incentives from dispensing antimicrobial drugs to test-negative cases and appropriate medicine use. It is also important to know if an iCCM intervention is good use of resources or not with regards to the emergent care-seeker utilization and drug seller treatment practices.
Hence, the authors set out to determine the effect of the iCCM intervention derived from the WHO/UNICEF recommended iCCM on appropriate testing and treatment for uncomplicated malaria, pneumonia symptoms and non-bloody diarrhoea among U5 children attending drug shops in two lower malaria transmission districts in South Western Uganda. In the intervention arm, drug seller adherence to the iCCM guidelines from three different data sources—care-seeker exit interviews, drug shop patient registries, and observation of drug seller–care-seeker encounters—and the trend of prescription of antimicrobials were evaluated.
A prospective evaluation of the iCCM intervention in registered drug shops was done in Mbarara (intervention) and Bushenyi (comparison) districts. These districts are located approximately 250 km South West of the Uganda commercial and administrative capital . They were chosen for this study because they had similar low malaria transmission rates, geographical terrain, and vegetation and a typical tropical climate with rainfall peaks in April and October during the year . Recent surveillance studies estimated the malaria parasite prevalence in the region to be between 4.1%  and 9.3% . By virtue of their location in South Western Uganda, the authors argued that the two districts were also similar on other factors that affect care-seeking from drug shops and drug seller treatment practices such as local beliefs, cultural practices, social economic status, health systems factors and medicine supply chain factors. The district drug inspector, district health educator, a cadre of CHWs and registered drug shops in both districts were available to participate.
Mbarara district had a population of 472,629 people served by 58 government health facilities, private medicine outlets and informal sector. Sheema district which is located between the two study districts served as a buffer area, minimizing the contamination of the comparison arm by the intervention. Bushenyi district had a population of 234,440 served by 36 government health facilities, private medicine outlets and numerous informal health providers [35, 37]. During the study, Bushenyi drug shops and community continued to receive the routine national malaria control and child health programme activities [18, 38, 39].
The investigation was a quasi-experimental design with one intervention (Mbarara with iCCM in drug shops) and one comparison area (Bushenyi without iCCM in drug shops). The study was conducted for 16 uninterrupted months from June 2013 to September 2015, divided into three stages: pre-intervention, intervention and post-intervention. Quantitative data collection was done before-, during and post-intervention to assess the effect of iCCM intervention.
It occurred from June to August 2013 with three major components:
Stakeholder mobilization and involvement
At national level, the study team consulted with the National Malaria Control Programme (NMCP) and the National Drug Authority (NDA) about implementation of the study. The approved protocol was shared with the NDA and NMCP prior to discussion and agreement on respective stakeholder roles. Subsequently, district level stakeholders including the district local government officials (Chief Administrative Officer, District Health Officer, District Health Educator, and the District Drug Inspector), pharmaceutical wholesalers, CHWs and drug sellers were consulted in inception meetings for advice on study design and extent of involvement.
Selection of drug shops
All registered drug shops on the list obtained from the district drug inspectors were subjected to the eligibility criteria, namely, having an up-to-date license status, location out of the municipality area (i.e. in the rural area) of the study arm, selling human medicines as opposed to veterinary medicines and providing consent to participate in the study. Only rural drug shops were eligible to participate in the study because the iCCM intervention is designed for implementation in rural and remote areas, underserved by formal health facilities, to increase adequate access to diagnostics and medicines for under-five fever management . A total of 217 drug shops in the study area were assessed. In the intervention arm, out of 152 drug shops, only 61 drug shops met the inclusion criteria. The other drug shops were either located in the municipality (60) or sold veterinary medicines (31). In the comparison arm, out of 65 drug shops, 23 were selected and recruited into the study. Similarly, the other drug shops were either located in the municipality (35) or sold veterinary medicines (7). Both the drug shop owner and attendant were recruited into the study.
Drug shop care-seeker exit interviews
The sample size for care-seeker drug shop exit interviews was estimated using the Bennett method for cluster-sample surveys . Each study drug shop was equivalent to a cluster in this study. The assumptions in this sample size estimation were: a prevalence of appropriate treatment for pneumonia symptoms of 15%, expected rate of homogeneity of proportions between clusters as compared to within clusters of 0.02 and expected responses per cluster of 26. A design effect of 1.5 was calculated. Taking this design effect, a two-sided test and level of significance of 5% gave a total of 12 clusters per study arm. However, baseline data was collected from 18 clusters in intervention and only 10 clusters in the comparison arm due to unforeseen variation in and non-performing clusters. At baseline, the number of clusters was increased from 12 to 18 in the intervention arm in order to achieve the desired sample size of exit interviews within the planned period.
Study drug shops were selected from the list provided by the district drug inspector. Directions to the first drug shop were obtained from district drug inspector. Subsequent drug shop(s) were located with directions provided by the recruited drug sellers till the sample size was achieved.
Baseline exit interviews were conducted from September 30th to November 8th 2013. Respondents in the exit interviews were care-seekers who sought care for U5 child with fever or history of fever, cough, fast breathing, difficult breathing or diarrhoea. Respondents were consecutively recruited as and when they exited the drug shop as long as they consented to participate in the study. Data was collected using a structured questionnaire by a trained interviewer, stationed at a convenient location outside the drug shop. Data on socio-demographic characteristics of the sick child and care-seeker, nature of current illness, if other health facilities were visited prior to that drug shop visit were collected. Other variables including medicines given at the drug shop, diagnostic tests conducted, time when that illness was noticed as well as reasons for seeking care at that drug shop were collected. The care-seeker exit interviews were conducted in Runyankole—the main local language used in the study area—and interviewers checked the medicines given to the care-seekers to verify responses to questionnaire items. A total of 212 and 216 care-seeker interviews were conducted in the intervention and comparison arms, respectively.
The intervention phase ran from February 2014 to September 2015. The iCCM intervention in Mbarara district consisted of four different components; namely (1) selection, training and work activities of drug sellers, (2) provision of information, education, information and communication (IEC), (3) supply mechanism by study team in partnership with pharmaceutical wholesalers in Mbarara for diagnostics (malaria RDT and respiratory rate counters) and medicines (ACT, amoxicillin dispersible tablets (DT) and zinc sulfate/ORS), (4) monthly support supervision done by study field supervisor trained in either pharmacy or clinical medicine, occasionally accompanied by the district drug inspector and district health educator. Table 1 provides details of the intervention components.
Towards the end of the study (May 4th to June 19th 2015), 285 and 268 drug shop care-seeker exit interviews were conducted in the intervention and comparison arms, respectively. These care-seeker exit interviews mirrored those conducted at baseline; they were conducted in drug shops that had participated in baseline exit interviews and the same structured questionnaire was used. In addition to exit interviews, trained interviewers used a pre-tested structured checklist to conduct 49 direct observations of drug seller–care-seeker interactions at intervention arm drug shops. Interviewers assessed actual verbal and nonverbal behavior of the drug sellers against the standard iCCM sick child job aid (treatment algorithm) to assess their quality of paediatric fever assessment and treatment of sick children .
In September 2015, the study field supervisor picked the final batch of the patient registries from all the intervention arm drug shops. At the completion of care-seeker exit interviews at respective drug shops, the field supervisor used the data collection dates and chronological order of observations to check the responses provided in care-seeker interviews against records of the same cases in the drug shop patient registry. Information of the malaria RDT results and respiratory rate was collected at this point for each interviewed care-seeker/child case.
The following outcome variables were derived as explained below.
The primary outcome variable was proportion of U5 children that received appropriate treatment (the right dose, frequency and duration for the right indication i.e. overall appropriate treatment) for each of uncomplicated malaria, pneumonia symptoms and non-bloody diarrhoea as assessed against the national treatment guidelines  and WHO definition of rational medicine use  as follows:
Appropriate treatment for uncomplicated malaria
A child with fever or history of fever was tested by malaria RDT, if positive received the right regimen of ACT and an afebrile child was neither tested nor prescribed ACT. Children with malaria RDT positive results should have received artemether/lumefantrine 20/120 mg DT as follows; 6 tablets in yellow pack for children aged 4–35 months (one tablet twice daily for 3 days), 12 tablets in blue pack for children aged 36–59 months (three tablets twice daily for 3 days).
Appropriate treatment for pneumonia symptoms
A child with cough and fast breathing (checked by respiratory timer to be 60 or more breaths per minute for a child 0–7 days, 50 or more breaths per minute for child 2–11 months and 40 or more breaths per minute for child 1–5 years) received right regimen of amoxicillin DT and child with cough and normal breathing was not prescribed amoxycillin DT. Children with cough and fast breathing should have received amoxicillin DT 125 mg as follows; 20 tablets in pink pack for children aged 2–11 months (two tablets twice daily for 5 days), 30 tablets in green pack for children aged 12–59 months (three tablets twice daily for 5 days).
Appropriate treatment for non-bloody diarrhoea
Child with non-bloody diarrhoea (loose stool with no visible presence of blood) received zinc 200 mg DT and ORS sachets as follows; 5 tablets for children aged 2–6 months (half tablet once a day for 10 days), 10 tablets for children aged 7–59 months (one tablet once a day for 10 days). Each of these children received two sachets of ORS and the drug seller demonstrated to care-seeker how to reconstitute. Each child was advised to drink at least half a 300 ml cup after every lose stool.
Secondary outcome variables included provision of ACT medicines, amoxicillin DT or diarrhoea treatment, as proportions of U5 febrile cases seeking treatment for fever or suspected malaria, pneumonia symptoms and diarrhoea who were sold ACT medicines, DT amoxicillin tablets or diarrhea treatment, respectively, regardless of whether or not they were examined for clinical signs (cough, fast-breathing, fever, diarrhoea) or were diagnostically tested.
The other secondary outcome variable was uptake of diagnostic tests as proportions of U5 febrile cases seeking treatment for fever or suspected malaria, pneumonia symptoms and diarrhoea that were tested with thermometer, malaria RDT or respiratory rate timer, respectively. Child cases with fever should have had their temperatures taken using a thermometer, those with fever or 24-h history of fever should have had blood sample taken off a finger-prick and tested by RDT for presence of malaria parasite antigens and children with cough should have had their respiratory rate counted with respiratory timer. Lastly, the monthly proportions of children that were prescribed any antimicrobial medicine were also derived.
The predictor variables were study (intervention) arm, pre- or post-intervention participation (time variable) of the care-seeker and child pair, interaction term between time variable and intervention. Extraneous variables including care-seeker/child characteristics and drug shop characteristics (see Table 3) were controlled for to obtain adjusted estimates of the primary outcomes.
Quantitative data from care-seeker exit interviews was collected using tablets with Open Data Kit (ODK) software, and were cleaned, checked, coded and then transferred to Stata version 13.0 (Stata Corp., College Station, TX, USA). The outcome variables were derived prior to analysis. The encounter between drug seller and the pair of care-seeker and respective child was the unit of observation and analysis. The data was analysed in five steps. First, the care-seeker/child characteristics of the exit interview sample population in the intervention and comparison arms at baseline were compared. Categorical variables were compared by calculating cluster-weighted Chi squares, while continuous variables were compared by calculating the adjusted t value using the Stata ‘CLTEST’ modules for performing cluster-adjusted Chi square and t tests . The adjustment for clustering used the intra cluster correlation estimated by large analysis of variance which applies a correction for imbalanced clusters . Similar comparison was made for the end-line care-seeker exit interviews (see Table 3).
Second, the effect of the iCCM intervention on the primary outcomes was determined by estimating the net intervention effect. The Stata ‘DIFF’ module was used to determine difference between intervention and comparison arm before the iCCM intervention subtracted from the difference between the intervention and comparison arm after (also known as difference-in-difference (DiD) analysis), with respect to appropriate treatment for uncomplicated malaria, pneumonia symptoms and non-bloody diarrhoea, respectively . The intervention effect(s) on the outcomes and their confidence intervals were estimated under the assumption of normally distributed residuals to facilitate presentation in percentage units. Given the small number (10–12) of clusters per study arm, bootstrapping was done in 50 replications to improve inference with clustered standard errors . Selection of covariates for multivariable analysis was by backward elimination stepwise regression, after removing collinear variables from the full model. A significant level to stay in the model of 0.2 was applied. Covariates of clinical importance to the primary outcomes such as care-seeker paying for diagnostic tests and time to get from home to drug shop were forced into the model and evaluated for their effect on R-squared . Thereafter, statistical significance for intervention effect was interpreted at p value less than 0.05. Similarly, DiD analysis was applied to secondary outcomes of provision of ACT, amoxicillin DT, and diarrhoea treatment, and uptake of diagnostic testing among under-five child cases presenting at the study drug shops.
Third, the primary outcomes in the intervention and comparison arms at baseline were tested for the balancing property using the “TEST” option in the Stata “DIFF” module, based on the two-sample t test . Covariates with p-value at significant level to stay in the model of 0.2 were retained in the model tested for the balancing property. This balancing t test was done to show that the distribution of the primary outcome variable between the intervention and comparison arms was the same regardless of the covariates .
Fourth, to determine drug seller adherence to the iCCM guidelines, the authors analysed the proportions of appropriate treatment for uncomplicated malaria and pneumonia symptoms among end-line care-seeker exit interviews in the intervention arm. Also proportions of child cases in whom respiratory rate was counted, and malaria RDT done were determined, respectively. Binomial proportion exact confidence intervals were calculated. These proportions were compared with corresponding proportions derived from drug shop patient registries and direct observation of drug seller–care-seeker interactions. This allowed for triangulation of proportions across evaluation components, thereby increasing robustness of findings and providing more in-depth understanding of the observed appropriate treatment for the childhood conditions.
Fifth and last, data from drug shop registries in intervention arm were summarized into monthly proportions of child visits prescribed any antimicrobial agent and analyzed using the STATA PTREND module for trend analysis for proportions . This nonparametric test calculated a Chi square statistic for trend based on regression and also analysed for departure from the trend line.
The research and ethics committees at Makerere University School of Public Health (IRB00011353), Uganda National Council of Science and Technology (HS1385) and World Health Organization approved the study. Written informed consent was obtained from drug sellers and care-seekers prior to their participation in the study.
All eligible drug shops in the intervention (61 drug shops) and comparison arm (23 drug shops) were recruited into the study. Table 2 shows the distribution of care-seeker–child pairs by study arm and survey.
At both baseline and end-line, the care-seeker/child characteristics in intervention and comparison arms were similar except for how care-seekers decided to buy medicine (Table 3). Also, there were differences between the two arms at baseline, on reasons for seeking care at that drug shop and a higher proportion of diarrhoea cases in the intervention arm. At end-line, there were differences in distance to drug shop, being friends with the drug seller and a higher proportion of cases with rapid or difficult breathing in the intervention arm. The test for the balancing property of the intervention and comparison arms, given a set of covariates (those selected for the multivariable models by backward elimination stepwise regression—see Table 4) showed a difference in appropriate treatment for uncomplicated malaria of 23.7% (p < 0.001) in favour of the comparison arm, no difference in appropriate treatment for pneumonia symptoms between the two arms and a difference of 5.6% (p 0.548) in appropriate treatment of non-bloody diarrhoea in favour of the intervention arm.
The largest intervention effect on appropriate treatment was recorded for uncomplicated malaria, 80.2% (95% CI 53.9, 106.5) followed by for pneumonia symptoms, 65.5% (95% CI 51.2, 79.8) and lastly for non-bloody diarrhoea, 31.4% (95% CI 0.8, 62.0). Controlling for extraneous variables (See Table 4) reduced the effect sizes as follows; appropriate treatment for uncomplicated malaria 34.5% (95% CI 8.6, 60.4), for pneumonia symptoms to 54.7% (95% CI 28.4, 81.0) and for non-bloody diarrhoea to − 11.2% (95% CI − 65.5, 43.1). Except for non-bloody diarrhoea, all percentage increases in appropriate treatment for the childhood conditions were statistically significant at p value < 0.05, even after controlling for extraneous variables (see Table 4). Also, a large negative change from 31.9 to 0.9% in appropriate treatment for uncomplicated malaria was observed in the comparison arm. Details of the proportions of child cases who presented with fever, pneumonia symptoms and non-bloody diarrhoea, proportions who were diagnostically tested and proportions who were given medicines are provided as supplementary materials (Additional file 1).
Effect on provision of ACT, amoxicillin DT and diarrhoea treatment and uptake of diagnostic testing
The largest intervention effect was on provision of amoxicillin DT to child cases with suspected pneumonia symptoms, 91.5% (95% CI 82.5, 100.5) (Table 5) followed by provision of ACT to child cases with suspected uncomplicated malaria, 24.8% (95% CI − 3.3, 51.1) and lastly for provision of diarrhoea treatment to child cases with non-bloody diarrhoea, 17.1% (95% CI − 22.3, 53.7%). Only the percentage increase in provision of amoxicillin DT to child cases with suspected pneumonia symptoms was statistically significant at p value < 0.05.
The reported intervention effect on uptake of diagnostic testing from smallest to largest improvement was on use of thermometer, 41.2% (95% CI 19.4, 63.0), followed by use of malaria RDTs, 52.6% (95% CI 27.3, 77.9) and use of respiratory timers, 60.1% (95% CI 47.6, 72.6) (Table 5). There was a reduction by half (-53.5%, 95% CI − 93.9, − 13.3) of child cases who were not subjected to any diagnostic test at all. These improvements or reduction were all statistically significant.
Drug seller adherence to iCCM guidelines in intervention arm drug shops in South Western Uganda
In the intervention arm, 285 care-seeker exit interviews were done at end-line, 5975 child cases were reported in drug shop patient registries during the intervention period from February 2014 to September 2015 and 49 drug seller–care-seeker encounters were directly observed at end-line.
Data from these three different sources was used to evaluate drug seller adherence to the national iCCM guidelines as presented in Fig. 1.
Generally, drug seller adherence to national iCCM guidelines was moderate for appropriate treatment of uncomplicated malaria, 57.4% (95% CI 50.3, 64.5), and pneumonia symptoms, 65.5% (95% CI 59.2, 71.8), and appropriate malaria RDT use, 73.9% (95% CI 68.6, 79.3), among the care-seeker exit interviews and highest for three outcomes among child cases reported in the drug shop patient registries, namely, appropriate treatment for uncomplicated malaria, 94.9% (95% CI 94.2, 95.6), appropriate treatment for pneumonia symptoms 93.6% (95% CI 92.8, 94.5) and appropriate malaria RDT use, 88.7% (95% CI 87.9, 89.5).
However, the proportion of child cases in whom respiratory rate was counted was lowest in drug shop patient registries, 57.5% (95% CI 56.3, 58.8), moderate among care-seeker exit interviews, 60.1% (95% CI 54.3, 65.9) and highest for direct observation 77.6% (95% CI 65.6, 89.7).
From the drug shop patient registry, 97% (3628/3738) child cases with fever had been tested for malaria using an RDT and the RDT positivity rate recorded was 47% (1957/4190) (95% CI 45–48). However, the proportion of child cases in whom treatment followed malaria RDT results was 73.9% among care-seeker exit interviews, highest (88.7%) in child cases reported in drug shop patient registries, and lowest (63.3%) in direct observation.
More details on how the drug sellers adhered with the national iCCM algorithm to detect and classify disease and treat the child cases in the intervention arm drug shops are available in the additional information (Additional file 2).
Effect of the adapted iCCM on antimicrobial medicine use in intervention drug shops from February 2014 to September 2015
Figure 2 is a plot of monthly proportions of child cases attending the intervention arm drug shops prescribed ACTs, amoxicillin DT or at least an antimicrobial medicine (either ACT or amoxicillin DT or both) from February 2014 to September 2015. The trend analysis for proportions by regression found; i) a linear trend with negative slope (− 0.009) with p value < 0.001 for “either ACT, amoxycillin DT or both” line, and a p value 0.769 for a nonlinear trend, ii) a linear trend with negative slope (− 0.006) with p value < 0.001 and a p value of 0.587 for nonlinear trend for “Amoxycillin DT” line, and iii) a linear trend negative slope (− 0.008) with p value < 0.001 and a p value of 0.674 for nonlinear trend for “ACT” line. In summary, statistically significant linear trend lines with slight negative gradient were observed for the three scenarios.
The authors demonstrated that implementing an iCCM intervention at retail drug shops increased appropriate treatment for uncomplicated malaria, pneumonia symptoms and non-bloody diarrhoea, which implied that a higher proportion of U5 children received the right medicine in the right dose, frequency and duration for the right indication in the intervention arm as compared to the comparison arm in South Western Uganda.
Also, an increase in diagnostic testing for malaria, pneumonia symptoms and high temperature among U5 children with fever or history of fever or cough was observed. An increase in proportion of children prescribed recommended child medicines (ACT medicines, amoxicillin DT and diarrhoea treatment) was found, thus indicating increased availability and provision of these medicines in the intervention arm as compared to comparison arm. Overall prescription rates of antimicrobials decreased over the 20-month study period.
Findings of this study in a lower malaria endemic setting are similar to those by Awor et al. in Eastern Uganda, notwithstanding the observed higher malaria RDT positivity of 78% as compared to 47% reported in the current study . Meanwhile an RDT positivity rate of 47% was higher than what was expected, given values of 5 and 9.3% malaria prevalence reported in studies conducted in the same area [34, 36]. This difference in malaria prevalence could be explained by the difference in the populations of U5 children tested, the current study tested sick children presenting at drug shops, while the 2014 malaria indicator survey  and the study by Kamya et al.  tested healthy children in their homes. However, it could also be related to possibilities of false malaria RDT positivity due to persistent HRP2 antigenaemia after malaria treatment , or falsification of the malaria RDT results in the patient registries by drug sellers so as to increase the sales of ACT medicines. A further study will attempt to validate the drug shop patient registry malaria RDT results.
Within the intervention arm drug shops, appropriate treatment for uncomplicated malaria was lower from care-seeker exit interviews than from drug shop patient registries and direct observation. The lower proportion observed in care-seeker interviews was likely due to increased non-adherence to malaria RDT-negative test results thereby prescribing of anti-malarial medicines to these cases. Similar findings have been reported elsewhere by Johansson et al.  and Burchett et al. , for failure to diagnose other causes of illness or to increase sale of ACT medicines.
Appropriate treatment for pneumonia symptoms was higher from drug shop patient registries than from care-seeker exit interviews and direct observation, which scored similar proportions of child cases. The higher proportion observed could be explained by over-diagnosis of pneumonia symptoms by drug sellers so as to increase antibiotic sales, or due to difficulties in applying respiratory rate cut-offs correctly, as shown by Mukanga et al. for community health workers .
The current study demonstrated a decline in the proportion of children prescribed either an anti-malarial or antibiotic medicine over study period of 20 months in the intervention arm. These findings contrast results of other studies that reported that testing with malaria RDTs reduced use of anti-malarial medicines but led to increased antibiotic use [53,54,55,56]. The authors believe that the integrated nature of the iCCM intervention in the current study, rather than vertical malaria intervention alone in the latter studies, contributed to this. This argument is supported by a synthesis of ten studies from multiple epidemiological and healthcare contexts by Burchett et al. That synthesis found that presence of options for alternative management of malaria negative cases promoted adherence to malaria RDT-negative results . Another study reported that presence of components that modify supply side factors such as price subsidies, consistent supply at the pharmaceutical distributor points, friendly relations with regulators and demand generation activities explained the mechanism of effect .
However, there are notable limitations to this study. First, the intervention effect on the outcome variables was assessed in a quasi-experimental design which is amenable to multiple threats to internal validity including selection and history , and hence may be less compelling to infer causality. Systematic differences were observed at baseline between the intervention and comparison arms for reasons for which care-seekers sought care from the drug shops. This could be explained by drug shops in the comparison arm being closer to care-seekers’ homes and more likely to be open all the time. This was further seen in care-seekers bringing in the under-five children sooner (within 24 h) in the comparison as compared with the intervention arm. Additionally, the comparison arm (Bushenyi district) occupied a smaller land area (942.3 square kilometres) than the intervention arm (Mbarara district, 1778.4 km2) . The authors argue that these differences do not explain the observed intervention effects. Backward elimination with stepwise regression identified only one drug shop characteristic—care-seeker being friends with drug seller—for multivariable analysis for appropriate treatment for uncomplicated malaria. And, the intervention effect remained even after controlling for the drug shop and care-seeker/child covariates. This implies the difference in appropriate treatment for uncomplicated malaria between the intervention and comparison arms is influenced by drug shop characteristics (testing child with malaria RDT, counting respiratory rate of child, care-seeker being friends with drug seller, if care-seeker paid for diagnostic tests, and time to get from home to drug shop), in addition to child presenting with fever or history of fever. The adjusted model for appropriate treatment for pneumonia symptoms showed an intervention effect while that for non-bloody diarhoea did not show any effect. Second, although the DiD analytic technique adjusts for time-invariant unobserved heterogeneity, it requires testing of the compliment of parallel paths of the outcome for the intervention and comparison arms . Given the availability of only two periods, pre- and post-intervention, in this study, this assumption could not be tested. However, the balancing t-test at baseline showed that there was no difference between appropriate treatment for pneumonia symptoms and diarrhoea, respectively, even after controlling for covariates that stayed in the multivariable model at significant level of 0.2 in the backward elimination stepwise regression procedure. There was a difference in appropriate treatment for uncomplicated malaria in favour of the comparison arm at baseline. Borrowing from the “parallel growth” extension of the balancing test, as advanced by Mora and Reggio , the authors argue that in the absence of the intervention, this outcome variable is orthogonal to the intervention effect. Since the parallel growth assumption is flexible and allows for differing trends before and after the intervention, the difference in appropriate treatment for uncomplicated malaria observed in the DID analysis is likely due to the intervention.
Third, in the current study, the comparison arm performed better at baseline as compared to post-intervention on uptake of diagnostic testing for malaria and fever, and provision of diarrhoea treatment, (see Table 5). The observed decline from pre- to post-intervention in the comparison arm, for uptake of malaria RDTs, could be explained by substantial reduction in availability of diagnostics in the private sector pharmaceutical supply chain of Uganda caused by the closure of the affordable medicine facility for malaria (AMFm), and its subsequent transition to the Global Fund private sector co-payment mechanism (in 2014 and 2015) . Similar studies have also argued that regulatory and policy barriers to implementation of malaria RDTs by drug sellers can cause problems to consistent availability of diagnostics in such retail health markets . Nevertheless, appropriate malaria treatment increased also in the comparison arm. The AMFm transition, regulatory and policy barriers would not explain the current study’s findings on pneumonia and diarrhoea treatment.
Fourth, an unequal number of clusters were observed at baseline with the comparison arm registering some non-performing clusters. In addition, the clusters had unequal observations within them. However, analytical methods applied in adjusting for clustering used the intra cluster correlation estimated by large analysis of variance which applies a correction for imbalanced clusters and computes robust standard errors . Fifth, the current study analysed the primary outcome at single time points before and after implementation of the adapted iCCM model in intervention arm drug shops. The observed iCCM intervention effect on assessing for and treating pneumonia symptoms could have been an over-estimation given the absence of respiratory timers and amoxycillin DT pre-intervention and in the comparison arm. However, the current study examined the uptake and utilization of these health technologies once available. This is in line with another study  that reported that presence of diagnostics at the drug shops was necessary but not sufficient to translate into their uptake. Other researchers [63, 64] have argued that behavioural change efforts and provider incentives are necessary. Hence, these findings infer that the iCCM intervention improved the appropriateness of drug seller treatment practices in the intervention arm as compared to the comparison arm. The strength of this study lies in the different data collection methods used and confidence in the findings is enhanced by Awor et al.  who observed similar improvements in quality of paediatric fever care at drug shops using multiple methods of assessing outcomes. A complete assessment of change in the trend of drug seller treatment practices would require measurements at multiple time points pre- and post-iCCM intervention and interrupted time series analysis [65,66,67].
Given the numerous challenges in scaling up government led iCCM in some contexts—including the lack of incentives to motivate CHWs, and national ownership of the intervention , drug shops could be a complementary source of healthcare going forward. However, several questions on how to construct supervision models for private sector interventions remain, as study–employed supervisors were used for the current and previous studies, rather than a scalable supervision model integrated in district health services . Addressing this question is important to determine if these interventions are reasonable at scale or not. More so, Visser et al. reported that training and more frequent supportive supervision were associated with higher adherence .
This study illustrates that implementing the iCCM intervention at retail drug shops improved appropriate treatment for uncomplicated malaria and reduced presumptive treatment of the febrile child, and promoted drug seller adherence to iCCM guidelines, also in lower malaria transmission settings of South Western Uganda. Furthermore, this study demonstrates overall declining prescription rates of anti-malarials as well as antibiotics as malaria RDTs were introduced and adhered to, contrary to other studies that introduced only malaria RDTs and led to increase in prescription of antibacterial medicines. The authors therefore argue that malaria control programmes in public as well as private sector should promote integrated case management of the sick child rather than single disease focus only. Further research should validate drug seller compliance with test results and assess the effects of the intervention in larger scale implementation, the sustainability and effect of different supervision modalities.
artemisinin-based combination therapy
access to life
affordable medicine facility for malaria
community health workers
- DiD analysis:
histidine rich protein 2 antigen
information, education and communication
integrated community case management
rapid diagnostic test
Uganda Ministry of Health
National Drug Authority
National Malaria Control Programme
oral rehydration salts
sustainable development goals
United Nations Children’s Fund
village health teams
World Health Organization
UNICEF. Levels and trends in child mortality. Report 2015. Estimates developed by the UN Interagency Group for child mortality estimation. New York, USA2015c.
Global Health Observatory (GHO) Data. under-five mortality http://www.who.int/gho/child_health/mortality/mortality_under_five_text/en/. Accessed 21 Apr 2017.
Child Survival. Under-five mortality—current status and progress. https://data.unicef.org/topic/child-survival/under-five-mortality/.
Uganda Ministry of Health. Annual health sector performance report—financial year 2015–2016. Kampala: Uganda Ministry of Health; 2016.
Young M, Wolfheim C, Marsh DR, Hammamy D. World Health Organization/United Nations Children’s Fund joint statement on integrated community case management: an equity-focused strategy to improve access to essential treatment services for children. Am J Trop Med Hyg. 2012;87:6–10.
Clinton Health Access Initiative & Coalition for Health Promotion and Social Development Uganda. Assessment of the management of diarrhoea, pneumonia and malaria in children under-5 in Uganda. Kampala, 2014.
Uganda Bureau of Statistics (UBOS) and ICF International Inc. Uganda Demographic and Health Survey 2011. Kampala, Calverton: Uganda Bureau of Statistics (UBOS) and ICF International Inc; 2012.
Uganda Ministry of Health. A promise renewed; reproductive maternal, newborn and child health sharpened plan for Uganda. Kampala: Uganda Ministry of Health; 2013.
WHO, UNICEF. An equity-focused strategy to improve access to essential treatment services for children. Geneva: WHO; 2012.
Uganda Ministry of Health. Integrated community case management of childhood malaria, pneumonia and diarrhoea; implementation guidelines. Uganda: Uganda Ministry of Health; 2010.
Hamer DH, Brooks ET, Semrau K, Pilingana P, MacLeod WB, Siazeele K, et al. Quality and safety of integrated community case management of malaria using rapid diagnostic tests and pneumonia by community health workers. Pathog Glob Health. 2012;106:32–9.
Kalyango JN, Alfven T, Peterson S, Mugenyi K, Karamagi C, Rutebemberwa E. Integrated community case management of malaria and pneumonia increases prompt and appropriate treatment for pneumonia symptoms in children under five years in Eastern Uganda. Malar J. 2013;12:340.
Kalyango JN, Lindstrand A, Rutebemberwa E, Ssali S, Kadobera D, Karamagi C, et al. Increased use of community medicine distributors and rational use of drugs in children less than five years of age in Uganda caused by integrated community case management of fever. Am J Trop Med Hyg. 2012;87:36–45.
Kalyango JN, Rutebemberwa E, Karamagi C, Mworozi E, Ssali S, Alfven T, et al. High adherence to antimalarials and antibiotics under integrated community case management of illness in children less than five years in eastern Uganda. PLoS ONE. 2013;8:e60481.
Miller NP, Amouzou A, Tafesse M, Hazel E, Legesse H, Degefie T, et al. Integrated community case management of childhood illness in Ethiopia: implementation strength and quality of care. Am J Trop Med Hyg. 2014;91:424–34.
Seidenberg PD, Hamer DH, Iyer H, Pilingana P, Siazeele K, Hamainza B, et al. Impact of integrated community case management on health-seeking behavior in rural Zambia. Am J Trop Med Hyg. 2012;87:105–10.
Uganda Ministry of Health. Integrated community case management in Uganda; Final National iCCM Review Report—December 2013 Kampala, 2013.
Uganda Ministry of Health. The Uganda malaria reduction strategic plan 2014–2020. Kampala: Uganda Ministry of Health; 2014.
Bennett S, George A, Rodriguez D, Shearer J, Diallo B, Konate M, et al. Policy challenges facing integrated community case management in sub-Saharan Africa. Trop Med Int Health. 2014;19:872–82.
Standing H, Chowdhury AM. Producing effective knowledge agents in a pluralistic environment: what future for community health workers? Soc Sci Med. 2008;66:2096–107.
Uganda Bureau of Statistics (UBOS) and ICF Macro. Uganda malaria indicator survey 2009. Kampala, Calverton: Uganda Bureau of Statistics (UBOS) and ICF Macro; 2010.
Licensing requirements and guidelines for 2015. http://www.masterlmis.com/uganda2/asset_uplds/files/LICENCING%20GUIDELINES%202015.pdf. Accessed 31 Oct 2016.
Girosi F, Olmsted SS, Keeler E, Hay Burgess DC, Lim YW, Aledort JE, et al. Developing and interpreting models to improve diagnostics in developing countries. Nature. 2006;444(Suppl 1):3–8.
Awor P, Wamani H, Bwire G, Jagoe G, Peterson S. Private sector drug shops in integrated community case management of malaria, pneumonia, and diarrhea in children in Uganda. Am J Trop Med Hyg. 2012;87:92–6.
Sudhinaraset M, Ingram M, Lofthouse HK, Montagu D. What is the role of informal healthcare providers in developing countries? A systematic review. PLoS ONE. 2013;8:e54978.
Gautham M, Shyamprasad KM, Singh R, Zachariah A, Singh R, Bloom G. Informal rural healthcare providers in North and South India. Health Policy Plan. 2014;29(Suppl 1):i20–9.
Konde-Lule J, Gitta SN, Lindfors A, Okuonzi S, Onama VO, Forsberg BC. Private and public health care in rural areas of Uganda. BMC Int Health Hum Rights. 2010;10:29.
Ansah EK, Gyapong M, Narh-Bana S, Bart-Plange C, Whitty CJM. Factors influencing choice of care-seeking for acute fever comparing private chemical shops with health centres and hospitals in Ghana: a study using case–control methodology. Malar J. 2016;15:290.
Awor P, Wamani H, Tylleskar T, Jagoe G, Peterson S. Increased access to care and appropriateness of treatment at private sector drug shops with integrated management of malaria, pneumonia and diarrhoea: a quasi-experimental study in Uganda. PLoS ONE. 2014;9:e115440.
Awor P, Wamani H, Tylleskar T, Peterson S. Drug seller adherence to clinical protocols with integrated management of malaria, pneumonia and diarrhoea at drug shops in Uganda. Malar J. 2015;14:277.
Oxfam Briefing Paper. Salt, sugar and malaria pills; how the affordable medicines facility—malaria (AMFm) endangers public health. 163 Oxfam Briefing Paper, 2012.
von Schoen-Angerer T. Learning all the lessons: expanding access to malaria diagnosis and treatment. J Public Health Policy. 2013;34:315–9.
Kabaghe AN, Visser BJ, Spijker R, Phiri KS, Grobusch MP, van Vugt M. Health workers’ compliance to rapid diagnostic tests (RDTs) to guide malaria treatment: a systematic review and meta-analysis. Malar J. 2016;15:163.
Uganda Bureau of Statistics (UBOS) and ICF International. Uganda malaria indicator survey 2014–15: key indicators. Kampala, Rockville: Uganda Bureau of Statistics (UBOS) and ICF International; 2015.
Uganda Bureau of Statistics (UBOS). The national population and housing census 2014—main report. Kampala: Uganda Bureau of Statistics (UBOS); 2014. p. 2016.
Kamya MR, Arinaitwe E, Wanzira H, Katureebe A, Barusya C, Kigozi SP, et al. Malaria transmission, infection, and disease at three sites with varied transmission intensity in Uganda: implications for malaria control. Am J Trop Med Hyg. 2015;92:903–12.
Uganda Bureau of Statistics (UBOS). Statistical abstract. Kampala: Uganda Bureau of Statistics (UBOS); 2014. p. 2014.
Sourcing and Procurement. Private sector co-payment mechanism-core components and key actors. http://www.theglobalfund.org/en/privatesectorcopayment/. Accessed 26 May 2016.
The Global Fund to fight AIDS TB and Malaria. Global fund information note: use of a private sector co-payment mechanism to improve access to ACTs in the new funding model (October 2013). Geneva: The Global Fund; 2013.
Bennett S, Woods T, Liyanage WM, Smith DL. A simplified general method for cluster-sample surveys of health in developing countries. World Health Stat Q. 1991;44:98–106.
Management Sciences for Health. Investigating medicine use. In: Embrey M, Ryan M, editors. MDS-3: managing access to medicines and health technologies. Arlington: Management Sciences for Health; 2012.
Promoting rational use of medicines. Core components. http://apps.who.int/medicinedocs/pdf/h3011e/h3011e.pdf. Accessed 30 May 2016.
Donner A, Klar N. Design and analysis of cluster randomization trials in health research. New York: Arnold; 2000.
Gleason JR. sg65: computing intraclass correlations and large ANOVAs. Stata Tech Bull. 1997;35:25–31.
Villa JM. DIFF: simplifying the estimation of difference in differences treatment effects. Stata J. 2016;16:52–71.
Cameron AC, Gelbach JB, Miller DL. Bootstrap-based improvements for inference with clustered errors. Rev Econ Stat. 2008;90:414–27.
Sun GW, Shook TL, Kay GL. Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis. J Clin Epidemiol. 1996;49:907–16.
Sasieni PD. snp12: stratified test for trend across ordered groups. Stata Tech Bull. 1996;33:24–7.
Kyabayinze DJ, Tibenderana JK, Odong GW, Rwakimari JB, Counihan H. Operational accuracy and comparative persistent antigenicity of HRP2 rapid diagnostic tests for Plasmodium falciparum malaria in a hyperendemic region of Uganda. Malar J. 2008;7:221.
Johansson EW, Kitutu FE, Mayora C, Awor P, Peterson SS, Wamani H, et al. It could be viral but you don’t know, you have not diagnosed it: health worker challenges in managing non-malaria paediatric fevers in the low transmission area of Mbarara District, Uganda. Malar J. 2016;15:197.
Burchett HE, Leurent B, Baiden F, Baltzell K, Bjorkman A, Bruxvoort K, et al. Improving prescribing practices with rapid diagnostic tests (RDTs): synthesis of 10 studies to explore reasons for variation in malaria RDT uptake and adherence. BMJ Open. 2017;7:e012973.
Mukanga D, Babirye R, Peterson S, Pariyo GW, Ojiambo G, Tibenderana JK, et al. Can lay community health workers be trained to use diagnostics to distinguish and treat malaria and pneumonia in children? Lessons from rural Uganda. Trop Med Int Health. 2011;16:1234–42.
Johansson EW, Gething PW, Hildenwall H, Mappin B, Petzold M, Peterson SS, et al. Effect of diagnostic testing on medicines used by febrile children less than five years in 12 malaria-endemic African countries: a mixed-methods study. Malar J. 2015;14:194.
Msellem MI, Martensson A, Rotllant G, Bhattarai A, Stromberg J, Kahigwa E, et al. Influence of rapid malaria diagnostic tests on treatment and health outcome in fever patients, Zanzibar: a crossover validation study. PLoS Med. 2009;6:e1000070.
D’Acremont V, Kahama-Maro J, Swai N, Mtasiwa D, Genton B, Lengeler C. Reduction of anti-malarial consumption after rapid diagnostic tests implementation in Dar es Salaam: a before-after and cluster randomized controlled study. Malar J. 2011;10:107.
Hopkins H, Bruxvoort KJ, Cairns ME, Chandler CI, Leurent B, Ansah EK, et al. Impact of introduction of rapid diagnostic tests for malaria on antibiotic prescribing: analysis of observational and randomised studies in public and private healthcare settings. BMJ. 2017;356:j1054.
Kitutu FE, Mayora C, Johansson EW, Peterson S, Wamani H, Bigdeli M, et al. Health system effects of implementing integrated community case management (iCCM) intervention in private retail drug shops in South Western Uganda: a qualitative study. BMJ Glob Health. 2017;2:e000334.
Shadish WR, Cook TD, Campbell DT. Quasi-experimental designs that use both control groups and pretests. In: Experimental and quasi-experimental designs for generalised causal inference. New York: Houghton Mifflin Company; 2002. p. 135–69.
Uganda Bureau of Statistics (UBOS). Statistical abstract. Kampala: Uganda Bureau of Statistics (UBOS); 2016. p. 2016.
Mora R, Reggio I. Treatment effect identification using alternative parallel assumptions—Working paper 12–33. Economic series 2012. p. 48.
ACTwatch Group and PACE. ACTwatch study reference document: the Republic of Uganda outlet survey 2015. Washington, DC: PSI; 2015.
Visser T, Bruxvoort K, Maloney K, Leslie T, Barat LM, Allan R, et al. Introducing malaria rapid diagnostic tests in private medicine retail outlets: a systematic literature review. PLoS ONE. 2017;12:e0173093.
Cohen J, Fink G, Maloney K, Berg K, Jordan M, Svoronos T, et al. Introducing rapid diagnostic tests for malaria to drug shops in Uganda: a cluster-randomized controlled trial. Bull World Health Organ. 2015;93:142–51.
Cohen J, Cox A, Dickens W, Maloney K, Lam F, Fink G. Determinants of malaria diagnostic uptake in the retail sector: qualitative analysis from focus groups in Uganda. Malar J. 2015;14:89.
Penfold RB, Zhang F. Use of interrupted time series analysis in evaluating health care quality improvements. Acad Pediatr. 2013;13:S38–44.
Fretheim A, Tomic O. Statistical process control and interrupted time series: a golden opportunity for impact evaluation in quality improvement. BMJ Qual Saf. 2015;24:748–52.
Interrupted time series (ITS) analyses. EPOC resources for review authors. http://epoc.cochrane.org/sites/epoc.cochrane.org/files/public/uploads/21%20Interrupted%20time%20series%20analyses%202013%2008%2012.pdf. Accessed 30 Jan 2017.
Miller E, Sikes HD. Addressing barriers to the development and adoption of rapid diagnostic tests in global health. Nanobiomedicine. 2015;2:6.
Acute respiratory infection (ARI)—timer manufactured for UNICEF. http://www.globalsources.com/si/AS/SINO-HEALTHCARE/6008844361082/pdtl/ARI-TIMER/1089412970.htm. Accessed 10 May 2016.
FEK, HW, and SP designed and conceptualized the study. FEK, CM & HW conducted the experiments. FEK entered, cleaned, prepared, analyzed, and interpreted data. FEK, KES JNK, and SSP contributed to data analysis. FEK, KES and SP contributed to interpretation of findings. FEK wrote the first draft of the paper. FEK, JNK, CM, KES, SP, and HW reviewed, revised and contributed to writing to the paper. FEK, JNK, CM, KES, SP, and HW read and met the ICMJE criteria for authorship. All authors read and approved the final manuscript.
The authors would like to thank drug shop owners, attendants and care-seekers who participated in the study, district health team of Mbarara and Bushenyi districts for supporting implementation of the study, the surveys’ enumerators and Brocher Foundation of Hermance, Switzerland for offering the main author a researcher residence to analyze and write the first draft of this paper.
The authors declare that they have no competing interests.
Availability of data and materials
The datasets during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Consent for publication
Ethics approval and consent to participate
The research and ethics committees at Makerere University School of Public Health (IRB00011353), Uganda National Council of Science and Technology (HS1385) and World Health Organization reviewed and approved the study. Informed consent was obtained from participants prior to care-seeker interviews.
The study was supported by WHO Alliance for Health Policy and Systems Research, Einhorn Family Foundation, Pehr Lagermans Family, Sweden, Swedish Science Council and Uppsala University. These organizations had no role in the design, implementation interpretation nor reporting of the findings. FEK, CM, HW and JNK receive salary support from Makerere University while KES, FEK and SP received salary support from Uppsala University.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.