Menu
Menu
Menu
Menu
Background: User fees have generally fallen out of favor across Africa, and they have been associated with reductions in access to healthcare. We examined the effects of the introduction and removal of user fees on outpatient attendances and new diagnoses of HIV, malaria, and tuberculosis in Neno District, Malawi where user fees were re-instated at three of 13 health centres in 2013 and subsequently removed at one of these in 2015. Methods: We conducted two analyses. Firstly, an unadjusted comparison of outpatient visits and new diagnoses over three periods between July 2012 and October 2015: during the period with no user fees, at the re-introduction of user fees at four centres, and after the removal of user fees at one centre. Secondly, we estimated a linear model of the effect of user fees on the outcome of interest that controlled for unobserved health centre effects, monthly effects, and a linear time trend. Results: The introduction of user fees was associated with a change in total attendances of -68 % [95 % CI: -89 %, -12 %], similar reductions were observed for new malaria and HIV diagnoses. The removal of user fees was associated with an increase in total attendances of 352 % [213 %, 554 %] with similar increases for malaria diagnoses. The results were not sensitive to control group or model specification. Conclusions: User fees for outpatient healthcare services present a barrier to patients accessing healthcare and reduce detection of serious infectious diseases.
In Neno District, Malawi the health care infrastructure consists of 13 different healthcare facilities: eight are operated by the Ministry of Health, one private facility operated by a local electric company, and four are administered by CHAM. In recent years, there have been several shifts in the implementation of user fees at CHAM facilities across the district. Partners In Health (PIH), an NGO known in Malawi by its Chichewa name Abwenzi Pa Za Umoyo, works with MOH to strengthen health systems and helped broker SLAs with the CHAM facilities (Matope, Matandani, and Nsambe). In July 2013, these three CHAM facilities terminated their Service Level Agreements and introduced user fees simultaneously for general outpatient visits. It was replaced with an SLA covering free maternal, neonatal, and HIV services, meaning user fees were instituted for all other outpatient visits. User fees comprise consultation fees for seeing a clinician, fees for laboratory tests, and fees for medications. This re-institution of user fees has been seen across Malawi, as previous funding for SLAs was withdrawn by health donors, prompting re-initiation of user fees at many CHAM facilities across the country. In 2016, the government of Malawi is pursuing an agreement with CHAM that will allow districts to independently proceed once again with these SLAs; however, funding limitations remain a significant barrier in most districts [19]. Because the assistance of PIH is available in Neno District, in July 2015, user fees were eliminated at one of these three centres (Matope) through a new SLA. One centre (Neno Parish) charged user fees for the duration of the study period. The remaining nine facilities did not charge user fees. Figure 1 shows the periods when each centre did or did not charge user fees. Figure 2 shows the location of each of the centres. Implementation of user fees across health centres in Neno District, Malawi Map of Neno District, Malawi and location of healthcare centres The aim of the study was to identify the effects of introducing or removing user fees on attendances and diagnoses of communicable diseases at health centres in Neno District, Malawi. In particular, the outcomes analysed were: total outpatient attendances, total number of new malaria diagnoses in patients aged under 5, total number of new malaria diagnoses in patients aged over 5, and total number of new confirmed HIV cases in patients aged 15 to 49. HIV care, once diagnosed, was free to patients throughout the entire time period. New HIV diagnoses were examined because of the opportunity for HIV case-finding during outpatient visits for acutely ill clients. Data were available on the number of new TB diagnoses, however this was not included in the regression analyses as the outcome was rare and could not be analysed. The use of routinely collected longitudinal data enabled us to take into account effects that may have confounded our analyses including secular trends in health care utilisation, seasonal effects, and unobserved health centre effects. The analysis presented here can be considered a generalisation of a difference in differences (DiD) regression model with multiple units in the treatment (user fee charging) and control (no user fees) groups and multiple time periods. We specified a linear model. The dependent variable was the natural logarithm of the number of health care attendances or new diagnoses of the nominated diseases. We included in the model different intercepts for each health care facility, monthly dummy variables, and a treatment group (user fees or no user fees) dummy variable. We also included a linear time trend interacted with the treatment group dummy variable: this allows for “correlated random trends”, which relaxes the parallel trends assumption normally required for DiD [20]. The user fee and non-user fee groups may have different trends over time in health services utilisation and these trends may be correlated with the introduction or removal of user fees. For example, user fees may have been introduced in response to declining attendances. We considered that the introduction and removal of user fees would have differential treatment effects. We therefore estimated the effects of the introduction and removal of user fees separately. The standard errors were clustered at the health centre level. The primary analysis may under or overestimate the effect of introducing user fees, since many users may travel to a different centre that does not charge a user fee. These individuals may not choose to attend a health centre had there been user fees implemented at all centres. As a sensitivity analysis we considered a different control group: the subset of non-user fee charging facilities separated from a user fee charging facility by another non-user fee charging facility (Magaleta, Chifunga, Luwani, and Nkula). We considered using a prior, formal rule to categorise centres in this regard, but chose simple discrimination based on visual inspection of the location of centres (see Fig. 2) since we discerned no ambiguous cases. We also considered different model specifications: a fixed-effects model that does not allow for “correlated random trends”, and a fixed effects Poisson regression. Finally, we excluded Nkula from the analyses since it was not operated by MOH or CHAM. Routinely collected data from the “HMIS-15” report were extracted from Malawi’s District Health Information Software 2 (DHIS2) for this analysis. The “HMIS-15” report summarizes core health service utilization at each facility including maternal health, antenatal care, HIV diagnoses, and outpatient department visits. No formal data quality assessments on the HMIS-15 report were performed during the study period to assess the accuracy and validity of these data; however spot checks on major outliers were conducted. The period for the analyses was July 2012 to October 2015. Ethics committee approval was obtained for analysis and publication of routinely collected data to evaluate services within Neno District from both the Malawi National Health Sciences Research Committee (Lilongwe, Malawi) and Partners Institutional Review Board (Boston, MA). Aggregated datasets were utilized, thus individual informed consent was not obtained.
