Study aims and design

The goal of this study was to evaluate the effectiveness of a local HSS initiative in improving population health in the rural district of Ifanadiana, Madagascar. Specific aims were to: (1) estimate changes in child mortality rates (U5, infant, neonatal) in the intervention and non-intervention areas between 2014 and 2023, as HSS activities were rolled out across the district; (2) evaluate the impact of the HSS intervention on healthcare utilization and care-seeking behaviors over time; and (3) measure impacts on economic and geographic inequalities in access to care over time. Our main hypothesis was that the HSS initiative would lead to reductions in mortality rates, increases in service coverage and utilization, and reductions in inequalities in access to care. For this, we used a quasi-experimental design that includes adjacent intervention and comparison (non-intervention) areas within the same district. The outcomes were pre-specified at baseline of the intervention [35], and the initial households included in the district-representative baseline survey were subsequently enrolled in an open prospective cohort study [9]. At baseline, there were no statistically significant differences in health system utilization or service coverage [35]. Over time, similar to a stepped-wedge trial but without randomization, the programs expanded and improved across a gradually widening intervention area until the entire district was covered by the HSS intervention.

The most fundamental outcome of interest is child mortality. However, the prospective cohort was established at the household level, which did not allow for individual child follow-up or clean statistical testing. Instead, five-year birth histories were analyzed from repeated cross-sections in the cohort using standard methods for demographic and health surveys (DHS). This enabled comparison of trends at the national and international level but resulted in many mortality observations that included periods both exposed and unexposed to the intervention. We present trends in child mortality in the intervention areas, the rest of the district (comparison areas), and the rest of Madagascar. To understand the potential role of health system interventions in explaining these trends, we analyze changes in health system performance over the same time period in the intervention and comparison areas. This included programs focused on the provision of health services (clinical programs and HSS pillars) and those focused on increasing access to those services (user-fee removal). We first analyze impacts on health system utilization from high resolution (but not population-representative) health facility data (i.e., consultations from the HMIS). We assess program impacts on service coverage outcomes using population-representative household surveys in the cohort. This is followed by an analysis of impacts on equity in access, with a particular focus on specific health program effects on geographic barriers to care. Details of the methods are described below. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).

Study site and HSS intervention

Ifanadiana is a rural district of approximately 200,000 people located in the region of Vatovavy, in Southeastern Madagascar. Per MoPH norms, Ifanadiana district has one reference hospital, one main primary care health center (PHC2) for each of its 15 communes (subdivisions of a district with approximately 15,000 people – two of which were created in 2016 and 2018), six additional basic health centers (PHC1) for its larger communes (larger communes have both a PHC2 and a PHC1), and one community health site (CHS) with two community health workers (CHWs) for each of its 195 fokontany (subdivision of a commune with approximately 1,000 population). The health system strengthening intervention is implemented by the NGO Pivot across all three levels of care in the district (community, health center, and hospital) and is structured around the integration of clinical programs with strengthened health system pillars and social protection. The clinical programs include child health with a focus on malnutrition and integrated management of childhood illness (IMCI), maternal and reproductive health services, and infectious disease programs targeting tuberculosis, malaria and emerging diseases. Strengthening of the health system pillars focuses on improving “readiness”, including infrastructure, trained staff, equipment, procurement systems, and an ambulance network. User fees are covered by the NGO and additional social support is provided to vulnerable patients to offset indirect costs of care. A summary of the intervention is available in Table 1, with further details in S1 Table.

The core activities were rolled out in the district in three waves. The first three years (2014–2016) covered four communes with approximately one third of the population of Ifanadiana (referred to as “initial HSS catchment”). In 2017–2020, HSS activities expanded to three additional communes, and then expanded again in 2021, such that by 2021 all PHC2s were supported (PHC1s have not yet been supported). Some activities, such as medical staff recruitments and strengthened information systems spanned the whole district (S1 Table). The HSS intervention, initially modelled after an earlier experience in HSS in Rwanda [31], has been progressively adapted over time in collaboration with the government to respond to coverage gaps and intervention deficiencies identified through an iterative learning process that uses health system and population-level data (S1 Table). The initial HSS intervention area and subsequent waves of expansion across the district were guided by operational considerations by the NGO’s implementation staff in consultation with the MoPH. In particular, implementation teams chose catchment populations where the health facility was accessible by 4 × 4 vehicles to facilitate the first years of operations. As a result, mortality rates were lower at baseline in the intervention area than in the rest of the district (Fig 1), and some socio-economic indicators were more favorable, while others were less so [35].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Under-five, infant and neonatal mortality improved in intervention areas while outcomes in comparison areas (as well as the country as a whole) generally worsened over the study period, 2013–2023.

(A) Each dot represents the average mortality rate in the 5 years preceding the survey, together with 95% confidence intervals. (B) Each dot represents the difference in mortality between each time point and the 2013–2014 baseline for each geographic area. Estimates for Ifanadiana district and HSS catchments were estimated from the IHOPE cohort. Estimates for Madagascar were obtained from reports of the national surveys conducted by INSTAT in Madagascar during the same period (ENSOMD 2013, MICS 2018, DHS 2021). These surveys do not report 95% confidence intervals of mortality estimates, but they have larger sample sizes than the IHOPE cohort. Abbreviations: DHS Demographic and Health Survey; ENSOMD Enquête de Suivi des Objectifs du Developpement du Millenaire à Madagascar; HSS Health Systems Strengthening; IHOPE Ifanadiana Health Outcomes and Prosperity longitudinal Evaluation; INSTAT Institut National des Statistiques; MICS Multiple Indicators Cluster Survey.

https://doi.org/10.1371/journal.pmed.1004549.g001

In addition to the HSS intervention, during this period, the population of Ifanadiana benefited from other notable programs that covered both the intervention and comparison areas. The PAUSENS project (2014–2017), funded by the World Bank, provided a basic package of services free of charge in all PHC2s through a voucher system for every woman attending the health center for antenatal, delivery, or postnatal care (first six weeks), and children under the age of five with any illness (see [43] for more details). The project also included training, support for child vaccination in remote areas, and the provision of some equipment to health centers. The Mikolo project (2014–2018), funded by USAID, provided support to a network of 150 CHWs in the fokontany located more than 5 km from a health center in eight communes in Ifanadiana, four of which were in the HSS catchment and four in the comparison area. The project provided annual trainings and periodic supervision, along with equipment, supplies and an initial stock of medicines to each CHW. The ACCESS project (2019–2023) funded by USAID, continued the Mikolo community health efforts, and included support to primary care centers, with a focus on IMCI and maternal and reproductive health services, as well as additional support for HMIS and the procurement pillars of the health system. The main difference between the initial HSS catchment and the rest of the district was the duration of implementation of the HSS intervention at endline by the MoPH-Pivot partnership (9 years versus 2–6 years). Specific differences between catchments are described in Table 1.

Population survey data: The IHOPE cohort

A longitudinal cohort study known as the Ifanadiana Health Outcomes and Prosperity longitudinal Evaluation (IHOPE cohort), was initiated in 2014 to collect demographic, health and socio-economic data from a representative sample of 1,600 households in Ifanadiana district over time [9]. Questionnaires were primarily adapted from the DHS [44]. The Madagascar National Institute of Statistics (INSTAT), which implements all major national health surveys in the country, was responsible for data collection, survey coordination, training, and oversight. A two-stage sampling process stratified the district by the initial HSS catchment and the rest of the district. Eighty clusters – half from each stratum – were selected at random from enumeration areas mapped during the 2009 census, and households were then mapped within each cluster. Twenty households were selected at random from each cluster.

The first wave of data collection was conducted between April and May of 2014. Individual face-to-face interviews were conducted with women aged 15–49 years and men aged 15–59 years (usual residents or visitors) in 1,522 of the sampled households (95.1% response rate). The original 1,600 households were revisited between August and September 2016 and again between April and June of 2018, 2021, and 2023; any missing or refused households were replaced with others from the same cluster using a predefined random replacement list. Response rates during follow-up survey waves varied between 94.5% and 95.9%. All residents, including children, had weight and height measured (or length, in the case of infants). Data collected in the questionnaires included, among other variables, household composition (size, gender, ages); indicators of socio-economic status (education, employment, household durable assets); recent illness (in the last 30 days) and care-seeking for all household members; preventive behavior (bed net ownership, access to water and sanitation); women’s reproductive history and care-seeking behavior for reproductive health; children’s health, development and care-seeking for illness; adult, maternal, and child mortality.

French and Malagasy questionnaires used in the cohort, along with data collection protocols, were standardized and validated for Madagascar through prior national surveys conducted by INSTAT. The study received ethical approval from the Madagascar National Ethics Committee and Harvard Medical School IRB. All adults (≥15 years) provided verbal consent for the in-person interview and anthropometric measurements. For children ≤ 5 years of age, consent was provided by a parent or guardian. Further details on data collection and survey design are available in [9]. Data were entered using CSPro.

Health system utilization data

We obtained aggregated monthly utilization data for the period of January 2013 to December 2023 from the MoPH for 15 PHC2 and 6 PHC1s in Ifanadiana district. These data capture the number of new individuals attending each health center for outpatient consultations or maternal care. Because the HSS intervention began in early 2014 (shortly after the baseline survey) and affected utilization rates that year, we included health system utilization data from 2013 to establish a true pre-intervention baseline. These data were available from the health centers’ monthly reports to the district (RMA), which were compiled by MoPH staff from the health centers’ registers every month. Starting in May 2015, the MoPH revised its method for estimating and reporting outpatient utilization rates. Thus, all subsequent outpatient utilization data were gathered directly from the registers to ensure consistent estimates across the full 2013–2023 period.

To assess changes in geographic access to primary care, we analyzed both district-wide health center catchment data and data disaggregated to the smallest administrative unit—fokontany—for the initial HSS catchment area during the 2018–2020 period. Fokontany comprise one or several villages, and are located at varying distances from the nearest health center (0–20 km). To obtain the number of consultations at the PHC-level by fokontany, we first digitized individual-level data from the health center registries for all outpatient visits in the district. For each patient (new visits; follow-up excluded), information included the age, fokontany of residence and some diagnosis data. We also obtained consultation data from the monthly reports of community health sites supported by the MoPH-Pivot partnership for the 80 fokontany in the initial HSS catchment, available at the fokontany level. As a component of the HSS intervention, health system data quality was ensured with support from the NGO’s monitoring and evaluation team and through joint MoPH-Pivot field supervisions [36].

Population estimates for each health center catchment area per year (for aggregated data) or each fokontany (for disaggregated data) were derived from the two national censuses, conducted in 1993 and 2018 by INSTAT, applying a constant annual growth rate. Consistent with MoPH estimates, populations of children under five, expected number of pregnant women, and expected number of deliveries were set at 18%, 4.5%, and 4% of the total population, respectively. Finally, information on key dates of the implementation of health system strengthening components in each supported commune (such as the start of the user-fee exemptions, support to the community health program, or support to the maternal health program) were obtained from internal records within the NGO.

Analysis of coverage and mortality rates from cohort data

Under-five, infant and neonatal mortality were estimated at the population level using the synthetic life-table method for DHS surveys [45]. These were defined as the probability of death before age 60 months, 12 months and 1 month per 1,000 children born alive, respectively. For each survey wave, we used information from children born in the five years prior to the survey. For each survey year, mortality rates were separately estimated for the entire district, and for the HSS catchments (initial HSS catchment versus rest of the district). From these estimates, absolute changes from the 2014 baseline were then calculated for each mortality indicator and population group.

Coverage indicators were estimated using standard definitions for DHS surveys [45]. Vaccination coverage was defined as the proportion of children aged 12–23 months who received all recommended vaccines: 3 doses of polio and diphtheria–tetanus–pertussis (DTP), one dose of Bacillus Calmette–Guérin (BCG) and measles. Care-seeking for illness was estimated as the proportion of children under five who had fever, acute respiratory infection, or diarrhea in the two weeks prior to the survey and sought medical care at a public health provider (hospital, health center or community health worker). For care seeking of individuals of all ages, a four week recall period for illness was used. To evaluate the impact of the HSS intervention on maternal health service coverage, indicators were estimated for the most recent pregnancy resulting in a live birth during the two years preceding the survey: prenatal care (first and fourth visit), delivery at a public health center and postnatal care. In addition, to track a summary indicator of maternal, neonatal and child (MNCH) coverage, we estimated a co-coverage index (five or more interventions) for women and children under five [46].

The impact of the HSS intervention on coverage indicators was modeled from individual-level data using multivariable logistic regressions (binomial model with logit link), controlling for relevant factors and accounting for survey weights, using the following formula:

(1)

where Y represents the predicted population-level coverage; year is the number of years since the baseline survey; catchment reflects whether the household of the individual was part of the initial HSS catchment, to account for baseline differences in these two areas; facility_HSS reflects whether the household was in a commune where the HSS intervention was implemented at the facility level (including, but not limited to, reimbursement of user fees and readiness support), which changed over time as the HSS catchment expanded; program_HSS reflects whether the household was in a commune receiving support to a specific program affecting a subset of indicators, namely the community health program, with potential impact on care seeking indicators, or the maternal health program, with potential impact on maternal care indicators – these also changed over time as the HSS catchment expanded but all were implemented after initiation of the facility-level HSS; voucher reflects whether the household was in a commune where the voucher program funded by the World Bank for maternal and child care was implemented (see “Study site and HSS intervention”); covid-19 reflects whether the covid-19 pandemic was occurring at the time of the survey (2021 follow-up) to account for a potential change in care seeking during this period. All these variables (except for year) were included as binary variables (0 when program or COVID-19 period not present in a particular commune and year, or 1 otherwise). The coefficient β₁ is the yearly change in the study area independent of the intervention, β₂ is the baseline difference between the initial HSS catchment and the rest of the district, β₃, β₄, β₅, and β₆ are the level of change associated with HSS support at the facility level, with HSS support to specific programs (community health or maternal health depending on the indicator), with the World Bank voucher program, and with the COVID-19 period, respectively. This design is equivalent to an interrupted time series analysis with control groups. Results are reported as adjusted odds ratios (OR).

All analyses were performed using R statistical software, version 4.2.1 [47]. All population-level analyses were done using survey commands available in the R-package survey and appropriate sampling weights [48], with the exception of population-level mortality rates and associated 95% confidence intervals, which were calculated with SAS 9.3 [49]. Sampling weights were calculated for household, women’s and men’s surveys, to adjust for unequal probability of selection due to stratification and non-response.

Analyses of healthcare utilization from health center and community health data

To study the impact of the HSS intervention on consultation rates, aggregated health center data for the 2013–2023 period were used to model changes in monthly per capita utilization rates for maternal health services (antenatal care, first and fourth visits; deliveries, and postnatal care) and for outpatient care for any illness (all patients and children under five). For this, a negative binomial mixed-effects regression model – analogous to the analyses of survey data described above – were conducted for each indicator, with a random intercept for each health center. We applied interrupted time-series analysis with control groups, incorporating explanatory variables for facility-level and program-specific HSS support (i.e., maternal care and community health programs, depending on the indicator). The model also controlled for secular trends in the study area, baseline differences between the initial HSS catchment and the rest of the district, for the effect of the World Bank voucher program (2014–2017), and the COVID-19 pandemic (2020–2021). Health center catchment population was log-transformed and included as an offset. A 1-month lag in consultation data for each health center was included to control for temporal autocorrelation. Supplementary analyses were conducted to examine the slope of changes (progressive change over time) associated with facility-level HSS support and the World Bank voucher program, after adjusting for associated level changes (S2 and S3 Tables). Results are reported as adjusted relative changes. Statistical models were performed using the R package lme4 [50].

To understand geographic trends in primary care utilization at different levels of care, individual-level data from health center registers were aggregated into consultations per capita-year for each of the 80 fokontany in the initial HSS catchment during 2018–2020, and compared with community health data from the same fokontany and period. The Open Source Routing Machine (OSRM) engine was used to accurately estimate the shortest path between all buildings in each fokontany and the nearest health center. For this, the entire district was previously mapped on OpenStreetMap, resulting in over 23,000 km of footpaths, more than 100,000 buildings and 5,000 residential areas mapped (see [42] for details).

Analysis of geographic and economic inequalities in healthcare coverage from cohort data

First, changes in the geographic distribution of coverage over time in Ifanadiana were assessed for selected indicators. Average values for the 80 geographic clusters in the IHOPE cohort were estimated, each of which included 20 households and approximately 100 individuals. Using the spatial location of each cluster, a raster surface of the entire district was generated to improve visualization of results. This was achieved through inverse distance weighted interpolation of the empirical Bayes estimates of the average coverage value of each cluster, using the R packages spdep and gstat.

Second, trends in economic and geographic inequalities within the HSS catchment were assessed for each coverage indicator over time. To estimate economic inequalities, a household wealth index was constructed using principal components analysis of household assets following standard DHS methods [8]. To estimate geographic inequalities, OSRM was used to estimate the shortest path between the villages in each cluster and the nearest health center, as explained above. For each indicator, we estimated wealth- and geographic-specific composite indicators of inequality, namely the relative concentration index (RCI) and slope index of inequality (SII) [51,52]. Lower values of the RCI or SII over time indicate reductions in relative or absolute inequalities, respectively.

Third, trends in self-reported barriers to seeking care were estimated in the initial HSS catchment and the rest of the district. Individuals of all ages in the IHOPE cohort surveys who reported being ill but did not seek care at a health facility were asked to provide their primary and secondary reasons for not seeking care (this information was not available for 2014). Reasons were classified into the following categories: no barrier (“not severe enough” or no reason reported); knowledge barrier (“did not think they could help me”, “did not know that a treatment existed”); health system barrier (“lack of confidence in health staff”, “health staff often absent”, “health staff is rude”, “health personnel asks for additional money”); financial barrier (“impossible to stop work”, “too expensive”); geographic barrier (“too far away or hard to reach”); or COVID-19 (“facility closed due to COVID-19”, only in 2021). Percentages of each barrier were estimated out of all primary and secondary reasons.

Changes in child mortality rates at the population level

The results show clear signals of reductions in neonatal, infant, and U5 mortality rates following implementation of the HSS intervention. The number of children under-five used in mortality analyses was constant across survey waves: 4,063 children for 2014, 4,037 children for 2016, 3,788 children for 2018, 4,380 for 2021 and 3,983 for 2023. Across the district as a whole, the average U5 mortality rate did not consistently decrease during 2014–2023, while infant and neonatal mortality increased, patterns which are consistent with national trends observed in Madagascar from surveys over the same period (Fig 1). However, when trends were split by the initial HSS catchment and the rest of the district, a striking difference emerged whereby measures of child mortality in the initial HSS catchment consistently decreased for all child mortality indicators by 20–30 deaths (per 1,000 live births) as compared to the 2014 baseline, with opposite trends for the rest of the district (Fig 1). While rates at baseline in the initial HSS catchment were roughly twice the national average, all child mortality rates had converged with national averages by the end of the study period (Fig 1).

The initial HSS catchment had lower mortality rates at baseline than the rest of the district. The initial HSS catchment was better-off in socio-economic characteristics at baseline (e.g., access to electricity, and sanitation), and worse-off in others (e.g., more single parent households and a higher proportion of the population in the lowest wealth quintile). However, these characteristics remained largely stable during the study period (S2 Table) and were not accompanied by significant differences in healthcare access at baseline between the two areas (S3 Table). The comparison area (“rest of the district”) consists of 11 communes that received HSS support primarily starting in 2021. Following this expansion, under-five and infant mortality declined in the rest of the district after 2021, but neonatal mortality increased.

Impact of HSS on health center utilization and population-level coverage of MNCH indicators

Statistical models of both health center utilization and population-level coverage yielded consistent results, revealing significant positive effects on indicators of child, adult, and maternal care after accounting for multiple confounders, including the expanding HSS catchment.

Time series analyses of monthly health center utilization (S1 Fig) revealed large and statistically significant effects of facility-level HSS (e.g., improvements in readiness, clinical programs, removal of user fees) on outpatient consultations (Table 2), both for children under five (rate ratio 1.48, 95% CI [1.37, 1.61]; p < 0.05) and individuals of all ages (rate ratio 2.14, 95% CI [1.98, 2.32]; p < 0.05). Specific support to the community health program (“Program-specific HSS support”) was associated with a small but significant reduction in health center utilization. For maternal care consultations, both facility-level HSS and program-specific support to the maternal care program were associated with statistically significant increases in utilization (Table 2), although the cumulative impact associated with both of these interventions (approximately 25% increase) was smaller than in analyses of population-level coverage (Table 3). No statistically different baseline differences in utilization were observed between the intervention and comparison areas.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Impact of health systems strengthening on per capita health center utilization rates for child, adult and maternal care from health system patient data (multivariable negative binomial mixed-effects model,¹ one per indicator). Results are expressed as rate ratio (95% Confidence intervals).

https://doi.org/10.1371/journal.pmed.1004549.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Impact of health systems strengthening on population-level coverage indicators of child, adult and maternal care from household survey data (multivariable binomial regression model accounting for survey weights,¹ one model for each indicator). Results are expressed as Odds Ratio (95% Confidence intervals).

https://doi.org/10.1371/journal.pmed.1004549.t003

Analyses of population-level service coverage from the IHOPE cohort (S2 Fig) showed that facility-level HSS had the largest effects on care-seeking for individuals of all ages and for children under five (Table 3), being associated with a near doubling of the odds of care seeking at a public provider (Odds Ratio [OR] 1.84, 95% CI [1.3, 2.59] and 1.89, 95% CI [1.19, 3], respectively; p < 0.05). For maternal care, facility-level HSS had positive but not statistically significant effects (e.g., OR for birth delivered at public health center of 1.75, 95% CI [0.98, 3.12]). Instead, program-specific HSS support to the maternal care program was associated with more than double the odds of having at least 1 prenatal care visit with a skilled provider, a birth delivered at a health facility and a postnatal care visit within 48 h – as well as a doubling of the co-coverage index, a summary index of maternal, child and neonatal care coverage. Program-specific HSS support to the community health program was associated with a significant increase in vaccination coverage for children 12–23 months (OR 1.96, 95% CI [1.14, 3.36]; p < 0.05) and an increase, although not statistically significant, in individual or child care seeking with any public provider (hospital, health center, or CHW) (Table 3).

Evolution of inequalities in population-level coverage indicators

While the results above indicate increases in average health service coverage, spatial patterns over time for coverage indicators provided further insights into the impact of HSS in Ifanadiana district, and the gaps that remained in 2023 (Fig 2). Improvements were more spatially homogeneous for prenatal care (4+ visits) and for care-seeking of children under 5 than for the other indicators assessed, highlighting the role of CHWs in supporting referral and care delivery, respectively. Significant geographic gaps persisted for births delivered at health facilities and for care seeking of individuals of all ages, which can only occur at health centers or a hospital. In particular, coverage remained low in remote areas along the southern and eastern limits of the district, and in the catchment areas of basic health centers (PHC1), which had not yet been supported by HSS efforts (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Spatial changes in selected population-level coverage indicators in Ifanadiana district, 2014–2023.

Each map shows the weighted average for each of the 80 survey clusters (circles) in Ifanadiana district. Inverse distance weighted interpolation was used between clusters to improve visualization. Maps were created with R, with boundary data from OCHA (https://data.humdata.org/dataset/cod-ab-mdg) under a CC BY 4.0 License. *Data collection for care-seeking behavior for individuals of all ages began in 2016. Abbreviations: OCHA United Nations Office for the Coordination of Humanitarian Affairs (OCHA).

https://doi.org/10.1371/journal.pmed.1004549.g002

In the initial HSS catchment, both economic and geographic inequalities in healthcare coverage (both relative and absolute) were reduced for all ill patients seeking care, as well as U5 specifically (Fig 3). For nearly all other coverage indicators, relative inequalities declined in the initial HSS catchment, while absolute inequalities had mixed results (Fig 3). These results were consistent with the reasons for not seeking care at a health facility reported by individuals of all ages, where geographic and financial barriers were cited progressively less often over time in the initial HSS catchment, but changed little in the rest of the district (S3 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Relative geographic and economic inequalities were reduced for nearly all coverage indicators in the initial HSS catchment, 2014–2023, while changes in absolute inequalities were mixed.

Figure shows the average annual change in relative inequalities (estimated as the relative concentration index) and absolute inequalities (estimated as the slope index of inequality) with respect to geography and wealth. For geographic inequalities, individuals were ranked according to their distance to the nearest health center. For wealth inequalities, individuals were ranked according to their household wealth score. The positive relative results compared to negative absolute results means that wealthier households and those that lived closer to health facilities experienced larger absolute increases in access to care, but poorer households that lived further from health facilities had greater proportional increases. *Data collection for care-seeking behavior for all ages began in 2016.

https://doi.org/10.1371/journal.pmed.1004549.g003

The role of community health in reducing geographic barriers to care for children under five

Geographic analyses of care-seeking behavior and health system utilization for children under five in the initial HSS catchment revealed an improvement in geographic access to care following the strengthening of the community health program (Fig 4). Survey results showed that, while care-seeking at a public provider (which includes both health facilities and community workers) initially (in 2014) declined with increasing household distance from health facilities, this effect disappeared by 2023 following HSS support, driven by larger improvements in care-seeking among more remote populations (Fig 4). Characterizing primary care consultations by level of care provided further insights into the relative contribution of CHWs and health facilities towards overall healthcare access for different geographic groups. Fig 4 shows that health center utilization declined exponentially as distance from health facilities increased. This distance decay was largely offset by increased CHW utilization along the same distance gradient. As a result, overall health system utilization for children under five remained above 2 consultations per capita-year across nearly all distances from health facilities (except for those further than 15 km, where it ranged from 1.5 to 2 consultations per capita-year). Health center consultations represented nearly 80% of primary care consultations for children under five living within 2.5 km of a health center, but less than 15% of primary care consultations for children living further than 15 km from a health center (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Household surveys (top row) and health system patient data (bottom row) indicate that reductions of geographic barriers through community health contributed to substantial increases in service utilization for children under 5 in remote areas.

Top panel shows changes in care seeking for child illness at a public provider (health facility or CHW) from 2014 to 2023. Each dot represents the weighted average for each of the 80 clusters in the household survey, and the solid line represents a LOESS non-linear smooth of the trends and 95% confidence intervals. Bottom panels represent the split between health center and CHW utilization, estimated from health system data for the 80 fokontany in the initial HSS catchment during 2018–2020. Each dot represents the average annual per capita consultation rate for each fokontany (left), which is aggregated by distance categories (right). Abbreviations: CHW community health worker; HSS health systems strengthening; LOESS locally estimated scatterplot smoothing; U5 under five.

https://doi.org/10.1371/journal.pmed.1004549.g004