Charting health system reconstruction in post-war Liberia: A comparison of rural vs. remote healthcare utilization

Background: Despite a growing global emphasis on universal healthcare, access to basic primary care for remote populations in post-conflict countries remains a challenge. To better understand health sector recovery in post-conflict Liberia, this paper seeks to evaluate changes in utilization of health services among rural populations across a 5-year time span. Methods: We assessed trends in healthcare utilization among the national rural population using the Liberian Demographic and Health Survey (DHS) from 2007 and 2013. We compared these results to results obtained from a two-staged cluster survey in 2012 in the district of Konobo, Liberia, to assess for differential health utilization in an isolated, remote region. Our primary outcomes of interest were maternal and child health service care seeking and utilization. Results: Most child and maternal health indicators improved in the DHS rural sub-sample from 2007 to 2013. However, this progress was not reflected in the remote Konobo population. A lower proportion of women received 4+ antenatal care visits (AOR 0.28, P < 0.001) or any postnatal care (AOR 0.25, P <0.001) in Konobo as compared to the 2013 DHS. Similarly, a lower proportion of children received professional care for common childhood illnesses, including acute respiratory infection (9 % vs. 52 %, P < 0.001) or diarrhea (11 % vs. 46 %, P < 0.001). Conclusions: Our data suggest that, despite the demonstrable success of post-war rehabilitation in rural regions, particularly remote populations in Liberia remain at disproportionate risk for limited access to basic health services. As a renewed effort is placed on health systems reconstruction in the wake of the Ebola-epidemic, a specific focus on solutions to reach isolated populations will be necessary in order to ensure extension of coverage to remote regions such as Konobo. We used two data sources for this analysis. First, we use data from a maternal and child health survey conducted by Last Mile Health for program planning purposes in 2012 in Konobo. Details on the design and implementation of the survey are provided below. Permission to utilize the data for research purposes was granted by Last Mile Health. We compared data from Konobo with data drawn from the rural sub-sample of the Liberian DHS from 2007 and 2013. The DHS is a nation-wide, cross-sectional study of socio-demographic and health indicators, which seeks to provide representative information for assessing health trends within the population. The variables, data collection, quality assurance, and analysis have been described previously, and the full DHS datasets are made publically available [23, 24]. Konobo District (which as a health district contains the Konobo and Glio-Twarbo administrative districts) has a total population of 33,000 and a population density of less than 50 persons per square mile [25]. The 2007 DHS estimated that the region containing Konobo had the lowest asset index distribution in Liberia, with approximately half of residents living in the bottom quintile of wealth nationally, as measured by a household index of consumer goods and dwelling characteristics [23]. The development, sampling methodology, and data collection of the Konobo survey have been previously described elsewhere [26]. Briefly, we assessed demographics and access to maternal healthcare for reproductively active women in Konobo. We also assessed episodes of illness and care seeking among respondents' children under-5 years old. For the Konobo population only, we summarized barriers to health care access, including distance and cost of transport to clinic. Participants were selected using two-stage representative cluster sampling, based on population data from the 2008 National Population and Housing Census [25]. We excluded Ziah Town, the district capital, because it met the Liberian definition of an urban area (i.e. population ≥ 2,000) [25]. An additional 25 villages were excluded because they had less than 20 households (n = 19), could only be reached on foot (n = 4) or could only be reached by canoe (n = 2). Together, residents from excluded villages comprised 15 % of Konobo’s rural population. Fever data was excluded from the final analysis due to very high reported prevalence, raising concern over question interpretation and result validity. Both DHS surveys had broader sampling criteria than the Konobo Health Survey and initially approached all women aged 15–49 for inclusion. The Konobo survey limited data collection to the woman in each household > 17 years of age who had most recently given birth. In all three surveys, questions regarding maternal health were only asked regarding the most recent pregnancy among women who had given birth within the last 5 years. Questions about child health were only asked regarding each of the respondent’s own children < 5 years currently living in the household. To ensure comparability between the DHS and Konobo sampling frames, the following exclusions were applied prior to data analysis 1) we excluded women who had not given birth within the last 5 years and therefore were not eligible for the maternal health questions in all three surveys and 2) we excluded women under 17 years old from the DHS dataset and over 49 from the Konobo dataset, even if they had given birth in the last five years. Thus the final population for analysis across all three surveys includes reproductively active women who have given birth within the last 5 years and who were between the ages of 17 and 49 at the time data was collected. We first conducted descriptive analyses using standard statistical techniques: means and confidence intervals for normally distributed continuous variables, medians and inter-quartile ranges for other continuous variables, and proportions with confidence intervals for categorical variables. We incorporated sampling structure and weights and produced design-corrected standard errors using Taylor series linearization. To make comparisons between survey years, we constructed a dataset that pooled data from the three surveys. We retained the complex sample design by keeping strata unique between surveys. Weights were rescaled so that relative weights for each observation were retained within each survey and each survey’s rural population contributed equally to the analysis. Our primary maternal health outcomes of interest were: 1) one or more antenatal care visits with a skilled provider (1+ ANC), 2) four or more ANC visits, at least one from a skilled provider (4+ ANC), 3) delivery in a health facility, and 4) post-natal care (PNC) from a skilled provider within 24 h of delivery. A skilled provider was defined as a doctor, nurse/midwife or physician’s assistant in the DHS and as a doctor, nurse or physician’s assistant in the Konobo survey. The combined term nurse/midwife was not used in Konobo as this was found to lack clarity in pilot testing. Traditional midwives were not considered “skilled providers,” as we have limited data on how these practitioners are trained in many regions of the country. Our primary child health outcomes were 1) prevalence of acute respiratory illness (ARI), defined as cough, difficulty breathing, and chest involvement within the past 2 weeks; 2) prevalence of diarrhea within the past 2 weeks, 3) receipt of care during an ARI or diarrheal episode from a health facility, and 4) receipt of care for either condition from any provider (including informal and traditional providers such as pharmacists and “tablet men,” individuals who sell common pharmaceuticals of unclear provenance in marketplaces). We fitted multivariable logistic regression models, using sampling weights and design-corrected standard errors, for each outcome of interest. We adjusted for potential confounders that have been hypothesized to influence care utilization, but which are outside the causal pathway for public health sector improvement. The maternal health model included variables for maternal age, marital status, age at first birth, education and birth order (defined as the first, second or third or greater birth). For the child health model, we also included child age and gender in addition to the maternal health variables listed above. For each outcome of interest, we then calculated predicted adjusted probabilities by survey, using post-regression marginal effects with other model covariates held at their mean values. Data analysis was performed with Stata Version 13.0 (Statacorp, College Station, Texas).