Background: Maternal health remains a concern in sub-Saharan Africa, where maternal mortality averages 680 per 100,000 live births and almost 50% of the approximately 350,000 annual maternal deaths occur. Improving access to skilled birth assistance is paramount to reducing this average, and user fee reductions could help. Objective. The aim of this research was to analyse the effect of user fee removal in rural areas of Zambia on the use of health facilities for childbirth. The analysis incorporates supply-side factors, including quantitative measures of service quality in the assessment. Method: The analysis uses quarterly longitudinal data covering 2003 (q1)-2008 (q4) and controls for unobserved heterogeneity, spatial dependence and quantitative supply-side factors within an Interrupted Time Series design. Results: User fee removal was found to initially increase aggregate facility-based deliveries. Drug availability, the presence of traditional birth attendants, social factors and cultural factors also influenced facility-based deliveries at the national level. Conclusion: Although user fees matter, to a degree, service quality is a relatively more important contributor to the promotion of facility-based deliveries. Thus, in the short-term, strengthening and improving community-based interventions could lead to further increases in facility-based deliveries.
This study uses routine quarterly data, collected within the Health Management and Information System (HMIS) administered by the Ministry of Health (MOH) in Zambia. It contains information on the supply and use of a wide range of health services at all public health facilities nationwide, aggregated to the district. Complete data was available for 46 out of 53 rural districts that abolished user fees in April 2006. However, data from the following districts was discarded: Chibombo, Kapiri Mposhi, Serenje, Chienge, Chavuma, Lukulu, Siavonga and Milenge, because there were multiple missing months of information. From the district level data, we compiled regional data for the 9 provinces from 2003 (q1) to 2008 (q4) (that is T=24 and N=9). Based on the available data and the previously discussed literature, we selected and included six quarterly time series: the proportion of FBDs (ID); average health centre client contacts per day (CC), which measures the staff workload (defined as the total number of patient visits divided by the total number of staff per day); traditional birth attendants per 1000 of the population (TBAs); the proportion of drugs available, based on the percentage of stock-outs of drugs on the essential drug list (DA); the average number of antenatal visits per quarter (ANC); and the population in the province (POP). CC and DA capture the quality of services, while cultural preferences and alternative options are captured by TBAs. The analysis is founded upon an Interrupted Time Series (ITS) design, complemented by a segmented regression analysis, which is adequate, when only retrospective longitudinal data, before and after an intervention, is available. We disaggregate the data to obtain regional level estimates from Seemingly Unrelated Regressions (SUR), addressing spatial dependence within an error component framework. Due to the strong persistence observed in FBDs, we specify a dynamic panel model, including one lag of the dependent variable, to assess the impact of the abolition of user fees on FBDs. where IDit is the vector of FBDs in the N=9 provinces; xit is a vector of explanatory variables and includes time dummies (t1, t2, t3) representing the first, second and third quarters of each year, to account for the cyclicality observed in FBDs, ai is the regional fixed effect; Timet is a vector of continuous values indicating time from the start to the end of the study period; Interventionit is a vector of indicators coded 0 for the pre-invention period and 1 for the post-intervention period; Postslopeit is a vector of indicators coded 0 up to the last point before the intervention and coded sequentially from 1 thereafter; and εit is a vector of disturbances. The analysis accounts for potential cross-sectional dependence. Pooled Ordinary Least Squares (POLS) serves as the baseline; however, Fixed Effects (FE) and Feasible Generalised Least Squares (FGLS) are also considered. FE control for time-invariant omitted variables that differ by province, such as the level of development, health infrastructure and health staff, allowing for intercept heterogeneity; an F-test (Pr>F=0.000) supports the existence of regional fixed effects. The FE method differences out the individual variability across regions. Thus, the FE estimator is a pooled OLS estimator on the demeaned equation (1) and yields unbiased estimates under the assumption of strict exogeniety. In the dynamic specification, the FE estimator with standard errors, robust to moderate levels of cross-sectional dependence in the error term, are implemented.49 However, the procedure does not correct for Nickel bias,50 an effect that can approach 20%, when T=30.51 We implemented a bias-correction procedure suitable for small T and moderate N (10<N N.55 Therefore, FGLS can be interpreted as pooled SUR, in which estimates represent the average values of the regional coefficients.
N/A