Menu
Menu
Menu
Menu
Background: Many sub-Saharan countries, including Ghana, have introduced policies to provide free medical care to pregnant women. The impact of these policies, particularly on access to health services among the poor, has not been evaluated using rigorous methods, and so the empirical basis for defending these policies is weak. In Ghana, a recent report also cast doubt on the current mechanism of delivering free care – the National Health Insurance Scheme. Longitudinal surveillance data from two randomized controlled trials conducted in the Brong Ahafo Region provided a unique opportunity to assess the impact of Ghana’s policies. Methods: We used time-series methods to assess the impact of Ghana’s 2005 policy on free delivery care and its 2008 policy on free national health insurance for pregnant women. We estimated their impacts on facility delivery and insurance coverage, and on socioeconomic differentials in these outcomes after controlling for temporal trends and seasonality. Results: Facility delivery has been increasing significantly over time. The 2005 and 2008 policies were associated with significant jumps in coverage of 2.3% (p = 0.015) and 7.5% (p<0.001), respectively after the policies were introduced. Health insurance coverage also jumped significantly (17.5%, p<0.001) after the 2008 policy. The increases in facility delivery and insurance were greatest among the poorest, leading to a decline in socioeconomic inequality in both outcomes. Conclusion: Providing free care, particularly through free health insurance, has been effective in increasing facility delivery overall in the Brong Ahafo Region, and especially among the poor. This finding should be considered when evaluating the impact of the National Health Insurance Scheme and in supporting the continuation and expansion of free delivery care. © 2012 Dzakpasu et al.
Data were obtained from a health and demographic surveillance system supporting the ObaapaVitA [17] and Newhints [18] cluster randomized controlled trials (RCTs) carried out in seven contiguous predominantly rural districts in the Brong Ahafo Region (Kintampo North and South, Nkoranza North and South, Wenchi, Techiman, Tain). Information was gathered through four-weekly home visits by resident fieldworkers. Approximately 120,000 women of reproductive age live in this area, with about 18,000 pregnancies and 15,000 live births a year. We based our analysis on deliveries from January 2004 to December 2009: one year before the 2005 policy to one year after the 2008 policy was introduced. The 2003 policy did not apply to Brong Ahafo. We examined trends in the percentage of deliveries taking place in a health facility, the percentage of delivered women enrolled with the NHIS, and socioeconomic differentials in both these outcomes. Facility delivery included hospital, health centre and maternity home births. Data on NHIS enrolment were collected from March 2008, so analysis of insurance coverage was only possible from then until December 2009. Data were based on women’s self-reports collected at the first fieldworker visit after the birth. For 76% of women, this occurred within 30 days of the delivery. Socioeconomic differentials were examined using wealth quintiles (estimated from household asset data) and concentration indices which summarize how an outcome varies across the entire socioeconomic distribution. We restricted our analysis to records for which asset data had been collected within a year of the delivery, assuming assets would not change substantially within this timeframe. For women with multiple deliveries over the six-year period, data were collected for each delivery. Principal component analysis was used to assign an asset score to each woman at the time of her delivery [19]. Reliability of these asset scores was confirmed by comparing them against individual asset ownership and educational levels of women. Women were then ranked from poorest to richest according to their scores, separately within each year of delivery from 2004 to 2009, and assigned to wealth quintiles, each representing a fifth (20%) of the women delivering within that year. Concentration indices were calculated for facility delivery and insurance coverage by plotting the cumulative percent of each against the cumulative percent of women ranked by their asset scores, and calculating the area between this curve and the line of equality. This area by definition ranges from −1 to+1, with positive values corresponding to the curve being below the line of equality and the outcome concentrated towards the richest, and negative values where the curve is above the line and the outcome concentrated towards the poorest [20]. 95% confidence intervals for concentration indices were calculated using the bootstrap method [21]. Monthly rates of facility delivery and insurance coverage were displayed graphically by wealth quintile using simple three-month moving averages [22]. We further studied temporal trends in these outcomes and in the monthly concentration indices of these outcomes using segmented linear regression models. We fitted separate temporal trends for three segments defined relative to the introduction of each policy: January 2004–March 2005, April 2005–June 2008, and July 2008–December 2009. We used a model of the form: Where: The immediate impact of each policy was calculated as the absolute difference between the predicted values just before and after the policy. The longer-term impact was assessed by comparing temporal trends before and after each policy, calculated as the absolute difference between the regression slopes for each period. The models estimated the impact of each policy after controlling for temporal trends, and seasonal variation in facility delivery. We re-ran all models taking into account that some women contributed more than one delivery; however this adjusting for clustering at the woman-level did not significantly impact model parameters and was therefore not done in the final models. We also did not adjust for other measured determinants of facility delivery, such as rural residence or maternal education, as there was no evidence of change in their distribution during the study period. We carried out all analysis using Stata version 11 [23].
N/A
