Is the National Health Insurance Scheme helping pregnant women in accessing health services? Analysis of the 2014 Ghana demographic and Health survey

We used data from the women recode file of the 2014 GDHS. This is the sixth version since the survey started in Ghana in 1988. It forms part of the Measure DHS Program which seeks to monitor core health indicators in LMICs. Two stage sample design was carried out. The initial stage involved the selection of 427 clusters constituting the enumeration areas (EAs). The enumeration areas emerged from urban (216) and rural (211) locations across all the ten regions at the time. The second phase involved the selection of 11,835 households from the EAs and this resulted in a total sample of 9396 women aged 15–49. The survey had 97.3% response rate [31]. For the purpose of our study, 4271 women with complete data were included.

The dependent variable was whether the NHIS is helping pregnant women for health services or not. The question was posed to women aged 15–49 during the 2014 GDHS. It was accompanied by two responses: “Yes” and “No”. This variable was chosen on the premise that one of the priorities of the pro-poor NHIS in Ghana is to ease the financial burden in accessing maternity services [32]. As a result, investigating the perception of women on whether this mandate is being achieved is essential for future health financing policy directions.

Eight independent variables were included in this study. Of these, the main independent variable was health insurance subscription. The other included variables were Age (15–19,20-24,25-29,30–34-35-39,40-44,45–49), education (No education, primary, secondary, tertiary), residence (rural,urban), antenatal care (ANC) visits (Below 4 Visits, At least 4 Visits), current pregnancy status (pregnant, not pregnant) and region (Western, Central, Greater Accra, Volta, Eastern, Ashanti, Brong Ahafo, Northern, Upper East, Upper West) and wealth quintile (poorest, poorer, middle, richer, richest). Wealth, in the DHS, is a composite measure computed by combining data on a household’s ownership of carefully identified assets including television, bicycle, materials used for house construction, sanitation facilities and type of water access. Principal component analysis was used to transform these variables into wealth index by placing individual households on a continuous measure of relative wealth. The DHS segregates households into five wealth quintiles; poorest, poorer, middle, richer and richest. These variables have been reported as essential for investigating NHIS [33, 34].

Stata version 13 was used to analyse the data using both descriptive and inferential statistics. In our descriptive analysis, we computed the proportion of women in each of the aforementioned independent variables. The proportion of women who indicated either “Yes” or “No” on whether the NHIS is helping pregnant women in health services was also calculated (see Table 1). Chi-square tests were conducted in order to ascertain the independent variables that had significant association with the dependent variable. With the exception of “current pregnancy status”, all the independent variables were significant and were included in our inferential analysis, where three Binary Logistic Regression models were fitted in all. This analytical approach was the most suitable option premised on the fact that our dependent variable had two outcomes. The first model (Model I) accounted for NHIS subscription and whether it helps pregnant women in accessing health services. In model two, we adjusted for the effect of ANC visit-as a woman needs to first access healthcare during pregnancy in order to know whether the NHIS helps in healthcare during pregnancy or otherwise. All the seven significant independent variables were fitted in the final model (Model III) after which post-estimation test (Linktest) was conducted to determine whether the model is devoid of model specification error and also to ensure that relevant variables have not been omitted. Multicollinearity was also checked and we found no evidence of multicollinearity. Results for Model I was reported as odds ratio (OR) whilst that of Model II and III were reported as adjusted odds ratios (aOR) with their respective confidence intervals which were considered statically significant at 95%. Samples were weighted to adjust for the sample design.

Table 1 presents the results on background characteristics and whether NHIS is helping pregnant women to access health services. Twenty four percent of the women were aged 25–29 years. Nearly half of the women (49.8%) had attained secondary level of education with 21.0% of them within the poorest wealth quintile. More than half (58.5%) of the women were found within the rural settings. Seventeen percent were in the Ashanti region, 16.3% in Greater Accra, 11.6% in Northern and 2.7% in Upper West region. The majority (87.7%) of the women made at least 4 visits. Ninety percent (90.1%) of the women were not pregnant at the time of the survey and 67.0% were subscribers of NHIS. Approximately 18% of these women were found in Ashanti region. The chi-square test showed that age (X² = 23.6, p < 0.001), education (X² = 122.6, p < 0.001), wealth quintile (X² = 183.9, p < 0.001), residence (25.8, p < 0.001), ANC visits (X² = 130, p < 0.001), NHIS subscription (X² = 81.7, p < 0.001) and region (X² = 595.5, p < 0.001) had statistically significant association with the perception that NHIS helps pregnant women to access health services (see Table 1).

Table 2 presents the logistic regression results on whether the NHIS is helping pregnant women to access health services or not. It was found that women who had subscribed to the NHIS had higher odds [aOR = 1.70; CI = 1.38–2.10] of indicating that NHIS is helping pregnant women in accessing health services compared to those who are not subscribed. The study showed that women who accessed ANC at least four times had higher odds [aOR = 1.99; CI = 1.43–2.77] to report that NHIS is helping to seek health services as compared to their counterparts who attained less than four ANC visits. Women with secondary level of education [aOR = 1.42; CI: 1.04–1.92] and those in the richest wealth quintile [OR = 3.51; CI = 1.94–6.34] had higher odds of indicating that NHIS is helping pregnant women in accessing healthcare. However, women aged 45–49 [aOR = 0.49; CI = 0.26–0.94], women in the Greater Accra [aOR = 0.29; CI = 0.16–0.53], Eastern region [aOR = 0.12; CI = 0.07–0.21], Northern [aOR = 0.29; CI = 0.12–0.66] and Upper East [aOR = 0.17; CI = 0.09–0.31] had lower odds of reporting that NHIS is helping them to access health services (see Table 2).

The study’s strength lies in the use of nationally representative survey data and the use of a broad sample size that represents the current situation in a very accurate  manner. We indeed, note that this study has a few limitations. First, causal inference is limited by the cross-sectional design of the study. Again, there may be other variables that may affect the perception of women that the NHIS helps pregnant women in  accessing health services that were not included in this study because of their unavailability in the dataset. Finally, the answer to the dependent variable–is NHIS helping pregnant women for health services or not–rely mainly on discretionary verbal responses of the women which did not consider a specific health service but looked at access to health serices in general.