Background: Ghana’s National Health Insurance Scheme (NHIS) was introduced in 2005 as a demand side intervention to remove financial barriers to accessing health services. After almost a decade of implementation, this study aims to investigate the association of NHIS membership with antenatal visits (ANC), postnatal visits (PNC) and under-five mortality, using data from the most recent Multiple Indicator Cluster Survey (MICS). Methods: The survey was nationally representative and used a two-stage sample design to produce separate estimates for key indicators for each of the ten regions in Ghana. A generalised linear model (GLM) with binomial-family logit-link was used to estimate the effect of NHIS membership on each of the MNCH service utilisation indicators, adjusting for relevant confounding factors. Using birth history data, the Cox proportional hazard regression model was used to estimate the effect of NHIS membership on the incidence of under-five deaths, adjusted for wealth quintiles and other potential confounders. Results: The results support the role of health insurance membership in improving access to maternal and child health services, including antenatal care (ANC4+ adjusted OR = 1.94; 95 % CI = [1.28, 2.95]; P < 0.01), and content of antenatal care (adjusted OR = 2.05; 95 % CI = (1.46, 2.90); P < 0.0001). However, the study failed to show evidence of association of NHIS membership and under-five mortality (adjusted hazard rate = 0.86; 95 % CI = [0.64, 1.14]; P = 0.30). Conclusions: National health insurance membership is associated with increased access to and utilisation of health care but not with under-five mortality.
Ethical approval for the Multiple Indicator Cluster Survey (MICS) was obtained from the Ghana Health Service. The data available for this study cannot be linked to an individual who participated in the study. The MICS 2011 was used for this analysis [20]. The choice of using MICS 2011 data was made to allow for a sufficient time period after the introduction of NHIS, yielding a greater likelihood of observing its potential effects on maternal and child health service utilisation and outcomes. The survey is nationally representative and used a two-stage sample design to produce separate estimates for key indicators for each of the ten regions in Ghana. The first stage involved systematically selecting clusters (called enumeration areas or EAs) with probability proportional to size from an updated master sampling frame constructed from the Ghana Population and Housing Census (2010) [21]. The second stage of selection involved the systematic sampling of the households listed in each cluster. The MICS (2011) duly interviewed 11,925 households. In these households, 10,627 women aged 15–49 years were duly interviewed giving a response rate of 97 percent. Complete responses were obtained on 7550 children under age 5 from their mother/caregiver. The number of women age 15–49 who had a live birth in the two years preceding the survey was 2,528. Further details of the sample design and questionnaire are described elsewhere [20]. In this analysis, the main exposure of interest was NHIS membership defined as having a valid insurance card, which was seen and confirmed by the interviewer. To examine if the membership to the NHIS contributes to increasing access and utilization of health services, the following maternal, newborn and child health (MNCH) service access and utilisation measures were selected: 1) ANC 4+, defined as the percentage of at least four antenatal care visits during pregnancy among women who had a live birth during the two years preceding the survey; 2) Content of antenatal care, defined as the percentage of comprehensive ANC (i.e. blood pressure measured, urine sample taken, and blood sample taken as part of antenatal care) among women who had a live birth during the two years preceding the survey; and 3) Post-natal health checks for newborns within 2 days of delivery, defined as percentage of newborns born in the last two years who received health checks and post-natal care (PNC) visits from any health provider within 2 days of delivery. All women who gave birth in the two years preceding the survey were included in the analysis. Pearson design-based F test was used to explore the association of NHIS membership and background characteristics of women aged 15–49 years dully interviewed. A generalised linear model (GLM) with binomial-family logit-link was used to estimate the effect of NHIS membership on each of the service utilisation indicators, adjusting for relevant confounding factors. Socioeconomic status measured by household wealth quintiles, using an asset index, was considered a priori as potential confounder for the NHIS membership-service utilisation relationship and so was adjusted for in all the analyses. Other maternal and household characteristics such as mother’s level of education and area of residence were explored for potential confounding. An adjusted Wald test was used to calculate the P-value as a measure of random error in the adjusted regression model. To examine if NHIS membership can contribute to improved health outcome for children, the study also considered under-five mortality, defined as the probability of dying before the fifth birthday. Every child recorded in the complete birth history dataset who was born within five years preceding the survey was included. The death before the child’s fifth birthday or the woman’s date of interview was estimated using the Kaplan-Meier failure method. The hazard function was estimated for key maternal and household characteristics as well as wealth quintiles. The Cox proportional hazard regression model was used to estimate the effect of NHIS membership on the incidence of under-five deaths adjusted for wealth quintiles and other potential confounders. The birth history data is suitable because it records the three key variables of survival data, i.e. date of birth of child, date of death of child (or age at death), and event status – death or alive. All analyses were adjusted for survey design characteristics (i.e. sampling weight, cluster sampling, and stratification). The analyses were performed using Stata version 13 (StatCorp, College Station, Texas, USA).
N/A