Improving maternal care through a state-wide health insurance program: A cost and cost-effectiveness study in rural Nigeria

Background While the Nigerian government has made progress towards the Millennium Development Goals, further investments are needed to achieve the targets of post-2015 Sustainable Development Goals, including Universal Health Coverage. Economic evaluations of innovative interventions can help inform investment decisions in resource-constrained settings. We aim to assess the cost and cost-effectiveness of maternal care provided within the new Kwara State Health Insurance program (KSHI) in rural Nigeria. Methods and Findings We used a decision analytic model to simulate a cohort of pregnant women. The primary outcome is the incremental cost effectiveness ratio (ICER) of the KSHI scenario compared to the current standard of care. Intervention cost from a healthcare provider perspective included service delivery costs and above-service level costs; these were evaluated in a participating hospital and using financial records from the managing organisations, respectively. Standard of care costs from a provider perspective were derived from the literature using an ingredient approach. We generated 95% credibility intervals around the primary outcome through probabilistic sensitivity analysis (PSA) based on a Monte Carlo simulation. We conducted one-way sensitivity analyses across key model parameters and assessed the sensitivity of our results to the performance of the base case separately through a scenario analysis. Finally, we assessed the sustainability and feasibility of this program’s scale up within the State’s healthcare financing structure through a budget impact analysis. The KSHI scenario results in a health benefit to patients at a higher cost compared to the base case. The mean ICER (US$46.4/disability-adjusted life year averted) is considered very cost-effective compared to a willingness-to-pay threshold of one gross domestic product per capita (Nigeria, US$ 2012, 2,730). Our conclusion was robust to uncertainty in parameters estimates (PSA: median US$49.1, 95% credible interval 21.9-152.3), during one-way sensitivity analyses, and when cost, quality, cost and utilization parameters of the base case scenario were changed. The sustainability of this program’s scale up by the State is dependent on further investments in healthcare. Conclusions This study provides evidence that the investment made by the KSHI program in rural Nigeria is likely to have been cost-effective; however, further healthcare investments are needed for this program to be successfully expanded within Kwara State. Policy makers should consider supporting financial initiatives to reduce maternal mortality tackling both supply and demand issues in the access to care. We used a decision analytic model to simulate a cohort of pregnant women, followed down a pathway of care during their current pregnancy until delivery. We defined two scenarios in our primary analysis: 1) current standard of care (base case scenario) where women do not have access to benefits from the insurance program; and 2) KSHI scenario (intervention scenario) where women have access to the insurance and to hospitals participating in the KSHI program. Alternative base case scenarios were defined in a scenario analysis, in addition to the primary analysis above, comparing: 3) an increased utilization of the standard of care clinics; 4) an increased cost and quality of care improvement in the standard of care clinics; and 5) increased utilization, increased cost and quality of care improvement in the standard of care clinics. The model’s primary outcome is the incremental cost per disability adjusted life year (DALY) averted in the KSHI scenario compared to the base case scenario. This incremental cost-effectiveness ratio (ICER) was calculated as the ratio of the difference in costs and DALYs averted between the intervention and base case scenarios. The ICER was then compared to a country-specific willingness-to-pay (WTP) threshold, defined as a country’s per capita gross domestic product (GDP) [23]. For Nigeria, the GDP per capita was US$ 2,730 in 2012 [24]. If the ICER is below this WTP threshold, the intervention is considered very cost-effective. Key model input parameters are shown in Table 1 and further details can be found in the S1 File. ANC, antenatal care; EOC, essential obstetric care; distr: probability distribution specified for each parameter in the Monte Carlo simulations; ref, reference; rr, relative risk; OL, obstructed labour; HTD, hypertensive disorder. Beta distributions are specified by mean (standard deviation); uniform distributions by minimum and maximum values; triangular distributions by average (minimum and maximum). *Own calculation (S1 File). The current standard of care in rural Nigeria (base case scenario) was characterized in two dimensions: utilization and quality of care. Kwara State has a health system with inadequate government funding, weak governance and legislation, and poor health infrastructure and service quality. The State is participating in the federally-funded National Health Insurance Scheme (NHIS). The majority of enrollees are individuals working in the formal sector. The NHIS started a rural community-based social health insurance program in 2010 but access to this scheme is limited [21]. Data collected during the baseline survey of the KSHI impact evaluation in 2009 showed that less than 1% of the population in the area was enrolled in any health insurance scheme [21]. The base case is therefore defined as a regionally-representative situation where functional health care facilities are mainly primary care clinics with limited access to secondary care (such as surgery, inpatient care). The assumptions on utilization and quality of care derived from regionally representative surveys, maternal health audits, and data collection as part of the baseline survey of the KSHI impact evaluation in 2009 [21]. All assumptions are described in the S1 File; key parameters and sources are presented in Table 1. The intervention modelled is the KSHI program. This includes a subsidized health insurance covering access to comprehensive health care, including primary care; treatment for malaria, tuberculosis, and HIV opportunistic infections; maternal and child care; surgeries; and care for chronic diseases. It also includes upgrades to facilities and technical assistance in program management by PharmAccess Foundation. In this context, the impact of the KSHI program is hypothesised to result from two pathways: 1) increased utilization of maternal services, defined as antenatal care (ANC) visits, delivery in health facilities, and emergency obstetric care (EOC) when complications during delivery arise; and 2) increased quality of care of maternal services provided (access to more facilities offering EOC and preventive treatment of hypertensive disorders complications during ANC) [11,25]. The model explicitly considers utilization and composition of ANC, the location of the care accessed, and the type of assistance provided during the delivery as well as availability of EOC. Essential obstetric care is defined as care including capacity to administrate parenteral antibiotics, parenteral oxytocic drugs, and parenteral anticonvulsants for pre-eclampsia and eclampsia; ability to perform manual removal of placenta and of retained products; ability to perform assisted vaginal delivery, surgery (C-section), and blood transfusions [26]. The schematic representation of the model structure is given in the S1 File. We considered five clinical outcomes of delivery: post-partum haemorrhage, obstructed labour, hypertensive disorder, sepsis, and uncomplicated delivery. The first four are responsible for the highest proportion of maternal mortality and morbidity in Nigeria [27–29]. We estimated a women’s probability of accessing treatment for these complications to be dependent on the location of care accessed during delivery and whether previous ANC visits were attended during the current pregnancy. Prevalence of adverse delivery outcomes were sourced from systematic reviews or cohort studies specific to Nigeria; when these were not available, we sourced estimates that were regionally representative. With regards to treatment outcome probabilities, all estimates were sourced from clinical trials or meta-analyses of clinical trials [30–34]. Mortality and morbidity outcomes were then translated into years of life lost due to premature mortality (YLLs) and years lived with disability (YLDs), respectively, to calculate total number of DALYs averted using standard methods [35], without age weighting [32]. We measured all costs and DALYs through a time horizon spanning the remaining life expectancy of the cohort (for a detailed description of assumptions, see S1 File). Costs were evaluated from a healthcare provider perspective. For the intervention scenario, we collected data at the Ogo Oluwa Hospital (OOH) in Kwara State. This is a private hospital participating in the KSHI program and serving the community of Bacita, part of Edu local government area (population estimated: 201,642 in 2006 [36]) in the North Central region of Nigeria [37]. The hospital provides ANC and perinatal care as well as EOC. The number of patients enrolled in the KSHI program registered in OOH was 9,738 for the period 2010–2011. These patients represented over 95% of the total number of patients accessing care in OOH (personal communication, medical director OOH). We measured service delivery costs including costs for building, overhead, staff, equipment and consumables, and maintenance at OOH. The resource use associated with each activity was estimated through observations of practice, a review of financial reporting, and interviews with staff. Resource use measurement took into account the allocation of fixed resources between maternal care and other services. Estimates of drugs and test prices were obtained from suppliers [38]. We extracted information on the total number of pregnancies, ANC visits, and deliveries from insurance claim data for the period covered by the costing exercise. We then calculated costs per ANC consultation and delivery care separately. Finally, we combined the utilisation data with unit costs to calculate the total costs of maternal services at OOH. A sensitivity analysis of assumptions where measurements of parameters were uncertain (percentage mark-up allocated to overheads, staff time, and medical equipment) was undertaken to estimate the impact of these assumptions on our cost estimates. We also included above-service program costs associated with the local operations of the insurer (Hygeia Nigeria Ltd) and program management at PharmAccess level. The operations at insurer level consist of administration of the package and marketing activities for scaling up of the project. Program management expenses at PharmAccess level consist of expenses related to upgrading of healthcare facilities and technical assistance concerning the health plan. In determining the cost-effectiveness of the program, these costs were taken into account from the beginning of the program in 2006 until 2018. After this date the program is expected to be transferred to the Kwara state Government. Expenses over the period 2006–2013 are audited, while from 2014 the amounts are based on projections. We added this as a mark-up to all patients in the intervention scenario, as this cohort was assumed to be insured. Detailed calculations are given in S1 File. Finally, we reviewed previous costing studies in Nigeria to validate our cost estimates and provide costs for treatment of morbidities associated with complicated deliveries. When estimates were missing, we used WHO guidelines and unit costs for outpatient visits sourced from WHO-CHOICE [39]. All prices were collected in local currency and are presented in 2012 US$ [40]. Cost information from previous studies was adjusted to account for inflation following standard methods [41,42]. All future costs and outcomes were discounted at 3% per year. To explore the sustainability of scaling up the intervention within the current health expenditures in Kwara state, we compared the cost of scaling up the program state-wide, over a five-year period, against current health expenditures for Kwara State, in a separate analysis [43,44]. The size of the population in need of maternal care (women aged 15 to 40 years) was estimated using available demographic data [36] and assuming a population growth equivalent to the rate of natural increase sourced from the World Bank [45]. The same cost assumptions were made for this analysis as in the primary analysis (Table 1). The level of insurance coverage was defined in terms of the proportion of the population in need that access the program. The annual cost of implementing the intervention was based on the estimated number of women in need accessing the program per year. The annual cost of scaling up the maternal care intervention to those in need was calculated using the following equation, excluding any financial gain from cost-sharing revenue collection: Population in need of ANC and EOC x unit cost of EOC and ANC (including above service level costs) given prevalence of different complications x insurance coverage—current estimated expenditure on ANC and EOC Only the resources and expenditures required above current spending levels were included. Given that the aim of the budget impact analysis is to explore the impact on the State’s health expenditure, the cost to households was not included in this analysis. Unit costs were inflated to 2012 prices when necessary, using the average inflation rate between 1996 and 2014 for Nigeria of 12.33% per annum [46]. Primary results are presented using a probabilistic sensitivity analysis (Monte Carlo simulation) to randomly sample parameters from their probability distributions repeatedly (10,000 times) to generate 95% credibility intervals around the incremental cost per DALY averted [47]. We assessed the sensitivity of our results to the performance of the base case in three ways (further information in S1 File): Finally, we conducted one-way sensitivity analyses across key model parameters to assess the robustness of our results, varying one parameter at a time between the outer limits of its confidence interval. In particular, we examined the sensitivity of our results to the probability of complications during delivery (by type of complication) as well as to the probabilities of mortality and morbidity from that complication. Similarly, we examined treatment costs for fistula and anaemia, duration of disability for all disabilities, estimates of ANC utilization and delivery at health facilities, and a large variation in the estimates of above-service program costs. The model was programmed using TreeAge Pro 2014 (TreeAge Software Inc., Williamstown MA), cost analyses were conducted using Microsoft Excel 2013 (Microsoft Corp., Redmond WA). We conducted and present this study following good reporting practices from published standards for reporting of economic evaluations of health interventions, the CHEERS statement, and the Bill and Melinda Gates Foundation, Methods for Economic Evaluation Project [49,50]. Empirical costing activities were conducted as part of ongoing evaluation efforts of the Health Insurance Fund program. The main project, QUality Improvement of Cardiovascular care in Kwara (QUICK), was approved on the 30th March 2010 by the ethical review committee at the University of Ilorin Teaching Hospital (reference number: UITH/CAT/189/13/13). We sought an extension of this ethics approval to include ANC and delivery care services data. This extension was granted on the 16th August 2012 by the same ethical review committee (ethical review committee at the University of Ilorin Teaching Hospital, reference number: UITH/CAT/189/15/450). No patient records/information were consulted and patients were not approached during this study. All data were aggregated and anonymized prior to analysis.