Menu
Menu
Menu
Menu
Background: Over two thirds of women who need contraception in Uganda lack access to modern effective methods. This study was conducted to estimate the potential cost-effectiveness of achieving universal access to modern contraceptives in Uganda by implementing a hypothetical new contraceptive program (NCP) from both societal and governmental (Ministry of Health (MoH)) perspectives. Methodology/Principal Findings: A Markov model was developed to compare the NCP to the status quo or current contraceptive program (CCP). The model followed a hypothetical cohort of 15-year old girls over a lifetime horizon. Data were obtained from the Uganda National Demographic and Health Survey and from published and unpublished sources. Costs, life expectancy, disability-adjusted life expectancy, pregnancies, fertility and incremental cost-effectiveness measured as cost per life-year (LY) gained, cost per disability-adjusted life-year (DALY) averted, cost per pregnancy averted and cost per unit of fertility reduction were calculated. Univariate and probabilistic sensitivity analyses were performed to examine the robustness of results. Mean discounted life expectancy and disability-adjusted life expectancy (DALE) were higher under the NCP vs. CCP (28.74 vs. 28.65 years and 27.38 vs. 27.01 respectively). Mean pregnancies and live births per woman were lower under the NCP (9.51 vs. 7.90 and 6.92 vs. 5.79 respectively). Mean lifetime societal costs per woman were lower for the NCP from the societal perspective ($1,949 vs. $1,987) and the MoH perspective ($636 vs. $685). In the incremental analysis, the NCP dominated the CCP, i.e. it was both less costly and more effective. The results were robust to univariate and probabilistic sensitivity analysis. Conclusion/Significance: Universal access to modern contraceptives in Uganda appears to be highly cost-effective. Increasing contraceptive coverage should be considered among Uganda’s public health priorities. © 2012 Babigumira et al.
A Markov cohort model was developed to assess the potential cost-effectiveness of the NCP compared to the CCP. The model projected the reproductive health experience of a hypothetical cohort of 15-year old girls over a lifetime horizon. The starting age of the hypothetical cohort was chosen to reflect as closely as possible the median age of sexual debut in Uganda – 16.6 years [1]. Figure 1 shows a schematic of the Markov model. The model illustrates the different states of contraception through which women between 15 and 49 years of age in Uganda transition. Each state is associated with a cost and a value of disability-adjusted life years lost. All states may progress to dead. The Markov model is suited to women’s reproductive experience because it spans many years and many events – pregnancies, miscarriages, abortions and births – that can occur multiple times. For instance, women face multiple opportunities to get pregnant with the probability of pregnancy diminishing with each subsequent cycle as the individual ages. The model had 7 states: (i) not sexually active (NSA); (ii) intentional non-contraception (INC); (iii) unintentional non-contraception (UNC); (iv) modern contraception (MOC); (v) traditional contraception (TRC); (vi) pregnant and (vii) dead. The INC state included women who were looking to get pregnant and the UNC state included women who lacked access to modern contraception. The cycle time was 9 months. The model assumed a constant modern contraceptive use mix across all ages for women on contraception. The model was checked (de-bugged) by varying transition probabilities between 0 and 1 to observe if responses were logical and setting costs and outcomes to 0 separately to examine if the expected values were identical. Validation was performed by comparing the predicted fertility to the published estimate for Uganda [1]. The analysis was performed from both the governmental (Ministry of Health (MoH)) and the societal perspectives and included direct and indirect costs. The MoH perspective included direct medical costs and direct non-medical costs that are incurred by the MoH which is the healthcare provider in Uganda and the societal perspective included, in addition to these, the direct non-medical costs incurred by patients (such as transportation) and indirect (productivity) costs. Costs and outcomes were discounted at 3% per year [9]. The NCP was compared to the CCP on the basis of costs, life expectancy and incremental cost-effectiveness analysis using cost per life-year (LY) saved and disability-adjusted life years (DALY) averted to capture both quality and quantity of life. The model was also used to compute other intermediate measures of cost-effectiveness: 1) cost per pregnancy averted; 2) cost per unit of fertility reduction; 3) cost per ectopic pregnancy averted; 4) cost per miscarriage averted; 5) cost per induced abortion averted; 6) cost per still birth averted 7) cost per neonatal death averted; 8) cost per infant death averted; and 9) cost per child death averted. A threshold for cost-effectiveness with ranges from 1 to 3 times per capita GDP per DALY averted has been suggested [10]–[12] and other studies have used this threshold in Uganda [13], [14]. Uganda’s GDP per capita was $474 in 2010 [15]. Therefore the NCP was judged to be highly cost-effective if the incremental cost-effectiveness ratio (ICER) was less than $474 per DALY and cost-effective if the ICER was less than $1423 per DALY (3 times per capita GDP). The proportion of 15-year olds who reported no sexual activity in the 3 months prior to the 2006 UDHS (80.3%) was started in the NSA state [1]. In the CCP, the remaining 19.7% who were sexually-active women were divided among the other states: 9.1% who used modern contraception started in the MOC state; 1.8% who used traditional contraception started in the TRC state; 6.6% who lacked access started in the UNC state; and the remaining 2.1%, considered to want to conceive, started in the INC state [1]. In the NCP, 17.6% (who started in the MOC, TRC, and the UNC states) were started in the MOC state, akin to universal access to modern contraception, and 0% was started in the INC and TRC states. Transition probabilities between states of contraceptive use over a woman’s life were computed using UDHS data and were age-specific within five-year age intervals (Table 2) [1]. The proportion of women who remained sexually inactive represented the probability (by age group) of staying in the NSA state (UDHS 2006; Table 7.7.1 (Page 93)) [1]. Sexual activity was defined as reported sexual activity within the previous 4 weeks, consistent with the UDHS definition [1]. The proportion of women using traditional and modern contraception (UDHS 2006; Table 6.2.1 (Page 67)) [1] represented the probability (by age group) of transition between both the NSA and UNC states and MOC and TRC states respectively. The proportion of sexually active women lacking access to modern contraception (UDHS 2006; Table 7.7.1 (Page 93) [1] represented the probability (by age group) of transition between the NSA and UNC states. The proportion of sexually active women who wish to conceive was calculated by subtracting the sum of the proportions (by age group) of women who use contraceptives and women who lack access to contraceptives from 1. The resulting proportion represented the probability of transition from the NSA state to the INC state. The probability of pregnancy without contraception (85%) [16] represented the probability of transition from the UNC and INC states to the pregnant state. This probability was adjusted for the age-specific prevalence of menopause (defined as last known menstrual period occurring 6 or more months prior to survey among non-pregnant and non-amenorhoeic women (UDHS 2006; Table 7.10 (Page 98)) [1], which increases from 2.4% between 30 and 34 years to 42.8% between 48 and 49 years of age. The rate of contraceptive failure on traditional contraception (20%) and modern contraception (3%) [16] represented the probability of transition between the TRC and MOC states and pregnancy. The failure rate for modern contraceptive use was weighted by the frequency of use of different modern methods in Uganda [1]. The probability of intentional and unintentional contraceptive discontinuation, estimated in an Eastern African study at 16.7% and 12.1% respectively [17], represented the probability of transition between the MOC state and INC and the MOC and TRC states respectively, assuming that women who lose access to modern contraception opt for traditional contraception. Women who had live births transitioned to the MOC state because the probability of pregnancy during lactation amenorrhea is similar to the probability of pregnancy on modern contraceptives [18]. Women who had non-live birth pregnancy outcomes transitioned to other states at the same rate as women in the NSA state. Transitions between the MOC and TRC states in a single cycle as well as movement from contraceptive use states to the NSA state were not allowed in a single cycle; women returned to the NSA state only after pregnancy. Pregnancy was a temporary state i.e. women did not spend more than a single cycle in this state. All estimates reported as annual probabilities were converted into 9-month transition probabilities to reflect the cycle time of the model. The changing probability of different events such as pregnancy over time as women age (time-dependency) was captured in the model by using age-group-specific tables in lieu of the relevant transition probabilities. This enabled members of the simulated cohort of different ages to be assigned their relevant age-specific transition probabilities. The non-age-specific transition probabilities are shown table 3. Age-specific mortality rates from all causes for women in Uganda were obtained from country-specific life tables published by the World Health Organization [19] and are shown in table 2. These were adjusted for the percentage of deaths due to maternal causes which is 13% [1]. Maternal mortality is 435 (345–524) deaths per 100,000 live births [1]. This estimate was adjusted for the proportion of pregnancies that result in live births. Neonatal, infant and child mortality estimates (table 3) were obtained from the UDHS [1] and are represented cumulatively i.e. infant mortality includes neonatal mortality and child mortality includes both neonatal and infant mortality. Annually, there are an estimated 498,000 DALYs lost due to maternal causes (pregnancy complications) in Uganda [20] and an estimated 1,830,000 pregnancies [21]. Therefore the average DALY loss due to pregnancy complications associated with a single pregnancy is 0.27. Costs were estimated for the pregnant (PRE) and modern contraception (MOC) Markov states only; we assumed that the other Markov states – NSA, INC, UNC, TRC and DEAD – bore no costs. For these two states, we estimated direct medical costs, direct non-medical costs, and indirect costs. In the MOC state, the direct medical costs included the cost of contraceptive technology, weighted by the prevalence of the use of the different methods [22], the cost of healthcare personnel [23], transportation costs [24], costs of upkeep [24], and out-of-pocket costs when patients seek contraceptive services [23]. The direct non-medical costs included overhead costs and capital costs for out-patient care in Uganda obtained from the WHO WHOCHOICE database [25]. The indirect costs included the costs of lost time when women seek health services [24]. The costs of different contraceptive different contraceptive technologies and their prevalence of use are shown in Table S1 and the costs of other inputs are shown in Table S2. In the PRE state, the different cost categories were estimated for antenatal care and the different potential outcomes of pregnancy – miscarriage, induced abortion, ectopic pregnancy, birth (still and live; vaginal and cesarean), obstetric hemorrhage, and eclampsia, weighted by their incidence [1], [21], [23], [26]–[28]. The direct medical costs included the costs of healthcare personnel [23] and other healthcare materials [23], transportation costs [24], costs of upkeep [24], and out-of-pocket costs when patients seek different services [23], [24]. The direct non-medical costs included the overhead and capital costs associated with different services [25] and the indirect costs included the productivity costs associated with lost time while seeking care for different services [24]. The costs of different inputs by pregnancy outcome are shown in Table S3 and the incidence of pregnancy outcomes, unit costs and total costs of pregnancy from the MoH and societal perspective are shown in Table S4. A detailed description of the estimation of the costs of modern contraception and pregnancy is given in Appendix S1. All costs were in 2010 US dollars. Sensitivity analyses were performed to determine which variables had substantial impact on costs and outcomes. All parameters were assigned a range of plausible values using 95% confidence intervals when available or +/−50% for costs and +/−20% for other parameters (table 2). To further test the robustness of our results, we conducted a probabilistic sensitivity analysis. We created probability distributions for all of the parameters in the model except those relating to the underlying methods of the analysis such as the discount rate [29]. For all other parameters, the base-case value was used for the mean, and the standard error was estimated based on the approximation that the range used for one-way sensitivity analyses represented a 95% confidence interval, with the range approximately equal to four times the standard error [29]. A Beta distribution was used for probabilities and DALYs because it is bounded on the interval 0–1 and resembles a normal distribution for some parameterisations and a normal distribution was used for costs because we were not overly concerned with skewness as to use a log-normal or gamma distribution [29]. Monte Carlo simulation was used to create 10,000 iterations for which the expected outcome values were calculated. The probability that either intervention was cost-effective was then calculated for a range of cost-effectiveness thresholds. Data analysis was performed using TreeAge Pro.
