Two recent studies conducted in Kenya suggest that voucher programs designed to subsidize poor women’s access to maternal health care are associated with increases in facility-based deliveries and in skilled attendance at delivery, though not in other maternal health services.^1,2 In the first study, women living in informal settlements in Nairobi were more likely to deliver in a facility and have a skilled attendant after implementation of a voucher program in 2006 than they had been before implementation (odds ratios, 1.4 and 1.2). Similarly, a survey conducted in six rural districts around the country found that women who lived in communities where the voucher program had been initiated were more likely to have facility-based deliveries and skilled birth attendance than were women in communities without the program (2.1 and 2.0, respectively). However, neither study found that the voucher program increased utilization of prenatal care.

Unattended delivery is one of the greatest risk factors for maternal mortality. Since 2006, as part of Kenya’s efforts to reduce maternal mortality, poor pregnant women in three rural Kenyan districts (Kiambu, Kisumu and Kitui) and two informal settlements in Nairobi (Korogocho and Viwandani) have been eligible to buy inexpensive safe motherhood vouchers that cover four prenatal care visits, delivery at a facility (including caesarean section, if necessary), treatment of maternal and neonatal complications, and post-natal care. Women can also purchase separate vouchers to cover family planning services.

To determine if the voucher program was associated with increases in use of maternal health services in the two Nairobi slums, researchers compared data collected through the Nairobi Urban Health Demographic Surveillance System before the voucher program was implemented with data collected after implementation. The first data set consisted of 1,927 women aged 12–54 who had had either a live birth or a stillbirth in 2004–2005; the second consisted of 2,448 similarly aged women with a live birth or stillbirth between September 2006 and December 2008. The data included information on participants’ social and demographic characteristics and birth history, as well as on place of delivery, skilled birth attendance, vaccination coverage and (for births in 2006–2008) voucher use. The four outcomes of interest were having any prenatal care, having four or more prenatal care visits, delivering in a facility and delivering with a skilled attendant. Logistic regression was used to estimate the odds of facility-based delivery before and after implementation of the voucher program.

Overall, 54% of participants lived in Korogocho. Forty-eight percent were aged 12–23 at the time of the birth, 45% were aged 24–34 and 8% were 35 or older; 30% had only one child. More than three-quarters of the women had at least a primary education. Among women who delivered in 2006–2008, 76% knew of the voucher program, and 45% had purchased the safe motherhood voucher. In 2004–2005, 65% of women delivered at a facility and 70% had the assistance of a skilled attendant; in 2006–2008, those figures were 72% and 74%, respectively.

In multivariate analyses, participants had greater odds of delivering in a facility after the voucher program began than before (odds ratio, 1.4). Women were more likely to have had a facility delivery if they were aged 24–34 than aged 12–23 (1.2), and if they had a secondary education rather than no education (1.6). Women residing in Viwandani were less likely than those in Korogocho to have a facility delivery (0.6), and the odds of delivering in a facility decreased as parity increased. Analyses of skilled birth attendance yielded similar findings, including elevated odds of usage after voucher implementation (1.2).

However, the association between voucher implementation and use of maternal health services did not extend to uptake of prenatal care. Analyses revealed no change in the likelihood that women had obtained at least one prenatal visit, and decreases occurred in the odds that women had had four or more prenatal visits (0.7) or had had a visit during the first trimester (0.8). The relationships between the three prenatal visit measures and respondent characteristics varied, but in general prenatal visits were negatively associated with parity and positively associated with education and socioeconomic status.

The researchers note that the biggest limitation of their study is the lack of contemporaneous control groups, which prevents the study from being a true impact evaluation. However, to explore whether the increase in the odds of a facility delivery and skilled attendance reflected a broader trend in health care utilization, the authors examined whether an unrelated type of care—vaccination among children aged 12–23 months—also increased between 2004–2005 and 2006–2008; they found no change in immunization rates. The researchers conclude that the voucher program "appears to be an effective mechanism to target low-income urban women and give them the financial means to overcome economic barriers to a facility delivery."

Rural Sites

To estimate the impact of vouchers on the use of reproductive health services (including family planning) in rural communities exposed to the program, researchers examined data collected in 2010 from 2,527 randomly selected women aged 15–49 who lived within 5 km of a health facility in three voucher sites (Kiambu, Kisumi and Kitui) or in three comparison districts with similar population characteristics and health services (Makueni, Nyandarua and Uasin Gishu). Women were eligible for the survey if they were pregnant or had given birth within the last 12 months; if no such woman was available in the selected household, a woman in the same age-group who was sexually active (and thus a potential contraceptive user) was interviewed. The information collected included demographic characteristics, health care utilization, birth history, and family planning knowledge and use; women who had given birth in the five years before the survey provided details on prenatal care, delivery and postnatal care services for all deliveries during that period. The outcomes of interest were use of any family planning method (ever and in the past 12 months); use of a long-acting or permanent method (ever and in past 12 months); having had four or more prenatal care visits; having had a first-trimester prenatal care visit; having delivered in a facility; having had a skilled birth attendant; and having used postnatal care services. Exposure to the voucher program was categorized according to whether women lived in an area where the program had been implemented in 2006, where the program had been implemented in 2010 (the same year that the survey was conducted) or where it had not been implemented at all (nonvoucher sites). The researchers estimated multilevel logit models to predict service utilization by program exposure.

Overall, more than three-quarters of the women were aged 15–34; most had at least a primary education, lived in a rural area, and were currently or formerly married or cohabiting. Forty percent were unemployed, and 74% were poor. Only 3% of women in the nonvoucher areas had ever heard of the vouchers, compared with 82–88% of women in the two intervention areas. Twenty-one percent of women in each exposure group, but none of the women in the comparison areas, had ever used a voucher.

Compared with women in the nonvoucher areas, those who lived in areas where the program had been in effect since 2006 were more likely to have ever used a long-acting or permanent method, though the two groups did not differ on the other three measures of contraceptive use (odds ratio, 1.5). The odds of having used a method ever and in the past year were lower among those not exposed to the voucher program until 2010 than among those not exposed at all (0.3 and 0.5, respectively).

For births that occurred after the voucher program began, women in communities where the program had been initiated in 2006 were more likely than those in unexposed communities to have had a facility-based delivery and to have been assisted by a skilled attendant (odds ratios, 2.1 and 2.0, respectively). However, the odds that a woman had received antenatal care (a first-trimester visit or four total visits) or postnatal care did not differ by voucher status. In addition, age was positively associated with having had four or more prenatal visits, having delivered in a facility and having had an attended birth, and parity was negatively associated with all safe motherhood services. Regardless of community exposure to the voucher program, nonpoor women were more likely than poor women to have delivered at health facilities, received skilled delivery care and used postnatal care services.

The researchers note a number of limitations. Although the vouchers were intended for poor women, some were sold to other women, which may have reduced the number available to those most in need. In addition, some aspects of the study design (e.g., the focus on women living within 5 km of a facility) may have affected estimates of the voucher program’s impact; the researchers were not able to determine the degree to which changes in service usage were attributable to the vouchers; and most of the variables used in models of service utilization referred to the time of the survey, not of the birth. The researchers note that although the voucher program "is associated with increased health facility deliveries and skilled delivery care," it has had little apparent impact on antenatal care and "has not eliminated the gap [in service utilization] between the poor and nonpoor." Thus, "there is need to consider other dimensions of access, such as availability and acceptability of services."—L. Melhado

REFERENCES

1. Bellows B et al., Increase in facility-based deliveries associated with a maternal health voucher programme in informal settlements in Nairobi, Kenya, Health Policy and Planning, 2013, 28(2):134–142.

2. Obare F et al., Community-level impact of the reproductive health vouchers programme on service utilization in Kenya, Health Policy and Planning, 2013, 28(2):165–175.