The economic burden of maternal mortality on households: Evidence from three sub-counties in rural western Kenya

Background: This study explores the consequences of a maternal death to households in rural Western Kenya focusing particularly on the immediate financial and economic impacts. Methods: Between September 2011 and March 2013 all households in the study area with a maternal death were surveyed. Data were collected on the demographic characteristics of the deceased woman; household socio-economic status; a history of the pregnancy and health care access and utilization; and disruption to household functioning due to the maternal death. These data were supplemented by in-depth and focus group discussions. Results: The health service utilization costs associated with maternal deaths were significantly higher, due to more frequent service utilization as well as due to the higher cost of each visit suggesting more involved treatments and interventions were sought with these women. The already high costs incurred by cases during pregnancy were further increased during delivery and postpartum mainly a result of higher facility-based fees and expenses. Households who experienced a maternal death spent about one-third of their annual per capita consumption expenditure on healthcare access and use as opposed to at most 12% among households who had a health pregnancy and delivery. Funeral costs were often higher than the healthcare costs and altogether forced households to dis-save, liquidate assets and borrow money. What is more, the surviving members of the households had significant redistribution of labor and responsibilities to make up for the lost contributions of the deceased women. Conclusion: Kenya is in the process of instituting free maternity services in all public facilities. Effectively implemented, this policy can lift a major economic burden experienced by a very large number of household who seek maternal health services which can be catastrophic in complicated cases that result in maternal death. There needs to be further emphasis on insurance schemes that can support households through catastrophic health spending. The study sample was selected from KEMRI/CDC’s Health and Demographic Surveillance System (HDSS), established in 2001. The HDSS currently includes a total population of 225,000 individuals in Rarieda, Gem and Siaya sub-counties who are visited quarterly. In 2008, 41.7 of the population was between the ages of 15-49 years, of which slightly over half (54) were women of reproductive age [24]. HDSS uses Global Positioning System (GPS) coordinates to map each compound in the surveillance area and assign each household and individual within these compounds a unique identification number. Information collected through the quarterly survey includes data on births, deaths and the causes of death (through verbal autopsy), pregnancy, pregnancy outcomes, morbidity, migration, education and socioeconomic status. The sampling frame inherent in this, as in other surveillance data removes many of the biases found in study designs that are not based on a whole population sample. Correct identification of maternal deaths is assured as the HDSS uses the WHO Verbal Autopsy method, and collects data on cause of death through an experienced team of community interviewers, village reporters and staff responsible for quality control. Since maternal death is a rare event, the study attempted to identify and interview respondents about all maternal deaths that occurred within the surveillance area in a period of 22 months, the duration of the data collection period of the study. To minimize recall issues, households were recruited on a rolling basis, after a period of at least 2 months after the maternal death to ensure they were respected during time of grief but as soon as possible after that death. Specifically, households were approached no sooner than two months after the maternal death, but no later than six months. Two control households were interviewed per each case household. Control households were defined as those where a woman had a healthy pregnancy and delivery in the same time period as the women from the case households. Due to ethical considerations around patient identification and findings of a preliminary scan of select health care facilities in the study areas that pointed at significant lack of documentation, the study chose not collect provider level data. Information on the medical course of the pregnancy, delivery and postpartum periods was collected at the household level and lacked clinical detail. Two survey questionnaires were used to collect data from all identified case and control households. The cost questionnaire collected detailed information on the types of care sought during pregnancy, childbirth and postpartum, and the costs associated with each incidence of help seeking, i.e. health care utilization. It was designed to capture expenditures associated with using a range of services; institutional services; hospitals, health centers and private clinics as well as services by non-institutional providers such as traditional birth attendants (TBAs) and informal medical practitioners. Expenditures associated with home-based delivery were also surveyed. For each incidence of care sought, cost information was collected on a detailed list of items, including spending on transportation and other medical and non/medical expenses related to the visit (see Table ​Table2).2). In households that experienced maternal mortality, an additional module was administered to collect information about funeral costs. While most cost data were collected in monetary terms, where in-kind payments are common, such as payment to informal sector providers and funeral expenses, households were asked about the monetary equivalent of the spending incurred. Classification of Costs The SES questionnaire collected socio-economic information on women and their household including household expenditure on food and non-food items and durable goods, household asset ownership and dwelling characteristics. Case households were also asked about the members’ employment and time use. Interviews were conducted with an adult household member aged 18 years or older who was most knowledgeable about the information sought. If one person was not knowledgeable in all topic areas, the questionnaires allowed for different respondents for different modules. Finally, even though enumerators were instructed to ask for receipts, most of the time, information collected relied on respondents’ memory. An important caveat to note is that while women in control groups report on their own experiences, the most knowledgeable people in the case households reported on the experiences of the deceased and the aftermath of her death. During the 22-month data collection period, 67 households that had a maternal death were identified using the sampling strategy described above. The sample size of the control households was 92, matched by timing of birth such that the control household for a maternal death household includes a woman who delivered in a period of up to two months after the maternal death. Of the identified, 59 cases and 86 controls were interviewed using the tools. The response rate for the SES questionnaire was slightly lower among case households with a total of 54 interviews fully or partially completed. The remaining households refused to be interviewed or had no appropriate respondent to take the survey Group discussions were also held with 11 of the case households. An attempt was made to select households to include a mix of socio-economic strata and households representing a variety in regard to characteristics such as number, age and sex of children. No group discussions were held with control households. These households were interviewed using an open-ended interview guide to collect information on the living and working arrangements of household members before and after the maternal death. The methodology integrated several approaches to respond to the specific features of the study context; a rural setting in a developing country with potential low service utilization, lagging administrative record-keeping in facilities, and where labor markets are highly informal. Overall, a similar methodological approach to Ye et al [3] was followed, particularly in the measurement of the direct costs of maternal costs. Specifically, similar to Ye et al, direct costs were measured in terms of the out-of-pocket expenditures related accessing and using health services throughout pregnancy, delivery and postpartum through an accounting methodology. For each incidence of service utilization, three cost categories were used– direct costs related to help seeking, transportation, and other medical and non-medical. Estimates of average per visit costs to case and control households were generated to ensure that the potential difference in number of incidents of help seeking/service utilization among case and control households is controlled for. On the other hand, average total costs were also reported across case and control households to fully understand the financial burden of these costs. In the absence of clinical information, the study could not compare the costs incurred by women who died and who survived the same morbidity to generate the “differential” financial costs of mortality. In order to assess the impact of these out-of-pocket costs on households, two measures of household economic status were generated. Household consumption expenditure estimates used data collected on expenditures on food and non-food items as well as durable goods (30 food items, 22 non-food consumption items and 13 durable items). The final estimates generated used the food and non-food items due to a high number of missing data in durables expenditures and complications in generating rigorous, reliable estimates. Household asset data as well as key variables of household dwelling characteristics were used to construct a wealth index and wealth groupings. Given the considerably small sample size, the wealth groupings assigned to the sample across three wealth strata are only modestly indicative of their placement. The study also attempted to capture the extent to which maternal death may trigger disruptions in productivity for an affected household, potentially exacerbating the effects of the financial impact of health and related expenditures. In their study, Ye et al use the income data collected from households to estimate the value of lost wages due to maternal death [3]. Meanwhile in this study, the number of days lost from productive activity and shifts in household division of labor and time use were reported without an attempt to monetize them. This decision was dictated by the near absence of wage labor in the communities the study took place. Specifically, baseline data were collected on the deceased women’s farming, other economic activity, and household-related work prior to their health deteriorating due to maternal causes. Questions were posed on the time use and task responsibilities of members of the deceased women’s household before and after the maternal death, as well as the changes in these members’ pre- and post-mortality economic activity. The aim of these questions was to understand how households re-arranged their responsibilities and economic activities to fill the gap created by the maternal death. While typical productivity analyses are in monetary terms, given the study site, no attempt was made to translate these changes into monetary value. This methodology extended typical time use studies by trying to document the shifts not only in time, but also in responsibility, before and after a key catastrophic event. The qualitative data complemented the quantitative by providing a more nuanced narrative of the productivity disruption described above, what it means for different household members’ workload and daily lives, and the emotional costs of the maternal death.