|
nif:isString
|
-
Type of study and sources of information: An ecological, cross-sectional study of several groups was conducted in Colombia at the state level. The study was performed based on information from the National Administrative Department of Statistics (DANE, Spanish acronym) [24,25], the National Planning Department (DNP, Spanish acronym), the Ministry of Health and Social Protection, and with data from the 2010 National Demographics Health Survey (ENDS, Spanish acronym) [26, 27]. The latter is representative of Colombia both nationally and at the state level, and included 101,482 women of reproductive age in 256 municipalities in the 33 states [27]. The ENDS collected information related to coverage and utilization of maternal health services in the areas of family planning, prenatal check-ups, delivery care and education of the expectant mother. Data concerning social determinants (unsatisfied basic needs) at the state level for 2010 were obtained from the DNP [28], the DANE, [25] and the Ministry of Health and Social Protection. S1 Appendix.
The response variable was the MMR for the year 2011. Maternal death was defined according to the International Classification of Diseases version 10 (ICD-10) [29]. All maternal deaths in 2011 classified with the following diagnoses were included: pregnancy with abortive outcome (O00-O08), edema, proteinuria and hypertensive disorders in pregnancy, childbirth and the puerperium (O10-O16), other maternal disorders predominantly related to pregnancy (O20-O29), maternal care related to the fetus and amniotic cavity and possible delivery problems (O30-O48), complications of labor and delivery (O60-O75), complications predominantly related to the puerperium (O85-O92), other obstetric conditions not classified elsewhere (O95-O99), and causes specified in other chapters (A34, B20-B24, C58, D392, E230, F530-F539, M830). Excluded were deaths that occurred in another country and cases in which information about the municipality or time of death was missing.
Variables related to coverage and utilization of services: The variables coverage and utilization of services for each state, for the year 2010, included: percentage of deliveries in healthcare institutions, percentage of reproductive-age women who at the time of the ENDS survey used some type of modern birth control, percentage of women who delivered within 5 years of the survey and did not receive prenatal care, percentage of deliveries by a healthcare professional, percentage of pregnant women for whom the cost of prenatal care was not covered due to not being enrolled, percentage of pregnant women for whom the cost of the delivery was not covered due to not being enrolled, percentage of pregnant women for whom the cost of postnatal care was not covered due to not being enrolled, percentage of deliveries by a midwife or other person or family member [27]. Indicators of social determinants were selected based on two criteria: 1) relevance and 2) feasibility. Relevance refers to scientific evidence that supports an association between social determinants of health and maternal health. Feasibility refers to the availability of state- and municipal-level information related to the indicators. Seven indicators were identified: Gini Coefficient, Human Development Index, unsatisfied basic needs, education of the mother, multidimensional poverty index, corruption index and effective access to health services. The corruption index was excluded as an indicator because it did not for meet the relevance criterion; the Gini coefficient, human development index, index of multidimensional poverty index, corruption index and the effective access to health services indicator did not meet the feasibility criterion due to a lack of data for all observation periods. The social variables included the education of the mother and the unsatisfied basic needs index. These two indicators have been evaluated in other regional and economic contexts in terms of their association with maternal health, child health outcomes and the performance of health systems in the areas of delivery and the quality of health services [30–32]. These indicators are generated periodically at the state level. Data from the year 2010 were used for the present study. The education of the mother was defined as the percentage of mothers who finished secondary school, standardized by municipality and state. The unsatisfied basic needs index is a compound indicator composed of indicators related to housing conditions, such as: housing materials, access to public services, extreme overcrowding, economic dependence and access to education. According to the unsatisfied basic needs index, a poor home is one whose basic needs are not met in terms of at least one indicator [25]. This index is constructed using a factor analysis and has been widely used in Colombia.
The units of analysis were the 33 states. The indicator of coverage and utilization of maternal health services was constructed using a polychoric correlation index with the maternal healthcare services variables mentioned earlier. A factor analysis was conducted using the principal component factor extraction method. The resulting indicator was then categorized by quartiles and the states were scored based on these categories. Finally, a path analysis (a type of structural equation model (SEM)) [30–32] was applied in order to evaluate the hypothesis that a relationship exists among social determinants, the indicator of the quality of healthcare services and health outcomes. The present study used a recursive path analysis given the characteristics of the proposed model, in which the health system is a mediating factor between the social determinants of health and health outcomes [33–35]. The main characteristics of the recursive path analysis are that there are no cycles or reciprocal associations among the variables and no correlation with errors related to the dependent variables. The theoretical model for the path analysis is given by: Y=BY+ΓX+ε Where: Y = vector of p x 1 dependent variables observedX = vector of q x 1 independent variables observedB = matrix of p x p coefficients corresponding to YΓ = matrix of p x q corresponding to Xε = vector of p x 1 errors.
Model of relationships among social determinants, health system performance and health outcomes: Fig 1 presents the initial model of the relationships among social determinants, where the healthcare system performance, which in this case refers to the indicator of coverage and utilization of maternal health services, is the mediating variable and the maternal mortality ratio is the final response variable. In this model, the variable used to construct the indicator of coverage and utilization of services was endogenous, and was mainly determined by the characteristics of the 33 states [35]. The path diagram shows that the underlying variables (social determinants) are exogenous. The proposed model is intended to reflect, in the most pragmatic way possible, the proposed conceptual model as well as the existing evidence of the problem under study [36].
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0180857.g001 Model of the relationship among social determinants, health systems and health outcomes. The equation associated with the analytical model is: Y1=γ11X1+γ12X2+ε1 Y2=γ21X1+β21Y1+ε2 The variables included in the model are described below. This group includes the underlying variables that are present in the social structure and are potentially related to the performance of maternal health systems. The study variables were a) percentage of mothers who finished secondary school and b) the unsatisfied basic needs index.
The indicator of the coverage and utilization of maternal healthcare services was constructed using variables that evaluate the coverage and use of healthcare services, from prenatal check-ups to delivery and postpartum check-ups, as described earlier. The final response variable was the MMR. All the data related to the variables were standardized according to the populations in the states. Associations with an alpha of 0.05 were considered statistically significant. The chi-square goodness-of-fit test was performed and the multiple correlation coefficient was estimated to validate the final model. All the analyses were performed using the Stata 13 statistical program (Stata Corporation, College Station, TX, USA).
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}