For a start, reducing child mortality is one of the MDGs (Millennium Development Goals). But, did you know that every year about 6.9 million children die around the world from mostly preventable causes? Among these, 40% represent neonatal (under 28 days old) children. To start looking at the problem from a closer perspective I decided to focus my attention in my own country, Mexico. Although the amount of child deaths within the country represent only about 2% of the total around the globe, the actual proportion of deaths in newborn children, 65%, is higher.
Next, I am sharing a study that I recently did based on data collected from here and here, which represent social gaps of different counties and (neonatal) child mortality rates given by the Mexican government. The granularity of the two files I chose is at the county level. The merged file was about 7.6 MB including 89 variables and 23,040 rows before filtering. My initial goal was to understand two simple questions:
– Do the social gaps of smaller villages can explain children mortality?
– What are specific factors to which we can attribute children mortality?
Following are two plots that I was able to obtain after performing a preliminary analysis using R. In my analysis I considered the correlation of 19 variables with respect to neonatal mortality rates for all the counties in Mexico with information available. The result were some interesting correlations with respect to a few variables (Pearson corr included in parentheses): medical attention during delivery (13%) , literacy level of the mother (31%), illiteracy rate (80%), social gap index (88%) and total population of the county (where the child’s death was registered and where the mother is assumed to reside, – 50%). When performing an Analysis Of Variance (AOV), medical attention during delivery, literacy level of the mother and illiteracy rate of the county resulted statistically significant. The following two multivariate plots show the relationship of two of such variables with respect to neonatal mortality rates at the county and state level (county is given by dots and states by the dot colors).
The figure on top confirms the positive relationship between high social gaps and mortality rates among new borns. However, there is a widening spanning range in mortality rates. That is, as the social gap increases (say above 0) we observe mortality rates ranging between about 20% and 50%, particularly for some states (e.g. color purple). This seems to indicate that there is something more than the social gap affecting neonatal mortality rates. The second graph shows the quality of medical attention during delivery (the highest being the best), which is related to medical infrastructure. Here, it is clear that having the highest quality of medical attention translates into mortality rates at or below 40%. Although this second variable explains further the causes of mortality rates in newborns, there are other factors missing. So, we considered variables as prenatal attention and number of medical visits during pregnancy which showed no significant impact. In summary, the few clues obtained so far pinpoint mainly to social gaps and educational factors as important descriptors of variation in neonatal mortality rates.
However, specific death causes were not considered in this study. Some studies have shown that surgical procedures can represent up to 26% of the causes of children deaths. This is a new variable directly controllable by health clinics that could lead to the better allocation of health related resources. Other variables related to specific death causes are the types of services and care that the woman receives (and/or provides to herself) during pregnancy. Given her educational level, we would be in a better position to understand how well informed she is about pregnancy risks at the time of giving birth.
Unfortunately, a big disadvantage regarding the type of data necessary to provide targeted solutions to this problem of global dimensions is the level of granularity available. The data available through public databases usually does not provide many details about specific needs not met at the local or personal level. Here, I believe, there is an important area of opportunity for mobile technologies. Collection of data through mobile devises can be cheaper than collecting information through traditional means. Moreover, it is possible that some of the data that can help explain the problems faced even in the remotest villages is already available. For instance, in Mexico penetration rates of mobile devises right now span 65% of the population and are expected to increase to 75% within three years. Geolocation data, which is already part of smartphone technology, can help understand the type of care received by mothers during pregnacy by indicating, for instance, the number of times the subject visited a given health care facility. This would provide a more accurate representation of what services were actually received before delivery that contributed to a child’s death. In addition to visits to health centers, information about the overall pattern of behavior could be collected based on the mother’s location during this period.
Well, these are just a few thoughts about potential solutions and questions we could ask when analyzing child mortality. Please share your thoughts!
(Interested in learning more reasons to start thinking about children mortality? See this revealing graph included in a recent Bill Gates article covering the topic.)