Research project The Impact of Early Interventions on Infant and Maternal Health: New Data, New Methods, New Findings
This research program aims to increase understanding of how health and interventions early in life affect education and health later in life using new data and new analytical methods.
In this research program the researchers aim to use new data and new methods to shed new light on the determinants and consequences of infant and maternal health, as well as documenting the short and long-run effects of interventions designed to improve infant and maternal health around the time of birth, and how such interventions help mitigate inequalities in health at birth and human capital more broadly.
To be able to shed new light on the role of infant and maternal health two new data sources that has not previously been explored by economists will be used. Since 2002, detailed information on all infants that receive care in the neonatal period has been gathered in the SNQ (Neonatalregistret). The SNQ data provides unprecedented detailed information on not only the interventions/medications that the neonate receives, but also, hard data on the determinants of why they are assigned to specific treatments. The second part of the research program focusing on the mothers health will instead explore another new data set, the PQR (Graviditetsregistret). The PQR covers all pregnancies in Sweden since 2014, and supersedes MBR by covering all pregnancies in Sweden, and not only live births. This is important innovation since a significant share of pregnancies are not carried to term. In addition, the PQR contains much more information about the mother, pregnancy, infant and the treatments received in gestation and afterwards. The plan is to link both of these (for economists) new data sources and link them to the rich and more standard administrative registers collected by SCB and National Board of Health and Welfare.
The research program will go beyond documenting causal effects of specific health shocks in the long-run, by also documenting effects of interventions designed to mitigate early health problems, as well as investigating how maternal health and interventions targeting mothers affects their health and labor market outcomes and in turn how such interventions affects the children in the long-run. Here the teams expertise in labor and health economics, in combination with the expertise of affiliated health professionals, will allow the program to add significantly to what is already known in the medical literature regarding the efficacy of specific treatments, by extending them to outcomes not considered by the medical profession.
Given the space limitations the researchers provide two specific questions that can be addressed with the new data.
1. The long-run health benefits and costs of health screening for, and treatment of, jaundice.
One third of all newborns in Sweden gets jaundice (icterus), and over 80 percent of those born prematurely are affected by jaundice. Since severe jaundice may damage the developing brain all newborn babies in Sweden are screened within the first few days after birth. Neonate jaundice is primarily treated using phototherapy. Assignment to treatment is based on the serum bilirubin blood level. If the neonate has a bilirubin level above 350 she is treated with phototherapy.
While the benefits of the phototherapy interventions on bilirubin level, clinical jaundice and other more immediate health outcomes is well-documented, we have very little information (i) on how it affects longterm human capital outcomes (e.g. school grades, college completion etc.), (ii) whether there are unintended health consequences of treatment in the long-run, and (iii) if treating infants with slightly lower levels of bilirubin levels (i.e. inframarginal infants) could benefit their long-term outcomes.
The SNQ data contains information on the observed bilirubin level and whether the neonate received treatment. The SNQ data can be used asses these questions by linking to other health and human capital outcomes data from SCB and Socialstyrelsen. Specifically, the known assignment cut offs will enables us to use a regression discontinuity design (RDD) to elicit the causal effect of being treated in the short and long-run. The idea with the RDD is to compare individuals who are just below and just above the treatment threshold. For example, children with a bilirubin level of 349 can be compared to children with a bilirubin level of 350. Because the exact threshold of 350 is set arbitrarily, we can compare the outcome of neonates with a very small difference in actual bilirubin level, where one are more likely to receive the treatment and the other are not. Given the small difference in bilirubin, the chances that the infant end up on either side of the cut-off will be as good as random. If the current follow-up and treatment criteria are set so that treatment is initiated before any negative health effects of jaundice occur, we do not expect a treatment effect of the treatment measures in place. In that case, children who are just over the limit and receiving treatment do not have better or worse outcomes than those children who are below the threshold and do not receive treatment.
However, it is possible that children who are just above the threshold and take part in the treatment will do better over the life cycle. This would indicate that health effects of high bilirubin levels occur at lower levels than the current threshold. Such health effects could – possibly – be detected only by studying broader outcomes later in life. The RDD allows the researchers to estimate whether the treatment actually may have negative effects; if the children who do not receive treatment fare better later in life compared to the treated children, this indicates that the threshold should be set higher or another treatment should be used, using methods developed by Angrist and Rokkanen (2015) can further assess the value of treatment away from the cut-off. For example, it is also possible that there is an overtreatment for jaundice among newborns where the threshold levels for treatments are set too low. In fact, light-treatment could, in theory, affect the DNA with the potential of increasing the risk for cancer later in life. IThe study will be able to assess these impacts as well.
2. Maternal health and post-partum health interventions
The second part of the research focuses on the health of the mother around the time of birth, birth complication and it impact on primarily the mother, but we also aim to test for spill-over effects on the child. Giving birth to a child can be a large shock to the mother’s health, and can have substantial long-term effects on the mother’s physical, mental and sexual health (SBU, 2021). There is an emerging literature on maternal morbidity and labor market outcomes. Most existing evidence from Sweden focuses on sickness benefit and find small effect of giving birth on health outcomes (Angelov et. al 2016, Fransson et. al 2021). However, a limitation of using sickness benefits as a measure of ill-health is that many of the health problems that women experience after giving birth do not qualify them for sickness benefit. In this part of the project, the researchers will start by documenting the effect of severe birth tears on subsequent outcomes for women in an event study design where we compare outcomes for women who experience a severe birth tear to their sister. To further address the problem of unobservable determinants of the risk of a severe birth tear, they are planning to exploit information from the PQR data that gives detailed information on which midwife/doctor that was in charge at the time of delivery. This, deidentified, information can be used in a fixed effects framework where the risk of tears are instrumented with the fixed effect of the individual midwife, estimated over the history of the midwifes recorded births. The analysis will provide information on not only of how mothers’ birth experiences shape subsequent health and labor market outcomes, but also if family stability and the child’s outcomes are affected by investments in maternal health around the time of birth.