Work in Progress
Long-term Effects of Growing Up with a Disabled Sibling (with Kevin Kloiber and Felicia Stokke)
Previous literature has shown that growing up alongside a disabled sibling may impose short-term challenges on educational outcomes, as reflected in indicators like 9th-grade GPA, potentially through altered family resource allocation or shifting responsibilities. Yet it remains unclear whether these early setbacks extend into adulthood or instead reflect transient adjustments to a household member’s special needs. Using comprehensive Swedish administrative data, this study addresses this gap by examining both the total exposure effect (across families) and the relative exposure effect (within families) of growing up with a disabled sibling on educational attainment, labor market outcomes, and occupational choices.
To define childhood disability, we adopt a data-driven approach informed by detailed medical records. First, we gather all diagnoses observed among children. For each diagnosis, we determine what fraction of these diagnosed children appear in the government support program for special needs. If this fraction exceeds a certain threshold, we classify that diagnosis as a disability.
For the total exposure effect, we analyze a sample of firstborns whose second sibling is either disabled or not. To isolate the causal effect of growing up with a disabled sibling, we make use of our comprehensive administrative data and employ a double machine learning approach. For the relative exposure effect, we exploit within-family variation by comparing firstborn and second-born siblings in families where the third child is (not) disabled. This allows to incorporate family-fixed effects to control for shared environmental and genetic factors.
Our within‐family estimates reveal that exposure to a disabled sibling leads to a statistically significant but small decline in 9th‐grade GPA but exerts no discernible impact on subsequent labor income or on the likelihood of entering social or medical occupations. By contrast, our across‐family (total exposure) estimates likewise document a negative effect on 9th‐grade GPA and uncover a pronounced reduction in labor income—an effect that is substantially attenuated once social transfers are included—and a very small but statistically significant rise in the probability of choosing a social sector career.
Improving Risk Assessment and Treatment Choice in Medical Decision-Making using Risk Assessment Tools (with Corinna Hartung, Charles F. Manski, Joachim Winter and Amelie Wuppermann).
Physicians frequently encounter situations where they must predict patients’ health outcomes under uncertainty. Existing assessment tools rely on prediction models derived from clinical data but often incorporate only a limited number of known risk factors. In clinical practice, physicians often observe additional risk factors and patient characteristics. Given enhanced access to individual electronic health records and advancements in machine-learning tools for health risk prediction, understanding physicians' interactions with these tools is increasingly crucial.
This study evaluates a new personalized risk assessment tool (PRAT) that combines an evidence-based risk assessment tool, which predicts health outcomes on a subset of observable risk factors, with additional patient characteristics observed by physicians in a structured way. We conduct an experiment with general practitioners and cardiologists to evaluate PRAT's effectiveness in predicting patient-specific 10-year risks of fatal cardiovascular events and guiding treatment decisions based on six risk factors. In our experiment, PRAT incorporates an established evidence-based tool, the Systematic COronary Risk Evaluation (SCORE), which includes only five risk factors.
Our experimental evidence shows that the utilization of SCORE leads to more accurate risk predictions compared to relying solely on clinical experience. Furthermore, the use of the evidence-based tool slightly reduces the probability of medication treatment and significantly decreases the likelihood of general practitioners referring patients to cardiologists. However, PRAT yields small further enhancements in risk assessment.
The Organization of Birth Care and Health Outcomes
Abstract coming soon