Doctors, Documents, and Diagnostic Disparities: Essays on the Economics of Mental Health
MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractMental health conditions are costly, both to the impacted individual and society as a whole. However, due to the subjectivity of mental health medical guidelines, these conditions are associated with high potential for diagnostic errors. By combining clinical data with quantitative economic modeling, I explore how individuals are diagnosed with mental health conditions in the presence of such uncertainty. My dissertation seeks to understand mechanisms underlying the clinical diagnostic process and sheds light on sources of mental health diagnostic errors and disparities. In the first chapter, I develop and estimate a structural model of diagnosis for the most prevalent child mental health condition, Attention Deficit Hyperactivity Disorder (ADHD). The model incorporates both patient and physician influences, highlighting four key mechanisms of mental health diagnosis: true underlying prevalence, patient stigma, diagnostic uncertainty, and physician costs from type I and type II diagnostic errors. I estimate sex-specific structural model parameters using novel electronic health record data on doctors’ notes together with machine learning and natural language processing techniques (later detailed in Chapter 2). In raw comparisons, males are 2.3 times more likely to be diagnosed with ADHD than females. Counterfactual simulations using model estimates show that less than one-half of this disparity can be explained by true differences in underlying ADHD prevalence, very little explained by patient preferences, and about 50% attributed to differences in physician decision-making. I show that physicians view missed diagnosis to be costlier than misdiagnosis, especially for their male patients. Back of the envelope calculations estimate the national economic impact of ADHD diagnostic errors to be $60-$117 billion US dollars, suggesting a need for public and/or medical policy responses aimed at increasing diagnostic accuracy and reducing disparities in mental health care. The second chapter further details the natural language processing algorithm used in Chapter 1 and provides another application for its use– estimating physician practice style. I first propose the text-analysis algorithm applied to digitized clinical doctor notes as a way to measure how closely the patient interview matches mental health diagnostic guidelines. This measure can then be used as a control in a reduced-form model to identify two components of physician practice style: diagnostic intensity (the mean propensity to diagnose) and diagnostic compliance (the weight that physicians place on official medical guidelines). As an application, I estimate and analyze physician practice style for ADHD diagnosis. I find significant variation in physician practice style, with physician gender and experience being the strongest predictors of this variation. I also discuss how mental health practice style estimates can be used to guide potential health care policies, and I provide a list of extensions and suggestions of how these methods can be used in future mental health care research. In the final chapter, I switch the focus away from physicians and onto the family. In this study, I use 16 waves of the Medical Expenditure Panel Survey to empirically explore the relationship between family structure shocks and child mental health, focusing on clinical diagnoses. Comparing cross-sectional estimates to within-family models, I show that over half of the positive relationship between family stability and child mental health can be explained by unobserved family selection. Using a child fixed-effect estimation approach, I find that the probability a child is diagnosed with a mental health condition increases by 3.5 percentage points following a parental divorce or parental death. I explore potential mechanisms, noting an increase in both behavioral symptoms and mental health care utilization. However, there is no significant change in overall health care, suggesting that due to time constraints, parents may be substituting general/physical health appointments for mental health related services. While the first two chapters highlight how physicians influence the diagnosis decision, this final chapter demonstrates that family structure and utilization of care are also significant determinants of child mental health.
Degree ProgramGraduate College