Preventive care: How comorbidities affect screening rates.

摘要

Many studies have concluded that beneficial preventive care is under-utilized. However, little is known about how utilization varies across different comorbid groups. Comorbid patients could have lower utilization rates because they have a lower life expectancy, which decreases the gains from long-term prevention. Alternatively, comorbid populations could place greater value on prevention because existing comorbidities often complicate the treatment of new conditions. Additionally, comorbid patients are more likely to interact with physicians, which makes them more likely to be offered (and receive) preventive services. Overall, it is not obvious whether the screening rates should be expected to be higher or lower than that of the healthy population.;This dissertation proposes a theoretical model that allows for variation in both marginal benefits and marginal costs of screening across comorbid groups. By allowing different comorbid populations to have different patterns of physician utilization and, therefore, different marginal costs of screening, the model predicts that screening rates will vary in ways that cannot be explained by variation in life expectancy and direct screening costs alone. Specifically, those comorbid groups which have more physician visits will also have higher screening rates, while those groups with fewer visits will have lower screening rates.;These predictions are tested with data from the 2001-2009 Medical Expenditure Panel Survey. First, the baseline results show that PSA and mammography rates vary significantly across different comorbid groups. Specifically, there are significantly higher screening rates among those with generalist-managed comorbidities (e.g. high blood pressure, high cholesterol), but lower screening rates among those who require more specialist care (e.g. previous stroke, heart attack). Since these comorbidities have similar negative effects on life expectancy, traditional economic models would have predicted that these conditions would also have similar effects on screening rates.;This dissertation explores the possibility that differences in physician visit patterns could be an intervening factor. Count models of annual physician visits confirm that there are differences in physician utilization rates across comorbid groups. Generalist-managed conditions are correlated with more physician visits on all margins while specialist-managed conditions had no statistically significant correlation with visits. The association between physician visits and screening rates is difficult to empirically control for because these variables are inherently simultaneously determined: physician visits increase the likelihood of cancer screening but screening tests necessarily require physician visits. This simultaneity would likely result in a positive bias of the estimated correlation between physician visits and screening utilization. Furthermore, this simultaneity bias would have second-order effects on other covariates that are correlated with physician visits. The empirical results show that controlling for physician utilization halved the estimates for simple, generalist-managed conditions, but there was no change in the estimates for specialist-managed conditions. Like what was found in the count models, this also suggests that these comorbid groups might have different patterns of physician utilization. Overall, these results suggest that patients with generalist-managed conditions have higher screening rates, partly due to increased physician visits, but patients with specialist-managed conditions have lower screening rates, with minor contribution from weakly fewer physician visits.;It is possible that these results are biased by individual heterogeneity. One way to eliminate all time-invariant endogeneity is by including individual fixed effects. However, the short panel contains little individual-level variation, so there is no statistical significance. The results suggest that PSA testing rates are minimally affected by newly diagnosed comorbidities but that mammography rates are affected differently by different comorbidities. Physician utilization controls did not change the results from these fixed effect models.;Overall, the disparities in screening rates between different comorbid populations suggest that efforts to improve health and/or contain cost by increasing preventive care will need to consider the varying determinants of demand for screening across different comorbid populations.