Working Paper: NBER ID: w15691
Authors: William Encinosa; Didem Bernard; Avi Dor
Abstract: We estimate the impact of diabetic drug adherence on hospitalizations, ER visits, and hospital costs, using insurance claims from MarketScan® employer data. However, it is often difficult to measure the impact of drug adherence on hospitalizations since both adherence and hospitalizations may be correlated with unobservable patient severity. We control for such unobservables using propensity score methods and instrumental variables for adherence such as drug coinsurance levels and direct-to- consumer-advertising. We find a significant bias due to unobservable severity in that patients with more severe health are more apt to comply with medications. Thus, the relationship between adherence and hospitalization will be underestimated if one does not control for unobservable severity. Overall, we find that increasing diabetic drug adherence from 50% to 100% reduced the hospitalization rate by 23.3% (p=0.02) from 15% to 11.5%. ER visits are reduces by 46.2% (p=.04) from 17.3% to 9.3%. While such an increase in adherence increases diabetic drug spending by $776 a year per diabetic, the annual cost savings for averted hospitalizations are $886 per diabetic, a cost offset of $110 (p=0.02), or $1.14 per $1 spent on drugs.
Keywords: Prescription drug adherence; Hospitalizations; Costs; Diabetes; Value-based insurance design
JEL Codes: H51; I11
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Unobservable patient severity (I12) | Biases the relationship between drug adherence and hospitalizations (I11) |
| Failure to control for unobservable patient severity (I12) | Leads to underestimation of the true effects of adherence on hospitalization rates and costs (I13) |
| Increasing diabetic drug adherence from 50% to 100% (I13) | Reduces hospitalization rates by 23.3% (I11) |
| Increasing diabetic drug adherence from 50% to 100% (I13) | Decreases emergency room visits by 46.2% (I11) |
| Increasing diabetic drug adherence from 50% to 100% (I13) | Raises diabetic drug spending by $776 per year per diabetic (H51) |
| Increasing diabetic drug adherence from 50% to 100% (I13) | Results in annual cost savings of $886 per diabetic due to averted hospitalizations (G52) |
| Increasing diabetic drug adherence from 50% to 100% (I13) | Leads to a cost offset of $1.10 for every dollar spent on drugs (H51) |