| Objectives. To investigate the medication taking patterns for patients who have both diabetes and hypertension and to examine the impact of multiple medication adherence on clinical outcomes (Hemoglobin A1C (HbA1C), blood pressure, micro- or macro-vascular complications) using instrumental variables to control for endogeneity bias.;Methods. We identified individuals newly starting oral diabetes or hypertension medication therapy between July 2006 and June 2009 with pre-existing comorbid hypertension or diabetes prescription history during six months of the pre-index period. We used administrative claims from a large physician group in Southern California (N=4,633). Multiple medication adherence was defined as a proportion of days covered for both diabetes and hypertension medications. Micro- or macro-vascular complications included myocardial infarction, stroke, renal failure, and diabetic retinopathy.;Using a six-month time interval, a dynamic panel model was constructed to estimate the direct impact of adherence to diabetes medications on HbA1C and adherence to hypertension medications on systolic and diastolic blood pressures, separately. Available lag dependent and independent variables were used as instrumental variables. To investigate multiple medication adherence on the occurrence of micro- or macro-vascular complications cross-sectionally, a propensity score weighted probit and an instrumental variable probit estimation using physician related variables were implemented for patients eligible during 33 months of post-index (N=2,351).;Results. Mean (SD) adherence was 0.63 (+/-0.01) for diabetes medications, 0.69 (+/-0.01) for hypertension medications, and 0.53 (+/-0.01) for both medications. Adherence to medications for DM and HTN differed when the patient had both conditions at the same time. Patients were more adherent to medications for their pre-existing condition (p<0.001). Patients, who were more clinically severe, were health service seeking, receiving care from the same physician, and receiving care from a physician who prescribed statin more often were more adherent to both disease medications.;After adjusting for state-dependence, time-varying age, hypoglycemia, hospitalization, micro- or macro-vascular complications at time t, we found a 10 percentage point increase in adherence was correlated with a 0.05% decrease in HbA1C, a 0.62 mmHg decrease in systolic blood pressure and a 0.47 mmHg decrease in diastolic blood pressure. The impact was greater compared to the ones from static models.;After adjusting for baseline characteristics such as age, gender, health plan type, clinical measures (HbA1C, blood pressure, and lipid), pre-existing condition, and Elixhauser comorbidity, multiple medication adherence was not significantly associated with decreased vascular complication rate (0.01+/-0.07, p=0.894) based on the probit model. The same patterns were observed after balancing patient's observable characteristics using propensity score weighting methods (0.04+/-0.07, p=0.541). However, after controlling for endogeneity, the impact of multiple medication adherence became statistically significant (-0.94+/-0.26, p<0.001). Instrumental variables satisfied identification conditions. An increase in multiple medication adherence from 50% to 80% reduced the average predicted probability of micro- or macro-vascular complications by 37.4%.;Conclusions. After controlling for observable and unobservable characteristics, multiple medication adherence was associated with the improvement of clinical mediators and overall reduction of micro- or macro-vascular complication rates. The magnitude of the adherence impact was greater after controlling for unobservables, which implies that without controlling for endogeneity, the adherence effect would be underestimated. However, we need to be cautious applying instruments as parameter estimates are unbiased only under the strict assumptions. |
