Impact of Geographic Variation, Disability, Socioeconomic Status and Risk Adjustment on High-Risk Medication Use among Elderly Medicare Beneficiaries

Abstract

BACKGROUND Inappropriate medication use is common and represents a substantial clinical and economic burden in the United States (US). The Center for Medicare and Medicaid Services (CMS) has adopted one of the Pharmacy Quality Alliance (PQA)’s quality measures to assess percentages of older adult beneficiaries receiving high-risk medications (HRM) in Medicare Advantage Prescription Drug plan and stand-alone Prescription Drug plan. Understanding geographic patterns of HRM use may help CMS and their partners develop and tailor prevention strategies (such as prior authorization) to be implemented in the areas of need. Furthermore, The HRM use measure was used to assess Medicare Advantage Prescription drug plan (MA-PD) and stand-alone Prescription Drug plan (PDP) performance and to provide guidance for practitioners to reduce the use of such medications. Limited evidence exists on how HRM use is associated with patient characteristics and whether risk adjustment is necessary to accurately evaluate health plan performance on the HRM measure. OBJECTIVES The overall objectives of this research were to understand regional and patient characteristics associated with HRM use measure to develop a risk adjustment model for the HRM measure to accurately evaluated health plan performance. The first specific aims were to examined HRM use patterns among Medicare beneficiaries enrolled in: (1) Medicare Advantage prescription drug plans (MA-PDs); and (2) stand-alone prescription drug plans (PDPs) across geographic areas over time in the United States. The second specific aims were to: (a) measure HRM use in MA-PD and PDP beneficiaries with disadvantaged characteristics, including low income and disability; and (b) examine the relationship between disadvantaged characteristics and HRM use given constant effect of health plans. The third specific aims were to examine the relations between patient risk factors and the HRM measure and develop risk adjustment tool for the HRM measure in older adults enrolled in MA-PDs and PDPs. METHODS This cross-sectional study used a 5% national Medicare sample (2011–2013 for the first aims and 2013 for the second and third aims). Among beneficiaries aged ³65 years who were continuously enrolled in MA-PDs or PDPs (~1.3 million each year), we identified those with ≥2 prescriptions for the same HRM (e.g., amitriptyline) during the year based on the HRM list provided by CMS and Pharmacy Quality Alliance. For the first specific aims, multivariable logistic regression was used to estimate adjusted annual HRM use rates across 306 Dartmouth Atlas of Health Care hospital referral regions (HRRs), adjusting for sociodemographic, health-status, and access-to-care factors. For the second aims, Multivariable generalized linear mixed models were used to assess the association of HRM use and disadvantage factors such as low-income subsidy (LIS)/dual eligibility status (DE) and disability after adjusting for health plan effect and patient-level confounding characteristics (i.e., sociodemographic, geographic, clinical complexity). For the third aims, multivariable generalized linear mixed models were used to assess the association of HRM use and patient risk factors (e.g., age, gender) and identify risk factors after adjusting for health plan effect. The identified risk factors were used as variables for regression-based risk adjustment for the HRM measure. Unadjusted and adjusted quality rankings among health plans were compared. RESULTS First, a total of 1,161,076, 1,237,653, and 1,402,861 beneficiaries satisfied the study criteria and were included in 2011, 2012, and 2013, respectively. Among our study sample, nearly 40% (39%, 39% and 37% in 2011, 2012 and 2013 respectively) were enrolled in MA-PD plans, whereas remaining 60% (61%, 61%, and 63% in 2011, 2012 and 2013 respectively) were enrolled in PDP plans. HRM use significantly decreased over time among Medicare beneficiaries enrolled in MA-PD (13.1% to 8.4%, p<0.001) and PDP (16.2% to 12.2%, p<0.001) plans. Among MA-PD beneficiaries, HRM users more frequently: female (70.4% vs. 59.9%, p<0.001); White (84.6% vs. 81.4%, p < 0.001); eligible for the Part D Low Income Subsidy or Medicaid benefits (22.3% vs. 16.6%, p<0.001); and disabled (15.6% vs 8.7%, p<0.001) compared to non-HRM users in 2013. Among PDP beneficiaries, HRM users had higher proportions of: females (72.8% vs. 62.5%, p < 0.001); Whites (86.6% vs. 85.3%, p<0.001); LIS/DEs (29.2% vs. 23.3%, p<0.001); and disabled people (15.4% vs 8.5%, p<0.001) compared to non-HRM users. In 2013, the ratios of 75th-to-25th percentile HRM use rates across HRRs were 1.42 (MAPDs) and 1.31 (PDPs). HRRs with the highest HRM use rates were: Casper, WY (20.4%), Waco, TX (16.7%), Lubbock, TX (15.7%), Santa Barbara, CA (15.2%), and Temple, TX (15.1%) (MA-PDs); and Lawton, OK (18.8%), Alexandria, LA (18.8%), Lake Charles, LA (18.6%), Oklahoma City, OK (18.0%), and Slidell, LA (18.0%) (PDPs). Second, there were a total of 520,019 MA-PD and 881,264 PDP beneficiaries who met the study criteria. Of the MA-PD beneficiaries, 88,693 (17.1%) were LIS/DE and 48,997 (9.4%) were disabled. Of PDP beneficiaries, 213,096 (24.2%) were LIS/DE, and 83,593(9.5%) were disabled. LIS/DE beneficiaries had a higher percent of HRM users compared to non-LIS/DE MA-PD (17.0% vs. 9.6%, p < 0.001) and PDP (17.1% vs. 13.2%, p < 0.001) beneficiaries. Disabled beneficiaries had a higher percent of HRM users compared to non-LIS/DE MA-PD (17.0% vs. 9.6%, p < 0.001)) and PDP (17.0% vs. 9.6%, p < 0.001) beneficiaries. Multivariable analyses showed LIS/DE (OR = 1.07; 95% CI: 1.04, 1.10) and disability (OR =1.38; 95% CI: 1.34, 1.42) were associated with HRM among the MA-PD population as well as in the PDP population (LIS/DE OR = 1.14; 95% CI: 1.12, 1.16 and disability OR = 1.37; 955 CI: 1.34, 1.40). Third, the HRM users were more likely to be younger (OR = 0.981, 95% CI, 0.980-0.983 for MA-PD and OR=0.982, 95% CI, 0.981-0.983 for PDP); women (OR = 1.545; 95% CI,1.514-1.576 for MA-PD and OR=1.606, 95% CI, 1.584-1.628); eligible to receive low-income subsidy (OR = 1.086, 95%CI, 1.057-1.115 for MA-PD and 1.170, 95% CI, 1.150–1.190 for PDP); disabled (OR = 1.380, 95%CI, 1.342 –1.420 for MA-PD and 1.378, 95%CI, 1.352–1.405 for PDP); seeing multiple prescibers (OR =1.076, 95%CI, 1.072, 1.081 for MA-PD and 1.072, 95%CI, 1.069-1.075); filling prescriptions at multiple pharmacies (OR = 1.092, 95%CI, 1.083-1.102 for MA-PD and OR = 1.092, 95%CI, 1.086, 1.099 for PDP); and had higher average modified RxRisk-V (OR = 1.176, 95%CI ,1.171 – 1.181 for MA-PD and OR = 1.173, 95%CI, 1.170-1.176 for PDP). Being older and white were protective against receipt of HRMs. These variables were recommended for the risk adjustment model. Unadjusted scores showed low levels of agreement (Cohen’s kappa < 0.7) with risk-adjusted scores in identifying statistical outliers suggesting risk adjustment is necessary. CONCLUSION Geographic variation in HRM use exists among older adults in Medicare, regardless of prescription drug plans. Areas with high HRM rates may benefit from targeted interventions to prevent potential adverse consequences. LIS/DE; disability; demographic such as age, gender, race; and clinical complexity were associated with higher HRM use in both the MA-PD and PDP populations even when controlling for health plan effects. Failure to adjust for beneficiaries case mix might penalize some truly high-quality MA-PD and PDP providers that serve sick beneficiaries or beneficiaries with poor socioeconomic conditions.