Publications

Short Physical Performance Battery (SPPB) assessment is a widely used measure of lower extremity function, strength, and balance. In the Lifestyles Interventions and Independence for Elders (LIFE) Study, baseline SPPB and changes throughout the trial were strongly associated with major mobility disability (MMD). This study further investigated this association by identifying trajectories of SPPB and evaluating the predictive validity of SPPB trajectories for future MMD. Participants (n = 1635) aged 70-89 years were randomized to a physical activity or health education intervention and assessed every 6 months for MMD. We used group-based trajectory models (GBTMs) to identify trajectories of a binary outcome for a decrease from baseline SPPB of ≥1. Multinomial logistic regression explored baseline factors associated with group membership. Survival analyses evaluated the association between trajectories with MMD. The GBTM identified a 3-group model which included a "No Decline" group (46.0%), "Late Decline" group (27.7%), and an "Early Decline" group (26.3%). Adjusting for all other baseline characteristics, group assignment during the previous follow-up visit was strongly associated with MMD at the subsequent period. Comparisons between groups showed a 2-to-3-fold increase in MMD comparing the "Late" to "No" decline group and a 4-to-5-fold increase in MMD comparing the "Early" to "No" decline group. Group membership and impact on MMD was not different between intervention arms. Group-based trajectories of SPPB scores identified distinct subgroups in LIFE Study participants. Using these group assignments in outcome models were highly associated with MMD. GBTMs have potential to identify and improve prediction of aging-related decline to better design and identify patients for interventions.

OBJECTIVE: To develop and validate a machine-learning algorithm to improve prediction of incident OUD diagnosis among Medicare beneficiaries with ≥1 opioid prescriptions.

METHODS: This prognostic study included 361,527 fee-for-service Medicare beneficiaries, without cancer, filling ≥1 opioid prescriptions from 2011-2016. We randomly divided beneficiaries into training, testing, and validation samples. We measured 269 potential predictors including socio-demographics, health status, patterns of opioid use, and provider-level and regional-level factors in 3-month periods, starting from three months before initiating opioids until development of OUD, loss of follow-up or end of 2016. The primary outcome was a recorded OUD diagnosis or initiating methadone or buprenorphine for OUD as proxy of incident OUD. We applied elastic net, random forests, gradient boosting machine, and deep neural network to predict OUD in the subsequent three months. We assessed prediction performance using C-statistics and other metrics (e.g., number needed to evaluate to identify an individual with OUD [NNE]). Beneficiaries were stratified into subgroups by risk-score decile.

RESULTS: The training (n = 120,474), testing (n = 120,556), and validation (n = 120,497) samples had similar characteristics (age ≥65 years = 81.1%; female = 61.3%; white = 83.5%; with disability eligibility = 25.5%; 1.5% had incident OUD). In the validation sample, the four approaches had similar prediction performances (C-statistic ranged from 0.874 to 0.882); elastic net required the fewest predictors (n = 48). Using the elastic net algorithm, individuals in the top decile of risk (15.8% [n = 19,047] of validation cohort) had a positive predictive value of 0.96%, negative predictive value of 99.7%, and NNE of 104. Nearly 70% of individuals with incident OUD were in the top two deciles (n = 37,078), having highest incident OUD (36 to 301 per 10,000 beneficiaries). Individuals in the bottom eight deciles (n = 83,419) had minimal incident OUD (3 to 28 per 10,000).

CONCLUSIONS: Machine-learning algorithms improve risk prediction and risk stratification of incident OUD in Medicare beneficiaries.

BACKGROUND: Understanding geographic patterns of high-risk medication (HRM) prescribed and dispensed among older adults may help the Centers for Medicare & Medicaid Services and their partners develop and tailor prevention strategies.

OBJECTIVE: To compare the geographic variation in the prevalence of HRM use among Medicare Part D beneficiaries from 2011 to 2013, for Medicare Advantage Prescription Drug (MA-PD) plans and stand-alone Prescription Drug Plans (PDPs).

METHODS: This retrospective study used the data of a 5% national Medicare sample (2011-2013). Beneficiaries were included in the study if they were aged ≥ 65 years, continuously enrolled in MA-PDs or PDPs ( 1.3 million each year), and had ≥ 2 prescriptions for the same HRM (e.g., amitriptyline) prescribed and dispensed during the year based on the Pharmacy Quality Alliance's (PQA) quality measures for HRM use. Multivariable logistic regression was used to estimate adjusted annual HRM use rates (i.e., adjusted predictions, average marginal predictions, or model-adjusted risk) across 306 Dartmouth Atlas of Health Care hospital referral regions (HRRs), controlling for sociodemographic, health-status, and access-to-care factors.

RESULTS: Among eligible beneficiaries each year (1,161,076 in 2011, 1,237,653 in 2012, and 1,402,861 in 2013), nearly 40% were enrolled in MA-PD plans, whereas the remaining 60% were in PDP plans. The adjusted prevalence of HRM use significantly decreased among Medicare beneficiaries enrolled in MA-PD (13.1%-8.4%, P < 0.001) and PDP (16.2%-12.2%, P < 0.001) plans from 2011 to 2013. For MA-PD and PDP beneficiaries, HRM users were more likely to be (all P < 0.001) the following: female (MA-PD: 70.4% vs. 59.9%; PDP: 72.8% vs. 62.5%); White (MA-PD: 84.6% vs. 81.4%; PDP: 86.6% vs. 85.3%); with low-income subsidy or dual eligibility for Medicaid (MA-PD: 22.3% vs. 16.6%; PDP: 29.2% vs. 23.3%); and disabled (MA-PD: 15.6% vs. 8.7%; PDP: 15.4% vs. 8.5%) compared with non-HRM users in 2013. In 2013, significant geographic variation existed, with the ratios of 75th-25th percentiles of HRM use rates across HRRs as 1.42 for MA-PDs and 1.31 for PDPs. For MA-PDs, the top 5 HRRs with the highest HRM use rates in 2013 were Casper, WY (20.4%), Waco, TX (16.7%), Lubbock, TX (15.7%), Santa Barbara, CA (15.2%), and Temple, TX (15.1%); for PDPs, they were Lawton, OK (18.8%), Alexandria, LA (18.8%), Lake Charles, LA (18.6%), Oklahoma City, OK (18.0%), and Slidell, LA (18.0%).

CONCLUSIONS: Substantial geographic variation exists in the prevalence of HRM use among older adults in Medicare, regardless of prescription drug plan. Areas with high prevalence of HRM use may benefit from targeted interventions (e.g., medication therapy management monitoring or alternative medication substitutions) to prevent potential adverse consequences.

DISCLOSURES: No outside funding supported this study. The authors have nothing to disclose. This study was presented as a poster at the International Society of Pharmacoeconomics and Outcomes Research (ISPOR) Asia Pacific Meeting; September 8-11, 2018; Tokyo, Japan.

OBJECTIVE: To examine associations between opioid prescriber specialty and patient likelihood of opioid use disorder (OUD), opioid misuse, and opioid overdose.

DESIGN: Longitudinal retrospective study using Pennsylvania Medicaid data (2007-2015).

METHODS: We constructed an incident cohort of 432,110 enrollees initiating prescription opioid use without a history of OUD or overdose six months before opioid initiation. We attributed patients to one of 10 specialties using the first opioid prescriber's specialty or, alternatively, the specialty of the dominant prescriber writing the majority of the patient's opioid prescriptions. We estimated adjusted rates for OUD, misuse, and overdose, adjusting for demographic variables and medical (including pain) and psychiatric comorbidities.

RESULTS: The unadjusted incidence rates of OUD, misuse, and overdose were 7.13, 4.73, and 0.69 per 100,000 person-days, respectively. Patients initiating a new episode of opioid treatment with Pain Medicine/Anesthesiology (6.7 events, 95% confidence interval [CI] = 5.5 to 8.2) or Physical Medicine and Rehabilitation (PM&R; 6.1 events, 95% CI = 5.1 to 7.2) had higher adjusted rates for OUD per 100,000 person-days compared with Primary Care practitioners (PCPs; 4.4 events, 95% CI = 4.1 to 4.7). Patients with index prescriptions from Pain Medicine/Anesthesiology (15.9 events, 95% CI = 13.2 to 19.3) or PM&R (15.8 events, 95% CI = 13.5 to 18.4) had higher adjusted rates for misuse per 100,000 person-days compared with PCPs (9.6 events, 95% CI = 8.8 to 10.6). Findings were largely similar when patients were attributed to specialty based on dominant prescriber.

CONCLUSIONS: Differences in opioid-related risks by specialty of opioid prescriber may arise from differences in patient risk factors, provider behavior, or both. Our findings inform targeting of opioid risk mitigation strategies to specific practitioner specialties.

INTRODUCTION: This study aimed to examine the association of race and ethnicity on the risk of lower extremity amputations among Medicare beneficiaries with diabetic foot ulcers (DFUs) and diabetic foot infections (DFIs).

RESEARCH DESIGN AND METHODS: A retrospective study included 2011-2015 data of a 5% sample of fee-for-service Medicare beneficiaries with a newly diagnosed DFU and/or DFI. The primary outcome was the time to the first major amputation episode after a DFU and/or DFI were identified using the diagnosis and procedure codes. We used multivariable Cox proportional hazards models to estimate the risk of time to the first major amputation across races, adjusting for sociodemographic and health status factors. Adjusted hazard ratios (aHRs) with a 95% CI were reported.

RESULTS: Among 92 929 Medicare beneficiaries newly diagnosed with DFUs and/or DFIs, 77% were whites, 14.3% African Americans (AAs), 3.3% Hispanics, 0.7% Native Americans (NAs), and 4.0% were other races. The incidence rates of major amputation were 0.02 person-years for NAs, 0.02 person-years for AAs, 0.01 person-years for Hispanics, 0.01 person-years for other races, and 0.01 person-years for whites (p<0.05). Multivariable analysis showed that AAs (aHR=1.9, 95% CI 1.7 to 2.2, p<0.0001) and NAs (aHR=1.8, 95% CI 1.3 to 2.6, p=0.001) were associated with an increased risk of major amputation compared with whites. Beneficiaries with DFUs and/or DFIs diagnosed by a podiatrist or primary care physician (aHR=0.7, 95% CI 0.6 to 0.8, p<0.0001, specialists as reference) or at an outpatient visit (aHR=0.3, 95% CI 0.3 to 0.3, p<0.0001, inpatient stay as reference) were associated with a decreased risk of major amputation.

CONCLUSIONS: Racial and ethnic disparities in the risk of lower extremity amputations appear to exist among fee-for-service Medicare beneficiaries with diabetic foot problems. AAs and NAs with DFUs and/or DFIs were associated with an increased risk of major amputations compared with white Medicare beneficiaries.

Because they contain detailed individual-level data on various patient characteristics including their medical conditions and treatment histories, electronic health record (EHR) systems have been widely adopted as an efficient source for health research. Compared to data from a single health system, real-world data (RWD) from multiple clinical sites provide a larger and more generalizable population for accurate estimation, leading to better decision making for health care. However, due to concerns over protecting patient privacy, it is challenging to share individual patient-level data across sites in practice. To tackle this issue, many distributed algorithms have been developed to transfer summary-level statistics to derive accurate estimates. Nevertheless, many of these algorithms require multiple rounds of communication to exchange intermediate results across different sites. Among them, the One-shot Distributed Algorithm for Logistic regression (termed ODAL) was developed to reduce communication overhead while protecting patient privacy. In this paper, we applied the ODAL algorithm to RWD from a large clinical data research network-the OneFlorida Clinical Research Consortium and estimated the associations between risk factors and the diagnosis of opioid use disorder (OUD) among individuals who received at least one opioid prescription. The ODAL algorithm provided consistent findings of the associated risk factors and yielded better estimates than meta-analysis.

OBJECTIVES: To examine health outcomes associated with adherence to Healthcare Effectiveness Data and Information Set (HEDIS) antidepressant medication management (AMM) during acute and continuation phases of depression treatment among older adults with dementia and major depressive disorder (MDD).

DESIGN: Retrospective cohort study.

SETTING: Medicare 5% sample data (2011-2013).

PARTICIPANTS: Older adults (aged 65 years or older) with dementia and MDD.

MEASUREMENTS: The first antidepressant prescription claim from 1 May 2011 through 30 April 2012 was considered the index prescription start date (IPSD). Adherence during acute- and continuation-phase AMM was based on HEDIS guidelines. Study outcomes included all-cause mortality, all-cause hospitalization, and falls/factures (with mortality being the competing event for hospitalization and falls/fractures) during follow-up from end of acute-/continuation-phase AMM adherence. Due to the proportionality assumption violation of Cox models, fully non-parametric approaches (Kaplan-Meier and modified Gray's test) were used for time-to-event analysis adjusting for the inverse probability of treatment weights.

RESULTS: Final study samples consisted of 4330 (adherent (N) = 3114 (71.92%)) and 3941 (adherent (N) = 2407 (61.08%)) older adults with dementia and MDD during acute- and continuation-phase treatments, respectively. No significant difference (p > 0.05) between adherent and non-adherent groups was observed for all-cause mortality and falls/fractures in both the acute and continuation phases. There was a significant difference in time to all-cause hospitalization during acute-phase treatment (p = 0.018), with median times of 530 (95% CI: 499-587) and 425 (95% CI: 364-492) days for adherent and non-adherent groups, respectively.

CONCLUSIONS: Acute-phase adherence to HEDIS AMM was associated with reductions in all-cause hospitalization risk among older adults with dementia and MDD.

BACKGROUND/OBJECTIVES: Guideline-based management of cardiovascular disease often involves prescribing multiple medications, which contributes to polypharmacy and risk for adverse drug events in older adults. Deprescribing is a potential strategy to mitigate these risks. We sought to characterize and compare clinician perspectives regarding deprescribing cardiovascular medications across three specialties.

DESIGN: National cross-sectional survey.

SETTING: Ambulatory.

PARTICIPANTS: Random sample of geriatricians, general internists, and cardiologists from the American College of Physicians.

MEASUREMENTS: Electronic survey assessing clinical practice of deprescribing cardiovascular medications, reasons and barriers to deprescribing, and choice of medications to deprescribe in hypothetical clinical cases.

RESULTS: In each specialty, 750 physicians were surveyed, with a response rate of 26% for geriatricians, 26% for general internists, and 12% for cardiologists. Over 80% of respondents within each specialty reported that they had recently considered deprescribing a cardiovascular medication. Adverse drug reactions were the most common reason for deprescribing for all specialties. Geriatricians also commonly reported deprescribing in the setting of limited life expectancy. Barriers to deprescribing were shared across specialties and included concerns about interfering with other physicians' treatment plans and patient reluctance. In hypothetical cases, over 90% of physicians in each specialty chose to deprescribe when patients experienced adverse drug reactions. Geriatricians were most likely and cardiologists were least likely to consider deprescribing cardiovascular medications in cases of limited life expectancy (all P < .001), such as recurrent metastatic cancer (84% of geriatricians, 68% of general internists, and 45% of cardiologists), Alzheimer dementia (92% of geriatricians, 81% of general internists, and 59% of cardiologists), or significant functional impairment (83% of geriatricians, 68% of general internists, and 45% of cardiologists).

CONCLUSIONS: While barriers to deprescribing cardiovascular medications are shared across specialties, reasons for deprescribing, especially in the setting of limited life expectancy, varied. Implementing deprescribing will require improved processes for both physician-physician and physician-patient communication. J Am Geriatr Soc 68:78-86, 2019.

IMPORTANCE: Elevated blood glucose levels are common in hospitalized older adults and may lead clinicians to intensify outpatient diabetes medications at discharge, risking potential overtreatment when patients return home.

OBJECTIVE: To assess how often hospitalized older adults are discharged with intensified diabetes medications and the likelihood of benefit associated with these intensifications.

DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study examined patients aged 65 years and older with diabetes not previously requiring insulin. The study included patients who were hospitalized in a Veterans Health Administration hospital for common medical conditions between 2011 and 2013.

MAIN OUTCOMES AND MEASURES: Intensification of outpatient diabetes medications, defined as receiving a new or higher-dose medication at discharge than was being taken prior to hospitalization. Mixed-effect logistic regression models were used to control for patient and hospitalization characteristics.

RESULTS: Of 16 178 patients (mean [SD] age, 73 [8] years; 15 895 [98%] men), 8535 (53%) had a preadmission hemoglobin A1c (HbA1c) level less than 7.0%, and 1044 (6%) had an HbA1c level greater than 9.0%. Overall, 1626 patients (10%) were discharged with intensified diabetes medications including 781 (5%) with new insulins and 557 (3%) with intensified sulfonylureas. Nearly half of patients receiving intensifications (49% [791 of 1626]) were classified as being unlikely to benefit owing to limited life expectancy or already being at goal HbA1c, while 20% (329 of 1626) were classified as having potential to benefit. Both preadmission HbA1c level and inpatient blood glucose recordings were associated with discharge with intensified diabetes medications. Among patients with a preadmission HbA1c level less than 7.0%, the predicted probability of receiving an intensification was 4% (95% CI, 3%-4%) for patients without elevated inpatient blood glucose levels and 21% (95% CI, 15%-26%) for patients with severely elevated inpatient blood glucose levels.

CONCLUSIONS AND RELEVANCE: In this study, 1 in 10 older adults with diabetes hospitalized for common medical conditions was discharged with intensified diabetes medications. Nearly half of these individuals were unlikely to benefit owing to limited life expectancy or already being at their HbA1c goal.