Patient Characteristics Associated With Using Transcatheter Left Atrial Appendage Occlusion Versus Oral Anticoagulants for Atrial Fibrillation.

BACKGROUND: Transcatheter left atrial appendage occlusion (LAAO) is an alternative to oral anticoagulants (OACs) for stroke prevention in patients with atrial fibrillation, but the predictors of LAAO use in routine care are unclear. We aimed to assess the utilization trends of LAAO and compare the change in characteristics of LAAO users versus OACs since its marketing. METHODS: Using the US Medicare claims database (March 15, 2015, to December 31, 2020), we identified patients with atrial fibrillation, ≥65 years, and CHA2DS2-VASc score ≥2 (men) or ≥3 (women), with either first implantation of an LAAO device or initiation of OACs, including apixaban, dabigatran, rivaroxaban, edoxaban, or warfarin. Patient characteristics, measured 365 days before the first LAAO or OAC use date, were compared using logistic regression. RESULTS: There were 30 058 LAAO recipients (mean age, 77.74 years; female, 42.1%) and 792 600 OAC initiators (mean age, 78.48; female, 53.3%). In 2020, patients had higher odds of initiating LAAO use than in 2015 (0.52 versus 9.32%; adjusted odds ratio [aOR], 13.64 [95% CI, 12.56-14.81]). Old age (ie, >85 versus 65-75 years; aOR, 0.84 [95% CI, 0.80-0.88]), female sex (aOR, 0.74 [95% CI, 0.71-0.76]), Black race (aOR, 0.63 [95% CI, 0.58-0.68]) versus White race, and Medicaid eligibility (aOR, 0.61 [95% CI, 0.58-0.64]) were associated with lower odds of receiving LAAO. Among clinical characteristics, frailty, cancer, fractures, and venous thromboembolism were associated with lower odds of LAAO use, while history of intracranial and extracranial bleeding, coagulopathy, and falls were associated with higher odds of receiving LAAO. CONCLUSIONS: Among patients with atrial fibrillation receiving stroke-preventive therapy, LAAO use increased rapidly from 2015 to 2020 and was positively associated with the risk factors for OAC complications but negatively associated with old age, advanced frailty, and cancer. Black race and female sex were associated with a lower likelihood of receiving LAAO.

Development and Validation of a Claims-Based Model to Predict Categories of Obesity.

We developed and validated a claims-based algorithm that classifies patients into obesity categories. Using Medicare (2007-2017) and Medicaid (2000-2014) claims data linked to 2 electronic health record (EHR) systems in Boston, Massachusetts, we identified a cohort of patients with an EHR-based body mass index (BMI) measurement (calculated as weight (kg)/height (m)2). We used regularized regression to select from 137 variables and built generalized linear models to classify patients with BMIs of ≥25, ≥30, and ≥40. We developed the prediction model using EHR system 1 (training set) and validated it in EHR system 2 (validation set). The cohort contained 123,432 patients in the Medicare population and 40,736 patients in the Medicaid population. The model comprised 97 variables in the Medicare set and 95 in the Medicaid set, including BMI-related diagnosis codes, cardiovascular and antidiabetic drugs, and obesity-related comorbidities. The areas under the receiver-operating-characteristic curve in the validation set were 0.72, 0.75, and 0.83 (Medicare) and 0.66, 0.66, and 0.70 (Medicaid) for BMIs of ≥25, ≥30, and ≥40, respectively. The positive predictive values were 81.5%, 80.6%, and 64.7% (Medicare) and 81.6%, 77.5%, and 62.5% (Medicaid), for BMIs of ≥25, ≥30, and ≥40, respectively. The proposed model can identify obesity categories in claims databases when BMI measurements are missing and can be used for confounding adjustment, defining subgroups, or probabilistic bias analysis.

Rates and costs of drug testing practices for private payors in the outpatient setting in the United States, 2015-2019.

INTRODUCTION: Clinical practice guidelines recommend drug testing patients who are receiving opioids chronically for pain or medication for a substance use disorder (SUD)-particularly opioid use disorder (OUD)-but practices vary due to a lack of consensus on testing frequency during follow-up. This study aimed to evaluate rates and costs of outpatient drug testing practices for patients receiving opioids for chronic pain or medication for an SUD. METHODS: Using claims data from a large de-identified claims data warehouse, we conducted a retrospective cohort study of chronic opioid, buprenorphine, and naltrexone users between January 2015 and December 2019. We identified two cohorts-chronic opioid medication cohort (CO) and SUD-indicated medication cohort (SUD). We assessed drug testing rates during follow-up using procedure codes and costs using copayment, deductible, co-insurance, and out-of-pocket data. RESULTS: Among 6,657,515 eligible claimants, 367,118 (5.5 %) received opioids chronically and 73,303 (1.1 %) received an SUD-indicated medication. The cumulative proportion of drug testing during follow-up was similar between cohorts (CO: 36 %; SUD: 35 %), but rate of testing was consistently twice as frequent for the SUD cohort. All cost variables for the first drug test were higher on average in the SUD cohort than the CO cohort except copay: deductible (SUD: $18.54; CO: $7.33); co-insurance (SUD: $10.36; CO: $2.53); out-of-pocket (SUD: $29.39; CO: $10.57); copay (CO: $0.71; SUD: $0.49) (all p < 0.001). CONCLUSIONS: Overall proportion of drug testing was similar between cohorts, but testing frequency was at least double during follow-up in the SUD cohort. Most cost variables were higher in the SUD cohort. Whether the high cost of drug testing is a barrier to medication use or is associated with treatment discontinuation should be evaluated.

Frailty and uptake of angiotensin receptor neprilysin inhibitor for heart failure with reduced ejection fraction.

BACKGROUND: Frail older adults may be less likely to receive guideline-directed medical therapy (GDMT)-renin-angiotensin blockers, beta-blockers, and mineralocorticoid receptor antagonists-for heart failure with reduced ejection fraction (HFrEF). We aimed to examine the uptake of angiotensin receptor neprilysin inhibitor (ARNI) and GDMT in frail older adults with HFrEF. METHODS: Using 2015-2019 Medicare data, we estimated the proportion of beneficiaries with HFrEF receiving ARNI and GDMT each year by frailty status, defined by a claims-based frailty index. Logistic regression was used to identify clinical characteristics associated with ARNI initiation. Cox proportional hazards regression was used to examine the association of GDMT use in 2015 and death or heart failure hospitalization in 2016-2019. RESULTS: Among 147,506-180,386 beneficiaries with HFrEF (mean age: 77 years; 27% women; 42.6-49.1% frail) in 2015-2019, the proportion of patients receiving ARNI increased in both non-frail (0.4%-16.4%) and frail (0.3%-13.7%) patients (p for yearly-trend-by-frailty = 0.970). Among those not receiving a renin-angiotensin system blocker, patients with age ≥ 85 years (odds ratio [95% CI], 0.89 [0.80-0.99]), dementia (0.88 [0.81-0.96]), and frailty (0.87 [0.81-0.94]) were less likely to initiate ARNI. The proportion of patients receiving all 3 GDMT classes increased in non-frail patients (22.0%-27.0%) but changed minimally in frail patients (19.6%-21.8%). Regardless of frailty status, treatment with at least 1 class of GDMT was associated with lower death or heart failure hospitalization than no GDMT medications (hazard ratio [95% CI], 0.94 [0.91-0.97], 0.92 [0.89-0.94], 0.94 [0.91-0.97] for 1, 2, and 3 classes, respectively). CONCLUSIONS: Our results suggest an evidence-practice gap in the use of ARNI and GDMT in Medicare beneficiaries with HFrEF, particularly those with frailty. Efforts to narrow this gap are needed to reduce the burden of HFrEF in older adults.

Scalable Feature Engineering from Electronic Free Text Notes to Supplement Confounding Adjustment of Claims-Based Pharmacoepidemiologic Studies.

Natural language processing (NLP) tools turn free-text notes (FTNs) from electronic health records (EHRs) into data features that can supplement confounding adjustment in pharmacoepidemiologic studies. However, current applications are difficult to scale. We used unsupervised NLP to generate high-dimensional feature spaces from FTNs to improve prediction of drug exposure and outcomes compared with claims-based analyses. We linked Medicare claims with EHR data to generate three cohort studies comparing different classes of medications on the risk of various clinical outcomes. We used "bag-of-words" to generate features for the top 20,000 most prevalent terms from FTNs. We compared machine learning (ML) prediction algorithms using different sets of candidate predictors: Set1 (39 researcher-specified variables), Set2 (Set1 + ML-selected claims codes), and Set3 (Set1 + ML-selected NLP-generated features), vs. Set4 (Set1 + 2 + 3). When modeling treatment choice, we observed a consistent pattern across the examples: ML models utilizing Set4 performed best followed by Set2, Set3, then Set1. When modeling the outcome risk, there was little to no improvement beyond models based on Set1. Supplementing claims data with NLP-generated features from free text notes improved prediction of prescribing choices but had little or no improvement on clinical risk prediction. These findings have implications for strategies to improve confounding using EHR data in pharmacoepidemiologic studies.

Associations Between Copays, Coverage Limits for Opioid Use Disorder Medications, and Prescribing in Medicaid, 2018.

BACKGROUND: Opioid use disorder (OUD) affects millions of Americans, but only a fraction receive treatment. Many patients with OUD are enrolled in Medicaid, but elements of different state Medicaid programs' drug benefit designs may impact patients' access to life-saving care. OBJECTIVE: To describe medication for OUD (mOUD) use in Medicaid and examine the relationship between mOUD use and state drug benefit design plans. DESIGN/SUBJECTS: Cross-sectional study using Medicaid State Drug Utilization Data from 2018 to quantify office-based mOUD and the Medicaid Behavioral Health Services Database to extract copay amounts and coverage limits for mOUD. We excluded states with <5% coverage and assessed for associations between copays or coverage limits and mOUD dispensing using simple linear regression. MEASURES: Proportion of mOUD prescriptions relative to all prescriptions, opioid prescriptions, and the state-level prevalence of pain reliever use disorder and association between copays, coverage limits and these proportions. RESULTS: There was substantial variability in mOUD use. Although state Medicaid drug benefit designs also varied, we found no significant relationship between copay requirements (yes/no), coverage limits (yes/no), copay amount ($0-$0.99 vs. $1 or more), and mOUD utilization measures. CONCLUSIONS: Substantial state-level variation exists in mOUD use, but we did not find a significant association between copays or coverage limits and use in Medicaid. Further research is needed to assess other potential impacts of mOUD drug benefit design elements in Medicaid.

Association of Medicare's Bundled Payment Reform With Changes in Use of Vitamin D Among Patients Receiving Maintenance Hemodialysis: An Interrupted Time-Series Analysis.

BACKGROUND & RATIONALE: Medicare's 2011 prospective payment system (PPS) was introduced to curb overuse of separately billable injectable drugs. After epoietin, intravenous (IV) vitamin D analogues are the biggest drug cost drivers in hemodialysis (HD) patients, but the association between PPS introduction and vitamin D therapy has been scarcely investigated. STUDY DESIGN: Interrupted time-series analyses. SETTING & PARTICIPANTS: Adult US HD patients represented in the US Renal Data System between 2008 and 2013. EXPOSURES: PPS implementation. OUTCOMES: The cumulative dose of IV vitamin D analogues (paricalcitol equivalents) per patient per calendar quarter in prevalent HD patients. The average starting dose of IV vitamin D analogues and quarterly rates of new vitamin D use (initiations/100 person-months) in incident HD patients within 90 days of beginning HD therapy. ANALYTICAL APPROACH: Segmented linear regression models of the immediate change and slope change over time of vitamin D use after PPS implementation. RESULTS: Among 359,600 prevalent HD patients, IV vitamin D analogues accounted for 99% of the total use, and this trend was unchanged over time. PPS resulted in an immediate 7% decline in the average dose of IV vitamin D analogues (average baseline dose = 186.5 µg per quarter; immediate change = -13.5 µg [P < 0.001]; slope change = 0.43 per quarter [P = 0.3]) and in the starting dose of IV vitamin D analogues in incident HD patients (average baseline starting dose = 5.22 µg; immediate change = -0.40 µg [P < 0.001]; slope change = -0.03 per quarter [P = 0.03]). The baseline rate of vitamin D therapy initiation among 99,970 incident HD patients was 44.9/100 person-months and decreased over time, even before PPS implementation (pre-PPS ß = -0.46/100 person-months [P < 0.001]; slope change = -0.19/100 person-months [P = 0.2]). PPS implementation was associated with an immediate change in initiation levels (by -4.5/100 person-months; P < 0.001). LIMITATIONS: Incident HD patients were restricted to those 65 years or older. CONCLUSION: PPS implementation was associated with a 7% reduction in the average dose and starting dose of IV vitamin D analogues and a 10% reduction in the rate of vitamin D therapy initiation.