Identifying Functional Status Impairment in People Living With Dementia Through Natural Language Processing of Clinical Documents: Cross-Sectional Study

BACKGROUND

Assessment of activities of daily living (ADLs) and instrumental ADLs (iADLs) is key to determining the severity of dementia and care needs among older adults. However, such information is often only documented in free-text clinical notes within the electronic health record and can be challenging to find.

OBJECTIVE

This study aims to develop and validate machine learning models to determine the status of ADL and iADL impairments based on clinical notes.

METHODS

This cross-sectional study leveraged electronic health record clinical notes from Mass General Brigham's Research Patient Data Repository linked with Medicare fee-for-service claims data from 2007 to 2017 to identify individuals aged 65 years or older with at least 1 diagnosis of dementia. Notes for encounters both 180 days before and after the first date of dementia diagnosis were randomly sampled. Models were trained and validated using note sentences filtered by expert-curated keywords (filtered cohort) and further evaluated using unfiltered sentences (unfiltered cohort). The model's performance was compared using area under the receiver operating characteristic curve and area under the precision-recall curve (AUPRC).

RESULTS

The study included 10,000 key-term-filtered sentences representing 441 people (n=283, 64.2% women; mean age 82.7, SD 7.9 years) and 1000 unfiltered sentences representing 80 people (n=56, 70% women; mean age 82.8, SD 7.5 years). Area under the receiver operating characteristic curve was high for the best-performing ADL and iADL models on both cohorts (>0.97). For ADL impairment identification, the random forest model achieved the best AUPRC (0.89, 95% CI 0.86-0.91) on the filtered cohort; the support vector machine model achieved the highest AUPRC (0.82, 95% CI 0.75-0.89) for the unfiltered cohort. For iADL impairment, the Bio+Clinical bidirectional encoder representations from transformers (BERT) model had the highest AUPRC (filtered: 0.76, 95% CI 0.68-0.82; unfiltered: 0.58, 95% CI 0.001-1.0). Compared with a keyword-search approach on the unfiltered cohort, machine learning reduced false-positive rates from 4.5% to 0.2% for ADL and 1.8% to 0.1% for iADL.

CONCLUSIONS

In this study, we demonstrated the ability of machine learning models to accurately identify ADL and iADL impairment based on free-text clinical notes, which could be useful in determining the severity of dementia.

Disentangling drug contributions: anticholinergic burden in older adults linked to individual medications: a cross-sectional population-based study

BACKGROUND

Medications with potent anticholinergic properties have well-documented adverse effects. A high cumulative anticholinergic burden may arise from the concurrent use of multiple medications with weaker anticholinergic effects. We sought to identify patterns of high anticholinergic burden and associated patient characteristics.

METHODS

We identified patients aged ≥ 65 who filled ≥ 1 medication with anticholinergic adverse effects in 2019 and had a cumulative Anticholinergic Burden score (ACB) ≥ 4 (i.e., high anticholinergic burden) in a large US health insurer. We classified patients based on how they attained high burden, as follows: 1) only filling strong or moderate anticholinergic medications (i.e., ACB = 2 or 3, "moderate/strong"), 2) only filling lightly anticholinergic medications (i.e., ACB = 1, "light/possible"), and 3) filling any combination ("mix"). We used multinomial logistic regression to assess the association between measured patient characteristics and membership in the three anticholinergic burden classifications, using the moderate/strong group as the referent.

RESULTS

In total, 83,286 eligible patients with high anticholinergic burden were identified (mean age: 74.3 years (SD:7.1), 72.9% female). Of these, 4.5% filled only strong/moderate anticholinergics, 4.3% filled only light/possible anticholinergics, and the rest filled a mix (91.2%). Within patients in the mixed group, 64.3% of medication fills were for light/possible anticholinergics, while 35.7% were for moderate/strong anticholinergics. Compared with patients in the moderate/strong anticholinergics group, patients filling only light/possible anticholinergics were more likely to be older (adjusted Odds Ratio [aOR] per 1-unit of age: 1.06, 95%CI: 1.05-1.07), less likely to be female (aOR: 0.56, 95%CI: 0.50-0.62 vs. male), more likely to have comorbidities (e.g., heart failure aOR: 3.18, 95%CI: 2.70-3.74 or depression aOR: 1.20, 95%CI: 1.09-1.33 vs. no comorbidity), and visited fewer physicians (aOR per 1-unit of change: 0.98, 95%CI: 0.97-0.98). Patients in the mixed group were older (aOR per 1-unit of age: 1.02, 95%CI: 1.02-1.03) and less likely to be female (aOR: 0.89, 95%CI: 0.82-0.97 vs. male) compared with those filling moderate/strong anticholinergics.

CONCLUSION

Most older adults accumulated high anticholinergic burden through a combination of light/possible and moderate/strong anticholinergics rather than moderate/strong anticholinergics, with light/possible anticholinergics being the major drivers of overall anticholinergic burden. These insights may inform interventions to improve prescribing in older adults.

Predicting Treatment Effects of a New-to-Market Drug in Clinical Practice Based on Phase III Randomized Trial Results

Trial results may not be generalizable to target populations treated in clinical practice with different distributions of baseline characteristics that modify the treatment effect. We used outcome models developed with trial data to predict treatment effects in Medicare populations. We used data from the Randomized Evaluation of Long-Term Anticoagulation Therapy trial (RE-LY), which investigated the effect of dabigatran vs. warfarin on stroke or systemic embolism (stroke/SE) among patients with atrial fibrillation. We developed outcome models by fitting proportional hazards models in trial data. Target populations were trial-eligible Medicare beneficiaries who initiated dabigatran or warfarin in 2010-2011 ("early") and 2010-2017 ("extended"). We predicted 2-year risk ratios (RRs) and risk differences (RDs) for stroke/SE, major bleeding, and all-cause death in the Medicare populations using the observed baseline characteristics. The trial and early target populations had similar mean (SD) CHADS2 scores (2.15 (SD 1.13) vs. 2.15 (SD 0.91)) but different mean ages (71 vs. 79 years). Compared with RE-LY, the early Medicare population had similar predicted benefit of dabigatran vs. warfarin for stroke/SE (trial RR = 0.63, 95% confidence interval (CI) = 0.50 to 0.76 and RD = -1.37%, -1.96% to -0.77%, Medicare RR = 0.73, 0.65 to 0.82 and RD = -0.92%, -1.26% to -0.59%) and risks for major bleeding and all-cause death. The time-extended target population showed similar results. Outcome model-based prediction facilitates estimating the average treatment effects of a drug in different target populations when treatment and outcome data are unreliable or unavailable. The predicted effects may inform payers' coverage decisions for patients, especially shortly after a drug's launch when observational data are scarce.

No association between initiation of phosphodiesterase-5 inhibitors and risk of incident Alzheimer's disease and related dementia: results from the Drug Repurposing for Effective Alzheimer's Medicines study

We evaluated the hypothesis that phosphodiesterase-5 inhibitors, including sildenafil and tadalafil, may be associated with reduced incidence of Alzheimer's disease and related dementia using a patient-level cohort study of Medicare claims and cell culture-based phenotypic assays. We compared incidence of Alzheimer's disease and related dementia after phosphodiesterase-5 inhibitor initiation versus endothelin receptor antagonist initiation among patients with pulmonary hypertension after controlling for 76 confounding variables through propensity score matching. Across four separate analytic approaches designed to address specific types of biases including informative censoring, reverse causality, and outcome misclassification, we observed no evidence for a reduced risk of Alzheimer's disease and related dementia with phosphodiesterase-5 inhibitors;hazard ratio (95% confidence interval): 0.99 (0.69-1.43), 1.00 (0.71-1.42), 0.67 (0.43-1.06), and 1.15 (0.57-2.34). We also did not observe evidence that sildenafil ameliorated molecular abnormalities relevant to Alzheimer's disease in most cell culture-based phenotypic assays. These results do not provide support to the hypothesis that phosphodiesterase-5 inhibitors are promising repurposing candidates for Alzheimer's disease and related dementia.

Using Data-Driven Approaches to Classify and Predict Health Care Spending in Patients With Gout Using Urate-Lowering Therapy

OBJECTIVE

Despite increasing overall health care spending over the past several decades, little is known about long-term patterns of spending among US patients with gout. Current approaches to assessing spending typically focus on composite measures or patients agnostic to disease state; in contrast, examining spending using longitudinal measures may better discriminate patients and target interventions to those in need. We used a data-driven approach to classify and predict spending patterns in patients with gout.

METHODS

Using insurance claims data from 2017-2019, we used group-based trajectory modeling to classify patients ages 40 years or older diagnosed with gout and treated with urate-lowering therapy (ULT) by their total health care spending over 2 years. We assessed the ability to predict membership in each spending group using logistic and generalized boosted regression with split-sample validation. Models were estimated using different sets of predictors and evaluated using C statistics.

RESULTS

In 57,980 patients, the mean ± SD age was 71.0 ± 10.5 years, and 17,194 patients (29.7%) were female. The best-fitting model included the following groups: minimal spending (13.2%), moderate spending (37.4%), and high spending (49.4%). The ability to predict groups was high overall (e.g., boosted C statistics with all predictors: minimal spending [0.89], moderate spending [0.78], and high spending [0.90]). Although average adherence was relatively high in the population, for the high-spending group, the most influential predictors were greater gout medication adherence and diabetes melllitus diagnosis.

CONCLUSION

We identified distinct long-term health care spending patterns in patients with gout using ULT with high accuracy. Several clinical predictors could be key areas for intervention, such as gout medication use or diabetes melllitus.

Comparative Risk of Alzheimer Disease and Related Dementia Among Medicare Beneficiaries With Rheumatoid Arthritis Treated With Targeted Disease-Modifying Antirheumatic Agents

IMPORTANCE

Cytokine signaling, including tumor necrosis factor (TNF) and interleukin (IL)-6, through the Janus-kinase (JAK)-signal transducer and activator of transcription pathway, was hypothesized to attenuate the risk of Alzheimer disease and related dementia (ADRD) in the Drug Repurposing for Effective Alzheimer Medicines (DREAM) initiative based on multiomics phenotyping.

OBJECTIVE

To evaluate the association between treatment with tofacitinib, tocilizumab, or TNF inhibitors compared with abatacept and risk of incident ADRD.

DESIGN, SETTING, AND PARTICIPANTS

This cohort study was conducted among US Medicare fee-for-service patients with rheumatoid arthritis aged 65 years and older from 2007 to 2017. Patients were categorized into 3 cohorts based on initiation of tofacitinib (a JAK inhibitor), tocilizumab (an IL-6 inhibitor), or TNF inhibitors compared with a common comparator abatacept (a T-cell activation inhibitor). Analyses were conducted from August 2020 to August 2021.

MAIN OUTCOMES AND MEASURES

The main outcome was onset of ADRD based on diagnosis codes evaluated in 4 alternative analysis schemes: (1) an as-treated follow-up approach, (2) an as-started follow-up approach incorporating a 6-month induction period, (3) incorporating a 6-month symptom to diagnosis period to account for misclassification of ADRD onset, and (4) identifying ADRD through symptomatic prescriptions and diagnosis codes. Hazard ratios (HRs) with 95% CIs were calculated from Cox proportional hazard regression after adjustment for 79 preexposure characteristics through propensity score matching.

RESULTS

After 1:1 propensity score matching to patients using abatacept, a total of 22 569 propensity score-matched patient pairs, including 4224 tofacitinib pairs (mean [SD] age 72.19 [5.65] years; 6945 [82.2%] women), 6369 tocilizumab pairs (mean [SD] age 72.01 [5.46] years; 10 105 [79.4%] women), and 11 976 TNF inhibitor pairs (mean [SD] age 72.67 [5.91] years; 19 710 [82.3%] women), were assessed. Incidence rates of ADRD varied from 2 to 18 per 1000 person-years across analyses schemes. There were no statistically significant associations of ADRD with tofacitinib (analysis 1: HR, 0.90 [95% CI, 0.55-1.51]; analysis 2: HR, 0.78 [95% CI, 0.53-1.13]; analysis 3: HR, 1.29 [95% CI, 0.72-2.33]; analysis 4: HR, 0.50 [95% CI, 0.21-1.20]), tocilizumab (analysis 1: HR, 0.82 [95% CI, 0.55-1.21]; analysis 2: HR, 1.05 [95% CI, 0.81-1.35]; analysis 3: HR, 1.21 [95% CI, 0.75-1.96]; analysis 4: HR, 0.78 [95% CI, 0.44-1.39]), or TNF inhibitors (analysis 1: HR, 0.93 [95% CI, 0.72-1.20]; analysis 2: HR, 1.02 [95% CI, 0.86-1.20]; analysis 3: HR, 1.13 [95% CI, 0.86-1.48]; analysis 4: 0.90 [95% CI, 0.60-1.37]) compared with abatacept. Results from prespecified subgroup analysis by age, sex, and baseline cardiovascular disease were consistent except in patients with cardiovascular disease, for whom there was a potentially lower risk of ADRD with TNF inhibitors vs abatacept, but only in analyses 2 and 4 (analysis 1: HR, 0.76 [95% CI, 0.50-1.16]; analysis 2: HR, 0.74 [95% CI, 0.56-0.99]; analysis 3: HR, 1.03 [95% CI, 0.65-1.61]; analysis 4: HR, 0.45 [95% CI, 0.21-0.98]).

CONCLUSIONS AND RELEVANCE

This cohort study did not find any association of risk of ADRD in patients treated with tofacitinib, tocilizumab, or TNF inhibitors compared with abatacept.

Non-Visit-Based and Non-Infection-Related Antibiotic Use in the US: A Cohort Study of Privately Insured Patients During 2016-2018

BACKGROUND

Ambulatory antibiotic prescriptions without a clinic visit or without documentation of infection could represent overuse and contribute to adverse outcomes. We aim to describe US ambulatory antibiotic prescribing, including those without an associated visit or infection diagnosis.

METHODS

We conducted an observational cohort study using data of all patients receiving antibacterial, antibiotic prescriptions from 04/01/2016 to 06/30/2018 in a large US private health insurance plan. We identified outpatient antibiotic prescriptions as (1) associated with a clinician visit and an infection-related diagnosis; (2) associated with a clinician visit but no infection-related diagnosis; or (3) not associated with an in-person clinician visit in the 7 days before the prescription (non-visit-based). We then assessed whether non-visit-based antibiotic prescriptions (NVBAPs) differed from visit-based antibiotics by patient, clinician, or antibiotic characteristics using multivariable models.

RESULTS

The cohort included 8.6M enrollees who filled 22.3M antibiotic prescriptions. NVBAP accounted for 31% (6.9M) of fills, and non-infection-related prescribing accounted for 22% (4.9M). NVBAP rates were lower for children than for adults (0-17 years old, 16%; 18-64 years old, 33%; >65 years old, 34%). Among most commonly prescribed antibiotic classes, NVBAP was highest for penicillins (36%) and lowest for cephalosporins (25%) and macrolides (25%). Specialist physicians had the highest rate of NVBAP (38%), followed by internists (28%), family medicine (20%), and pediatricians (10%). In multivariable models, NVBAP was associated with increasing age, and NVBAP was less likely for patients in the South, those with more baseline clinical visits, or those with chronic lung disease.

CONCLUSIONS

Over half of ambulatory antibiotic use was either non-visit-based or non-infection-related. Particularly given health care changes due to the coronavirus disease 2019 pandemic, efforts to improve antibiotic prescribing must account for non-visit-based and non-infection-related prescribing.

External validation of a claims-based model to predict left ventricular ejection fraction class in patients with heart failure

Background

Ejection fraction (EF) is an important prognostic factor in heart failure (HF), but administrative claims databases lack information on EF. We previously developed a model to predict EF class from Medicare claims. Here, we evaluated the performance of this model in an external validation sample of commercial insurance enrollees.

Methods

Truven MarketScan claims linked to electronic medical records (EMR) data (IBM Explorys) containing EF measurements were used to identify a cohort of US patients with HF between 01-01-2012 and 10-31-2019. By applying the previously developed model, patients were classified into HF with reduced EF (HFrEF) or preserved EF (HFpEF). EF values recorded in EMR data were used to define gold-standard HFpEF (LVEF ≥45%) and HFrEF (LVEF<45%). Model performance was reported in terms of overall accuracy, positive predicted values (PPV), and sensitivity for HFrEF and HFpEF.

Results

A total of 7,001 HF patients with an average age of 71 years were identified, 1,700 (24.3%) of whom had HFrEF. An overall accuracy of 0.81 (95% CI: 0.80–0.82) was seen in this external validation sample. For HFpEF, the model had sensitivity of 0.96 (95%CI, 0.95–0.97) and PPV of 0.81 (95% CI, 0.81–0.82); while for HFrEF, the sensitivity was 0.32 (95%CI, 0.30–0.34) and PPV was 0.73 (95%CI, 0.69–0.76). These results were consistent with what was previously published in US Medicare claims data.

Conclusions

The successful validation of the Medicare claims-based model provides evidence that this model may be used to identify patient subgroups with specific EF class in commercial claims databases as well.

Effectiveness of angiotensin-neprilysin inhibitor treatment versus renin-angiotensin system blockade in older adults with heart failure in clinical care

OBJECTIVE

To evaluate the effectiveness of angiotensin receptor-neprilysin inhibitor (ARNI) versus renin-angiotensin system (RAS) blockade alone in older adults with heart failure with reduced ejection fraction (HFrEF).

METHODS

We conducted a cohort study using US Medicare fee-for-service claims data (2014-2017). Patients with HFrEF ≥65 years were identified in two cohorts: (1) initiators of ARNI or RAS blockade alone (ACE inhibitor, ACEI; or angiotensin receptor blocker, ARB) and (2) switchers from an ACEI to either ARNI or ARB. HR with 95% CI from Cox proportional hazard regression and 1-year restricted mean survival time (RMST) difference with 95% CI were calculated for a composite outcome of time to first worsening heart failure event or all-cause mortality after adjustment for 71 pre-exposure characteristics through propensity score fine-stratification weighting. All analyses of initiator and switcher cohorts were conducted separately and then combined using fixed effects.

RESULTS

51 208 patients with a mean age of 76 years were included, with 16 193 in the ARNI group. Adjusted HRs comparing ARNI with RAS blockade alone were 0.92 (95% CI 0.84 to 1.00) among initiators and 0.79 (95% CI 0.74 to 0.85) among switchers, with a combined estimate of 0.84 (95% CI 0.80 to 0.89). Adjusted 1-year RMST difference (95% CI) was 4 days in the initiator cohort (-1 to 9) and 12 days (8 to 17) in the switcher cohort, resulting in a pooled estimate of 9 days (6 to 12) favouring ARNI.

CONCLUSION

ARNI treatment was associated with lower risk of a composite effectiveness endpoint compared with RAS blockade alone in older adults with HFrEF.

Epidemiologic Characterization of Heart Failure with Reduced or Preserved Ejection Fraction Populations Identified Using Medicare Claims

BACKGROUND

Administrative claims do not contain ejection fraction information for heart failure patients. We recently developed and validated a claims-based model to predict ejection fraction subtype.

METHODS

Heart failure patients aged 65 years or above from US Medicare fee-for-service claims were identified using diagnoses recorded after a 6-month baseline period of continuous enrollment, which was used to identify predictors and to apply the claims-based model to distinguish heart failure with reduced or preserved ejection fraction (HFrEF or HFpEF). Patients were followed for the composite outcome of time to first worsening heart failure event (heart failure hospitalization or outpatient intravenous diuretic treatment) or all-cause mortality.

RESULTS

A total of 3,134,414 heart failure patients with an average age of 79 years were identified, of which 200,950 (6.4%) were classified as HFrEF. Among those classified as HFrEF, men comprised a larger proportion (68% vs 41%) and the average age was lower (76 vs 79 years) compared with HFpEF. History of myocardial infarction was more common in HFrEF (32% vs 13%), while hypertension was more common in HFpEF (71% vs 77%). One-year cumulative incidence of the composite endpoint was 42.6% for HFrEF and 36.9% for HFpEF. One-year all-cause mortality incidence was similar between the groups (27.4% for HFrEF and 26.4% for HFpEF), however, cardiovascular mortality was higher for HFrEF (15.6% vs 11.3%), whereas noncardiovascular mortality was higher for HFpEF (11.8% vs 15.1%).

CONCLUSION

We replicated well-documented differences in key patient characteristics and cause-specific outcomes between HFrEF and HFpEF in populations identified based on the application of a claims-based model.

Rationale and design of the Novel Uses of adaptive Designs to Guide provider Engagement in Electronic Health Records (NUDGE-EHR) pragmatic adaptive randomized trial: a trial protocol

BACKGROUND

The prescribing of high-risk medications to older adults remains extremely common and results in potentially avoidable health consequences. Efforts to reduce prescribing have had limited success, in part because they have been sub-optimally timed, poorly designed, or not provided actionable information. Electronic health record (EHR)-based tools are commonly used but have had limited application in facilitating deprescribing in older adults. The objective is to determine whether designing EHR tools using behavioral science principles reduces inappropriate prescribing and clinical outcomes in older adults.

METHODS

The Novel Uses of Designs to Guide provider Engagement in Electronic Health Records (NUDGE-EHR) project uses a two-stage, 16-arm adaptive randomized pragmatic trial with a "pick-the-winner" design to identify the most effective of many potential EHR tools among primary care providers and their patients ≥ 65 years chronically using benzodiazepines, sedative hypnotic ("Z-drugs"), or anticholinergics in a large integrated delivery system. In stage 1, we randomized providers and their patients to usual care (n = 81 providers) or one of 15 EHR tools (n = 8 providers per arm) designed using behavioral principles including salience, choice architecture, or defaulting. After 6 months of follow-up, we will rank order the arms based upon their impact on the trial's primary outcome (for both stages): reduction in inappropriate prescribing (via discontinuation or tapering). In stage 2, we will randomize (a) stage 1 usual care providers in a 1:1 ratio to one of the up to 5 most promising stage 1 interventions or continue usual care and (b) stage 1 providers in the unselected arms in a 1:1 ratio to one of the 5 most promising interventions or usual care. Secondary and tertiary outcomes include quantities of medication prescribed and utilized and clinically significant adverse outcomes.

DISCUSSION

Stage 1 launched in October 2020. We plan to complete stage 2 follow-up in December 2021. These results will advance understanding about how behavioral science can optimize EHR decision support to improve prescribing and health outcomes. Adaptive trials have rarely been used in implementation science, so these findings also provide insight into how trials in this field could be more efficiently conducted.

TRIAL REGISTRATION

Clinicaltrials.gov ( NCT04284553 , registered: February 26, 2020).

Targeting abnormal metabolism in Alzheimer's disease: The Drug Repurposing for Effective Alzheimer's Medicines (DREAM) study

Drug discovery for disease-modifying therapies for Alzheimer's disease and related dementias (ADRD) based on the traditional paradigm of experimental animal models has been disappointing. We describe the rationale and design of the Drug Repurposing for Effective Alzheimer's Medicines (DREAM) study, an innovative multidisciplinary alternative to traditional drug discovery. First, we use a systems biology perspective in the "hypothesis generation" phase to identify metabolic abnormalities that may either precede or interact with the accumulation of ADRD neuropathology, accelerating the expression of clinical symptoms of the disease. Second, in the "hypothesis refinement" phase we propose use of large patient cohorts to test whether drugs approved for other indications that also target metabolic drivers of ADRD pathogenesis might alter the trajectory of the disease. We emphasize key challenges in population-based pharmacoepidemiologic studies aimed at quantifying the association between medication use and ADRD onset and outline robust causal inference principles to safeguard against common pitfalls. Candidate ADRD treatments emerging from this approach will hold promise as plausible disease-modifying therapies for evaluation in randomized controlled trials.

Clinical outcomes following bone marrow transplantation in patients with sickle cell disease: A cohort study of US Medicaid enrollees

OBJECTIVES

Bone marrow transplantation (BMT) is currently the only curative therapy available for patients with sickle cell disease (SCD), but clinical outcomes in routine care are not well understood. We describe the rates of vaso-occlusive crises (VOCs), transplant complications, and mortality in SCD patients after BMT.

METHODS

A cohort study of SCD patients who underwent BMT was designed using US Medicaid claims data (2000-2013).

RESULTS

A total of 204 SCD patients undergoing BMT were identified with a mean (SD) age of 10.6 (7.3) years, with 52.9% male and 67.6% African American. The overall VOC rate was 0.99 per person-year (95% CI: 0.91-1.07) over a median follow-up time of 2.1 years (IQR: 0.8-4.3 years). A total of 138 (67.6%) remained free of VOCs. The mortality rate was 1.7 (95% CI: 0.9-3.1) per 100 person-years, transplant-related complications occurred among 113 (55.4%) patients with an incidence rate of 38.2 (95% CI: 31.7-45.9) per 100 person-years, while 47 (23%) patients had GvHD with an incidence rate of 8.0 (95% CI: 6.0-10.7) per 100 person-years.

CONCLUSION

Two thirds of the BMT recipients remained VOC-free over 2 years of follow-up, but transplant-related complications, including GvHD occurred with high frequency. This highlights a continuing unmet need for alternative curative interventions in SCD.

Application of a Medicare claims-based model predicting left ventricular ejection fraction subtype to investigate the epidemiology of heart failure in the US Medicare program

Introduction

Administrative claims do not contain ejection fraction (EF) information for heart failure (HF) patients. To address this limitation, we recently developed a claims-based model to classify HF patients into reduced EF (rEF) or preserved EF (pEF) using 35 predictors.

Purpose

To report distribution of key patient characteristics and rates of HF decompensation and mortality in model-identified rEF and pEF patients from nationwide Medicare claims (2012–2016) and compare with estimates from the literature.

Methods

We identified HF patients ≥65 years from US Medicare claims using recorded diagnosis after ≥6 months of continuous enrollment. The date of HF diagnosis was the cohort entry date. The 6-month baseline period prior to the cohort entry date was used to identify predictors and apply the claims-based model to distinguish rEF and pEF. Patients were followed for the composite outcome of time to first HF decompensation (HF hospitalization or outpatient IV diuretic treatment) or all-cause mortality. Descriptive statistics were used to summarize baseline patient characteristics. Cumulative incidence estimates along with 95% confidence intervals (CI) were calculated for the composite endpoint as well as all-cause and cause-specific mortality (derived from National Death Index linkage) using the Kaplan-Meier method.

Results

A total of 3,134,414 HF patients with an average age of 79 years were identified, of which 200,950 (6.4%) were classified as rEF. Among those classified as rEF, men comprised a larger proportion (68% vs 41%), the average age was lower (76 vs 79 years), and history of myocardial infarction was more frequent (32% vs 13%) compared to pEF. One-year cumulative incidence (95% CI) of the composite endpoint was 42.6% (42.4–42.8%) for rEF and 36.9% (36.7–37.0%) for pEF. One-year all-cause mortality incidence was similar between the groups (27.4% [27.2–27.6%] for rEF and 26.4% [26.3–26.4%] for pEF), however, cardiovascular mortality was higher for rEF (16.7% [16.5–16.8%] vs 12.3% [12.2–12.3%]), whereas non-cardiovascular mortality was higher for pEF (12.9% [12.7–13.1%] vs 16.0% (16.0–16.1%) (Figure 1). These results were in line with estimates from other well-established cardiovascular cohorts including the Get With The Guidelines-HF cohort and the Olmstead County HF epidemiology cohort.

Conclusion

We replicated well-documented differences in key patient characteristics and endpoints between rEF and pEF in these populations identified based on application of a claims-based model. Our results support use of this model for identifying cohorts of rEF and pEF to conduct subtype specific investigations of treatment outcomes. However, a notably lower proportion of patients were identified as having rEF compared to previous reports indicating low sensitivity of this approach for rEF and suggesting that model-based classification may not be useful in tracking subtype specific incidence or prevalence of HF using Medicare claims.

Cumulative incidence of outcomes in HF

Funding Acknowledgement

Type of funding source: Private grant(s) and/or Sponsorship. Main funding source(s): Novartis Inc.

Use of Data-Driven Methods to Predict Long-term Patterns of Health Care Spending for Medicare Patients

IMPORTANCE

Current approaches to predicting health care costs generally rely on a single composite value of spending and focus on short time horizons. By contrast, examining patients' spending patterns using dynamic measures applied over longer periods may better identify patients with different spending and help target interventions to those with the greatest need.

OBJECTIVE

To classify patients by their long-term, dynamic health care spending patterns using a data-driven approach and assess the ability to predict spending patterns, particularly using characteristics that are potentially modifiable through intervention.

DESIGN, SETTING, AND PARTICIPANTS

This cohort study used a retrospective cohort design from a random nationwide sample of Medicare fee-for-service administrative claims data to identify beneficiaries aged 65 years or older with continuous eligibility from 2011 to 2013. Statistical analysis was performed from August 2018 to December 2019.

MAIN OUTCOMES AND MEASURES

Group-based trajectory modeling was applied to the claims data to classify the Medicare beneficiaries by their total health care spending patterns over a 2-year period. The ability to predict membership in each trajectory spending group was assessed using generalized boosted regression, a data mining approach to model building and prediction, with split-sample validation. Models were estimated using (1) prior-year predictors and (2) prior-year predictors potentially modifiable through intervention measured in the claims data. These models were evaluated using validated C-statistics. The relative influence of individual predictors in the models was evaluated.

RESULTS

Among the 329 476 beneficiaries, the mean (SD) age was 76.0 (7.2) years and 190 346 (57.8%) were female. This final 5-group model included a minimal-user group (group 1, 37 572 individuals [11.4%]), a low-cost group (group 2, 48 575 individuals [14.7%]), a rising-cost group (group 3, 24 736 individuals [7.5%]), a moderate-cost group (group 4, 83 338 individuals [25.3%]), and a high-cost group (group 5, 135 255 individuals [41.2%]). Potentially modifiable characteristics strongly predicted these patterns (C-statistics range: 0.68-0.94). For groups with progressively increasing spending in particular, the most influential factors were number of medications (relative influence: 29.2), number of office visits (relative influence: 30.3), and mean medication adherence (relative influence: 33.6).

CONCLUSIONS AND RELEVANCE

Using a data-driven approach, distinct spending patterns were identified with high accuracy. The potentially modifiable predictors of membership in the rising-cost group represent important levers for early interventions that may prevent later spending increases. This approach could be adapted by organizations to target quality improvement interventions, particularly because numerous health care organizations are increasingly using these routinely collected data.

Not there yet: using data-driven methods to predict who becomes costly among low-cost patients with type 2 diabetes

BACKGROUND

Diabetes is a leading cause of Medicare spending; predicting which individuals are likely to be costly is essential for targeting interventions. Current approaches generally focus on composite measures, short time-horizons, or patients who are already high utilizers, whose costs may be harder to modify. Thus, we used data-driven methods to classify unique clusters in Medicare claims who were initially low utilizers by their diabetes spending patterns in subsequent years and used machine learning to predict these patterns.

METHODS

We identified beneficiaries with type 2 diabetes whose spending was in the bottom 90% of diabetes care spending in a one-year baseline period in Medicare fee-for-service data. We used group-based trajectory modeling to classify unique clusters of patients by diabetes-related spending patterns over a two-year follow-up. Prediction models were estimated with generalized boosted regression, a machine learning method, using sets of all baseline predictors, diabetes predictors, and predictors that are potentially-modifiable through interventions. Each model was evaluated through C-statistics and 5-fold cross-validation.

RESULTS

Among 33,789 beneficiaries (baseline median diabetes spending: $4153), we identified 5 distinct spending patterns that could largely be predicted; of these, 68.1% of patients had consistent spending, 25.3% had spending that rose quickly, and 6.6% of patients had spending that rose progressively. The ability to predict these groups was moderate (validated C-statistics: 0.63 to 0.87). The most influential factors for those with progressively rising spending were age, generosity of coverage, prior spending, and medication adherence.

CONCLUSIONS

Patients with type 2 diabetes who were initially low spenders exhibit distinct subsequent long-term patterns of diabetes spending; membership in these patterns can be largely predicted with data-driven methods. These findings as well as applications of the overall approach could potentially inform the design and timing of diabetes or cost-containment interventions, such as medication adherence or interventions that enhance access to care, among patients with type 2 diabetes.

Utilization Patterns of Oral Disease-Modifying Drugs in Commercially Insured Patients with Multiple Sclerosis

BACKGROUND

The approval of new oral disease-modifying drugs (DMDs), such as fingolimod, dimethyl fumarate (DMF), and teriflunamide, has considerably expanded treatment options for relapsing forms of multiple sclerosis (MS). However, data describing the use of these agents in routine clinical practice are limited.

OBJECTIVE

To describe time trends and identify factors associated with oral DMD treatment initiation and switching among individuals with MS.

METHODS

Using data from a large sample of commercially insured patients, we evaluated changes over time in the proportion of MS patients who initiated treatment with an oral DMD and who switched from an injectable DMD to an oral DMD between 2009 and 2014 in the United States. We evaluated predictors of oral DMD use using conditional logistic regression in 2 groups matched on calendar time: oral DMD initiators matched to injectable DMDs initiators and oral DMD switchers matched to those who switched to a second injectable DMD.

RESULTS

Our cohort included 7,576 individuals who initiated a DMD and 1,342 who switched DMDs, of which oral DMDs accounted for 6% and 39%, respectively. Oral DMD initiation and switching steadily increased from 5% to 16% and 35% to 84%, respectively, between 2011 and 2014, with DMF being the most commonly used agent. Of the potential predictors with clinical significance, a recent neurologist consultation (OR = 1.60; 95% CI = 1.20-2.15) and emergency department visit (OR = 1.43; 95% CI = 1.01-2.01) were significantly associated with oral DMD initiation. History of depression was noted to be a potential predictor of oral DMD initiation; however, the estimate for this predictor did not reach statistical significance (OR = 1.35; 95% CI = 0.99-1.84). No clinically relevant factors measured in our data were associated with switching to an oral DMD.

CONCLUSIONS

Oral DMDs were found to be routinely used as second-line treatment. However, we identified few factors predictive of oral DMD initiation or switching, which implies that their selection is driven by patient and/or physician preferences.

DISCLOSURES

This study was funded by CVS Caremark through an unrestricted research grant to Brigham and Women's Hospital. Shrank and Matlin were employees of, and shareholders in, CVS Health at the time of the study; they report no financial interests in products or services that are related to the subject of this study. Spettell is an employee of, and shareholder in, Aetna. Chitnis serves on clinical trial advisory boards for Novartis and Genzyme-Sanofi; has consulted for Bayer, Biogen Idec, Celgene, Novartis, Merck-Serono, and Genentech-Roche; and has received research support from NIH, National Multiple Sclerosis Society, Peabody Foundation, Consortium for MS Centers, Guthy Jackson Charitable Foundation, EMD-Serono, Novartis Biogen, and Verily. Desai reports receiving a research grant from Merck for unrelated work. Gagne is principal investigator of a research grant from Novartis Pharmaceuticals Corporation to the Brigham and Women's Hospital and has received grant support from Eli Lilly, all for unrelated work. He is also a consultant to Aetion and Optum. Minden reports grants from Biogen and other fees from Genentech, EMD Serano, Avanir, and Novartis, unrelated to this study. The other authors have no conflicts to report. This study was presented as a poster at the International Society for Pharmacoepidemiology 32nd Annual Meeting; August 25-28, 2016; Dublin, Ireland.

Longitudinal Trends and Variation in Antipsychotic Use in Older Adults After Cardiac Surgery

OBJECTIVES

To evaluate temporal trends and between-hospital variation in off-label antipsychotic medication (APM) use in older adults undergoing cardiac surgery.

DESIGN

Retrospective cohort study.

SETTING

National administrative database including 465 U.S. hospitals.

PARTICIPANTS

Individuals aged 65 and older without known indications for APMs who underwent cardiac surgery from 2004 to 2014 (N=293,212).

MEASUREMENTS

Postoperative exposure to any APMs and potentially excessive dosing were examined. Hospital-level APM prescribing intensity was defined as the proportion of individuals newly treated with APMs in the postoperative period.

RESULTS

The rate of APM use declined from 8.8% in 2004 to 6.2% in 2014 (p<.001). Use of haloperidol (parenteral 7.0% to 4.5%, p<.001; oral: 1.9% to 0.5%, p<.001), and risperidone (1.1% to 0.3%, p<.001) declined, whereas quetiapine use tripled (0.6% to 1.9%, p=.03). Hospital APM prescribing intensity varied widely, from 0.3% to 35.6%, across 465 hospitals. Treated individuals at higher-prescribing hospitals were more likely to receive APMs on the day of discharge (highest vs lowest quintile: 15.1% vs 9.6%; p<.001) and for a longer duration (4.8 vs 3.7 days; p<.001) than those at lower-prescribing hospitals. Delirium was the strongest risk factor for APM exposure (odds ratio=9.73, 95% confidence interval=9.02-10.5), whereas none of the hospital characteristics were significantly associated. The rate of potentially excessive dosing declined (60.7% to 44.9%, p<.001), and risk factors for potentially excessive dosing were similar to those for any APM exposure.

CONCLUSIONS

Our findings suggest highly variable prescribing cultures and raise concerns about inappropriate use, highlighting the need for better evidence to guide APM prescribing in hospitalized older adults after cardiac surgery.

Association of Osteoporosis Medication Use After Hip Fracture With Prevention of Subsequent Nonvertebral Fractures: An Instrumental Variable Analysis

IMPORTANCE

Osteoporosis medication treatment is recommended after hip fracture, yet contemporary estimates of rates of initiation and clinical benefit in the patient population receiving routine care are not well documented.

OBJECTIVES

To report osteoporosis treatment initiation rates between January 1, 2004, and September 30, 2015, and to estimate the risk reduction in subsequent nonvertebral fractures associated with treatment initiation in patients with hip fracture.

DESIGN, SETTING, AND PARTICIPANTS

In this cohort study, data from a commercial insurance claims database from the United States were analyzed. Patients 50 years and older who had a hip fracture and were not receiving treatment with osteoporosis medications before their fracture were included.

EXPOSURE

Prescription dispensing of an osteoporosis medication within 180 days of a hip fracture hospitalization.

MAIN OUTCOMES AND MEASURES

Each initiation episode was matched with 10 nonuse episodes on person-time after the index hip fracture event to preclude immortal time bias and followed up for the outcome of nonvertebral fracture until change in exposure or a censoring event. An instrumental variable analysis using 2-stage residual inclusion method was conducted using calendar year, specialist access, geographical variation in prescribing patterns, and hospital preference.

RESULTS

Among 97 169 patients with a hip fracture identified, the mean (SD) age was 80.2 (10.8) years, and 64 164 (66.0%) were women. A continuous decline over the study years was observed in osteoporosis medication initiation rates from 9.8% (95% CI, 9.0%-10.6%) in 2004 to 3.3% (95% CI, 2.9%-3.8%) in 2015. In the effectiveness analyses, the hospital preference instrumental variable had a stronger association with treatment (pseudo R2 = 0.20) than the other 3 instrumental variables (specialist access: pseudo R2 = 0.04; calendar year: pseudo R2 = 0.05; and geographic variation: pseudo R2 = 0.07). Instrumental variable analysis with hospital preference suggested a rate difference of 4.2 events (95% CI, 1.1-7.3) per 100 person-years in subsequent fractures associated with osteoporosis treatment initiation compared with nonuse in an additive hazard model.

CONCLUSIONS AND RELEVANCE

Low rates of osteoporosis treatment initiation after a hip fracture in recent years were observed. Clinically meaningful reduction in subsequent nonvertebral fracture rates associated with treatment suggests that improving prescriber adherence to guidelines and patient adherence to prescribed regimens may result in notable public health benefit.

A Randomized Controlled Feasibility Trial in Behavioral Weight Management for Underserved Postpartum African American Women: The RENEW Study

We aimed to test the feasibility of an in-person behavioral weight-loss intervention for underserved postpartum African American women with overweight or obesity in an urban hospital setting. Participants were randomized to an intervention of a culturally tailored adaptation of the Diabetes Prevention Program or usual care. The primary outcome was program satisfaction. Women who completed the intervention reported higher levels of satisfaction with the program, despite low attendance rates at group meetings. The intervention was not feasible because of these low rates of attendance and high rates of attrition after randomization. Offering the program electronically and off-site for convenience and more psychosocial support for postpartum women with obesity may improve feasibility.

International trends in clozapine use: a study in 17 countries

OBJECTIVE

There is some evidence that clozapine is significantly underutilised. Also, clozapine use is thought to vary by country, but so far no international study has assessed trends in clozapine prescribing. Therefore, this study aimed to assess clozapine use trends on an international scale, using standardised criteria for data analysis.

METHOD

A repeated cross-sectional design was applied to data extracts (2005-2014) from 17 countries worldwide.

RESULTS

In 2014, overall clozapine use prevalence was greatest in Finland (189.2/100 000 persons) and in New Zealand (116.3/100 000), and lowest in the Japanese cohort (0.6/100 000), and in the privately insured US cohort (14.0/100 000). From 2005 to 2014, clozapine use increased in almost all studied countries (relative increase: 7.8-197.2%). In most countries, clozapine use was highest in 40-59-year-olds (range: 0.6/100 000 (Japan) to 344.8/100 000 (Finland)). In youths (10-19 years), clozapine use was highest in Finland (24.7/100 000) and in the publicly insured US cohort (15.5/100 000).

CONCLUSION

While clozapine use has increased in most studied countries over recent years, clozapine is still underutilised in many countries, with clozapine utilisation patterns differing significantly between countries. Future research should address the implementation of interventions designed to facilitate increased clozapine utilisation.

Can purchasing information be used to predict adherence to cardiovascular medications? An analysis of linked retail pharmacy and insurance claims data

OBJECTIVE

The use of retail purchasing data may improve adherence prediction over approaches using healthcare insurance claims alone.

DESIGN

Retrospective.

SETTING AND PARTICIPANTS

A cohort of patients who received prescription medication benefits through CVS Caremark, used a CVS Pharmacy ExtraCare Health Care (ECHC) loyalty card, and initiated a statin medication in 2011.

OUTCOME

We evaluated associations between retail purchasing patterns and optimal adherence to statins in the 12 subsequent months.

RESULTS

Among 11 010 statin initiators, 43% were optimally adherent at 12 months of follow-up. Greater numbers of store visits per month and dollar amount per visit were positively associated with optimal adherence, as was making a purchase on the same day as filling a prescription (p<0.0001 for all). Models to predict adherence using retail purchase variables had low discriminative ability (C-statistic: 0.563), while models with both clinical and retail purchase variables achieved a C-statistic of 0.617.

CONCLUSIONS

While the use of retail purchases may improve the discriminative ability of claims-based approaches, these data alone appear inadequate for adherence prediction, even with the addition of more complex analytical approaches. Nevertheless, associations between retail purchasing behaviours and adherence could inform the development of quality improvement interventions.

GLYCEMIC OUTCOMES 3 YEARS AFTER IMPLEMENTATION OF A PERI-OPERATIVE GLYCEMIC CONTROL ALGORITHM IN AN ACADEMIC INSTITUTION

OBJECTIVE

While hyperglycemia in the postoperative setting has been linked to an increase in surgical complications, limited data are available to inform the management of patients with diabetes in the operating room and the immediate peri-operative period. We describe the results of a peri-operative glycemic control program that standardized intravenous insulin with a target glucose (BG) range of 120 to 180 mg/dL for patients with diabetes presenting with a BG level >180 mg/dL and included transition to subcutaneous insulin.

METHODS

Patients with known diabetes and a BG >180 mg/dL who underwent surgery were included. The control group included 260 patients from March 2, 2008 through December 31, 2008. The intervention group included 588 patients following protocol implementation from April 1, 2009 through December 31, 2012. Data included demographic information, hospital BG values, length of stay (LOS), mortality, and wound infections.

RESULTS

The intervention group had significantly lower BG on arrival in the postoperative care unit (182.2 vs 194.9 mg/dL, P = .012). Mean BG during the first 24 hours after surgery was lower in the intervention group (182.1 vs. 190.5 mg/dL), and there were fewer BG values >200 mg/dL in the intervention group (P = .005). The percentage of BG values <70 mg/dL during the hospital stay was lower in the intervention group (1.94 vs. 2.43%, P<.01). There was no significant difference in mortality, LOS, or wound infections.

CONCLUSION

Following implementation of a hospital-wide peri-operative glycemic control algorithm, we found a reduction in peri-operative BG levels and hypoglycemia rates. Ongoing research is needed to assess the impact on clinical outcomes.

ABBREVIATIONS

BG = blood glucose CCI = Charlson comorbidity index EHR = electronic health record ICD-9 = International Classification of Disease-9 IV = intravenous LOS = length of stay OR = operating room PACU = postoperative care unit POC = point-of-care.