Introduction: Relative impacts of coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI) on mortality and end-stage kidney disease (ESKD) in chronic kidney disease (CKD) are uncertain.
Methods: Data from Massachusetts residents with CKD undergoing CABG or PCI from 2003 to 2012 were linked to the United States Renal Data System. Associations with death, ESKD, and combined death and ESKD were analyzed in propensity score-matched multivariable survival models.
Results: We identified 6805 CABG and 17,494 PCI patients. Among 3775 matched-pairs, multi-vessel disease was present in 97%, and stage 4 CKD was present in 11.9% of CABG and 12.2% of PCI patients. One-year mortality (CABG 7.7%, PCI 11.0%) was more frequent than ESKD (CABG 1.4%, PCI 1.7%). Overall survival was improved and ESKD risk decreased with CABG compared to PCI, but effects differed in the presence of left main disease and prior myocardial infarction (MI). Survival was worse following PCI than following CABG among patients with left main disease and without MI (hazard ratio = 3.7, 95% confidence interval = 1.3-10.5). ESKD risk was higher with PCI for individuals with left main disease and prior infarction (hazard ratio = 8.1, 95% confidence interval = 1.7-39.2).
Conclusion: Risks following CABG and PCI were modified by left main disease and prior MI. In individuals with CKD, survival was greater after CABG than after PCI in patients with left main disease but without MI, whereas ESKD risk was lower with CABG in those with left main and MI. Absolute risks of ESKD were markedly lower than for mortality, suggesting prioritizing mortality over ESKD in clinical decision making.
BACKGROUND/OBJECTIVES: Gabapentinoids are commonly prescribed to relieve pain. The development of edema, an established adverse effect of gabapentinoids, may lead to a potentially harmful prescribing cascade whereby individuals are subsequently prescribed diuretics and exposed to diuretic-induced adverse events. The frequency of this prescribing cascade is unknown. Our objective was to measure the association between new dispensing of a gabapentinoid and the subsequent dispensing of a diuretic in older adults with new low back pain.
DESIGN: Population-based cohort study.
SETTING: Ontario, Canada.
PARTICIPANTS: A total of 260,344 community-dwelling adults aged 66 years or older, newly diagnosed with low back pain between April 1, 2011, and March 31, 2019.
MEASUREMENTS: Exposure status was assigned using dispensed medications in the 1 week after low back pain diagnosis. Older adults newly dispensed a gabapentinoid (N = 7867) were compared with older adults who were not newly dispensed a gabapentinoid (N = 252,477). Hazard ratios (HRs) with 95% confidence intervals (CIs) for dispensing of a diuretic within 90 days of follow-up among older adults prescribed gabapentin relative to those who were not.
RESULTS: Older adults newly dispensed a gabapentinoid had a higher risk of being subsequently dispensed a diuretic within 90 days compared with older adults who were not prescribed a gabapentinoid (2.0% vs. 1.3%). After covariate adjustment, new gabapentinoid users had a higher rate of being dispensed a diuretic compared with those not prescribed a gabapentinoid (HR: 1.44, 95% CI: 1.23, 1.70). The rate of diuretic prescription among new gabapentinoid users increased with increasing gabapentinoid dosages.
CONCLUSIONS: We have demonstrated the presence of a potentially inappropriate and harmful prescribing cascade. Given the widespread use of gabapentinoids, the population-based scale of this problem may be substantial. Increased awareness of this prescribing cascade is required to reduce the unnecessary use of diuretics and the exposure of patients to additional adverse drug events.
OBJECTIVE: Off-label utilization of second-generation antipsychotic medications may expose patients to significant risks. The authors examined the prevalence, temporal trends, and factors associated with off-label utilization of second-generation antipsychotics among publicly insured adults.
METHODS: A retrospective repeated panel was used to examine monthly off-label utilization of second-generation antipsychotics among fee-for-service Medicare, Medicaid, and dually eligible White, Black, and Latino adult beneficiaries filling prescriptions for second-generation antipsychotics in California, Georgia, Mississippi, and Oklahoma from July 2008 through June 2013.
RESULTS: Among 301,367 users of second-generation antipsychotics, between 36.5% and 41.9% had utilization that was always off-label. Payer did not modify effects of race-ethnicity on off-label utilization. Compared with Whites, Blacks had lower monthly odds of off-label utilization in all four states, and Latinos had lower odds of utilization in California and Georgia. Payer was associated with off-label utilization in California, Mississippi, and Oklahoma. California Medicaid beneficiaries were 1.12 (95% confidence interval=1.10-1.13) times as likely as dually eligible beneficiaries to have off-label utilization. Off-label utilization increased relative to the baseline year in all states, but a downward trend followed in three states.
CONCLUSIONS: Off-label utilization of second-generation antipsychotics was prevalent despite the drugs' cardiometabolic risks and little evidence of their effectiveness. The lower likelihood of off-label utilization among patients from racial-ethnic minority groups might stem from prescribers' efforts to minimize risks, given a higher baseline risk for these groups, or from disparities-associated factors. Variation among payers suggests that payer policies can affect off-label utilization.
BACKGROUND/OBJECTIVES: No data exist regarding the validity of International Classification of Disease (ICD)-10 dementia diagnoses against a clinician-adjudicated reference standard within Medicare claims data. We examined the accuracy of claims-based diagnoses with respect to expert clinician adjudication using a novel database with individual-level linkages between electronic health record (EHR) and claims.
DESIGN: In this retrospective observational study, two neurologists and two psychiatrists performed a standardized review of patients' medical records from January 2016 to December 2018 and adjudicated dementia status. We measured the accuracy of three claims-based definitions of dementia against the reference standard.
SETTING: Mass-General-Brigham Healthcare (MGB), Massachusetts, USA.
PARTICIPANTS: From an eligible population of 40,690 fee-for-service (FFS) Medicare beneficiaries, aged 65 years and older, within the MGB Accountable Care Organization (ACO), we generated a random sample of 1002 patients, stratified by the pretest likelihood of dementia using administrative surrogates.
MEASUREMENTS: We evaluated the accuracy (area under receiver operating curve [AUROC]) and calibration (calibration-in-the-large [CITL] and calibration slope) of three ICD-10 claims-based definitions of dementia against clinician-adjudicated standards. We applied inverse probability weighting to reconstruct the eligible population and reported the mean and 95% confidence interval (95% CI) for all performance characteristics, using 10-fold cross-validation (CV).
RESULTS: Beneficiaries had an average age of 75.3 years and were predominately female (59%) and non-Hispanic whites (93%). The adjudicated prevalence of dementia in the eligible population was 7%. The best-performing definition demonstrated excellent accuracy (CV-AUC 0.94; 95% CI 0.92-0.96) and was well-calibrated to the reference standard of clinician-adjudicated dementia (CV-CITL <0.001, CV-slope 0.97).
CONCLUSION: This study is the first to validate ICD-10 diagnostic codes against a robust and replicable approach to dementia ascertainment, using a real-world clinical reference standard. The best performing definition includes diagnostic codes with strong face validity and outperforms an updated version of a previously validated ICD-9 definition of dementia.
OBJECTIVE: To examine whether quality of dental care varies by age and over time and whether community-level characteristics explain these patterns.
DATA SOURCE: Deidentified medical and dental claims from a commercial insurer from January 2015 to December 2019.
STUDY DESIGN: A retrospective cohort study. The primary outcome was a composite quality score, derived from seven dental quality measures (DQMs), with higher values corresponding to better quality. Hierarchical regression models identified person- and zip code-level factors associated with the quality.
DATA COLLECTION/EXTRACTION METHODS: Continuously enrolled US dental insurance beneficiaries younger than 21 years of age.
PRINCIPAL FINDINGS: Quality was assessed for 4.88 million person-years covering 1.31 million persons. Overall quality slightly improved over time, mostly driven by substantial improvements among children aged 0-5 years by 0.153 points/year (95% confidence interval [CI]:0.151, 0.156). Quality was poorest and declined over time among adolescents with only 20.5% of DQMs met as compared to 42.6% among aged 0-5 years in 2019. Dental professional shortage, median household income, percentages of African Americans, unemployed, and less-educated populations at the zip code level were associated with the composite score.
CONCLUSION: Quality of dental care among adolescents remains low, and place of residence influenced the quality. Increasing the supply of dentists and oral health promotion strategies targeting adolescents and low-performing localities should be explored.
BACKGROUND: New methods such as machine learning techniques have been increasingly used to enhance the performance of risk predictions for clinical decision-making. However, commonly reported performance metrics may not be sufficient to capture the advantages of these newly proposed models for their adoption by health care professionals to improve care. Machine learning models often improve risk estimation for certain subpopulations that may be missed by these metrics.
METHODS AND RESULTS: This article addresses the limitations of commonly reported metrics for performance comparison and proposes additional metrics. Our discussions cover metrics related to overall performance, discrimination, calibration, resolution, reclassification, and model implementation. Models for predicting acute kidney injury after percutaneous coronary intervention are used to illustrate the use of these metrics.
CONCLUSIONS: We demonstrate that commonly reported metrics may not have sufficient sensitivity to identify improvement of machine learning models and propose the use of a comprehensive list of performance metrics for reporting and comparing clinical risk prediction models.
Machine learning algorithms are increasingly used in the clinical literature, claiming advantages over logistic regression. However, they are generally designed to maximize the area under the receiver operating characteristic curve. While area under the receiver operating characteristic curve and other measures of accuracy are commonly reported for evaluating binary prediction problems, these metrics can be misleading. We aim to give clinical and machine learning researchers a realistic medical example of the dangers of relying on a single measure of discriminatory performance to evaluate binary prediction questions. Prediction of medical complications after surgery is a frequent but challenging task because many post-surgery outcomes are rare. We predicted post-surgery mortality among patients in a clinical registry who received at least one aortic valve replacement. Estimation incorporated multiple evaluation metrics and algorithms typically regarded as performing well with rare outcomes, as well as an ensemble and a new extension of the lasso for multiple unordered treatments. Results demonstrated high accuracy for all algorithms with moderate measures of cross-validated area under the receiver operating characteristic curve. False positive rates were <1%, however, true positive rates were <7%, even when paired with a 100% positive predictive value, and graphical representations of calibration were poor. Similar results were seen in simulations, with the addition of high area under the receiver operating characteristic curve (>90%) accompanying low true positive rates. Clinical studies should not primarily report only area under the receiver operating characteristic curve or accuracy.
Antipsychotic polypharmacy (APP) lacks evidence of effectiveness in the care of schizophrenia or other disorders for which antipsychotic drugs are indicated, also exposing patients to more risks. Authors assessed APP prevalence and APP association with beneficiary race/ethnicity and payer among publicly-insured adults regardless of diagnosis. Retrospective repeated panel study of fee-for-service (FFS) Medicare, Medicaid, and dually-eligible white, black, and Latino adults residing in California, Georgia, Iowa, Mississippi, Oklahoma, South Dakota, or West Virginia, filling antipsychotic prescriptions between July 2008 and June 2013. Primary outcome was any monthly APP utilization. Across states and payers, 11% to 21% of 397,533 antipsychotic users and 12% to 19% of 9,396,741 person-months had some APP utilization. Less than 50% of person-months had a schizophrenia diagnosis and up to 19% had no diagnosed mental illness. Payer modified race/ethnicity effects on APP utilization only in CA; however, the odds of APP utilization remained lower for minorities than for whites. Elsewhere, the odds varied by race/ethnicity only in OK, with Latinos having lower odds than whites (odds ratio 0.76; 95% confidence interval 0.60-0.96). The odds of APP utilization varied by payer in several study states, with odds generally higher for Dual eligibles, although the differences were generally small; the odds also varied by year (lower at study end). APP was frequently utilized but mostly declined over time. APP utilization patterns varied across states, with no consistent association with race/ethnicity and small payer effects. Greater use of APP-reducing strategies are needed, particularly among non-schizophrenia populations.
Importance: Accurate prediction of adverse outcomes after acute myocardial infarction (AMI) can guide the triage of care services and shared decision-making, and novel methods hold promise for using existing data to generate additional insights.
Objective: To evaluate whether contemporary machine learning methods can facilitate risk prediction by including a larger number of variables and identifying complex relationships between predictors and outcomes.
Design, Setting, and Participants: This cohort study used the American College of Cardiology Chest Pain-MI Registry to identify all AMI hospitalizations between January 1, 2011, and December 31, 2016. Data analysis was performed from February 1, 2018, to October 22, 2020.
Main Outcomes and Measures: Three machine learning models were developed and validated to predict in-hospital mortality based on patient comorbidities, medical history, presentation characteristics, and initial laboratory values. Models were developed based on extreme gradient descent boosting (XGBoost, an interpretable model), a neural network, and a meta-classifier model. Their accuracy was compared against the current standard developed using a logistic regression model in a validation sample.
Results: A total of 755 402 patients (mean [SD] age, 65  years; 495 202 [65.5%] male) were identified during the study period. In independent validation, 2 machine learning models, gradient descent boosting and meta-classifier (combination including inputs from gradient descent boosting and a neural network), marginally improved discrimination compared with logistic regression (C statistic, 0.90 for best performing machine learning model vs 0.89 for logistic regression). Nearly perfect calibration in independent validation data was found in the XGBoost (slope of predicted to observed events, 1.01; 95% CI, 0.99-1.04) and the meta-classifier model (slope of predicted-to-observed events, 1.01; 95% CI, 0.99-1.02), with more precise classification across the risk spectrum. The XGBoost model reclassified 32 393 of 121 839 individuals (27%) and the meta-classifier model reclassified 30 836 of 121 839 individuals (25%) deemed at moderate to high risk for death in logistic regression as low risk, which were more consistent with the observed event rates.
Conclusions and Relevance: In this cohort study using a large national registry, none of the tested machine learning models were associated with substantive improvement in the discrimination of in-hospital mortality after AMI, limiting their clinical utility. However, compared with logistic regression, XGBoost and meta-classifier models, but not the neural network, offered improved resolution of risk for high-risk individuals.
Importance: Little is known about how new and expensive drugs diffuse into practice affects health care costs.
Objective: To describe the variation in second-generation diabetes drug use among Medicare enrollees between 2007 and 2015.
Design, Setting, and Participants: This population-based, cross-sectional study included data from 100% of Medicare Parts A, B, and D enrollees who first received diabetes drug therapy from January 1, 2007, to December 31, 2015. Patients with type 1 diabetes were excluded. Data were analyzed beginning in the spring of 2018, and revisions were completed in 2019.
Exposures: For each patient, the initial diabetes drug choice was determined; drugs were classified as first generation (ie, approved before 2000) or second generation (ie, approved after 2000, including dipeptidyl peptidase 4 [DPP-4] inhibitors, glucagon-like peptide-1 [GLP-1] receptor agonists, and sodium-glucose cotransporter-2 [SGLT-2] inhibitors).
Main Outcomes and Measures: The primary outcome was the between-practice variation in use of second-generation diabetes drugs between 2007 and 2015. Practices with use rates of second-generation diabetes drugs more than 1 SD above the mean were considered high prescribing, while those with use rates more than 1 SD below the mean were considered low prescribing.
Results: Among 1 182 233 patients who initiated diabetes drug therapy at 42 977 practices between 2007 and 2015, 1 104 718 (93.4%) were prescribed a first-generation drug (mean [SD] age, 75.4 [6.7] years; 627 134 [56.8%] women) and 77 515 (6.6%) were prescribed a second-generation drug (mean [SD] age, 76.5 [7.2] years; 44 697 [57.7%] women). By December 2015, 22 457 practices (52.2%) had used DPP-4 inhibitors once, compared with 3593 practices (8.4%) that had used a GLP-1 receptor agonist once. Furthermore, 17 452 practices (40.6%) were using DPP-4 inhibitors in 10% of eligible patients, while 1286 practices (3.0%) were using GLP-1 receptor agonists in 10% of eligible patients, and SGLT-2 inhibitors, available after March 2013, were used at least once by 1716 practices (4.0%) and used in 10% of eligible patients by 872 practices (2.0%) by December 2015. According to Poisson random-effect regression models, beneficiaries in high-prescribing practices were more than 3-fold more likely to receive DPP-4 inhibitors (relative risk, 3.55 [95% CI, 3.42-3.68]), 24-fold more likely to receive GLP-1 receptor agonists (relative risk, 24.06 [95% CI, 14.14-40.94]) and 60-fold more likely to receive SGLT-2 inhibitors (relative risk, 60.41 [95% CI, 15.99-228.22]) compared with beneficiaries in low-prescribing practices.
Conclusions and Relevance: These findings suggest that there was substantial between-practice variation in the use of second-generation diabetes drugs between 2007 and 2015, with a concentration of use among a few prescribers and practices responsible for much of the early diffusion.
Importance: Studies have shown that adverse events are associated with increasing inpatient care expenditures, but contemporary data on the association between expenditures and adverse events beyond inpatient care are limited.
Objective: To evaluate whether hospital-specific adverse event rates are associated with hospital-specific risk-standardized 30-day episode-of-care Medicare expenditures for fee-for-service patients discharged with acute myocardial infarction (AMI), heart failure (HF), or pneumonia.
Design, Setting, and Participants: This cross-sectional study used the 2011 to 2016 hospital-specific risk-standardized 30-day episode-of-care expenditure data from the Centers for Medicare & Medicaid Services and medical record-abstracted in-hospital adverse event data from the Medicare Patient Safety Monitoring System. The setting was acute care hospitals treating at least 25 Medicare fee-for-service patients for AMI, HF, or pneumonia in the United States. Participants were Medicare fee-for-service patients 65 years or older hospitalized for AMI, HF, or pneumonia included in the Medicare Patient Safety Monitoring System in 2011 to 2016. The dates of analysis were July 16, 2017, to May 21, 2018.
Main Outcomes and Measures: Hospitals' risk-standardized 30-day episode-of-care expenditures and the rate of occurrence of adverse events for which patients were at risk.
Results: The final study sample from 2194 unique hospitals included 44 807 patients (26.1% AMI, 35.6% HF, and 38.3% pneumonia) with a mean (SD) age of 79.4 (8.6) years, and 52.0% were women. The patients represented 84 766 exposures for AMI, 96 917 exposures for HF, and 109 641 exposures for pneumonia. Patient characteristics varied by condition but not by expenditure category. The mean (SD) risk-standardized expenditures were $22 985 ($1579) for AMI, $16 020 ($1416) for HF, and $16 355 ($1995) for pneumonia per hospitalization. The mean risk-standardized rates of occurrence of adverse events for which patients were at risk were 3.5% (95% CI, 3.4%-3.6%) for AMI, 2.5% (95% CI, 2.5%-2.5%) for HF, and 3.0% (95% CI, 2.9%-3.0%) for pneumonia. An increase by 1 percentage point in the rate of occurrence of adverse events was associated with an increase in risk-standardized expenditures of $103 (95% CI, $57-$150) for AMI, $100 (95% CI, $29-$172) for HF, and $152 (95% CI, $73-$232) for pneumonia per discharge.
Conclusions and Relevance: Hospitals with high adverse event rates were more likely to have high 30-day episode-of-care Medicare expenditures for patients discharged with AMI, HF, or pneumonia.
Background Patients who survive acute myocardial infarction (AMI) are at high risk for recurrence. We determined whether rehospitalizations after AMI further increased risk of recurrent AMI. Methods and Results The study included Medicare fee-for-service patients aged ≥65 years discharged alive after AMI from acute-care hospitals in fiscal years 2009-2014. The outcome was recurrent AMI within 1 year of the index AMI. The Clinical Classifications Software (CCS) was used to classify rehospitalizations into disease categories. A Cox regression model was fit accounting for CCS-specific hospitalizations as time-varying variables and patient characteristics at discharge for the index AMI, adjusting for the competing risk of death. The rate of 1-year recurrent AMI was 5.3% (95% CI, 5.27%-5.41%), and median (interquartile range) time from discharge to recurrent AMI was 115 (34-230) days. Eleven disease categories (diabetes mellitus, anemia, hypertension, coronary atherosclerosis, chest pain, heart failure, pneumonia, chronic obstructive pulmonary disease, gastrointestinal hemorrhage, renal failure, complication of implant or graft) were associated with increased risk of recurrent AMI. Septicemia was associated with lower recurrence risk. Hazard ratios ranged from 1.6 (95% CI, 1.55-1.70, heart failure) to 1.1 (95% CI, 1.04-1.25, pneumonia) to 0.6 (95% CI, 0.58-0.71, septicemia). Conclusions Patient risk of recurrent AMI changed based on the occurrence of hospitalizations after the index AMI. Improving post-acute care to prevent unplanned rehospitalizations, especially rehospitalizations for chronic diseases, and extending the focus of outcomes measures to condition-specific rehospitalizations within 30 days and beyond is important for the secondary prevention of AMI.
OBJECTIVE: To investigate risk-adjusted, 30-day postdischarge heart failure mortality and readmission rates stratified by hospital teaching intensity.
DATA SOURCES AND STUDY SETTING: A total of 709 221 Medicare fee-for-service beneficiaries discharged from 3135 US hospitals between 1/1/2013 and 11/30/2014 with a principal diagnosis of heart failure.
STUDY DESIGN: Hospitals were classified as Council of Teaching Hospitals and Health Systems (COTH) major teaching hospitals, non-COTH teaching hospitals, and nonteaching hospitals. Hospital teaching status was linked with MedPAR patient data and FY2016 Hospital Readmission Reduction Program penalties. Index hospitalization survival probabilities were estimated with hierarchical logistic regression and used to stratify index hospitalization survivors into severity deciles. Decile-specific models were estimated for 30-day postdischarge readmission and mortality. Thirty-day postdischarge outcomes were estimated by teaching intensity and penalty categories.
PRINCIPAL FINDINGS: Averaged across deciles, adjusted 30-day COTH hospital readmission rates were, on a relative scale ([COTH minus nonteaching] ÷ nonteaching), 1.63 percent higher (95% CI: 0.89 percent, 2.25 percent) than at nonteaching hospitals, but their average adjusted 30-day postdischarge mortality rates were 11.55 percent lower (95% CI: -13.78 percent, -9.37 percent). Penalized COTH hospitals had the highest readmission rates of all categories (23.99 percent [95% CI: 23.50 percent, 24.49 percent]) but the lowest 30-day postdischarge mortality (8.30 percent [95% CI: 7.99 percent, 8.57 percent] vs 9.84 percent [95% CI: 9.69 percent, 9.99 percent] for nonpenalized, nonteaching hospitals).
CONCLUSIONS: Heart failure readmission penalties disproportionately impact major teaching hospitals and inadequately credit their better postdischarge survival.
Background More than 600 000 coronary stents are implanted during percutaneous coronary interventions (PCIs) annually in the United States. Because no real-world surveillance system exists to monitor their long-term safety, claims data are often used for this purpose. The extent to which adverse events identified with claims data can be reasonably attributed to a specific medical device is uncertain. Methods and Results We used deterministic matching to link the NCDR (National Cardiovascular Data Registry) CathPCI Registry to Medicare fee-for-service claims for patients aged ≥65 years who underwent PCI with drug-eluting stents (DESs) between July 1, 2009 and December 31, 2013. We identified subsequent PCIs within 1 year of the index procedure in Medicare claims as potential safety events. We linked these subsequent PCIs back to the NCDR CathPCI Registry to ascertain how often the revascularization could be reasonably attributed to the same coronary artery as the index PCI (ie, target vessel revascularization). Of 415 306 DES placements in 368 194 patients, 33 174 repeat PCIs were identified in Medicare claims within 1 year. Of these, 28 632 (86.3%) could be linked back to the NCDR CathPCI Registry; 16 942 (51.1% of repeat PCIs) were target vessel revascularizations. Of these, 8544 (50.4%) were within a previously placed DES: 7652 for in-stent restenosis and 1341 for stent thrombosis. Of 16 176 patients with a claim for acute myocardial infarction in the follow-up period, 4446 (27.5%) were attributed to the same coronary artery in which the DES was implanted during the index PCI (ie, target vessel myocardial infarction). Of 24 288 patients whose death was identified in claims data, 278 (1.1%) were attributed to the same coronary artery in which the DES was implanted during the index PCI. Conclusions Most repeat PCIs following DES stent implantation identified in longitudinal claims data could be linked to real-world registry data, but only half could be reasonably attributed to the same coronary artery as the index procedure. Attribution among those with acute myocardial infarction or who died was even less frequent. Safety signals identified using claims data alone will require more in-depth examination to accurately assess stent safety.
New prescription medications are a primary driver of spending growth in the United States. For patients with severe mental illnesses, second-generation antipsychotic (SGA) medications feature prominently. However, many SGAs are costly, particularly before generic entry, and some may increase the risk of diabetes. Because physicians play a prominent role in new prescription adoption, understanding their prescribing behaviors is policy-relevant. Several features of prescription data, such as different antipsychotic choice sets over time, variable physician prescription volumes, and correlation among drug choices within physicians, complicate inferences. We propose a multivariate Bayesian hierarchical model with piecewise random effects to characterize the diffusion of new antipsychotic drugs. This model captures the complex prescriber-specific relationships among the different diffusion processes and takes advantage of the Bayesian paradigm to quantify uncertainty for all parameters straightforwardly. To evaluate the prescribing patterns for each physician, we propose various indices to identify early new SGA adopters. A sample of nearly 17,000 US physicians whose antipsychotic drug prescribing information was collected between January 1, 1997 and December 31, 2007 illustrates the methods. Determinants of high prescription rates and adoption speeds of new SGAs included physician sex, age, hospital affiliation, physician specialty, and office location. Large within- and between-provider variations in prescribing patterns of new SGAs were identified. Early adopters for one drug were not early adopters for another drug.
In late 2018, the Food and Drug Administration (FDA) outlined a framework for evaluating the possible use of real-world evidence (RWE) to support regulatory decision-making. This framework was created to facilitate studies that would generate high-quality RWE, including pragmatic clinical trials (PCTs), which are randomized trials designed to inform clinical or policy decisions by assessing the real-world effectiveness of an intervention. There is general agreement among experts that the use of existing healthcare and patient-generated data holds promise for making randomized trials more efficient, less costly, and more generalizable. Yet the benefits of relying on real-world data sources must be weighed against difficulties with ensuring data integrity and completeness. Additionally, appropriately monitoring patient safety in randomized trials of new drugs using healthcare system data that might not be available in real time can be quite difficult. Recognizing that these and other concerns are critical to the development and acceptability of PCTs, a group of stakeholders from academia, industry, professional organizations, regulatory bodies, government agencies, and patient advocates discussed a path forward for PCT growth and sustainability at a think tank meeting entitled "Monitoring and Analyzing Data from Pragmatic Streamlined Randomized Clinical Trials," which took place in January 2019 (Washington, DC). The goals of this meeting were to: (1) evaluate study design and methodological options specific to PCTs that have the potential to yield high-quality evidence; (2) discuss best practices to ensure data quality in PCTs; and (3) identify appropriate methods for study monitoring. Proceedings from the think tank meeting are summarized in this manuscript.
OBJECTIVE: Examine patterns of alcohol use disorder (AUD) medication use and identify factors associated with prescription fill among commercially insured individuals with an index AUD visit.
DESIGN: Using 2008-2018 claims data from a large national insurer, estimate days to first AUD medication using cause-specific hazards approach to account for competing risk of benefits loss.
PARTICIPANTS: Aged 17-64 with ≥ 1 AUD visit.
MAIN MEASURE: Days to AUD medication fill.
KEY RESULTS: A total of 13.3% of the 151,128 with an index visit filled an AUD prescription after that visit, while 69.8% lost benefits before filling and 17.0% remained enrolled but did not fill (median days observed = 305). Almost half (46.3%) of those who filled a prescription received substance use disorder (SUD) inpatient care within 7 days before the fill, and 63.4% received SUD outpatient care. Likelihood of medication use was higher for those aged 26-35, 36-45, and 46-55 years relative to 56-64 years (e.g., 26-35: hazard ratio = 1.29 [95% confidence interval 1.23-1.36]); those diagnosed with moderate/severe AUD (2.05 [1.98-2.12]), co-occurring opioid use disorder (OUD) (1.33 [1.26-1.39]), or severe mental illness (1.31 [1.27-1.35]); those with a chronic alcohol-related diagnosis (1.08 [1.04-1.12]); and those whose index visit was in an inpatient/emergency department (1.27 [1.23-1.31]) or intermediate care setting (1.13 [1.07-1.20]) relative to outpatient. Likelihood of use was higher in later years relative to 2008 (e.g., 2018:2.02 [1.89-2.15]) and higher for those who received the majority of AUD care in a practice with a psychiatrist/addiction medicine specialist (1.13 [1.10-1.16]). Likelihood of use was lower for those diagnosed with a SUD other than AUD or OUD (0.88 [0.85-0.92]), those with an acute alcohol-related condition (0.79 [0.75-0.84]), and males (0.71 [0.69-0.73]).
CONCLUSIONS: While AUD medication use increased and was more common among individuals with greater severity, few patients who could benefit from medications are using them. More efforts are needed to identify and treat individuals in non-acute care settings earlier in their course of AUD.
Percutaneous coronary interventions (PCIs) are nonsurgical procedures to open blocked blood vessels to the heart, frequently using a catheter to place a stent. The catheter can be inserted into the blood vessels using an artery in the groin or an artery in the wrist. Because clinical trials have indicated that access via the wrist may result in fewer post procedure complications, shortening the length of stay, and ultimately cost less than groin access, adoption of access via the wrist has been encouraged. However, patients treated in usual care are likely to differ from those participating in clinical trials, and there is reason to believe that the effectiveness of wrist access may differ between males and females. Moreover, the choice of artery access strategy is likely to be influenced by patient or physician unmeasured factors. To study the effectiveness of the two artery access site strategies on hospitalization charges, we use data from a state-mandated clinical registry including 7,963 patients undergoing PCI. A hierarchical Bayesian likelihood-based instrumental variable analysis under a latent index modeling framework is introduced to jointly model outcomes and treatment status. Our approach accounts for unobserved heterogeneity via a latent factor structure, and permits nonparametric error distributions with Dirichlet process mixture models. Our results demonstrate that artery access in the wrist reduces hospitalization charges compared to access in the groin, with a higher mean reduction for male patients.
BACKGROUND: Evidence-based outpatient treatment for opioid use disorder (OUD) consists of medications that treat OUD (MOUD) and psychosocial treatments (e.g., psychotherapy or counseling, case management). Prior studies have not examined the use of these components of care in a commercially insured population.
METHODS: We analyzed claims data from a large national commercial insurer of enrollees age 17-64 identified with OUD (2008-2016, N = 87,877 persons and 122,708 person-years). Multinomial logistic regression models identified factors associated with receiving in a given year: 1) both MOUD and psychosocial visits, 2) MOUD without psychosocial visits, 3) psychosocial visits without MOUD, or 4) neither. We estimated predicted probabilities for key variables of interest.
RESULTS: Identification of OUD nearly tripled during the observation period (0.17% in 2008, 0.45% in 2016). Among person-years identified as having OUD, 36.3% included MOUD (8.1% both MOUD and psychosocial visits and 28.2% MOUD without psychosocial visits). In adjusted analyses, women had a lower probability of receiving either treatment alone or in combination (e.g.,MOUD plus psychosocial visits: women = 6.7% [6.5%-6.9%] vs. men = 9.2% [9.0%-9.4%]). Moderate/severe vs. mild OUD was associated with a higher probability of receiving MOUD (e.g., MOUD plus psychosocial visits: 8.7% [8.6%-8.9%] vs. 0.9% [0.7%-1.0%]). In contrast, an OUD overdose was associated with a greater probability of receiving neither treatment (78.2% [77.4%-79.0%] vs. 55.5% [55.2%-55.8%]). Over time, the probability of receiving each MOUD and psychosocial treatment category increased relative to 2008, but reached a peak and then plateaued or declined, by the end of the study period.
CONCLUSIONS: A significant treatment gap exists among individuals identified with OUD in this commercially insured population, with greater risks of receiving no treatment for women and for individuals with mild versus moderate or severe OUD. Overdose is associated with receiving neither MOUD nor psychosocial treatment. While treated prevalence initially increased relative to 2008, rates of treatment subsequently plateaued. Additional study and monitoring to elucidate barriers to OUD treatment in commercially insured populations are warranted.