Identifying treatment heterogeneity in atrial fibrillation using a novel causal machine learning method.

BACKGROUND: Lifelong oral anticoagulation is recommended in patients with atrial fibrillation (AF) to prevent stroke. Over the last decade, multiple new oral anticoagulants (OACs) have expanded the number of treatment options for these patients. While population-level effectiveness of OACs has been compared, it is unclear if there is variability in benefit and risk across patient subgroups. METHODS: We analyzed claims and medical data for 34,569 patients who initiated a nonvitamin K antagonist oral anticoagulant (non-vitamin K antagonist oral anticoagulant (NOAC); apixaban, dabigatran, and rivaroxaban) or warfarin for nonvalvular AF between 08/01/2010 and 11/29/2017 from the OptumLabs Data Warehouse. A machine learning (ML) method was applied to match different OAC groups on several baseline variables including, age, sex, race, renal function, and CHA2DS2 -VASC score. A causal ML method was then used to discover patient subgroups characterizing the head-to-head treatment effects of the OACs on a primary composite outcome of ischemic stroke, intracranial hemorrhage, and all-cause mortality. RESULTS: The mean age, number of females and white race in the entire cohort of 34,569 patients were 71.2 (SD, 10.7) years, 14,916 (43.1%), and 25,051 (72.5%) respectively. During a mean follow-up of 8.3 (SD, 9.0) months, 2,110 (6.1%) of patients experienced the composite outcome, of whom 1,675 (4.8%) died. The causal ML method identified 5 subgroups with variables favoring apixaban over dabigatran; 2 subgroups favoring apixaban over rivaroxaban; 1 subgroup favoring dabigatran over rivaroxaban; and 1 subgroup favoring rivaroxaban over dabigatran in terms of risk reduction of the primary endpoint. No subgroup favored warfarin and most dabigatran vs warfarin users favored neither drug. The variables that most influenced favoring one subgroup over another included Age, history of ischemic stroke, thromboembolism, estimated glomerular filtration rate, Race, and myocardial infarction. CONCLUSIONS: Among patients with AF treated with a NOAC or warfarin, a causal ML method identified patient subgroups with differences in outcomes associated with OAC use. The findings suggest that the effects of OACs are heterogeneous across subgroups of AF patients, which could help personalize the choice of OAC. Future prospective studies are needed to better understand the clinical impact of the subgroups with respect to OAC selection.

Association of vitamins, minerals, and lead with lipoprotein(a) in a cross-sectional cohort of US adults.

Lipoprotein(a)(Lp[a]) is a low-density lipoprotein-cholesterol (LDL-C)-like particle with potent pro-atherothrombotic properties. The association of Lp(a) with several circulating factors, including vitamins, remains unresolved. We performed an observational analysis using the National Health and Nutrition Examination Survey III cohort, a cohort used to monitor the nutrition status of US-citizens. We used multivariable linear regression to test associations of Lp(a) and LDL-C with levels of serum vitamins and minerals and whole-blood lead. Analyses controlled for factors known to associate with Lp(a) (age, sex, race/ethnicity, statin use, hemoglobin A1c, body mass index, hypertension, diabetes, glomerular filtration rate, alcohol intake, and saturated fat intake). LDL-C was corrected for Lp(a) mass. Multiple sensitivity tests were performed, including considering factors as categorical variables (deficient, normal, elevated). Among 7,662 subjects, Lp(a) correlated (ß-coefficient) positively (change per 1 conventional unit increase) with carotenoids (lycopene (0.17(0.06,0.28), p=0.005), lutein (0.19(0.07,0.30), p=0.002), ß-cryptoxanthin (0.21(0.05,0.37), p=0.01), ß-carotene (0.05(0.02,0.09), p=0.003), and α-carotene (0.15(0.01,0.30), p=0.04)) and lead (0.54(0.03,1.05), p=0.04) levels when tested as continuous variables. LDL-C had similar associations. Lp(a) did not associate with vitamins A, B12, C, or E retinyl esters, folate, RBC-folate, selenium, ferritin, transferrin saturation, or calcium. With factors as categorical variables, Lp(a) but not LDL-C negatively associated with elevated vitamin B12 (-5.41(-9.50, -1.53), p=0.01) and folate (-2.86(-5.09, -0.63), p=0.01). In conclusion, Lp(a) associated similarly to LDL-C when vitamins, minerals, and lead were tested as continuous variables, while only Lp(a) correlated with vitamin B12 and folate when tested as categorical variables. These observations are hypotheses generating and require further studies to determine causality.

Treatment Changes, Healthcare Resource Utilization, and Costs Among Patients with Symptomatic Obstructive Hypertrophic Cardiomyopathy: A Claims Database Study.

INTRODUCTION: There is limited evidence on therapies for obstructive hypertrophic cardiomyopathy (HCM), and data regarding treatment patterns and cost are scarce. This study assessed treatment patterns and economic outcomes in patients with symptomatic obstructive HCM. METHODS: Adults with symptomatic obstructive HCM as per study design and treated with pharmacotherapies [beta blockers (BBs), calcium channel blockers (CCBs), BB + CCB, or disopyramide] or procedures (septal reduction therapy, heart transplantation, implantable cardioverter-defibrillator, and pacemaker implantation) were identified from the IBM® MarketScan® Commercial and Medicare Supplemental database (January 2009-March 2019). Patients had 12-month continuous eligibility before and after (study period) treatment initiation (index treatment). Healthcare resource utilization (HRU), costs, and treatment changes were assessed. RESULTS: Of the 4883 patients included in the analysis, 85% received pharmacotherapies (BB 52.5%; CCB 11.7%; BB + CCB 17.7%; disopyramide 2.4%) and 15.7% underwent procedures. During the study period, 38, 34, and 100% of all patients had ≥ 1 inpatient stay, emergency room (ER) visit, and outpatient visit, respectively; mean total healthcare costs were US$53,053. Patients undergoing procedures had the highest HRU and costs across groups. Among patients receiving pharmacotherapies, HRU was lowest with BBs and highest with disopyramide. Treatment changes were observed in 43.8% of patients receiving pharmacotherapies. CONCLUSIONS: Patients experienced high rates of treatment changes, and the economic burden associated with symptomatic obstructive HCM increased as therapy escalated. More effective therapies are needed to stabilize or decrease the economic burden of obstructive HCM.

Rationale and design of a cluster-randomized pragmatic trial aimed at improving use of guideline directed medical therapy in outpatients with heart failure: PRagmatic trial of messaging to providers about treatment of heart failure (PROMPT-HF).

Heart failure with reduced ejection fraction (HFrEF) is one of the most common chronic illnesses in the United States and carries significant risk of morbidity and mortality. Use of guideline-directed medical therapy (GDMT) for patients with HFrEF has been shown to dramatically improve outcomes, but adoption of these treatments remains generally low. Possible explanations for poor GDMT uptake include lack of knowledge about recommended management strategies and provider reluctance due to uncertainties regarding application of said guidelines to real-world practice. One way to overcome these barriers is by harnessing the electronic health record (EHR) to create patient-centered "best practice alerts" (BPAs) that can guide clinicians to prescribe appropriate medical therapies. If found to be effective, these low-cost interventions can be rapidly applied across large integrated healthcare systems. The PRagmatic Trial Of Messaging to Providers about Treatment of Heart Failure (PROMPT-HF) trial is a pragmatic, cluster randomized controlled trial designed to test the hypothesis that tailored and timely alerting of recommended GDMT in heart failure (HF) will result in greater adherence to guidelines when compared with usual care. PROMPT-HF has completed enrollment of 1,310 ambulatory patients with HFrEF cared for by 100 providers who were randomized to receive a BPA vs usual care. The BPA alerted providers to GDMT recommended for their patients and displayed current left ventricular ejection fraction (LVEF) along with the most recent blood pressure, heart rate, serum potassium and creatinine levels, and estimated glomerular filtration rate. It also linked to an order set customized to the patient that suggests medications within each GDMT class not already prescribed. Our goal is to examine whether tailored EHR-based alerting for outpatients with HFrEF will lead to higher rates of GDMT at 30 days post randomization when compared with usual care. Additionally, we are assessing clinical outcomes such as hospital readmissions and death between the alert versus usual care group. Trial Registration: Clinicaltrials.gov NCT04514458.

COVID-19 infections and outcomes in a live registry of heart failure patients across an integrated health care system.

BACKGROUND: Patients with comorbid conditions have a higher risk of mortality with SARS-CoV-2 (COVID-19) infection, but the impact on heart failure patients living near a disease hotspot is unknown. Therefore, we sought to characterize the prevalence and outcomes of COVID-19 in a live registry of heart failure patients across an integrated health care system in Connecticut. METHODS: In this retrospective analysis, the Yale Heart Failure Registry (NCT04237701) that includes 26,703 patients with heart failure across a 6-hospital integrated health care system in Connecticut was queried on April 16th, 2020 for all patients tested for COVID-19. Sociodemographic and geospatial data as well as, clinical management, respiratory failure, and patient mortality were obtained via the real-time registry. Data on COVID-19 specific care was extracted by retrospective chart review. RESULTS: COVID-19 testing was performed on 900 symptomatic patients, comprising 3.4% of the Yale Heart Failure Registry (N = 26,703). Overall, 206 (23%) were COVID- 19+. As compared to COVID-19-, these patients were more likely to be older, black, have hypertension, coronary artery disease, and were less likely to be on renin angiotensin blockers (P<0.05, all). COVID-19- patients tended to be more diffusely spread across the state whereas COVID-19+ were largely clustered around urban centers. 20% of COVID-19+ patients died, and age was associated with increased risk of death [OR 1.92 95% CI (1.33-2.78); P<0.001]. Among COVID-19+ patients who were ≥85 years of age rates of hospitalization were 87%, rates of death 36%, and continuing hospitalization 62% at time of manuscript preparation. CONCLUSIONS: In this real-world snapshot of COVID-19 infection among a large cohort of heart failure patients, we found that a small proportion had undergone testing. Patients found to be COVID-19+ tended to be black with multiple comorbidities and clustered around lower socioeconomic status communities. Elderly COVID-19+ patients were very likely to be admitted to the hospital and experience high rates of mortality.

National Quality Assessment of Early Clopidogrel Therapy in Chinese Patients With Acute Myocardial Infarction (AMI) in 2006 and 2011: Insights From the China Patient-Centered Evaluative Assessment of Cardiac Events (PEACE)-Retrospective AMI Study.

BACKGROUND: Early clopidogrel administration to patients with acute myocardial infarction (AMI) has been demonstrated to improve outcomes in a large Chinese trial. However, patterns of use of clopidogrel for patients with AMI in China are unknown. METHODS AND RESULTS: From a nationally representative sample of AMI patients from 2006 and 2011, we identified 11 944 eligible patients for clopidogrel therapy and measured early clopidogrel use, defined as initiation within 24 hours of hospital admission. Among the patients eligible for clopidogrel, the weighted rate of early clopidogrel therapy increased from 45.7% in 2006 to 79.8% in 2011 (P<0.001). In 2006 and 2011, there was significant variation in early clopidogrel use by region, ranging from 1.5% to 58.0% in 2006 (P<0.001) and 48.7% to 87.7% in 2011 (P<0.001). While early use of clopidogrel was uniformly high in urban hospitals in 2011 (median 89.3%; interquartile range: 80.1% to 94.5%), there was marked heterogeneity among rural hospitals (median 50.0%; interquartile range: 11.5% to 84.4%). Patients without reperfusion therapy and those admitted to rural hospitals were less likely to be treated with clopidogrel. CONCLUSIONS: Although the use of early clopidogrel therapy in patients with AMI has increased substantially in China, there is notable wide variation across hospitals, with much less adoption in rural hospitals. Quality improvement initiatives are needed to increase consistency of early clopidogrel use for patients with AMI. CLINICAL TRIAL REGISTRATION: URL: https://www.clinicaltrials.gov/. Unique identifier: NCT01624883.

Patterns of initiation of oral anticoagulants in patients with atrial fibrillation- quality and cost implications.

BACKGROUND: Dabigatran, rivaroxaban, and apixaban have been approved for use in patients with atrial fibrillation based upon randomized trials demonstrating their comparable or superior efficacy and safety relative to warfarin. Little is known about their adoption into clinical practice, whether utilization is consistent with the controlled trials on which their approval was based, and how their use has affected health spending for patients and insurers. METHODS: We used medical and prescription claims data from a large insurer to identify patients with nonvalvular atrial fibrillation who were prescribed an oral anticoagulant in 2010-2013. We plotted trends in medication initiation over time, assessed corresponding insurer and patient out-of-pocket spending, and evaluated the cumulative number and cost of anticoagulants. We identified predictors of novel anticoagulant initiation using multivariable logistic models. Finally, we estimated the difference in total drug expenditures over 6 months for patients initiating warfarin versus a novel anticoagulant. RESULTS: There were 6893 patients with atrial fibrillation that initiated an oral anticoagulant during the study period. By the end of the study period, novel anticoagulants accounted for 62% of new prescriptions and 98% of anticoagulant-related drug costs. Female sex, lower household income, and higher CHADS2, CHA2DS2-VASC, and HAS-BLED scores were significantly associated with lower odds of receiving a novel anticoagulant (P <.001 for each). Average combined patient and insurer anticoagulant spending in the first 6 months after initiation was more than $900 greater for patients initiating a novel anticoagulant. CONCLUSIONS: This study demonstrates rapid adoption of novel anticoagulants into clinical practice, particularly among patients with lower CHADS2 and HAS-BLED scores, and high health care cost consequences. These findings provide important directions for future comparative and cost-effectiveness research.