Identifying Potential Factors Associated With Racial Disparities in COVID-19 Outcomes: Retrospective Cohort Study Using Machine Learning on Real-World Data.

BACKGROUND: Racial disparities in COVID-19 incidence and outcomes have been widely reported. Non-Hispanic Black patients endured worse outcomes disproportionately compared with non-Hispanic White patients, but the epidemiological basis for these observations was complex and multifaceted. OBJECTIVE: This study aimed to elucidate the potential reasons behind the worse outcomes of COVID-19 experienced by non-Hispanic Black patients compared with non-Hispanic White patients and how these variables interact using an explainable machine learning approach. METHODS: In this retrospective cohort study, we examined 28,943 laboratory-confirmed COVID-19 cases from the OneFlorida Research Consortium's data trust of health care recipients in Florida through April 28, 2021. We assessed the prevalence of pre-existing comorbid conditions, geo-socioeconomic factors, and health outcomes in the structured electronic health records of COVID-19 cases. The primary outcome was a composite of hospitalization, intensive care unit admission, and mortality at index admission. We developed and validated a machine learning model using Extreme Gradient Boosting to evaluate predictors of worse outcomes of COVID-19 and rank them by importance. RESULTS: Compared to non-Hispanic White patients, non-Hispanic Blacks patients were younger, more likely to be uninsured, had a higher prevalence of emergency department and inpatient visits, and were in regions with higher area deprivation index rankings and pollutant concentrations. Non-Hispanic Black patients had the highest burden of comorbidities and rates of the primary outcome. Age was a key predictor in all models, ranking highest in non-Hispanic White patients. However, for non-Hispanic Black patients, congestive heart failure was a primary predictor. Other variables, such as food environment measures and air pollution indicators, also ranked high. By consolidating comorbidities into the Elixhauser Comorbidity Index, this became the top predictor, providing a comprehensive risk measure. CONCLUSIONS: The study reveals that individual and geo-socioeconomic factors significantly influence the outcomes of COVID-19. It also highlights varying risk profiles among different racial groups. While these findings suggest potential disparities, further causal inference and statistical testing are needed to fully substantiate these observations. Recognizing these relationships is vital for creating effective, tailored interventions that reduce disparities and enhance health outcomes across all racial and socioeconomic groups.

Trajectories of Sacubitril/Valsartan Adherence Among Medicare Beneficiaries With Heart Failure.

Background: Sacubitril/valsartan, an angiotensin receptor/neprilysin inhibitor (ARNi), improves heart failure (HF) outcomes, yet real-world adherence patterns are not well understood. Objectives: The purpose of this study was to analyze longitudinal patterns of adherence to ARNis in patients with HF and to identify factors associated with adherence patterns. Methods: Using Medicare beneficiaries from 2015 to 2018, we included patients diagnosed with HF who initiated an ARNi. A group-based trajectory model was constructed to identify adherence patterns during follow-up. We used multivariable logistic regression to investigate factors associated with membership in each adherence trajectory group. Results: Among 9,475 eligible beneficiaries (age 77 ± 7 years, 34% female), we identified 5 distinct ARNi adherence trajectories, characterized as: immediate discontinuers, who discontinued treatment within the first 3 months (12%); early discontinuers, who discontinued treatment in months 4 to 7 (10%); late discontinuers, who discontinued treatment in months 7 to 10 (12%); intermittently adherent patients (12%); and consistently adherent patients (54%). The first 4 groups were collectively categorized as nonconsistent adherents. Living in a socioeconomically disadvantaged area, ie, a county with the top 20% of Area Deprivation Index (adjusted OR [aOR]: 1.12 [95% CI: 1.00-1.24]) and Black race (aOR: 1.36, [95% CI: 1.18-1.56]) were associated with a higher likelihood of being nonconsistently adherent. Receiving prescriptions from a cardiologist (aOR: 0.64 [95% CI: 0.57-0.73]) was associated with a lower likelihood of suboptimal ARNi adherence. Conclusions: Half of ARNi users were not consistently adherent to the drug in the first year after treatment initiation. There exist significant racial and socioeconomic inequities in longitudinal adherence to ARNi.

Predictors of left atrial appendage thrombus in atrial fibrillation patients undergoing cardioversion.

BACKGROUND: Atrial fibrillation and atrial flutter represent the most prevalent clinically significant cardiac arrhythmias. While the CHA2DS2-VASc score is commonly used to inform anticoagulation therapy decisions for patients with these conditions, its predictive power is limited. Therefore, we sought to improve risk prediction for left atrial appendage thrombus (LAAT), a known risk factor for stroke in these patients. METHODS: We developed and validated an explainable machine learning model using the eXtreme Gradient Boosting algorithm with 5 × 5 nested cross-validation. The primary outcome was to predict the probability of LAAT in patients with atrial fibrillation and atrial flutter who underwent transesophageal echocardiogram prior to cardioversion. Our algorithm used 37 demographic, comorbid, and transthoracic echocardiographic variables. RESULTS: A total of 795 patients were included in our analysis. LAAT was present in 11.3% of the patients. The average age of patients was 63.3 years and 34.7% were women. Patients with LAAT had significantly lower left ventricular ejection fraction (29.9% vs 43.5%; p < 0.001), lower E' lateral velocity (5.7 cm vs. 7.9 cm; p < 0.001) and higher E/A ratio (2.6 vs 1.8; p = 0.002). Our machine learning model achieved a high AUC of 0.79, with a high specificity of 0.82, and modest sensitivity of 0.57. Left ventricular ejection fraction was the most important variable in predicting LAAT. Patients were split into 10 buckets based on the percentile of their predicted probability of having thrombus. The lower the percentile (e.g., 10%), the lower the probability of having thrombus. Using a cutoff point of 0.16 which includes 10.0% of the patients, we can rule out thrombus with 100% confidence. CONCLUSION: Using machine learning, we refined the predictive power of predicting LAAT and explained the model. These results show promise in providing better guidance for anticoagulation therapy and cardioversion in AF and AFL patients.

Sodium-glucose cotransporter-2 inhibitors and incidence of atrial fibrillation in older adults with type 2 diabetes: a retrospective cohort analysis.

Objectives: To investigate the risk of atrial fibrillation (AF) with sodium-glucose cotransporter-2 inhibitors (SGLT2is) compared to dipeptidyl peptidase-4 inhibitor (DPP4i) use in older US adults and across diverse subgroups. Methods: We conducted a retrospective cohort analysis using claims data from 15% random samples of Medicare fee-for-service beneficiaries. Patients were adults with type 2 diabetes (T2D), no preexisting AF, and were newly initiated on SGLT2i or DPP4i. The outcome was the first incident AF. Inverse probability treatment weighting (IPTW) was used to balance the baseline covariates between the treatment groups including sociodemographics, comorbidities, and co-medications. Cox regression models were used to assess the effect of SGLT2i compared to DPP4i on incident AF. Results: Of the 97,436 eligible individuals (mean age 71.2 ± 9.8 years, 54.6% women), 1.01% (n = 983) had incident AF over a median follow-up of 361 days. The adjusted incidence rate was 8.39 (95% CI: 6.67-9.99) and 11.70 (95% CI: 10.9-12.55) per 1,000 person-years in the SGLT2i and DPP4i groups, respectively. SGLT2is were associated with a significantly lower risk of incident AF (HR 0.73; 95% CI, 0.57 to 0.91; p = 0.01) than DPP4is. The risk reduction of incident AF was significant in non-Hispanic White individuals and subgroups with existing atherosclerotic cardiovascular diseases and chronic kidney disease. Conclusion: Compared to the use of DPP4i, that of SGLT2i was associated with a lower risk of AF in patients with T2D. Our findings contribute to the real-world evidence regarding the effectiveness of SGLT2i in preventing AF and support a tailored therapeutic approach to optimize treatment selection based on individual characteristics.

Confidence score: a data-driven measure for inclusive systematic reviews considering unpublished preprints.

OBJECTIVES: COVID-19, since its emergence in December 2019, has globally impacted research. Over 360 000 COVID-19-related manuscripts have been published on PubMed and preprint servers like medRxiv and bioRxiv, with preprints comprising about 15% of all manuscripts. Yet, the role and impact of preprints on COVID-19 research and evidence synthesis remain uncertain. MATERIALS AND METHODS: We propose a novel data-driven method for assigning weights to individual preprints in systematic reviews and meta-analyses. This weight termed the "confidence score" is obtained using the survival cure model, also known as the survival mixture model, which takes into account the time elapsed between posting and publication of a preprint, as well as metadata such as the number of first 2-week citations, sample size, and study type. RESULTS: Using 146 preprints on COVID-19 therapeutics posted from the beginning of the pandemic through April 30, 2021, we validated the confidence scores, showing an area under the curve of 0.95 (95% CI, 0.92-0.98). Through a use case on the effectiveness of hydroxychloroquine, we demonstrated how these scores can be incorporated practically into meta-analyses to properly weigh preprints. DISCUSSION: It is important to note that our method does not aim to replace existing measures of study quality but rather serves as a supplementary measure that overcomes some limitations of current approaches. CONCLUSION: Our proposed confidence score has the potential to improve systematic reviews of evidence related to COVID-19 and other clinical conditions by providing a data-driven approach to including unpublished manuscripts.

Dynamic prediction modeling of postoperative mortality among patients undergoing surgical aortic valve replacement in a statewide cohort over a 12-year period.

Objective: Clinical prediction models for surgical aortic valve replacement mortality, are valuable decision tools but are often limited in their ability to account for changes in medical practice, patient selection, and the risk of outcomes over time. Recent research has identified methods to update models as new data accrue, but their effect on model performance has not been rigorously tested. Methods: The study population included 44,546 adults who underwent an isolated surgical aortic valve replacement from January 1, 1999, to December 31, 2018, statewide in Pennsylvania. After chronologically splitting the data into training and validation sets, we compared calibration, discrimination, and accuracy measures amongst a nonupdating model to 2 methods of model updating: calibration regression and the novel dynamic logistic state space model. Results: The risk of mortality decreased significantly during the validation period (P < .01) and the nonupdating model demonstrated poor calibration and reduced accuracy over time. Both updating models maintained better calibration (Hosmer-Lemeshow χ2 statistic) than the nonupdating model: nonupdating (156.5), calibration regression (4.9), and dynamic logistic state space model (8.0). Overall accuracy (Brier score) was consistently better across both updating models: dynamic logistic state space model (0.0252), calibration regression (0.0253), and nonupdating (0.0256). Discrimination improved with the dynamic logistic state space model (area under the curve, 0.696) compared with the nonupdating model (area under the curve, 0.685) and calibration regression method (area under the curve, 0.687). Conclusions: Dynamic model updating can improve model accuracy, discrimination, and calibration. The decision as to which method to use may depend on which measure is most important in each clinical context. Because competing therapies have emerged for valve replacement models, updating may guide clinical decision making.

Dynamic logistic state space prediction model for clinical decision making.

Prediction modeling for clinical decision making is of great importance and needed to be updated frequently with the changes of patient population and clinical practice. Existing methods are either done in an ad hoc fashion, such as model recalibration or focus on studying the relationship between predictors and outcome and less so for the purpose of prediction. In this article, we propose a dynamic logistic state space model to continuously update the parameters whenever new information becomes available. The proposed model allows for both time-varying and time-invariant coefficients. The varying coefficients are modeled using smoothing splines to account for their smooth trends over time. The smoothing parameters are objectively chosen by maximum likelihood. The model is updated using batch data accumulated at prespecified time intervals, which allows for better approximation of the underlying binomial density function. In the simulation, we show that the new model has significantly higher prediction accuracy compared to existing methods. We apply the method to predict 1 year survival after lung transplantation using the United Network for Organ Sharing data.

A Conceptual Model for Sources of Differential Selection in Lung Transplant Allocation.

Rationale: In the United States, donor lungs are allocated to transplant candidates on the basis of lung allocation scores (LAS). However, additional factors beyond the LAS can impact who is transplanted, including listing and donor-organ acceptance practices. These factors can result in differential selection, undermining the objectivity of lung allocation. Yet their impact on the lung transplant pathway has been underexplored. Objectives: We sought to systematically examine sources of differential selection in lung transplantation via qualitative methods. Methods: We conducted semistructured qualitative interviews with lung transplant surgeons and pulmonologists in the United States between June 2019 and June 2020 to understand clinician perspectives on differential selection in lung transplantation and the LAS. Results: A total of 51 respondents (30 surgeons and 21 pulmonologists) identified many sources of differential selection arising throughout the pathway from referral to transplantation. We synthesized these sources into a conceptual model with five themes: 1) transplant center's degree of risk tolerance and accountability; 2) successfulness and fairness of the LAS; 3) donor-organ availability and regional competition; 4) patient health versus program health; and 5) access to care versus responsible stewardship of organs. Conclusions: Our conceptual model demonstrates how differential selection can arise throughout lung transplantation and facilitates the further study of such selection. As new organ allocation models are developed, differential selection should be considered carefully to ensure that these models are more equitable.

Cost-effectiveness of bivalirudin in pediatric ventricular assist devices.

BACKGROUND: Despite recent data suggesting improved outcomes with bivalirudin vs heparin in pediatric Ventricular assist devices (VAD), higher costs remain a barrier. This study quantified trends in bivalirudin use and compared outcomes, resource utilization, and cost-effectiveness associated with bivalirudin vs heparin. METHODS: Children age 0 to 6 year who received VAD from 2009 to 2021 were identified in Pediatric Health Information System. Bivalirudin use was evaluated using trend analysis and outcomes were compared using Fine-Gray subdistrubtion hazard ratios (SHR). Daily-level hospital costs were compared due to differences in length of stay. Cost-effectiveness was evaluated using incremental cost-effectiveness ratio (ICER). RESULTS: Of 691 pediatric VAD recipients (median age 1 year, IQR 0-2), 304 (44%) received bivalirudin with 90% receiving bivalirudin in 2021 (trend p-value <0.01). Bivalirudin had lower hospital mortality (26% vs 32%; adjusted SHR 0.57, 95% CI 0.40-0.83) driven by lower VAD mortality (20% vs 27%; adjusted SHR 0.46, 95% CI 0.32-0.77) after adjusting for year, age, diagnosis, and center VAD volume. Post-VAD length of stay was longer for bivalirudin than heparin (median 91 vs 64 days, respectively, p < 0.001). Median daily-level costs were lower among bivalirudin (cost ratio 0.87, 95% CI 0.79-0.96) with higher pharmacy costs offset by lower imaging, laboratory, supply, and room/board costs. Estimated ICER for bivalirudin vs heparin was $61,192 per quality-adjusted life year gained with a range of $27,673 to $131,243. CONCLUSIONS: Bivalirudin use significantly increased over the past decade and is now used in 90% young pediatric VAD recipients. Bivalirudin was associated with significantly lower hospital mortality and an ICER <$65,000, making it a cost-effective therapy for pediatric VAD recipients.

Underprescribing vs underfilling to oral anticoagulation: An analysis of linked medical record and claims data for a nationwide sample of patients with atrial fibrillation.

BACKGROUND: Oral anticoagulants (OAC) is indicated for stroke prevention in patients with atrial fibrillation (AF) with a moderate or high risk of stroke. Despite the benefits of stroke prevention, only 50%-60% of Americans with nonvalvular AF and a moderate or high risk of stroke receive OAC medication. OBJECTIVE: To understand the extent to which low OAC use by patients with AF is attributed to underprescribing or underfilling once the medication is prescribed. METHODS: This is a retrospective cohort study that used linked claims data and electronic health records from Optum Integrated data. Participants were adults (aged ≥ 18 years) with first AF between January 2013 and June 2017. The outcomes included (1) being prescribed OACs within 180 days of AF diagnosis or not and (2) filling an OAC prescription or not among patients with AF who were prescribed an OAC within 150 days of AF diagnosis. Multivariable logistic regression models were constructed to determine factors associated with underprescribing and underfilling. RESULTS: Of the 6,141 individuals in the study cohort, 51% were not prescribed OACs within 6 months of their AF diagnosis. Of the 2,956 patients who were prescribed, 19% did not fill it at the pharmacy. In the final adjusted model, younger age, location (Northeast and South), a low CHA2DS2-VASc score, and a high HAS-BLED score were associated with a lower likelihood of being prescribed OACs. Among patients who were prescribed, Medicare enrollment (odds ratio [OR] [95% CI] = 2.2 [1.3-3.7]) and having a direct oral anticoagulant prescription (1.5 [1.2-1.9]) were associated with a lower likelihood of filling the prescription. CONCLUSIONS: Both underprescribing and underfilling are major drivers of low OAC use among patients with AF, and solutions to increase OAC use must address both prescribing and filling. DISCLOSURES: Research reported in this study was supported by the National Heart, Lung and Blood Institute (K01HL142847 and R01HL157051). Dr Guo is supported by the National Institute of Diabetes and Digestive and Kidney Diseases (R01DK133465), PhMRA Foundation Research Starter Award, and the University of Florida Research Opportunity Seed Fund. Dr Hernandez reports scientific advisory board fees from Pfizer and Bristol Myers Squibb, outside of the submitted work.

Clinical impact of a modified lung allocation score that mitigates selection bias.

BACKGROUND: The Lung Allocation Score (LAS) is used in the U.S. to prioritize lung transplant candidates. Selection bias, induced by dependent censoring of waitlisted candidates and prediction of posttransplant survival among surviving, transplanted patients only, is only partially addressed by the LAS. Recently, a modified LAS (mLAS) was designed to mitigate such bias. Here, we estimate the clinical impact of replacing the LAS with the mLAS. METHODS: We considered lung transplant candidates waitlisted during 2016 and 2017. LAS and mLAS scores were computed for each registrant at each observed organ offer date; individuals were ranked accordingly. Patient characteristics associated with better priority under the mLAS were investigated via logistic regression and generalized linear mixed models. We also determined whether differences in rank were explained more by changes in predicted pre- or posttransplant survival. Simulations examined how 1-year waitlist, posttransplant, and overall survival might change under the mLAS. RESULTS: Diagnosis group, 6-minute walk distance, continuous mechanical ventilation, functional status, and age demonstrated the highest impact on differential allocation. Differences in rank were explained more by changes in predicted pretransplant survival than changes in predicted posttransplant survival, suggesting that selection bias has more impact on estimates of waitlist urgency. Simulations suggest that for every 1000 waitlisted individuals, 12.8 (interquartile range: 5.2-24.3) fewer waitlist deaths per year would occur under the mLAS, without compromising posttransplant and overall survival. CONCLUSIONS: Implementing a mLAS that mitigates selection bias into clinical practice can lead to important differences in allocation and possibly modest improvement in waitlist survival.

Modified Brier score for evaluating prediction accuracy for binary outcomes.

The Brier score has been a popular measure of prediction accuracy for binary outcomes. However, it is not straightforward to interpret the Brier score for a prediction model since its value depends on the outcome prevalence. We decompose the Brier score into two components, the mean squares between the estimated and true underlying binary probabilities, and the variance of the binary outcome that is not reflective of the model performance. We then propose to modify the Brier score by removing the variance of the binary outcome, estimated via a general sliding window approach. We show that the new proposed measure is more sensitive for comparing different models through simulation. A standardized performance improvement measure is also proposed based on the new criterion to quantify the improvement of prediction performance. We apply the new measures to the data from the Breast Cancer Surveillance Consortium and compare the performance of predicting breast cancer risk using the models with and without its most important predictor.

Outcomes of the National Heart, Lung, and Blood Institute K12 program in emergency care research: 7-year follow-up.

BACKGROUND: Long-term follow-up for clinician-scientist training programs is sparse. We describe the outcomes of clinician-scientist scholars in the National Heart Lung and Blood Institute (NHLBI) K12 program in emergency care research up to 8.7 years after matriculation in the program. METHODS: This was a cohort study of faculty clinician-scientist scholars enrolled in a NHLBI K12 research training program at 6 sites across the US, with median follow-up 7.7 years (range 5.7-8.7 years) from the date of matriculation. Scholars completed electronic surveys in 2017 and 2019, with the 2019 survey collecting information for their current work setting, percent time for research, and grant funding from all sources. We used NIH RePorter and online resources to verify federal grants through March 2021. The primary outcome was a funded career development award (CDA) or research project grant (RPG) where the scholar was principal investigator. We included funding from all federal sources and national foundations. RESULTS: There were 43 scholars, including 16 (37%) women. Over the follow-up period, 32 (74%) received an individual CDA or RPG, with a median of 36 months (range 9-83 months) after entering the program. Of the 43 scholars, 23 (54%) received a CDA and 22 (51%) received an RPG, 7 (16%) of which were R01s. Of the 23 scholars who received a CDA, 13 (56%) subsequently had an RPG funded. Time to CDA or RPG did not differ by sex (women vs. men log-rank test p = 0.27) or specialty training (emergency medicine versus other specialties, p = 0.59). CONCLUSIONS: After 7 years of follow-up for this NHLBI K12 emergency care research training program, three quarters of clinician-scientist scholars had obtained CDA or RPG funding, with no notable differences by sex or clinical training.

The development and initial validation of the PROMIS®+HF-27 and PROMIS+HF-10 profiles.

AIMS: Heart failure (HF) is a common and morbid condition impacting multiple health domains. We previously reported the development of the PROMIS®-Plus-HF (PROMIS+HF) profile measure, including universal and HF-specific items. To facilitate use, we developed shorter, PROMIS+HF profiles intended for research and clinical use. METHODS AND RESULTS: Candidate items were selected based on psychometric properties and symptom range coverage. HF clinicians (n = 43) rated item importance and clinical actionability. Based on these results, we developed the PROMIS+HF-27 and PROMIS+HF-10 profiles with summary scores (0-100) for overall, physical, mental, and social health. In a cross-sectional sample (n = 600), we measured internal consistency reliability (Cronbach's alpha and Spearman-Brown), test-retest reliability (intraclass coefficient; n = 100), known-groups validity via New York Heart Association (NYHA) class, and convergent validity with Kansas City Cardiomyopathy Questionnaire (KCCQ) scores. In a longitudinal sample (n = 75), we evaluated responsiveness of baseline/follow-up scores by calculating mean differences and Cohen's d and comparing with paired t-tests. Internal consistency was good to excellent (α 0.82-0.94) for all PROMIS+HF-27 scores and acceptable to good (α/Spearman-Brown 0.60-0.85) for PROMIS+HF-10 scores. Test-retest intraclass coefficients were acceptable to excellent (0.75-0.97). Both profiles demonstrated known-groups validity for the overall and physical health summary scores based on NYHA class, and convergent validity for nearly all scores compared with KCCQ scores. In the longitudinal sample, we demonstrated responsiveness for PROMIS+HF-27 and PROMIS+HF-10 overall and physical summary scores. For the PROMIS+HF overall summary scores, a group-based increase of 7.6-8.3 points represented a small to medium change (Cohen's d = 0.40-0.42). For the PROMIS+HF physical summary scores, a group-based increase of 5.0-5.9 points represented a small to medium change (Cohen's d = 0.29-0.35). CONCLUSIONS: The PROMIS+HF-27 and PROMIS+HF-10 profiles demonstrated good psychometric characteristics with evidence of responsiveness for overall and physical health. These new measures can facilitate patient-centred research and clinical care, such as improving care quality through symptom monitoring, facilitating shared decision-making, evaluating quality of care, assessing new interventions, and monitoring during the initiation and titration of guideline-directed medical therapy.

Identifying Patients at Risk of Acute Kidney Injury Among Medicare Beneficiaries With Type 2 Diabetes Initiating SGLT2 Inhibitors: A Machine Learning Approach.

Introduction: To predict acute kidney injury (AKI) risk in patients with type 2 diabetes (T2D) prescribed sodium-glucose cotransporter two inhibitors (SGLT2i). Methods: Using a 5% random sample of Medicare claims data, we identified 17,694 patients who filled ≥1 prescriptions for canagliflozin, dapagliflozin and empagliflozin in 2013-2016. The cohort was split randomly and equally into training and testing sets. We measured 65 predictor candidates using claims data from the year prior to SGLT2i initiation. We then applied three machine learning models, including random forests (RF), elastic net and least absolute shrinkage and selection operator (LASSO) for risk prediction. Results: The incidence rate of AKI was 1.1% over a median 1.5 year follow up. Among three machine learning methods, RF produced the best prediction (C-statistic = 0.72), followed by LASSO and elastic net (both C-statistics = 0.69). Among individuals classified in the top 10% of the RF risk score (i.e., high risk group), the actual incidence rate of AKI was as high as 3.7%. In the logistic regression model including 14 important risk factors selected by LASSO, use of loop diuretics [adjusted odds ratio (95% confidence interval): 3.72 (2.44-5.76)] had the strongest association with AKI incidence. Disscusion: Our machine learning model efficiently identified patients at risk of AKI among Medicare beneficiaries with T2D undergoing SGLT2i treatment.

Comparable Cardiorenal Benefits of SGLT2 Inhibitors and GLP-1RAs in Asian and White Populations: An Updated Meta-analysis of Results From Randomized Outcome Trials.

BACKGROUND: Whether the cardiorenal benefits of sodium-glucose cotransporter 2 (SGLT2) inhibitors and glucagon-like peptide 1 receptor agonists (GLP-1RAs) are comparable between White and Asian populations remains unclear. PURPOSE: To compare the cardiorenal benefits of SGLT2 inhibitors and GLP-1RAs between White and Asian populations and to compare the cardiorenal benefits between the two agents in Asian patients. DATA SOURCES: Electronic databases were searched up to 28 March 2021. STUDY SELECTION: We included the cardiovascular (CV) and renal outcome trials of SGLT2 inhibitors and GLP-1RAs where investigators reported major adverse CV events (MACE), CV death/hospitalization for heart failure (HHF), or composite renal outcomes with stratification by race. DATA EXTRACTION: We extracted the hazard ratio of each outcome stratified by race (Asian vs. White populations). DATA SYNTHESIS: In 10 SGLT2 inhibitor trials, there was no significant difference between Asian and White populations for MACE (P = 0.55), CV death/HHF (P = 0.87), or composite renal outcomes (P = 0.97). In seven GLP-1RA trials, we observed a similar MACE benefit between Asian and White populations (P = 0.10). In our networkmeta-analysis we found a comparable benefit for MACE between SGLT2 inhibitors and GLP-1RAs in Asian patients. LIMITATIONS: The data were from stratified analyses. CONCLUSIONS: There appear to be comparable cardiorenal benefits of SGLT2 inhibitors and GLP-1RAs between Asian and White participants enrolled in CV and renal outcome trials; the two therapies seem to have similar CV benefits for Asian participants.

A Comparison of Methods to Detect Changes in Prediction Models.

BACKGROUND: Prediction models inform decisions in many areas of medicine. Most models are fitted once and then applied to new (future) patients, despite the fact that model coefficients can vary over time due to changes in patients' clinical characteristics and disease risk. However, the optimal method to detect changes in model parameters has not been rigorously assessed. METHODS: We simulated data, informed by post-lung transplant mortality data and tested the following two approaches for detecting model change: (1) the "Direct Approach," it compares coefficients of the model refit on recent data to those at baseline; and (2) "Calibration Regression," it fits a logistic regression model of the log-odds of the observed outcomes versus the linear predictor from the baseline model (i.e., the log-odds of the predicted probabilities obtained from the baseline model) and tests whether the intercept and slope differ from 0 and 1, respectively. Four scenarios were simulated using logistic regression for binary outcomes as follows: (1) we fixed all model parameters, (2) we varied the outcome prevalence between 0.1 and 0.2, (3) we varied the coefficient of one of the ten predictors between 0.2 and 0.4, and (4) we varied the outcome prevalence and coefficient of one predictor simultaneously. RESULTS: Calibration regression tended to detect changes sooner than the Direct Approach, with better performance (e.g., larger proportion of true claims). When the sample size was large, both methods performed well. When two parameters changed simultaneously, neither method performed well. CONCLUSION: Neither change detection method examined here proved optimal under all circumstances. However, our results suggest that if one is interested in detecting a change in overall incidence of an outcome (e.g., intercept), the Calibration Regression method may be superior to the Direct Approach. Conversely, if one is interested in detecting a change in other model covariates (e.g., slope), the Direct Approach may be superior.

Comparative Safety of Dipeptidyl Peptidase-4 Inhibitors and Sudden Cardiac Arrest and Ventricular Arrhythmia: Population-Based Cohort Studies.

In vivo studies suggest that arrhythmia risk may be greater with less selective dipeptidyl peptidase-4 inhibitors, but evidence from population-based studies is missing. We aimed to compare saxagliptin, sitagliptin, and linagliptin with regard to risk of sudden cardiac arrest (SCA)/ventricular arrhythmia (VA). We conducted high-dimensional propensity score (hdPS) matched, new-user cohort studies. We analyzed Medicaid and Optum Clinformatics separately. We identified new users of saxagliptin, sitagliptin (both databases), and linagliptin (Optum only). We defined SCA/VA outcomes using emergency department and inpatient diagnoses. We identified and then controlled for confounders via a data-adaptive, hdPS approach. We generated marginal hazard ratios (HRs) via Cox proportional hazards regression using a robust variance estimator while adjusting for calendar year. We identified the following matched comparisons: saxagliptin vs. sitagliptin (23,895 vs. 96,972) in Medicaid, saxagliptin vs. sitagliptin (48,388 vs. 117,383) in Optum, and linagliptin vs. sitagliptin (36,820 vs. 78,701) in Optum. In Medicaid, use of saxagliptin (vs. sitagliptin) was associated with an increased rate of SCA/VA (adjusted HR (aHR), 2.01, 95% confidence interval (CI) 1.24-3.25). However, in Optum data, this finding was not present (aHR, 0.79, 95% CI 0.41-1.51). Further, we found no association between linagliptin (vs. sitagliptin) and SCA/VA (aHR, 0.65, 95% CI 0.36-1.17). We found discordant results regarding the association between SCA/VA with saxagliptin compared with sitagliptin in two independent datasets. It remains unclear whether these findings are due to heterogeneity of treatment effect in the different populations, chance, or unmeasured confounding.

Comparison of dynamic updating strategies for clinical prediction models.

BACKGROUND: Prediction models inform many medical decisions, but their performance often deteriorates over time. Several discrete-time update strategies have been proposed in the literature, including model recalibration and revision. However, these strategies have not been compared in the dynamic updating setting. METHODS: We used post-lung transplant survival data during 2010-2015 and compared the Brier Score (BS), discrimination, and calibration of the following update strategies: (1) never update, (2) update using the closed testing procedure proposed in the literature, (3) always recalibrate the intercept, (4) always recalibrate the intercept and slope, and (5) always refit/revise the model. In each case, we explored update intervals of every 1, 2, 4, and 8 quarters. We also examined how the performance of the update strategies changed as the amount of old data included in the update (i.e., sliding window length) increased. RESULTS: All methods of updating the model led to meaningful improvement in BS relative to never updating. More frequent updating yielded better BS, discrimination, and calibration, regardless of update strategy. Recalibration strategies led to more consistent improvements and less variability over time compared to the other updating strategies. Using longer sliding windows did not substantially impact the recalibration strategies, but did improve the discrimination and calibration of the closed testing procedure and model revision strategies. CONCLUSIONS: Model updating leads to improved BS, with more frequent updating performing better than less frequent updating. Model recalibration strategies appeared to be the least sensitive to the update interval and sliding window length.