Multi-Source Conformal Inference Under Distribution Shift.

Recent years have experienced increasing utilization of complex machine learning models across multiple sources of data to inform more generalizable decision-making. However, distribution shifts across data sources and privacy concerns related to sharing individual-level data, coupled with a lack of uncertainty quantification from machine learning predictions, make it challenging to achieve valid inferences in multi-source environments. In this paper, we consider the problem of obtaining distribution-free prediction intervals for a target population, leveraging multiple potentially biased data sources. We derive the efficient influence functions for the quantiles of unobserved outcomes in the target and source populations, and show that one can incorporate machine learning prediction algorithms in the estimation of nuisance functions while still achieving parametric rates of convergence to nominal coverage probabilities. Moreover, when conditional outcome invariance is violated, we propose a data-adaptive strategy to upweight informative data sources for efficiency gain and downweight non-informative data sources for bias reduction. We highlight the robustness and efficiency of our proposals for a variety of conformal scores and data-generating mechanisms via extensive synthetic experiments. Hospital length of stay prediction intervals for pediatric patients undergoing a high-risk cardiac surgical procedure between 2016-2022 in the U.S. illustrate the utility of our methodology.

Detecting univariate, bivariate, and overall effects of drug mixtures using Bayesian kernel machine regression.

Background: Innovative analytic approaches to drug studies are needed to understand better the co-use of opioids with non-opioids among people using illicit drugs. One approach is the Bayesian kernel machine regression (BKMR), widely applied in environmental epidemiology to study exposure mixtures but has received far less attention in substance use research.Objective: To describe the utility of the BKMR approach to study the effects of drug substance mixtures on health outcomes.Methods: We simulated data for 200 individuals. Using the Vale and Maurelli method, we simulated multivariate non-normal drug exposure data: xylazine (mean = 300 ng/mL, SD = 100 ng/mL), fentanyl (mean = 200 ng/mL, SD = 71 ng/mL), benzodiazepine (mean = 300 ng/mL, SD = 55 ng/mL), and nitazene (mean = 200 ng/mL, SD = 141 ng/mL) concentrations. We performed 10,000 MCMC sampling iterations with three Markov chains. Model diagnostics included trace plots, r-hat values, and effective sample sizes. We also provided visual relationships of the univariate and bivariate exposure-response and the overall mixture effect.Results: Higher levels of fentanyl and nitazene concentrations were associated with higher levels of the simulated health outcome, controlling for age. Trace plots, r-hat values, and effective sample size statistics demonstrated BKMR stability across multiple Markov chains.Conclusions: Our understanding of drug mixtures tends to be limited to studies of single-drug models. BKMR offers an innovative way to discern which substances pose a greater health risk than other substances and can be applied to assess univariate, bivariate, and cumulative drug effects on health outcomes.

Sensitivity Analyses of Clinical Trial Designs: Selecting Scenarios and Summarizing Operating Characteristics.

The use of simulation-based sensitivity analyses is fundamental for evaluating and comparing candidate designs of future clinical trials. In this context, sensitivity analyses are especially useful to assess the dependence of important design operating characteristics with respect to various unknown parameters. Typical examples of operating characteristics include the likelihood of detecting treatment effects and the average study duration, which depend on parameters that are unknown until after the onset of the clinical study, such as the distributions of the primary outcomes and patient profiles. Two crucial components of sensitivity analyses are (i) the choice of a set of plausible simulation scenarios and (ii) the list of operating characteristics of interest. We propose a new approach for choosing the set of scenarios to be included in a sensitivity analysis. We maximize a utility criterion that formalizes whether a specific set of sensitivity scenarios is adequate to summarize how the operating characteristics of the trial design vary across plausible values of the unknown parameters. Then, we use optimization techniques to select the best set of simulation scenarios (according to the criteria specified by the investigator) to exemplify the operating characteristics of the trial design. We illustrate our proposal in three trial designs.

TAXN: Translate Align Extract Normalize, a Multilingual Extraction Tool for Clinical Texts.

Several studies have shown that about 80% of the medical information in an electronic health record is only available through unstructured data. Resources such as medical terminologies in languages other than English are limited and restrain the NLP tools. We propose here to leverage English based resources in other languages using a combination of translation, word alignment, entity extraction and term normalization (TAXN). We implement this extraction pipeline in an open-source library called "medkit". We demonstrate the interest of this approach through a specific use-case: enriching a phenotypic dictionary for post-acute sequelae in COVID-19 (PASC). TAXN proved to be efficient to propose new synonyms of UMLS terms using a corpus of 70 articles in French with 356 terms enriched with at least one validated new synonym. This study was based on freely available deep-learning models.

Robust approach to combining multiple markers to improve surrogacy.

Identifying effective and valid surrogate markers to make inference about a treatment effect on long-term outcomes is an important step in improving the efficiency of clinical trials. Replacing a long-term outcome with short-term and/or cheaper surrogate markers can potentially shorten study duration and reduce trial costs. There is sizable statistical literature on methods to quantify the effectiveness of a single surrogate marker. Both parametric and nonparametric approaches have been well developed for different outcome types. However, when there are multiple markers available, methods for combining markers to construct a composite marker with improved surrogacy remain limited. In this paper, building on top of the optimal transformation framework of Wang et al. (2020), we propose a novel calibrated model fusion approach to optimally combine multiple markers to improve surrogacy. Specifically, we obtain two initial estimates of optimal composite scores of the markers based on two sets of models with one set approximating the underlying data distribution and the other directly approximating the optimal transformation function. We then estimate an optimal calibrated combination of the two estimated scores which ensures both validity of the final combined score and optimality with respect to the proportion of treatment effect explained by the final combined score. This approach is unique in that it identifies an optimal combination of the multiple surrogates without strictly relying on parametric assumptions while borrowing modeling strategies to avoid fully nonparametric estimation which is subject to the curse of dimensionality. Our identified optimal transformation can also be used to directly quantify the surrogacy of this identified combined score. Theoretical properties of the proposed estimators are derived, and the finite sample performance of the proposed method is evaluated through simulation studies. We further illustrate the proposed method using data from the Diabetes Prevention Program study.

Breaking Free from the Hazard Ratio: Embracing the Restricted Mean Survival Time in Clinical Trials.

Clinicians consider the Cox proportional hazards model to be the default tool of choice for analyzing right-censored time-to-event outcomes in randomized clinical trials.1 The Cox model estimates the ratio of the hazard of the outcome of interest (e.g., the time to death) between two treatment groups. The hazard is defined as the instantaneous probability of experiencing the event of interest in the next time interval among individuals who had not yet experienced the event. As a statistical measure, the hazard ratio can be difficult to interpret.

Identifying surrogate markers in real-world comparative effectiveness research.

In comparative effectiveness research (CER), leveraging short-term surrogates to infer treatment effects on long-term outcomes can guide policymakers evaluating new treatments. Numerous statistical procedures for identifying surrogates have been proposed for randomized clinical trials (RCTs), but no methods currently exist to evaluate the proportion of treatment effect (PTE) explained by surrogates in real-world data (RWD), which have become increasingly common. To address this knowledge gap, we propose inverse probability weighted (IPW) and doubly robust (DR) estimators of an optimal transformation of the surrogate and the corresponding PTE measure. We demonstrate that the proposed estimators are consistent and asymptotically normal, and the DR estimator is consistent when either the propensity score model or outcome regression model is correctly specified. Our proposed estimators are evaluated through extensive simulation studies. In two RWD settings, we show that our method can identify and validate surrogate markers for inflammatory bowel disease (IBD).

Evaluating and volunteering for crowdsourced interventions: Cross-sectional data on COVID-19 safety from a University Survey.

Despite many innovative ideas generated in response to COVID-19, few studies have examined community preferences for these ideas. Our study aimed to determine university community members' preferences for three novel ideas identified through a crowdsourcing open call at the University of North Carolina (UNC) for making campus safer in the pandemic, as compared to existing (i.e. pre-COVID-19) resources. An online survey was conducted from March 30, 2021 -May 6, 2021. Survey participants included UNC students, staff, faculty, and others. The online survey was distributed using UNC's mass email listserv and research directory, departmental listservs, and student text groups. Collected data included participant demographics, COVID-19 prevention behaviors, preferences for finalist ideas vs. existing resources in three domains (graduate student supports, campus tours, and online learning), and interest in volunteering with finalist teams. In total 437 survey responses were received from 228 (52%) staff, 119 (27%) students, 78 (18%) faculty, and 12 (3%) others. Most participants were older than age 30 years (309; 71%), women (332, 78%), and white (363, 83.1%). Five participants (1%) were gender minorities, 66 (15%) identified as racial/ethnic minorities, and 46 (10%) had a disability. Most participants preferred the finalist idea for a virtual campus tour of UNC's lesser-known history compared to the existing campus tour (52.2% vs. 16.0%). For graduate student supports, 41.4% of participants indicated no preference between the finalist idea and existing supports; for online learning resources, the existing resource was preferred compared to the finalist idea (41.6% vs. 30.4%). Most participants agreed that finalists' ideas would have a positive impact on campus safety during COVID-19 (81.2%, 79.6%, and 79.2% for finalist ideas 1, 2 and 3 respectively). 61 (14.1%) participants indicated interest in volunteering with finalist teams. Together these findings contribute to the development and implementation of community-engaged crowdsourced campus safety interventions during COVID-19.

Joint international consensus statement on crowdsourcing challenge contests in health and medicine: results of a modified Delphi process.

OBJECTIVES: To develop a consensus statement to provide advice on designing, implementing and evaluating crowdsourcing challenge contests in public health and medical contexts. DESIGN: Modified Delphi using three rounds of survey questionnaires and one consensus workshop. SETTING: Uganda for face-to-face consensus activities, global for online survey questionnaires. PARTICIPANTS: A multidisciplinary expert panel was convened at a consensus-development conference in Uganda and included 21 researchers with experience leading challenge contests, five public health sector workers, and nine Ugandan end users. An online survey was sent to 140 corresponding authors of previously published articles that had used crowdsourcing methods. RESULTS: A subgroup of expert panel members developed the initial statement and survey. We received responses from 120 (85.7%) survey participants, which were presented at an in-person workshop of all 21 panel members. Panelists discussed each of the sections, revised the statement, and participated in a second round of the survey questionnaire. Based on this second survey round, we held detailed discussions of each subsection with workshop participants and further revised the consensus statement. We then conducted the third round of the questionnaire among the 21 expert panelists and used the results to finalize the statement. This iterative process resulted in 23 final statement items, all with greater than 80% consensus. Statement items are organised into the seven stages of a challenge contest, including the following: considering the appropriateness, organising a community steering committee, promoting the contest, assessing contributions, recognising contributors, sharing ideas and evaluating the contest (COPARSE). CONCLUSIONS: There is high agreement among crowdsourcing experts and stakeholders on the design and implementation of crowdsourcing challenge contests. The COPARSE consensus statement can be used to organise crowdsourcing challenge contests, improve the rigour and reproducibility of crowdsourcing research and enable large-scale collaboration.

Worldwide differences in primary prevention implantable cardioverter defibrillator utilization and outcomes in hypertrophic cardiomyopathy.

AIMS: Risk stratification algorithms for sudden cardiac death (SCD) in hypertrophic cardiomyopathy (HCM) and regional differences in clinical practice have evolved over time. We sought to compare primary prevention implantable cardioverter defibrillator (ICD) implantation rates and associated clinical outcomes in US vs. non-US tertiary HCM centres within the international Sarcomeric Human Cardiomyopathy Registry. METHODS AND RESULTS: We included patients with HCM enrolled from eight US sites (n = 2650) and five non-US (n = 2660) sites and used multivariable Cox-proportional hazards models to compare outcomes between sites. Primary prevention ICD implantation rates in US sites were two-fold higher than non-US sites (hazard ratio (HR) 2.27 [1.89-2.74]), including in individuals deemed at high 5-year SCD risk (≥6%) based on the HCM risk-SCD score (HR 3.27 [1.76-6.05]). US ICD recipients also had fewer traditional SCD risk factors. Among ICD recipients, rates of appropriate ICD therapy were significantly lower in US vs. non-US sites (HR 0.52 [0.28-0.97]). No significant difference was identified in the incidence of SCD/resuscitated cardiac arrest among non-recipients of ICDs in US vs. non-US sites (HR 1.21 [0.74-1.97]). CONCLUSION: Primary prevention ICDs are implanted more frequently in patients with HCM in US vs. non-US sites across the spectrum of SCD risk. There was a lower rate of appropriate ICD therapy in US sites, consistent with a lower-risk population, and no significant difference in SCD in US vs. non-US patients who did not receive an ICD. Further studies are needed to understand what drives malignant arrhythmias, optimize ICD allocation, and examine the impact of different ICD utilization strategies on long-term outcomes in HCM.

Assessment of a Crowdsourcing Open Call for Approaches to University Community Engagement and Strategic Planning During COVID-19.

Importance: Reimagining university life during COVID-19 requires substantial innovation and meaningful community input. One method for obtaining community input is crowdsourcing, which involves having a group of individuals work to solve a problem and then publicly share solutions. Objective: To evaluate a crowdsourcing open call as an approach to COVID-19 university community engagement and strategic planning. Design, Setting, and Participants: This qualitative study assessed a crowdsourcing open call offered from June 16 to July 16, 2020, that sought ideas to inform safety in the fall 2020 semester at the University of North Carolina at Chapel Hill (UNC). Digital methods (email and social media) were used for promotion, and submissions were collected online for 4 weeks. Participation was open to UNC students, staff, faculty, and others. Main Outcomes and Measures: Submissions were evaluated for innovation, feasibility, inclusivity, and potential to improve safety and well-being. Demographic data were collected from submitting individuals, and submissions were qualitatively analyzed for emergent themes on challenges with and solutions for addressing safety and well-being in the fall semester. Data were shared with UNC leadership to inform decision-making. Results: The open call received 82 submissions from 110 participants, including current UNC students (56 submissions [68%]), people younger than 30 years (67 [82%]), women (55 [67%]), and individuals identifying as a racial/ethnic minority or as multiracial/ethnic (49 [60%]). Seven submissions were identified as finalists and received cash prizes with the encouragement to use these funds toward idea development and implementation. Seventeen runner-up teams were linked to university resources for further development. Thematic analysis of submissions regarding challenges with the fall semester revealed not only physical health concerns and the limitations of remote learning but also challenges that have been exacerbated by the pandemic, such as a lack of mental health support, structural racism and inequality, and insufficient public transportation. Solutions included novel ideas to support mental health among specific populations (eg, graduate students and racial/ethnic minorities), improve health equity, and increase transit access. All 24 finalists and runners-up indicated interest in implementation after being notified of the open call results. Conclusions and Relevance: This study suggests that open calls are a feasible strategy for university community engagement on COVID-19, providing a stakeholder-driven approach to identifying promising ideas for enhancing safety and well-being. Open calls could be formally incorporated into university planning processes to develop COVID-19 safety strategies that are responsive to diverse community members' concerns.

Clinical characteristics and outcomes in childhood-onset hypertrophic cardiomyopathy.

AIMS: Childhood-onset hypertrophic cardiomyopathy (HCM) is far less common than adult-onset disease, thus natural history is not well characterized. We aim to describe the characteristics and outcomes of childhood-onset HCM. METHODS AND RESULTS: We performed an observational cohort study of 7677 HCM patients from the Sarcomeric Human Cardiomyopathy Registry (SHaRe). Hypertrophic cardiomyopathy patients were stratified by age at diagnosis [<1 year (infancy), 1-18 years (childhood), >18 years (adulthood)] and assessed for composite endpoints reflecting heart failure (HF), life-threatening ventricular arrhythmias, atrial fibrillation (AF), and an overall composite that also included stroke and death. Stratifying by age of diagnosis, 184 (2.4%) patients were diagnosed in infancy; 1128 (14.7%) in childhood; and 6365 (82.9%) in adulthood. Childhood-onset HCM patients had an ∼2%/year event rate for the overall composite endpoint, with ventricular arrhythmias representing the most common event in the 1st decade following baseline visit, but HF and AF becoming more common by the end of the 2nd decade. Sarcomeric variants were more common in childhood-onset HCM (63%) and carried a worse prognosis than non-sarcomeric disease, including a greater than two-fold increased risk of HF [HRadj 2.39 (1.36-4.20), P = 0.003] and 67% increased risk of the overall composite outcome [HRadj 1.67 (1.16-2.41), P = 0.006]. When compared with adult-onset HCM, childhood-onset was 36% more likely to develop life-threatening ventricular arrhythmias [HRadj 1.36 (1.03-1.80)] and twice as likely to require transplant or ventricular assist device [HRadj 1.99 (1.23-3.23)]. CONCLUSION: Patients with childhood-onset HCM are more likely to have sarcomeric disease, carry a higher risk of life-threatening ventricular arrythmias, and have greater need for advanced HF therapies. These findings provide insight into the natural history of disease and can help inform clinical risk stratification.

Associations Between Female Sex, Sarcomere Variants, and Clinical Outcomes in Hypertrophic Cardiomyopathy.

BACKGROUND: The impact of sex on phenotypic expression in hypertrophic cardiomyopathy (HCM) has not been well characterized in genotyped cohorts. METHODS: Retrospective cohort study from an international registry of patients receiving care at experienced HCM centers. Sex-based differences in baseline characteristics and clinical outcomes were assessed. RESULTS: Of 5873 patients (3788 genotyped), 2226 (37.9%) were women. At baseline, women were older (49.0±19.9 versus 42.9±18.4 years, P<0.001) and more likely to have pathogenic/likely pathogenic sarcomeric variants (HCM patients with a sarcomere mutation; 51% versus 43%, P<0.001) despite equivalent utilization of genetic testing. Age at diagnosis varied by sex and genotype despite similar distribution of causal genes. Women were 3.6 to 7.1 years older at diagnosis (P<0.02) except for patients with MYH7 variants where age at diagnosis was comparable for women and men (n=492; 34.8±19.2 versus 33.3±16.8 years, P=0.39). Over 7.7 median years of follow-up, New York Heart Association III-IV heart failure was more common in women (hazard ratio, 1.87 [CI, 1.48-2.36], P<0.001), after controlling for their higher burden of symptoms and outflow tract obstruction at baseline, reduced ejection fraction, HCM patients with a sarcomere mutation, age, and hypertension. All-cause mortality was increased in women (hazard ratio, 1.50 [CI, 1.13-1.99], P<0.01) but neither implantable cardioverter-defibrillator utilization nor ventricular arrhythmia varied by sex. CONCLUSIONS: In HCM, women are older at diagnosis, partly modified by genetic substrate. Regardless of genotype, women were at higher risk of mortality and developing severe heart failure symptoms. This points to a sex-effect on long-term myocardial performance in HCM, which should be investigated further.

Ventricular tachycardia in cardiolaminopathy: Characteristics and considerations for device programming.

BACKGROUND: Mutations in LMNA cause an arrhythmogenic cardiomyopathy (cardiolaminopathy) with high risk of ventricular tachycardia (VT). The natural history of VT among patients with cardiolaminopathy is incompletely understood. OBJECTIVE: The purpose of this study was to determine the longitudinal burden and progression of VT, including change in tachycardia cycle length (TCL), response to antitachycardia pacing (ATP), and prognostic significance of high-burden VT (>5 episodes of VT at any device interrogation) in cardiolaminopathy patients. METHODS: Patients with cardiolaminopathy and an implantable cardioverter-defibrillator (ICD) were identified from a single-center database. Serial device interrogations and medical records were used to collect data on VT burden, TCL, and response to ATP. RESULTS: Cardiolaminopathy patients with primary (n = 27) or secondary prevention (n = 16) ICDs were followed for 2 years (interquartile range [IQR] 1-5). VT burden was substantially higher in patients receiving secondary prevention ICDs (28 ± 40.9 vs 3.6 ± 7.3 episodes per 100 patient-years; P <.001). ATP was highly effective (94%) at terminating VT except for short TCL (<250 ms), for which ATP failed in 60%. Among patients with recurrent VT, TCL increased by 112 ± 93.6 ms during follow-up. Inappropriate shocks were rare (0.4% of all therapies). Median time to transplantation, ventricular assist device, or death was 18 months (IQR 0.7-27.1) in patients with high-burden VT. CONCLUSION: In patients with cardiolaminopathy, VT is recurrent and highly responsive to ATP, which supports the use of transvenous ICDs iteratively programmed to manage VT of various TCLs. Onset of high-burden VT indicates poor prognosis and should warrant referral to a heart failure specialist.

Hypertrophic Cardiomyopathy With Left Ventricular Systolic Dysfunction: Insights From the SHaRe Registry.

BACKGROUND: The term "end stage" has been used to describe hypertrophic cardiomyopathy (HCM) with left ventricular systolic dysfunction (LVSD), defined as occurring when left ventricular ejection fraction is <50%. The prognosis of HCM-LVSD has reportedly been poor, but because of its relative rarity, the natural history remains incompletely characterized. METHODS: Data from 11 high-volume HCM specialty centers making up the international SHaRe Registry (Sarcomeric Human Cardiomyopathy Registry) were used to describe the natural history of patients with HCM-LVSD. Cox proportional hazards models were used to identify predictors of prognosis and incident development. RESULTS: From a cohort of 6793 patients with HCM, 553 (8%) met the criteria for HCM-LVSD. Overall, 75% of patients with HCM-LVSD experienced clinically relevant events, and 35% met the composite outcome (all-cause death [n=128], cardiac transplantation [n=55], or left ventricular assist device implantation [n=9]). After recognition of HCM-LVSD, the median time to composite outcome was 8.4 years. However, there was substantial individual variation in natural history. Significant predictors of the composite outcome included the presence of multiple pathogenic/likely pathogenic sarcomeric variants (hazard ratio [HR], 5.6 [95% CI, 2.3-13.5]), atrial fibrillation (HR, 2.6 [95% CI, 1.7-3.5]), and left ventricular ejection fraction <35% (HR, 2.0 [95% CI, 1.3-2.8]). The incidence of new HCM-LVSD was ≈7.5% over 15 years. Significant predictors of developing incident HCM-LVSD included greater left ventricular cavity size (HR, 1.1 [95% CI, 1.0-1.3] and wall thickness (HR, 1.3 [95% CI, 1.1-1.4]), left ventricular ejection fraction of 50% to 60% (HR, 1.8 [95% CI, 1.2, 2.8]-2.8 [95% CI, 1.8-4.2]) at baseline evaluation, the presence of late gadolinium enhancement on cardiac magnetic resonance imaging (HR, 2.3 [95% CI, 1.0-4.9]), and the presence of a pathogenic/likely pathogenic sarcomeric variant, particularly in thin filament genes (HR, 1.5 [95% CI, 1.0-2.1] and 2.5 [95% CI, 1.2-5.1], respectively). CONCLUSIONS: HCM-LVSD affects ≈8% of patients with HCM. Although the natural history of HCM-LVSD was variable, 75% of patients experienced adverse events, including 35% experiencing a death equivalent an estimated median time of 8.4 years after developing systolic dysfunction. In addition to clinical features, genetic substrate appears to play a role in both prognosis (multiple sarcomeric variants) and the risk for incident development of HCM-LVSD (thin filament variants).

Crowdsourcing in health and medical research: a systematic review.

BACKGROUND: Crowdsourcing is used increasingly in health and medical research. Crowdsourcing is the process of aggregating crowd wisdom to solve a problem. The purpose of this systematic review is to summarize quantitative evidence on crowdsourcing to improve health. METHODS: We followed Cochrane systematic review guidance and systematically searched seven databases up to September 4th 2019. Studies were included if they reported on crowdsourcing and related to health or medicine. Studies were excluded if recruitment was the only use of crowdsourcing. We determined the level of evidence associated with review findings using the GRADE approach. RESULTS: We screened 3508 citations, accessed 362 articles, and included 188 studies. Ninety-six studies examined effectiveness, 127 examined feasibility, and 37 examined cost. The most common purposes were to evaluate surgical skills (17 studies), to create sexual health messages (seven studies), and to provide layperson cardio-pulmonary resuscitation (CPR) out-of-hospital (six studies). Seventeen observational studies used crowdsourcing to evaluate surgical skills, finding that crowdsourcing evaluation was as effective as expert evaluation (low quality). Four studies used a challenge contest to solicit human immunodeficiency virus (HIV) testing promotion materials and increase HIV testing rates (moderate quality), and two of the four studies found this approach saved money. Three studies suggested that an interactive technology system increased rates of layperson initiated CPR out-of-hospital (moderate quality). However, studies analyzing crowdsourcing to evaluate surgical skills and layperson-initiated CPR were only from high-income countries. Five studies examined crowdsourcing to inform artificial intelligence projects, most often related to annotation of medical data. Crowdsourcing was evaluated using different outcomes, limiting the extent to which studies could be pooled. CONCLUSIONS: Crowdsourcing has been used to improve health in many settings. Although crowdsourcing is effective at improving behavioral outcomes, more research is needed to understand effects on clinical outcomes and costs. More research is needed on crowdsourcing as a tool to develop artificial intelligence systems in medicine. TRIAL REGISTRATION: PROSPERO: CRD42017052835. December 27, 2016.

Association of Race With Disease Expression and Clinical Outcomes Among Patients With Hypertrophic Cardiomyopathy.

Importance: Racial differences are recognized in multiple cardiovascular parameters, including left ventricular hypertrophy and heart failure, which are 2 major manifestations of hypertrophic cardiomyopathy. The association of race with disease expression and outcomes among patients with hypertrophic cardiomyopathy is not well characterized. Objective: To assess the association between race, disease expression, care provision, and clinical outcomes among patients with hypertrophic cardiomyopathy. Design, Setting, and Participants: This retrospective cohort study included data on black and white patients with hypertrophic cardiomyopathy from the US-based sites of the Sarcomeric Human Cardiomyopathy Registry from 1989 through 2018. Exposures: Self-identified race. Main Outcomes and Measures: Baseline characteristics; genetic architecture; adverse outcomes, including cardiac arrest, cardiac transplantation or left ventricular assist device implantation, implantable cardioverter-defibrillator therapy, all-cause mortality, atrial fibrillation, stroke, and New York Heart Association (NYHA) functional class III or IV heart failure; and septal reduction therapies. The overall composite outcome consists of the first occurrence of any component of the ventricular arrhythmic composite end point, cardiac transplantation, left ventricular assist device implantation, NYHA class III or IV heart failure, atrial fibrillation, stroke, or all-cause mortality. Results: Of 2467 patients with hypertrophic cardiomyopathy at the time of analysis, 205 (8.3%) were black (130 male [63.4%]; mean [SD] age, 40.0 [18.6] years) and 2262 (91.7%) were white (1351 male [59.7%]; mean [SD] age, 45.5 [20.5] years). Compared with white patients, black patients were younger at the time of diagnosis (mean [SD], 36.5 [18.2] vs 41.9 [20.2] years; P < .001), had higher prevalence of NYHA class III or IV heart failure at presentation (36 of 205 [22.6%] vs 174 of 2262 [15.8%]; P = .001), had lower rates of genetic testing (111 [54.1%] vs 1404 [62.1%]; P = .03), and were less likely to have sarcomeric mutations identified by genetic testing (29 [26.1%] vs 569 [40.5%]; P = .006). Implantation of implantable cardioverter-defibrillators did not vary by race; however, invasive septal reduction was less common among black patients (30 [14.6%] vs 521 [23.0%]; P = .007). Black patients had less incident atrial fibrillation (35 [17.1%] vs 608 [26.9%]; P < .001). Black race was associated with increased development of NYHA class III or IV heart failure (hazard ratio, 1.45; 95% CI, 1.08-1.94) which persisted on multivariable Cox proportional hazards regression (hazard ratio, 1.97; 95% CI, 1.34-2.88). There were no differences in the associations of race with stroke, ventricular arrhythmias, all-cause mortality, or the overall composite outcome. Conclusions and Relevance: The findings suggest that black patients with hypertrophic cardiomyopathy are diagnosed at a younger age, are less likely to carry a sarcomere mutation, have a higher burden of functionally limited heart failure, and experience inequities in care with lower use of invasive septal reduction therapy and genetic testing compared with white patients. Further study is needed to assess whether higher rates of heart failure may be associated with underlying ancestry-based disease pathways, clinical management, or structural inequities.

Stratifying risk for dementia onset using large-scale electronic health record data: A retrospective cohort study.

INTRODUCTION: Preventing dementia, or modifying disease course, requires identification of presymptomatic or minimally symptomatic high-risk individuals. METHODS: We used longitudinal electronic health records from two large academic medical centers and applied a validated natural language processing tool to estimate cognitive symptomatology. We used survival analysis to examine the association of cognitive symptoms with incident dementia diagnosis during up to 8 years of follow-up. RESULTS: Among 267,855 hospitalized patients with 1,251,858 patient years of follow-up data, 6516 (2.4%) received a new diagnosis of dementia. In competing risk regression, an increasing cognitive symptom score was associated with earlier dementia diagnosis (HR 1.63; 1.54-1.72). Similar results were observed in the second hospital system and in subgroup analysis of younger and older patients. DISCUSSION: A cognitive symptom measure identified in discharge notes facilitated stratification of risk for dementia up to 8 years before diagnosis.

Is seniority of emergency physician associated with the weekend mortality effect? An exploratory analysis of electronic health records in the UK.

OBJECTIVE: Admission to hospital over a weekend is associated with increased mortality, but the underlying causes of the weekend effect are poorly understood. We explore to what extent differences in emergency department (ED) admission and discharge processes, severity of illness and the seniority of the treating physician explain the weekend effect. METHODS: We analysed linked ED attendances to hospital admissions to Cambridge University Hospital over a 7-year period from 1 January 2007 to 31 December 2013, with 30-day in-hospital death as the primary outcome and discharge as a competing risk. The primary exposure was day of the week of arrival. Subdistribution hazards models controlled for multiple confounders, including physician seniority, calendar year, mode of arrival, triage category, referral from general practice, sex, arrival time, prior attendances and admissions, diagnosis group and age. RESULTS: 229 401 patients made 424 845 ED attendances, of which 158 396 (37.3%) were admitted to the hospital. The case-mix of admitted patients was more ill at weekends: 2530 (6.4%) admitted at a weekend required immediate resuscitation compared with 6450 (5.4%) admitted on a weekday (p<0.0001). Senior doctors admitted 24.8% of patients on weekdays and 24.0% at weekends, but junior doctors admitted 61.7% of patients on weekdays and 44.2% at weekends. 3947 (3.3%) patients admitted on a weekday and 1454 (3.7%) patients admitted at a weekend died within 30 days. In the adjusted subdistribution hazards model, the HR of in-hospital death was 1.11 (95% CI 1.04 to 1.18) for weekend arrivals. After controlling for confounders, the in-hospital mortality of patients admitted by junior doctors was greater at the weekend (adjusted HR (aHR) 1.15, 95% CI 1.06 to 1.24). In-hospital mortality for patients admitted by senior doctors was not statistically different at the weekend (aHR 1.08, 95% CI 0.98 to 1.19). CONCLUSIONS: Our findings suggest that the weekend effect was driven by a higher proportion of admitted patients requiring immediate resuscitation at the weekend. Junior doctors admitted a lower proportion of relatively healthy patients at the weekend compared with the weekday, thus diluting the risk pool of weekday admissions and contributing to the weekend effect. Senior doctors' admitting behaviour did not change at the weekend, and the corresponding weekend effect was reduced.