Assessing calibration and bias of a deployed machine learning malnutrition prediction model within a large healthcare system.

Malnutrition is a frequently underdiagnosed condition leading to increased morbidity, mortality, and healthcare costs. The Mount Sinai Health System (MSHS) deployed a machine learning model (MUST-Plus) to detect malnutrition upon hospital admission. However, in diverse patient groups, a poorly calibrated model may lead to misdiagnosis, exacerbating health care disparities. We explored the model's calibration across different variables and methods to improve calibration. Data from adult patients admitted to five MSHS hospitals from January 1, 2021 - December 31, 2022, were analyzed. We compared MUST-Plus prediction to the registered dietitian's formal assessment. Hierarchical calibration was assessed and compared between the recalibration sample (N = 49,562) of patients admitted between January 1, 2021 - December 31, 2022, and the hold-out sample (N = 17,278) of patients admitted between January 1, 2023 - September 30, 2023. Statistical differences in calibration metrics were tested using bootstrapping with replacement. Before recalibration, the overall model calibration intercept was -1.17 (95% CI: -1.20, -1.14), slope was 1.37 (95% CI: 1.34, 1.40), and Brier score was 0.26 (95% CI: 0.25, 0.26). Both weak and moderate measures of calibration were significantly different between White and Black patients and between male and female patients. Logistic recalibration significantly improved calibration of the model across race and gender in the hold-out sample. The original MUST-Plus model showed significant differences in calibration between White vs. Black patients. It also overestimated malnutrition in females compared to males. Logistic recalibration effectively reduced miscalibration across all patient subgroups. Continual monitoring and timely recalibration can improve model accuracy.

Real-time prognostic biomarkers for predicting in-hospital mortality and cardiac complications in COVID-19 patients.

Hospitalized patients with Coronavirus disease 2019 (COVID-19) are highly susceptible to in-hospital mortality and cardiac complications such as atrial arrhythmias (AA). However, the utilization of biomarkers such as potassium, B-type natriuretic peptide, albumin, and others for diagnosis or the prediction of in-hospital mortality and cardiac complications has not been well established. The study aims to investigate whether biomarkers can be utilized to predict mortality and cardiac complications among hospitalized COVID-19 patients. Data were collected from 6,927 hospitalized COVID-19 patients from March 1, 2020, to March 31, 2021 at one quaternary (Henry Ford Health) and five community hospital registries (Trinity Health Systems). A multivariable logistic regression prediction model was derived using a random sample of 70% for derivation and 30% for validation. Serum values, demographic variables, and comorbidities were used as input predictors. The primary outcome was in-hospital mortality, and the secondary outcome was onset of AA. The associations between predictor variables and outcomes are presented as odds ratio (OR) with 95% confidence intervals (CIs). Discrimination was assessed using area under ROC curve (AUC). Calibration was assessed using Brier score. The model predicted in-hospital mortality with an AUC of 90% [95% CI: 88%, 92%]. In addition, potassium showed promise as an independent prognostic biomarker that predicted both in-hospital mortality, with an AUC of 71.51% [95% Cl: 69.51%, 73.50%], and AA with AUC of 63.6% [95% Cl: 58.86%, 68.34%]. Within the test cohort, an increase of 1 mEq/L potassium was associated with an in-hospital mortality risk of 1.40 [95% CI: 1.14, 1.73] and a risk of new onset of AA of 1.55 [95% CI: 1.25, 1.93]. This cross-sectional study suggests that biomarkers can be used as prognostic variables for in-hospital mortality and onset of AA among hospitalized COVID-19 patients.

Trumpet plots: visualizing the relationship between allele frequency and effect size in genetic association studies.

Recent advances in genome-wide association and sequencing studies have shown that the genetic architecture of complex traits and diseases involves a combination of rare and common genetic variants distributed throughout the genome. One way to better understand this architecture is to visualize genetic associations across a wide range of allele frequencies. However, there is currently no standardized or consistent graphical representation for effectively illustrating these results. Here we propose a standardized approach for visualizing the effect size of risk variants across the allele frequency spectrum. The proposed plots have a distinctive trumpet shape: with the majority of variants having high frequency and small effects, and a small number of variants having lower frequency and larger effects. To demonstrate the utility of trumpet plots in illustrating the relationship between the number of variants, their frequency, and the magnitude of their effects in shaping the genetic architecture of complex traits and diseases, we generated trumpet plots for more than one hundred traits in the UK Biobank. To facilitate their broader use, we developed an R package, 'TrumpetPlots' (available at the Comprehensive R Archive Network) and R Shiny application, 'Shiny Trumpets' (available at https://juditgg.shinyapps.io/shinytrumpets/) that allows users to explore these results and submit their own data.

Predictors of Prolonged Length of Stay After Pituitary Adenoma Resection: A Large Cohort Analysis Using the National Inpatient Sample (2016-2019).

OBJECTIVE: In this nationwide retrospective study, the authors aimed to identify demographic, clinical, and baseline health risk factors predictive of a prolonged length of stay (PLOS) for patients with pituitary adenomas (PAs). METHODS: The National Inpatient Sample dataset from 2016 to 2019 was utilized to identify all included hospitalizations for PA resection as identified by the appropriate diagnosis-related group code. Comorbidities were classified based on the Charlson Comorbidity Index mapping of ICD-10 codes, and PLOS was identified as any stay longer than 3 days. Univariable and multivariable logistic regression models, accounting for the sample design, were built to determine factors associated with PLOS and emergent surgery. RESULTS: Overall, 30 945 patients were included in this study with 10 535 patients having PLOS. Female patients experienced an increased odds of PLOS (odds ratio [OR]: 1.29; P < .001). Black patients (OR: 1.49; P < .001) and Hispanic patients (OR: 1.30; P = .003) had 1.49 times and 1.30 times the odds of PLOS compared to White patients, respectively. Compared to patients insured by Medicare, patients insured by Medicaid had an increased odds of PLOS (OR: 1.36; P = .007) as well as emergent surgery (OR: 5.40; P < .001). When stratified by emergent surgeries, Black patients (OR: 1.89; P < .001), Hispanic patients, (OR: 2.14; P < .001), and patients on Medicaid insurance (OR: 1.71; P < .001) were at an increased risk of emergent procedures. However, female sex (OR: 0.65; P < .001), upper third quartile (OR: 0.73; P = .017), and fourth quartile (OR: 0.69; P = .014) of patients categorized by zip code income were at decreased odds of an emergent procedure. CONCLUSIONS: Black and Hispanic patients, patients with Medicaid insurance, and patients of low socioeconomic status patients are at significantly higher risk of emergent PA resection and PLOS. Efforts to prevent emergent surgeries and shorten hospitalization after pituitary surgery may need to primarily focus on patient groups with select sociodemographic characteristics.

Utilization of Palliative Care Services Among Patients With Malignant Brain Tumors: An Analysis of the National Inpatient Sample (2016-2019).

BACKGROUND: Recent advances in treatment of malignant brain tumors have improved outcomes. However, patients continue to experience significant disability. Palliative care helps patients with advanced illnesses improve their quality of life. There is a paucity of clinical studies examining palliative care usage among patients with malignant brain tumors. OBJECTIVE: To assess if there were any patterns in palliative care utilization among patients hospitalized with malignant brain tumors. METHODS: A retrospective cohort representing hospitalizations for malignant brain tumors was created from The National Inpatient Sample (2016-2019). Palliative care utilization was identified by ICD-10 code. Univariable and multivariable logistic regression models, accounting for the sample design, were built to evaluate the demographic variables associated with palliative care consultation in all patients and fatal hospitalizations. RESULTS: 375 010 patients admitted with a malignant brain tumor were included in this study. Over the whole cohort, 15.0% of patients used palliative care. In fatal hospitalizations, Black and Hispanic patients had 28% lower odds of receiving a palliative care consultation compared with White patients (odds ratio for both = 0.72; P = .02). For fatal hospitalizations, patients insured privately were 34% more likely to use palliative care services compared with patients insured with Medicare (odds ratio = 1.34, P = .006). CONCLUSION: Palliative care is underutilized among all patients with malignant brain tumors. Within this population, disparities in utilization are exacerbated by sociodemographic factors. Prospective studies investigating utilization disparities across race and insurance status are necessary to improve access to palliative care services for this population.

Racial disparities in post-transplant stroke and mortality following stroke in adult cardiac transplant recipients in the United States.

Black heart transplant recipients have a higher mortality rate than white recipients 6-12 months after transplant. Whether there are racial disparities in post-transplant stroke incidence and all-cause mortality following post-transplant stroke among cardiac transplant recipients is unknown. Using a nationwide transplant registry, we assessed the association between race and incident post-transplant stroke using logistic regression and the association between race and mortality among adults who survived a post-transplant stroke using Cox proportional hazards regression. We found no evidence of an association between race and the odds of post-transplant stroke (OR = 1.00, 95% CI: 0.83-1.20). The median survival time of those with a post-transplant stroke in this cohort was 4.1 years (95% CI: 3.0, 5.4). There were 726 deaths among the 1139 patients with post-transplant stroke, including 127 deaths among 203 Black patients and 599 deaths among 936 white patients. Among post-transplant stroke survivors, Black transplant recipients experienced a 23% higher rate of mortality compared to white recipients (HR = 1.23, 95% CI: 1.00-1.52). This disparity is strongest in the period beyond the first 6 months and appears to be mediated by differences in the post-transplant setting of care between Black and white patients. The racial disparity in mortality outcomes was not evident in the past decade. The improved survival of Black patients in the recent decade may reflect overall protocol improvements for heart transplant recipients irrespective of race, such as advancements in surgical techniques and immediate postoperative care as well as increased awareness about reducing racial disparities.

Global FDR control across multiple RNAseq experiments.

MOTIVATION: While classical approaches for controlling the false discovery rate (FDR) of RNA sequencing (RNAseq) experiments have been well described, modern research workflows and growing databases enable a new paradigm of controlling the FDR globally across RNAseq experiments in the past, present and future. The simplest analysis strategy that analyses each RNAseq experiment separately and applies an FDR correction method can lead to inflation of the overall FDR. We propose applying recently developed methodology for online multiple hypothesis testing to control the global FDR in a principled way across multiple RNAseq experiments. RESULTS: We show that repeated application of classical repeated offline approaches has variable control of global FDR of RNAseq experiments over time. We demonstrate that the online FDR algorithms are a principled way to control FDR. Furthermore, in certain simulation scenarios, we observe empirically that online approaches have comparable power to repeated offline approaches. AVAILABILITY AND IMPLEMENTATION: The onlineFDR package is freely available at http://www.bioconductor.org/packages/onlineFDR. Additional code used for the simulation studies can be found at https://github.com/latlio/onlinefdr_rnaseq_simulation. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Racial and Ethnic Disparities in Incidence and Prognosis of Perioperative Stroke Among Pediatric Cardiac Transplant Recipients.

Background In the general population, Black children have a higher incidence of stroke and all-cause mortality after stroke than White children. Beginning 6 months following cardiac transplantation, Black children have higher mortality than White children. However, whether there are racial and ethnic disparities in incidence and all-cause mortality following perioperative stroke among pediatric cardiac transplant recipients is unknown. Methods and Results Using the Scientific Registry of Transplant Recipients, we studied children who underwent their first heart transplant in the United States between January 1994 and September 2019. Using multivariable logistic regression, we assessed the association between race and ethnicity and perioperative stroke. We used multivariable piecewise Cox regression to examine the association between race and ethnicity and mortality among survivors of perioperative stroke. Among 8224 children who had a first cardiac transplant, 255 (3%) had a perioperative stroke. Black children had 32% lower odds of perioperative stroke compared with White children (adjusted odds ratio, 0.68 [95% CI, 0.46-0.996]). Following perioperative stroke, mortality rates were similar for Black and White children in the first 6 months (adjusted hazard ratio [HR], 0.99 [95% CI, 0.44-2.26]). However, Black children had a higher mortality rate than White children beyond 6 months (adjusted HR, 3.36 [95% CI, 1.22-9.22]). Conclusions Among pediatric cardiac transplant recipients, Black children have a lower incidence of perioperative stroke than White children. Among survivors of perioperative stroke, mortality is initially similar by race and ethnicity, but beyond 6 months, Black children have over a 3-fold higher mortality rate than White children. Identifying and intervening on potential differences in care is essential to addressing these disparities.

Post-transplant mortality and graft failure after induction immunosuppression among Black heart transplant recipients in the United States.

Black heart transplant recipients are more likely to receive induction immunosuppression compared to other races because of higher rates of acute rejection, graft failure, and mortality. However, it is not known whether contemporary induction immunosuppression improves their post-transplant outcomes. To evaluate whether Black patients who were prescribed induction immunosuppression therapy have lower all-cause mortality or graft-failure rates compared to those who were not, we studied Black U.S. adult heart transplant recipients in the Scientific Registry of Transplant Recipients database (2008-2018). We used multivariable Cox proportional hazards regression analysis to compare the hazards of all-cause mortality or graft failure as a composite, for patients who were prescribed induction immunosuppression and those who were not. Among 5160 recipients, 2787 (54.0%) were prescribed induction immunosuppression and 2373 (46.0%) were not. There was no evidence of survival differences according to induction immunosuppression for the composite of all-cause mortality or graft failure (aHR = 1.13, 95% CI 0.96-1.32), mortality (aHR = 1.14, 95% CI 0.97-1.34), graft failure (aHR = 1.05, 95% CI 0.82-1.34) and acute rejection (aHR = 1.00, 95% CI 0.89-1.12). Given the side effects of treatment, future guidelines should reconsider the recommendation for induction immunosuppression among Black patients.

Genomic risk prediction of coronary artery disease in women with breast cancer: a prospective cohort study.

BACKGROUND: Advancements in cancer therapeutics have resulted in increases in cancer-related survival; however, there is a growing clinical dilemma. The current balancing of survival benefits and future cardiotoxic harms of oncotherapies has resulted in an increased burden of cardiovascular disease in breast cancer survivors. Risk stratification may help address this clinical dilemma. This study is the first to assess the association between a coronary artery disease-specific polygenic risk score and incident coronary artery events in female breast cancer survivors. METHODS: We utilized the Studies in Epidemiology and Research in Cancer Heredity prospective cohort involving 12,413 women with breast cancer with genotype information and without a baseline history of cardiovascular disease. Cause-specific hazard ratios for association of the polygenic risk score and incident coronary artery disease (CAD) were obtained using left-truncated Cox regression adjusting for age, genotype array, conventional risk factors such as smoking and body mass index, as well as other sociodemographic, lifestyle, and medical variables. RESULTS: Over a median follow-up of 10.3 years (IQR: 16.8) years, 750 incident fatal or non-fatal coronary artery events were recorded. A 1 standard deviation higher polygenic risk score was associated with an adjusted hazard ratio of 1.33 (95% CI 1.20, 1.47) for incident CAD. CONCLUSIONS: This study provides evidence that a coronary artery disease-specific polygenic risk score can risk-stratify breast cancer survivors independently of other established cardiovascular risk factors.

Association of small, dense LDL-cholesterol concentration and lipoprotein particle characteristics with coronary heart disease: A systematic review and meta-analysis.

OBJECTIVES: The aim of this study was to systematically collate and appraise the available evidence regarding the associations between small, dense low-density lipoprotein (sdLDL) and incident coronary heart disease (CHD), focusing on cholesterol concentration (sdLDL-C) and sdLDL particle characteristics (presence, density, and size). BACKGROUND: Coronary heart disease (CHD) is the leading cause of death worldwide. Small, dense low-density lipoprotein (sdLDL) has been hypothesized to induce atherosclerosis and subsequent coronary heart disease (CHD). However, the etiological relevance of lipoprotein particle size (sdLDL) versus cholesterol content (sdLDL-C) remains unclear. METHODS: PubMed, MEDLINE, Web of Science, and EMBASE were systematically searched for studies published before February 2020. CHD associations were based on quartile comparisons in eight studies of sdLDL-C and were based on binary categorization in fourteen studies of sdLDL particle size. Reported hazards ratios (HR) and odds ratios (OR) with 95% confidence interval (CI) were standardized and pooled using a random-effects meta-analysis model. RESULTS: Data were collated from 21 studies with a total of 30,628 subjects and 5,693 incident CHD events. The average age was 67 years, and 53% were men. Higher sdLDL and sdLDL-C levels were both significantly associated with higher risk of CHD. The pooled estimate for the high vs. low categorization of sdLDL was 1.36 (95% CI: 1.21, 1.52) and 1.07 (95% CI: 1.01, 1.12) for comparing the top quartiles versus the bottom of sdLDL-C. Several studies suggested a dose response relationship. CONCLUSIONS: The findings show a positive association between sdLDL or sdLDL-C levels and CHD, which is supported by an increasing body of genetic evidence in favor of its causality as an etiological risk factor. Thus, the results support sdLDL and sdLDL-C as a risk marker, but further research is required to establish sdLDL or sdLDL-C as a potential therapeutic marker for incident CHD risk reduction.

Precision-weighted estimates of neonatal, post-neonatal and child mortality for 640 districts in India, National Family Health Survey 2016.

BACKGROUND: The conventional indicators of infant and under-five mortality are aggregate deaths occurring in the first year and the first five years, respectively. Monitoring deaths by <1 month (neonatal), 1-11 months (post-neonatal), and 12-59 months (child) can be more informative given various etiological causes that may require different interventions across these three mutually exclusive periods. For optimal resource allocation, it is also necessary to track progress in robust estimates of child survival at a smaller geographic and administrative level. METHODS: Data on 259 627 children came from the 2015-2016 Indian National Family Health Survey. We used a random effects model to account for the complex survey design and sampling variability, and predicted district-specific probabilities of neonatal, post-neonatal, and child mortality. The resulting precision-weighted estimates are more reliable as they pool information and borrow strength from other districts that share the same state membership. The Pearson correlation and Spearman's rank correlation were assessed for the three mortality estimates, and the Moran's I measure was used to detect spatial clustering of high burden districts for each outcome. RESULTS: The majority of under-five deaths was disproportionately concentrated in the neonatal period. Across all districts, the predicted probability of neonatal, post-neonatal, and child mortality varied from 6.0 to 63.9 deaths, 3.8 to 47.6 deaths, and 1.7 to 11.8 deaths per 1000 live births, respectively. The overall correlation between district-wide probabilities of mortality for the three mutually exclusive periods was moderate (Pearson correlation = 0.47-0.58, Spearman's rank correlation = 0.58-0.64). For each outcome, a relatively strong spatial clustering was detected across districts that transcended state boundaries (Moran's I = 0.61-0.76). CONCLUSIONS: Sufficiently breaking down the under-five mortality to distinct age groups and using the precision-weighted estimations to monitor performances at smaller geographic and administrative units can inform more targeted interventions and foster accountability to improve child survival.

Inequalities in life expectancy: An analysis of 201 countries, 1950-2015.

With global improvements in life expectancy, one important concern is to understand whether there is reduction in inequalities or greater cross-country convergence in expected length of life at various age thresholds. Insights on convergence patterns can help governments and other stakeholders decide upon health investments across age groups. This paper applies a novel econometric approach to test convergence and identify convergent clubs in life expectancy at various age groups for 201 countries/areas between 1950 and 2015. Life expectancy estimates for 201 countries/areas (1950 and 2015) from United Nations Department of Economic and Social Affairs (UNDESA) World Population Prospects (2015 Revision) are used for the analysis. We find global convergence in life expectancy at birth, but do not observe grand convergence for any other age groups. In the case of life expectancy at younger ages, most countries are moving in the same direction, but significant cross-country variations and convergence clubs are noted for older adults and elderly. Most of the better performing countries/areas are from Western Europe, Northern Europe and North America, the average performers are from South America, Eastern Europe, Southern Europe, South Asia, Central Asia, Eastern Africa, Central Africa, and the Caribbean Islands whereas the poor-performing ones are mainly Western Africa, Southern African and Oceania. In addition, we observe increasing between-country variance in life expectancy for older adults and elderly. The analysis reveals increasing global heterogeneity in the survival experience of older adults and the elderly population which has remained a neglected aspect in the discussions on global life expectancy improvements. Data, research and policy focus on life-expectancy at older ages is therefore critical to accelerate survival gains among older adults and elderly, particularly from the developing world.

Identifying geospatial patterns in wealth disparity in child malnutrition across 640 districts in India.

We assessed district-level geospatial trends in precision weighted prevalence and absolute wealth disparity in stunting, underweight, wasting, low birthweight, and anemia among children under five in India. The largest wealth disparities were found for anthropometric failures and substantial variation existed across states. We identified statistically significant (p < 0.001) geospatial patterns in district-wide wealth disparities for all outcomes, which differed from geospatial patterns for the overall prevalence. We characterized each district as either a "Disparity", "Pitfall", "Intensity", or "Prosperity" area based on its overall burden and wealth disparity, as well as discuss the importance of considering both measures for geographically-targeted public health interventions to improve health equity.