In-hospital outcomes in patients with and without epilepsy diagnosed with COVID-19-A cohort study.

OBJECTIVES: Coronavirus disease 2019 (COVID-19) is associated with mortality in persons with comorbidities. The aim of this study was to evaluate in-hospital outcomes in patients with COVID-19 with and without epilepsy. METHODS: We conducted a retrospective study of patients with COVID-19 admitted to a multicenter health system between March 15, 2020, and May 17, 2021. Patients with epilepsy were identified using a validated International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)/ICD-10-CM case definition. Logistic regression models and Kaplan-Meier analyses were conducted for mortality and non-routine discharges (i.e., not discharged home). An ordinary least-squares regression model was fitted for length of stay (LOS). RESULTS: We identified 9833 people with COVID-19 including 334 with epilepsy. On univariate analysis, people with epilepsy had significantly higher ventilator use (37.70% vs 14.30%, p < .001), intensive care unit (ICU) admissions (39.20% vs 17.70%, p < .001) mortality rate (29.60% vs 19.90%, p < .001), and longer LOS (12 days vs 7 days, p < .001). and fewer were discharged home (29.64% vs 57.37%, p < .001). On multivariate analysis, only non-routine discharge (adjusted odds ratio [aOR] 2.70, 95% confidence interval [CI] 2.00-3.70; p < .001) and LOS (32.50% longer, 95% CI 22.20%-43.60%; p < .001) were significantly different. Factors associated with higher odds of mortality in epilepsy were older age (aOR 1.05, 95% CI 1.03-1.08; p < .001), ventilator support (aOR 7.18, 95% CI 3.12-16.48; p < .001), and higher Charlson comorbidity index (CCI) (aOR 1.18, 95% CI 1.04-1.34; p = .010). In epilepsy, admissions between August and December 2020 or January and May 2021 were associated with a lower odds of non-routine discharge and decreased LOS compared to admissions between March and July 2020, but this difference was not statistically significant. SIGNIFICANCE: People with COVID-19 who had epilepsy had a higher odds of non-routine discharge and longer LOS but not higher mortality. Older age (≥65), ventilator use, and higher CCI were associated with COVID-19 mortality in epilepsy. This suggests that older adults with epilepsy and multimorbidity are more vulnerable than those without and should be monitored closely in the setting of COVID-19.

Mesenchymal stromal cell therapy for acute respiratory distress syndrome due to coronavirus disease 2019.

BACKGROUND AIMS: The acute respiratory distress syndrome (ARDS) resulting from coronavirus disease 2019 (COVID-19) is associated with a massive release of inflammatory cytokines and high mortality. Mesenchymal stromal cells (MSCs) have anti-inflammatory properties and have shown activity in treating acute lung injury. Here the authors report a case series of 11 patients with COVID-19-associated ARDS (CARDS) requiring mechanical ventilation who were treated with remestemcel-L, an allogeneic MSC product, under individual patient emergency investigational new drug applications. METHODS: Patients were eligible if they were mechanically ventilated for less than 72 h prior to the first infusion. Patients with pre-existing lung disease requiring supplemental oxygen or severe liver or kidney injury were excluded. Each patient received two infusions of remestemcel-L at a dose of 2 million cells/kg per infusion given 48-120 h apart. RESULTS: Remestemcel-L infusions were well tolerated in all 11 patients. At the end of the 28-day follow-up period, 10 (91%, 95% confidence interval [CI], 59-100%) patients were extubated, nine (82%, 95% CI, 48-97%) patients remained liberated from mechanical ventilation and were discharged from the intensive care unit and two (18%, 95 CI%, 2-52%) patients died. The median time to extubation was 10 days. Eight (73%, 95% CI, 34-100%) patients were discharged from the hospital. C-reactive protein levels significantly declined within 5 days of MSC infusion. CONCLUSIONS: The authors demonstrate in this case series that remestemcel-L infusions to treat moderate to severe CARDS were safe and well tolerated and resulted in improved clinical outcomes.

A flexible approach for variable selection in large-scale healthcare database studies with missing covariate and outcome data.

BACKGROUND: Prior work has shown that combining bootstrap imputation with tree-based machine learning variable selection methods can provide good performances achievable on fully observed data when covariate and outcome data are missing at random (MAR). This approach however is computationally expensive, especially on large-scale datasets. METHODS: We propose an inference-based method, called RR-BART, which leverages the likelihood-based Bayesian machine learning technique, Bayesian additive regression trees, and uses Rubin's rule to combine the estimates and variances of the variable importance measures on multiply imputed datasets for variable selection in the presence of MAR data. We conduct a representative simulation study to investigate the practical operating characteristics of RR-BART, and compare it with the bootstrap imputation based methods. We further demonstrate the methods via a case study of risk factors for 3-year incidence of metabolic syndrome among middle-aged women using data from the Study of Women's Health Across the Nation (SWAN). RESULTS: The simulation study suggests that even in complex conditions of nonlinearity and nonadditivity with a large percentage of missingness, RR-BART can reasonably recover both prediction and variable selection performances, achievable on the fully observed data. RR-BART provides the best performance that the bootstrap imputation based methods can achieve with the optimal selection threshold value. In addition, RR-BART demonstrates a substantially stronger ability of detecting discrete predictors. Furthermore, RR-BART offers substantial computational savings. When implemented on the SWAN data, RR-BART adds to the literature by selecting a set of predictors that had been less commonly identified as risk factors but had substantial biological justifications. CONCLUSION: The proposed variable selection method for MAR data, RR-BART, offers both computational efficiency and good operating characteristics and is utilitarian in large-scale healthcare database studies.

Postpartum psychiatric readmissions: A nationwide study in women with and without epilepsy.

OBJECTIVE: To assess whether epilepsy is associated with increased odds of 30-day readmission due to psychiatric illness during the postpartum period. METHODS: The 2014 Nationwide Readmissions Database and the International Classification of Disease, Ninth Revision, Clinical Modification codes were used to identify postpartum women up to 50 years old in the United States, including the subgroup with epilepsy. The primary outcome was 30-day readmission and was categorized as (1) readmission due to psychiatric illness, (2) readmission due to all other causes, or (3) no readmission. Secondary outcome was diagnosis at readmission. The association of the primary outcome and presence of epilepsy was examined using multinomial logistic regression. RESULTS: Of 1 558 875 women with admissions for delivery identified, 6745 (.45%) had epilepsy. Thirteen of every 10 000 women had 30-day psychiatric readmissions in the epilepsy group compared to one of every 10 000 in the no-epilepsy group (p < .0001). Of every 10 000 women with epilepsy, 256 had 30-day readmissions due to other causes compared to 115 of every 10 000 women in the no-epilepsy group (p < .0001). The odds ratio for readmission due to psychiatric illness was 10.13 (95% confidence interval = 5.48-18.72) in those with epilepsy compared to those without. Top psychiatric causes for 30-day readmissions among women with epilepsy were mood disorders, schizophrenia and other psychotic disorders, and substance-related disorders. SIGNIFICANCE: This large-scale study demonstrated that postpartum women with epilepsy have higher odds of readmission due to a psychiatric illness compared to women without epilepsy. Postpartum treatment strategies and interventions to prevent psychiatric readmissions are necessary in this vulnerable population.

Hospital readmissions in older adults with epilepsy in the US - A population-based study.

OBJECTIVE: Our objective was to determine proportions, causes, and predictors of 30-day readmissions among older adults with epilepsy. Understanding predictors of readmissions may inform future interventions aimed at reducing avoidable hospitalizations in this vulnerable population. METHODS: Individuals 65 years or older with epilepsy were identified using previously validated ICD-9-CM codes in any diagnostic position in the 2014 Nationwide Readmissions Database. Proportions of 30-day readmissions and causes of readmissions in older adults with epilepsy were compared to both older adults without and younger adults (18-64 years old) with epilepsy. We identified predictors of readmission in older adults with epilepsy using logistic regression. RESULTS: There were 92,030 older adults with, 3,166,852 older adults without, and 168,622 younger adults with epilepsy. Proportions of readmissions were higher in older adults with (16.2%) than older adults without (12.5%) and younger adults with epilepsy (15.1%). The main cause of readmission for older adults with and without epilepsy was septicemia, and epilepsy/seizure in younger adults with epilepsy. Predictors of 30-day readmissions in older adults with epilepsy were: non-elective admissions (OR 1.37, 95%CI 1.27-1.48), public insurance (Medicaid vs. private insurance OR 1.19, 95%CI 1.02-1.39; Medicare vs. private insurance OR 1.11, 95%CI 1.00-1.22), lower median household income for patient's zip code ($1-$39,999 vs. $66,000 + OR 1.15, 95% CI 1.08-1.22), hospital location in large metropolitan areas (OR 1.22, 95%CI 1.05-1.42), higher Charlson-Deyo comorbidity index (OR 1.11, 95%CI 1.10-1.02), and male sex (OR 1.04, 95%CI 1.00-1.09). SIGNIFICANCE: Our findings suggest that targeted interventions to reduce the risk of infection may potentially reduce readmission in older people with epilepsy, similarly to those without. Provision of coordinated care and appropriate discharge planning may reduce readmissions particularly in those who are males, are of lower socioeconomic status and with more comorbidities.

The co-occurrence of dementia in those with epilepsy is associated with 30-day readmission - A population-based study.

OBJECTIVE: Dementia and epilepsy often co-occur and are associated with poor health outcomes and increased healthcare utilization. The literature on the association between readmission and co-occurrence of dementia and epilepsy is scant. Our objective was to determine if dementia in patients with epilepsy >40 years old is associated with 30-day hospital readmission, in-hospital mortality, discharge disposition, and length-of-stay. METHODS: This retrospective cohort study used the 2014 Nationwide Readmissions Database, containing data from hospital discharges across the US and readmissions. Epilepsy and dementia were identified using previously validated ICD-9-CM codes. Primary outcome was 30-day readmission, analyzed with univariable and multivariable logistic regressions. Secondary outcomes were discharge disposition, in-hospital mortality, and length-of-stay, analyzed with univariable multinomial logistic, univariable logistic, and univariable ordinary least squared regressions, respectively. The top ten causes of readmission in each group were compared as well. All analyses accounted for survey weights, cluster, and stratum. RESULTS: Patients with epilepsy with dementia (n = 15,588) had longer hospital stays [15% (95%CI 10-20%)], and higher odds of readmission [OR 1.11 (95%CI 1.05-1.17)], transfer to another facility [OR 2.18 (95%CI 1.93-2.46)], and in-hospital mortality [OR 1.50 (95%CI 1.25-1.79)] compared to those without dementia (n = 186,289).The top two causes of readmission were septicemia (dementia: 14.81%; no dementia: 9.45%) and epilepsy/convulsions (dementia: 5.91%; no dementia: 6.25%). Other top 10 causes of readmissions in those with epilepsy and dementia which were not present in those without dementia included delirium (5.21%), urinary tract infections (4.98%), and aspiration pneumonitis (4.29%). SIGNIFICANCE: Dementia in epilepsy is associated with worse outcomes, including higher in-hospital mortality and higher readmissions. Potentially preventable causes of readmission in those with epilepsy and dementia were identified, including septicemia, delirium, urinary tract infection, and aspiration pneumonitis. Future studies are needed to inform interventions aimed at decreasing premature mortality and reducing potentially preventable readmissions in this vulnerable population.

Nonelective hospital admissions, discharge disposition, and health services utilization in epilepsy patients: A population-based study.

OBJECTIVE: Identifying adverse outcomes and examining trends and causes of nonelective admissions among persons with epilepsy would be beneficial to optimize patient care and reduce health services utilization. We examined the association of epilepsy with discharge status, in-hospital mortality, length-of-stay, and charges. We also examined 10-year trends and causes of hospital admissions among those with and without epilepsy. METHODS: Nonelective hospital admission in persons with epilepsy was identified in the 2005-2014 National Inpatient Sample (NIS) using a validated International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) case definition. The NIS is the largest US all-payer database including patient and hospital-level variables, and represents hospitalizations in the general population. Descriptive statistics on trends and causes of admissions and multivariable regression analysis summarizing the association of epilepsy with the outcomes of interest are presented. RESULTS: Of 4 718 178 nonelective admissions in 2014, 3.80% (n = 179 461) were in persons with epilepsy. Admissions in persons with epilepsy increased from 14 636 to 179 461 (P < .0001) between 2005 and 2014. As compared to persons without epilepsy, hospital admissions in persons with epilepsy had higher odds of transfer to other facilities (odds ratio [OR] = 1.77, 95% confidence interval [CI]: 1.72-1.81, P < .0001), being discharged against medical advice (OR = 1.48, 95% CI: 1.38-1.59, P < .0001), and incurring 4% greater total charges (P < .0001). Epilepsy, convulsions, pneumonia, mood disorders, cerebrovascular disease, and septicemia were the top causes for admissions in those with epilepsy. SIGNIFICANCE: Future research should focus on designing targeted health care interventions that decrease the number of hospital admissions, reduce health services utilization, and increase the odds of discharge home in people with epilepsy.

Comparison of statistical and machine learning models for healthcare cost data: a simulation study motivated by Oncology Care Model (OCM) data.

BACKGROUND: The Oncology Care Model (OCM) was developed as a payment model to encourage participating practices to provide better-quality care for cancer patients at a lower cost. The risk-adjustment model used in OCM is a Gamma generalized linear model (Gamma GLM) with log-link. The predicted value of expense for the episodes identified for our academic medical center (AMC), based on the model fitted to the national data, did not correlate well with our observed expense. This motivated us to fit the Gamma GLM to our AMC data and compare it with two other flexible modeling methods: Random Forest (RF) and Partially Linear Additive Quantile Regression (PLAQR). We also performed a simulation study to assess comparative performance of these methods and examined the impact of non-linearity and interaction effects, two understudied aspects in the field of cost prediction. METHODS: The simulation was designed with an outcome of cost generated from four distributions: Gamma, Weibull, Log-normal with a heteroscedastic error term, and heavy-tailed. Simulation parameters both similar to and different from OCM data were considered. The performance metrics considered were the root mean square error (RMSE), mean absolute prediction error (MAPE), and cost accuracy (CA). Bootstrap resampling was utilized to estimate the operating characteristics of the performance metrics, which were described by boxplots. RESULTS: RF attained the best performance with lowest RMSE, MAPE, and highest CA for most of the scenarios. When the models were misspecified, their performance was further differentiated. Model performance differed more for non-exponential than exponential outcome distributions. CONCLUSIONS: RF outperformed Gamma GLM and PLAQR in predicting overall and top decile costs. RF demonstrated improved prediction under various scenarios common in healthcare cost modeling. Additionally, RF did not require prespecification of outcome distribution, nonlinearity effect, or interaction terms. Therefore, RF appears to be the best tool to predict average cost. However, when the goal is to estimate extreme expenses, e.g., high cost episodes, the accuracy gained by RF versus its computational costs may need to be considered.

Using evolutionary computation on GPS position correction.

More and more devices are equipped with global positioning system (GPS). However, those handheld devices with consumer-grade GPS receivers usually have low accuracy in positioning. A position correction algorithm is therefore useful in this case. In this paper, we proposed an evolutionary computation based technique to generate a correction function by two GPS receivers and a known reference location. Locating one GPS receiver on the known location and combining its longitude and latitude information and exact poisoning information, the proposed technique is capable of evolving a correction function by such. The proposed technique can be implemented and executed on handheld devices without hardware reconfiguration. Experiments are conducted to demonstrate performance of the proposed technique. Positioning error could be significantly reduced from the order of 10 m to the order of 1 m.

Emotional and Financial Stressors in New York City During the COVID-19 Pandemic: A Consecutive Cross-Sectional Analysis.

Mental and financial hardship during the COVID-19 pandemic in New York City was severe, but how vulnerable groups have been disproportionately impacted is incompletely understood. In partnership with community stakeholders, we administered a web-based survey to a convenience sample of New York City residents (18 + years) from May 2020 to April 2021 to evaluate their financial and emotional stressors. We analyzed outcomes by race, ethnicity, and education level. A total of 1854 adults completed the survey across three consecutive non-overlapping samples. Fifty-five percent identified other than non-Latinx White. Sixty-four percent reported emotional stress; 38%, 32%, and 32% reported symptoms of anxiety, depression, and post-traumatic stress disorder respectively; and 21% reported a large adverse financial impact. The leading unmet needs were mental health and food services (both 19%), and health services (18%). Need for both resources grew over time. Adverse financial impact directly correlated with presence of all four adverse mental health outcomes above. In multivariate analysis, non-White race and lack of college degree were associated with adverse financial impact, whereas LGBT identity and lack of college degree were associated with mental health impact. Throughout the COVID-19 pandemic, participants in this research demonstrated a large and growing mental and financial strain, disproportionately associated with lower education level, non-White race, and LGBT status. Our findings suggest an urgent need to differentially target COVID-19 mental health and resource support in New York City to persons in these vulnerable communities.

Assessment of systemic and gastrointestinal tissue damage biomarkers for GVHD risk stratification.

We used a rigorous PRoBE (prospective-specimen collection, retrospective-blinded-evaluation) study design to compare the ability of biomarkers of systemic inflammation and biomarkers of gastrointestinal (GI) tissue damage to predict response to corticosteroid treatment, the incidence of clinically severe disease, 6-month nonrelapse mortality (NRM), and overall survival in patients with acute graft-versus-host disease (GVHD). We prospectively collected serum samples of newly diagnosed GVHD patients (n = 730) from 19 centers, divided them into training (n = 352) and validation (n = 378) cohorts, and measured TNFR1, TIM3, IL6, ST2, and REG3α via enzyme-linked immunosorbent assay. Performances of the 4 strongest algorithms from the training cohort (TNFR1 + TIM3, TNFR1 + ST2, TNFR1 + REG3α, and ST2 + REG3α) were evaluated in the validation cohort. The algorithm that included only biomarkers of systemic inflammation (TNFR1 + TIM3) had a significantly smaller area under the curve (AUC; 0.57) than the AUCs of algorithms that contained ≥1 GI damage biomarker (TNFR1 + ST2, 0.70; TNFR1 + REG3α, 0.73; ST2 + REG3α, 0.79; all P < .001). All 4 algorithms were able to predict short-term outcomes such as response to systemic corticosteroids and severe GVHD, but the inclusion of a GI damage biomarker was needed to predict long-term outcomes such as 6-month NRM and survival. The algorithm that included 2 GI damage biomarkers was the most accurate of the 4 algorithms for all endpoints.

Variable selection with missing data in both covariates and outcomes: Imputation and machine learning.

Variable selection in the presence of both missing covariates and outcomes is an important statistical research topic. Parametric regression are susceptible to misspecification, and as a result are sub-optimal for variable selection. Flexible machine learning methods mitigate the reliance on the parametric assumptions, but do not provide as naturally defined variable importance measure as the covariate effect native to parametric models. We investigate a general variable selection approach when both the covariates and outcomes can be missing at random and have general missing data patterns. This approach exploits the flexibility of machine learning models and bootstrap imputation, which is amenable to nonparametric methods in which the covariate effects are not directly available. We conduct expansive simulations investigating the practical operating characteristics of the proposed variable selection approach, when combined with four tree-based machine learning methods, extreme gradient boosting, random forests, Bayesian additive regression trees, and conditional random forests, and two commonly used parametric methods, lasso and backward stepwise selection. Numeric results suggest that, extreme gradient boosting and Bayesian additive regression trees have the overall best variable selection performance with respect to the F1 score and Type I error, while the lasso and backward stepwise selection have subpar performance across various settings. There is no significant difference in the variable selection performance due to imputation methods. We further demonstrate the methods via a case study of risk factors for 3-year incidence of metabolic syndrome with data from the Study of Women's Health Across the Nation.

Estimating heterogeneous survival treatment effects of lung cancer screening approaches: A causal machine learning analysis.

The National Lung Screening Trial (NLST) found that low-dose computed tomography (LDCT) screening provided lung cancer (LC) mortality benefit compared to chest radiography (CXR). Considerable research concerns identifying the differential treatment effects that may exist in certain subpopulations. We shed light on several important issues in existing research and highlight the need for further investigation of the heterogeneous comparative effect of LDCT versus CXR, using more flexible and rigorous statistical approaches. We used a high-performance Bayesian machine learning approach designed for censored survival data, accelerated failure time Bayesian additive regression trees model (AFT-BART), to flexibly capture the relationships between the failure time and predictors. We then used the counterfactual framework to draw Markov chain Monte Carlo samples of the individual treatment effect for each participant. Using these posterior samples, we explored the possible treatment effect heterogeneity via a stepwise binary tree approach. When re-analyzed with AFT-BART, LDCT did not have a statistically significant LC or overall mortality benefit compared to CXR. The Asian and Black (particularly those with pack-year ≥ 37 years and without emphysema) NLST population were shown to have enhanced overall mortality benefit from LDCT than the population average. Although inconclusive for LC mortality benefit, Asians, Blacks and Whites with history of chronic obstructive pulmonary disease showed a small trend towards benefit from LDCT. Causal inference with flexible machine learning modeling can provide valuable knowledge for informing treatment decision and planning targeted clinical trials emphasizing personalized medicine approaches.

A single-institutional analysis of racial disparities in clinicopathologic characteristics, treatment selections, and outcomes in advanced-stage pancreatic cancer patients.

BACKGROUND: To investigate racial disparities among unresectable/metastatic pancreatic ductal adenocarcinoma (PDA) patients treated with contemporary chemotherapy regimens at an urban center. METHODS: Retrospective review of all PDA patients treated at a single institution between 2012-2017. Continuous and categorical variables were tested using t-test, Mann-Whitney U, chi-squared or Fisher's exact test as appropriate. Kaplan-Meier curves were generated and Cox proportional hazards models were used to analyze survival outcomes. RESULTS: One hundred and forty-five patients identified as: White [69], African American (AA, 34), Asian [15], and Other [27]. Fifty-five-point-seven percent of patients received gemcitabine-based therapy vs. 36.6% received fluorouracil (5-FU) based therapy, specifically 26.1% received FOLFIRINOX and 43.7% received gemcitabine/nab-paclitaxel. In a univariable model, Asians had significantly worse overall survival (OS) than Whites [hazard ratio (HR) 2.74, P=0.013], but there were no OS differences between AA vs. Whites (HR 1.51, P=0.297) nor Other vs. Whites (HR 2.05, P=0.062). On multivariable analysis, Asians had worse OS compared to Whites (HR 2.62, P=0.018), and gemcitabine-based therapy was inferior to 5-FU-based therapy (HR 2.65, P=0.005). There were no OS differences between AA vs. Whites nor Other vs. Whites (HR 1.12, P=0.769 and HR 0.8, P=0.763, respectively). CONCLUSIONS: In this series of advanced PDA patients treated with contemporary chemotherapy, AA and White patients had comparable outcomes, but Asians had worse OS than White patients.

Evaluating Race and Time to Transplantation in Multiple Myeloma: The Mount Sinai Hospital Experience.

BACKGROUND: Previous studies have found that Black patients with multiple myeloma undergo autologous stem-cell transplantation (ASCT) less frequently than their white counterparts, although the factors leading to decreased access and utilization have not been fully elucidated. PATIENTS AND METHODS: To identify whether racial differences in transplantation timing played a role in these disparities, we retrospectively analyzed 410 Black and white patients who received their first transplant at The Mount Sinai Hospital between 2011 and 2016 (260 white and 150 Black patients). We compared the time from initial diagnosis to stem-cell collection and the time from collection to transplantation between the 2 races while controlling for age, socioeconomic status, and functional status. RESULTS: Between Blacks and whites, time from diagnosis to collection was higher in Black patients (median 238, vs. 195 days, respectively, P = .051). Functional status, socioeconomic status, and age were also significantly associated with time to collection, and after controlling for these covariates, the effect of race was not significant (P = .0625). Conversely, time from collection to transplantation was increased in white patients compared to Black. CONCLUSION: Increased time from diagnosis to stem-cell collection for Black patients was driven in part by socioeconomic status and baseline functional status.

Derivation and validation of a novel comorbidity-based delirium risk index to predict postoperative delirium using national administrative healthcare database.

OBJECTIVE: To derive and validate a comorbidity-based delirium risk index (DRI) to predict postoperative delirium. DATA SOURCE/STUDY SETTING: Data of 506 438 hip fracture repair surgeries from 2006 to 2016 were collected to derive DRI and perform internal validation from the Premier Healthcare Database, which provided billing information on 20-25 percent of hospitalizations in the USA. Additionally, data of 1 130 569 knee arthroplasty surgeries were retrieved for external validation. STUDY DESIGN: Thirty-six commonly seen comorbidities were evaluated by logistic regression with the outcome of postoperative delirium. The hip fracture repair surgery cohort was separated into a training dataset (60 percent) and an internal validation (40 percent) dataset. The least absolute shrinkage and selection operator (LASSO) procedure was applied for variable selection, and weights were assigned to selected comorbidities to quantify corresponding risks. The newly developed DRI was then compared to the Charlson-Deyo Index for goodness-of-fit and predictive ability, using the Akaike information criterion (AIC), Bayesian information criterion (BIC), area under the ROC curve (AUC) for goodness-of-fit, and odds ratios for predictive performance. Additional internal validation was performed by splitting the data by four regions and in 4 randomly selected hospitals. External validation was conducted in patients with knee arthroplasty surgeries. DATA COLLECTION: Hip fracture repair surgeries, knee arthroplasty surgeries, and comorbidities were identified by using ICD-9 codes. Postoperative delirium was defined by using ICD-9 codes and analyzing billing information for antipsychotics (specifically haloperidol, olanzapine, and quetiapine) typically recommended to treat delirium. PRINCIPAL FINDINGS: The derived DRI includes 14 comorbidities and assigns comorbidities weights ranging from 1 to 6. The DRI outperformed the Charlson-Deyo Comorbidity Index with better goodness-of-fit and predictive performance. CONCLUSIONS: Delirium risk index is a valid comorbidity index for covariate adjustment and risk prediction in the context of postoperative delirium. Future work is needed to test its performance in different patient populations and varying definitions of delirium.

A nationally representative study on discharge against medical advice among those living with epilepsy.

PURPOSE: Discharges against medical advice (DAMA) are associated with adverse patient outcomes among those with epilepsy. Our goal was to examine trends and factors associated with DAMA among those living with epilepsy. METHODS: A retrospective cross-sectional study was performed using the 2003-2014 National Inpatient Sample database. ICD-9-CM diagnosis codes were used to identify admissions of patients with epilepsy. Following outcomes were examined among epilepsy patients: proportion and predictors of DAMA, 12-year DAMA trends and causes of admissions. RESULTS: In 2014, of the 187,850 admissions in patients with epilepsy, 3783 (2.01 %) were DAMA. Male sex, Black race, younger age, lower household income, Medicaid/self-pay/other as primary payer, lower Elixhauser comorbidities index, weekend admission, non-elective admission, hospital in northeast region, and urban nonteaching hospital were all associated with DAMA. There was a significant increase in the proportion of DAMA in people with epilepsy from 2003 to 2014 (1.13 %-2.01 %, p < 0.0001). The top reasons of admissions for epilepsy patients who were DAMA were: epilepsy/convulsion (21.02 %), alcohol- (8.86 %) and substance-related disorders (3.75 %), and diabetes mellitus with complications (3.33 %). CONCLUSIONS: Our findings provide opportunities to understand DAMA among those living with epilepsy, which is more prevalent in socially-disadvantaged populations. This study highlight the need to develop electronic medical records-based prediction tools that could be used at the point-of-care to enable the early identification of people at risk for DAMA, since it is often likely preventable. Future mixed methods studies are recommended to identify facilitators of DAMA and strategies for prevention.

Impact of pathological response after neoadjuvant chemotherapy on adjuvant therapy decisions and patient outcomes in gastrointestinal cancers.

BACKGROUND: Neoadjuvant chemotherapy (NAC) is frequently used in gastrointestinal cancers (GIC), and pathological, radiological, and tumor marker responses are assessed during and after NAC. AIM: To evaluate the relationship between pathologic, radiologic, tumor marker responses and recurrence-free survival (RFS), overall survival (OS), adjuvant chemotherapy (AC) decisions, and the impact of changing to a different AC regimen after poor response to NAC. METHODS AND RESULTS: Medical records of GIC patients treated with NAC at Mount Sinai between 1/2012 and 12/2018 were reviewed. One hundred fifty-six patients (58.3% male, mean age 63 years) were identified. Primary tumor sites were: 43 (27.7%) pancreas, 62 (39.7%) gastroesophageal, and 51 (32.7%) colorectal. After NAC, 31 (19.9%) patients had favorable pathologic response (FPR; defined as College of American Pathologists [CAP] score 0-1). Of 107 patients with radiological data, 59 (55.1%) had an objective response, and of 113 patients with tumor marker data, 61 (54.0%) had a ≥50% reduction post NAC. FPR, but not radiographic or serological responses, was associated with improved RFS (HR 0.28; 95% CI 0.11-0.72) and OS (HR 0.13; 95% CI 0.2-0.94). Changing to a different AC regimen from initial NAC, among all patients and specifically among those with unfavorable pathological response (UPR; defined as CAP score 2-3) after NAC, was not associated with improved RFS or OS. CONCLUSIONS: GIC patients with FPR after NAC experienced significant improvements in RFS and OS. Patients with UPR did not benefit from changing AC. Prospective studies to better understand the role of pathological response in AC decisions and outcomes in GIC patients are needed.

Antacid exposure and immunotherapy outcomes among patients with advanced hepatocellular carcinoma.

BACKGROUND: Antibiotic exposure has been associated with worse outcomes with immune checkpoint inhibitors (ICIs) in cancer patients, likely due to disruption of the gut microbiome. Other commonly prescribed medications, such as proton pump inhibitors (PPIs) and histamine-2-receptor antagonists (H2RAs), are also known to disrupt the microbiome, but data on their association with ICI outcomes are conflicting. METHODS: We conducted a retrospective, multicenter, international cohort study including 314 hepatocellular carcinoma (HCC) patients treated with ICIs from 2017 to 2019 to assess the association between PPI or H2RA exposure (up to 30 days before ICI) and overall survival. Secondary outcomes included overall response rate (ORR) and development of any treatment-related adverse events (AEs). RESULTS: Baseline PPI/H2RA exposure was not associated with overall survival in univariable (HR 1.01, 95% CI 0.75-1.35) or multivariable analysis (HR 0.98, 95% CI 0.71-1.36). Baseline PPI/H2RA exposure was not associated with either ORR (OR 1.32, 95% CI 0.66-2.65) or AEs (OR 1.07, 95% CI 0.54-2.12) in multivariable analysis. CONCLUSIONS: Our results suggest that exposure to PPI/H2RA prior to ICIs does not adversely affect outcomes in HCC patients.

Predictive value of neutrophil to lymphocyte ratio and platelet to lymphocyte ratio in advanced hepatocellular carcinoma patients treated with anti-PD-1 therapy.

BACKGROUND: Currently, there are no recognized or validated biomarkers to identify hepatocellular carcinoma patients (HCC) likely to benefit from anti-PD-1 therapy. We evaluated the relationship between neutrophil-lymphocyte ratio (NLR) and platelet-lymphocyte ratio (PLR) and survival outcomes, pretreatment and after three doses (posttreatment) of nivolumab in HCC patients. METHODS: Medical records of HCC patients treated with nivolumab between June 2016 and July 2018 were reviewed. Kaplan-Meier analysis and the log-rank test were used to calculate and compare overall survival between NLR < 5 Vs ≥ 5 and among PLR tertiles. RESULTS: A total of 103 patients were identified. Median age was 66 (29-89) years. Median treatment duration was 26 (2-149) weeks. Sixty-four (62%) patients had Child-Pugh class A (CP-A) liver function. Barcelona Clinic Liver Cancer stage was B in 20 (19%) and C in 83 (81%) patients. CP-A patients who achieved a partial or complete response had significantly lower posttreatment NLR and PLR (P < .001 for both) compared to patients who had stable disease or progression of disease. No relationship was observed between response and pretreatment NLR and PLR. NLR < 5 was associated with improved OS compared to NLR ≥ 5 both pretreatment (23 Vs10 months, P = .004) and posttreatment (35 Vs 9 months, P < .0001). Survival also differed significantly among PLR tertiles both pre- (P = .05) and posttreatment (P = .013). In a multivariable model, posttreatment NLR (HR = 1.10, P < .001) and PLR (HR = 1.002, P < .001) were strongly associated with survival. In a composite model of posttreatment NLR and PLR, a combination of high NLR and PLR was associated with an eightfold increased risk of death (HR = 8.3, P < .001). CONCLUSIONS: This study suggests a strong predictive role of these inflammatory cell ratios in the posttreatment setting in HCC patients treated with anti anti-PD-1 therapy and should be evaluated in a larger cohort.