Machine Learning to Predict Mortality and Critical Events in a Cohort of Patients With COVID-19 in New York City: Model Development and Validation.

BACKGROUND: COVID-19 has infected millions of people worldwide and is responsible for several hundred thousand fatalities. The COVID-19 pandemic has necessitated thoughtful resource allocation and early identification of high-risk patients. However, effective methods to meet these needs are lacking. OBJECTIVE: The aims of this study were to analyze the electronic health records (EHRs) of patients who tested positive for COVID-19 and were admitted to hospitals in the Mount Sinai Health System in New York City; to develop machine learning models for making predictions about the hospital course of the patients over clinically meaningful time horizons based on patient characteristics at admission; and to assess the performance of these models at multiple hospitals and time points. METHODS: We used Extreme Gradient Boosting (XGBoost) and baseline comparator models to predict in-hospital mortality and critical events at time windows of 3, 5, 7, and 10 days from admission. Our study population included harmonized EHR data from five hospitals in New York City for 4098 COVID-19-positive patients admitted from March 15 to May 22, 2020. The models were first trained on patients from a single hospital (n=1514) before or on May 1, externally validated on patients from four other hospitals (n=2201) before or on May 1, and prospectively validated on all patients after May 1 (n=383). Finally, we established model interpretability to identify and rank variables that drive model predictions. RESULTS: Upon cross-validation, the XGBoost classifier outperformed baseline models, with an area under the receiver operating characteristic curve (AUC-ROC) for mortality of 0.89 at 3 days, 0.85 at 5 and 7 days, and 0.84 at 10 days. XGBoost also performed well for critical event prediction, with an AUC-ROC of 0.80 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. In external validation, XGBoost achieved an AUC-ROC of 0.88 at 3 days, 0.86 at 5 days, 0.86 at 7 days, and 0.84 at 10 days for mortality prediction. Similarly, the unimputed XGBoost model achieved an AUC-ROC of 0.78 at 3 days, 0.79 at 5 days, 0.80 at 7 days, and 0.81 at 10 days. Trends in performance on prospective validation sets were similar. At 7 days, acute kidney injury on admission, elevated LDH, tachypnea, and hyperglycemia were the strongest drivers of critical event prediction, while higher age, anion gap, and C-reactive protein were the strongest drivers of mortality prediction. CONCLUSIONS: We externally and prospectively trained and validated machine learning models for mortality and critical events for patients with COVID-19 at different time horizons. These models identified at-risk patients and uncovered underlying relationships that predicted outcomes.

Peri-implantitis prevalence, incidence rate, and risk factors: A study of electronic health records at a U.S. dental school.

OBJECTIVES: We assessed peri-implantitis prevalence, incidence rate, and associated risk factors by analyzing electronic oral health records (EHRs) in an educational institution. METHODS: We used a validated reference cohort comprising all patients receiving dental implants over a 3.5-year period (2,127 patients and 6,129 implants). Electronic oral health records of a random 10% subset were examined for an additional follow-up of ≥2.5 years to assess the presence of radiographic bone loss, defined as >2 mm longitudinal increase in the distance between the implant shoulder and the supporting peri-implant bone level (PBL) between time of placement and follow-up. "Intact" implants had no or ≤2 mm PBL increase from baseline. Electronic oral health record notes were reviewed to corroborate a definitive peri-implantitis diagnosis at implants with progressive bone loss. A nested case-control analysis of peri-implantitis-affected implants randomly matched by age with "intact" implants from peri-implantitis-free individuals identified putative risk factors. RESULTS: The prevalence of peri-implantitis over an average follow-up of 2 years was 34% on the patient level and 21% on the implant level. Corresponding incidence rates were 0.16 and 0.10 per patient-year and implant-year, respectively. Multiple conditional logistic regression identified ill-fitting fixed prosthesis (OR = 5.9; 95% CI: 1.6-21.1), cement-retained prosthesis (OR = 4.5; 2.1-9.5), and radiographic evidence of periodontitis (OR = 3.6; 1.7-7.6) as statistically associated with peri-implantitis. Implant location in the mandible (OR = 0.02; 0.003-0.2) and use of antibiotics in conjunction with implant surgery (OR = 0.19; 0.05-0.7) emerged as protective exposures. CONCLUSIONS: Approximately 1/3 of the patients and 1/5 of all implants experienced peri-implantitis. Ill-fitting/ill-designed fixed and cement-retained restorations, and history of periodontitis emerged as the principal risk factors for peri-implantitis.

Consumer Health Information Technology in the Prevention of Substance Abuse: Scoping Review.

BACKGROUND: Addiction is one of the most rapidly growing epidemics that currently plagues nations around the world. In the United States, it has cost the government more than US $700 billion a year in terms of health care and other associated costs and is also associated with serious social, physical, and mental consequences. Increasing efforts have been made to tackle this issue at different levels, from primary prevention to rehabilitation across the globe. With the use of digital technology rapidly increasing, an effort to leverage the consumer health information technologies (CHITs) to combat the rising substance abuse epidemic has been underway. CHITs are identified as patient-focused technological platforms aimed to improve patient engagement in health care and aid them in navigating the complex health care system. OBJECTIVE: This review aimed to provide a holistic and overarching view of the breadth of research on primary prevention of substance abuse using CHIT conducted over nearly past five decades. It also aimed to map out the changing landscape of CHIT over this period. METHODS: We conducted a scoping review using the Arksey and O'Malley's modified methodological framework. We searched 4 electronic databases (PubMed, Cochrane, Scopus, and EMBASE). Papers were included if the studies addressed the use of CHIT for primary prevention of substance abuse and were published in English between 1809 and 2018. Studies that did not focus solely on primary prevention or assessed additional comorbid conditions were eliminated. RESULTS: Forty-two papers that met our inclusion criteria were included in the review. These studies were published between 1970 and 2018 and were not restricted by geography, age, race, or sex. The review mapped studies using the most commonly used CHIT platforms for substance abuse prevention from mass media in the 1970s to mobile and social media in 2018. Moreover, 191 studies that were exclusively focused on alcohol prevention were excluded and will be addressed in a separate paper. The studies included had diverse research designs although the majority were randomized controlled trials (RCT) or review papers. Many of the RCTs used interventions based on different behavioral theories such as family interactions, social cognitive theories, and harm-minimization framework. CONCLUSIONS: This review found CHIT platforms to be efficacious and cost-effective in the real-world settings. We also observed a gradual shift in the types and use of CHIT platforms over the past few decades and mapped out their progression. In addition, the review detected a shift in consumer preferences and behaviors from face-to-face interactions to technology-based platforms. However, the studies included in this review only focused on the aspect of primary prevention. Future reviews could assess the effectiveness of platforms for secondary prevention and for prevention of substance abuse among comorbid populations.

Introducing Contactless Blood Pressure Assessment Using a High Speed Video Camera.

Recent studies demonstrated that blood pressure (BP) can be estimated using pulse transit time (PTT). For PTT calculation, photoplethysmogram (PPG) is usually used to detect a time lag in pulse wave propagation which is correlated with BP. Until now, PTT and PPG were registered using a set of body-worn sensors. In this study a new methodology is introduced allowing contactless registration of PTT and PPG using high speed camera resulting in corresponding image-based PTT (iPTT) and image-based PPG (iPPG) generation. The iPTT value can be potentially utilized for blood pressure estimation however extent of correlation between iPTT and BP is unknown. The goal of this preliminary feasibility study was to introduce the methodology for contactless generation of iPPG and iPTT and to make initial estimation of the extent of correlation between iPTT and BP "in vivo." A short cycling exercise was used to generate BP changes in healthy adult volunteers in three consecutive visits. BP was measured by a verified BP monitor simultaneously with iPTT registration at three exercise points: rest, exercise peak, and recovery. iPPG was simultaneously registered at two body locations during the exercise using high speed camera at 420 frames per second. iPTT was calculated as a time lag between pulse waves obtained as two iPPG's registered from simultaneous recoding of head and palm areas. The average inter-person correlation between PTT and iPTT was 0.85 ± 0.08. The range of inter-person correlations between PTT and iPTT was from 0.70 to 0.95 (p < 0.05). The average inter-person coefficient of correlation between SBP and iPTT was -0.80 ± 0.12. The range of correlations between systolic BP and iPTT was from 0.632 to 0.960 with p < 0.05 for most of the participants. Preliminary data indicated that a high speed camera can be potentially utilized for unobtrusive contactless monitoring of abrupt blood pressure changes in a variety of settings. The initial prototype system was able to successfully generate approximation of pulse transit time and showed high intra-individual correlation between iPTT and BP. Further investigation of the proposed approach is warranted.

Deep Learning Approaches to Predict Exercise Exertion Levels Using Wearable Physiological Data.

Using physiological data from wearable devices, the study aimed to predict exercise exertion levels by building deep learning classification and regression models. Physiological data were obtained using an unobtrusive chest-worn ECG sensor and portable pulse oximeter from healthy individuals who performed 16-minute cycling exercise sessions. During each session, real-time ECG, pulse rate, oxygen saturation, and revolutions per minute (RPM) data were collected at three intensity levels. Subjects' ratings of perceived exertion (RPE) were collected once per minute. Each 16-minute exercise session was divided into eight 2-minute windows. The self-reported RPEs, heart rate, RPMs, and oxygen saturation levels were averaged for each window to form the predictive features. In addition, heart rate variability (HRV) features were extracted from the ECG for each window. Different feature selection algorithms were used to choose top-ranked predictors. The best predictors were then used to train and test deep learning models for regression and classification analysis. Our results showed the highest accuracy and F1 score of 98.2% and 98%, respectively in training the models. For testing the models, the highest accuracy and F1 score were 80%.

Pneumococcal Vaccination Lowers the Risk of Alzheimer's Disease: A Study Utilizing Data from the IBM® MarketScan® Database.

Previous studies demonstrated an association between influenza vaccination and the likelihood of developing Alzheimer's disease. This study was aimed at assessing whether pneumococcal vaccinations are associated with a lower risk of Alzheimer's disease based on analysis of data from the IBM® MarketScan® Database. Vaccinated and unvaccinated matched cohorts were generated using propensity-score matching with the greedy nearest-neighbor matching algorithm. The conditional logistic regression method was used to estimate the relationship between pneumococcal vaccination and the onset of Alzheimer's disease. There were 142,874 subjects who received the pneumococcal vaccine and 14,392 subjects who did not. The conditional logistic regression indicated that the people who received the pneumococcal vaccine had a significantly lower risk of developing Alzheimer's disease as compared to the people who did not receive any pneumococcal vaccine (OR=0.37; 95%CI: 0.33-0.42; P-value < .0001). Our findings demonstrated that the pneumococcal vaccine was associated with a 63% reduction in the risk of Alzheimer's disease among US adults aged 65 and older.

Identifying Determinants of Survival Disparities in Multiple Myeloma Patients Using Electronic Health Record Data.

Multiple myeloma (MM) is one of the most common hematological malignancies. The goal of this study was to analyze the sociodemographic, economic, and genetic characteristics of long-term and short-term survival of multiple myeloma patients using EHR data from an academic medical center in New York City. The de-identified analytical dataset comprised 2,111 patients with MM who were stratified based on the length of survival into two groups. Demographic variables, cancer stage, income level, and genetic mutations were analyzed using descriptive statistics and logistic regression. Age, race, and cancer stage were all significant factors that affected the length of survival of multiple myeloma patients. In contrast, gender and income level were not significant factors based on the multivariate adjusted analysis. Older adults, African American patients, and patients who were diagnosed with stage III of multiple myeloma were the people most likely to exhibit short-term survival after the MM diagnosis.

Towards a Patient-Centered Design of a Cancer Telerehabilitation System.

Successful implementation of telehealth platforms requires a detailed understanding of patient's needs, preferences, and attitudes toward a home-based platform. The goal of this study was to identify patient-centered characteristics of a cancer rehabilitation system based on cognitive evaluation of user interface and semi-structured qualitative interviews. Quantitative and qualitative feedback from 29 patients with metastatic urogenital cancer was collected after using a cancer telerehabilitation system. Heuristic evaluation, cognitive walkthrough, and analysis of qualitative interviews demonstrated a high level of support for the concept of home-based cancer telerehabilitation by cancer patients. Post-task surveys demonstrated sufficient usability and satisfaction scores from the participants. The patients provided valuable and insightful comments on how to further improve the functionality and interface of the platform. Further improvement of the system usability, consistency, and accessibility based on the patient-centered design principles will significantly facilitate the implementation of cancer telerehabilitation in clinical practice.

Assessing the Barriers and Facilitators to Pulmonary Rehabilitation Referrals Using the Consolidated Framework for Implementation Research (CFIR).

Chronic obstructive pulmonary disease (COPD) is a global health issue causing significant illness and death. Pulmonary Rehabilitation (PR) offers non-pharmacological treatment, including education, exercise, and psychological support which was shown to improve clinical outcomes. In both stable COPD and after an acute exacerbation, PR has been demonstrated to increase exercise capacity, decrease dyspnea, and enhance quality of life. Despite these benefits, referrals for PR for COPD treatment remain low. This study aims to evaluate the perceptions of healthcare providers for referring a COPD patient to PR. Semi-structured qualitative interviews were conducted with pulmonary specialists, hospitalists, and emergency department physicians. Domains and constructs from the Consolidated Framework for Implementation Research (CFIR) were applied to the qualitative data to organize, analyze, and identify the barriers and facilitators to referring COPD patients. The findings from this study will help guide strategies to improve the referral process for PR.

NLP-Assisted Differential Diagnosis of Chronic Obstructive Pulmonary Disease Exacerbation.

Chronic Obstructive Pulmonary Disease (COPD) frequently coincides with other comorbidities such as congestive heart failure, hypertension, coronary artery disease, or atrial fibrillation. The exhibition of overlapping sets of symptoms associated with these conditions prevents early identification of an acute exacerbation upon admission to a hospital. Early identification of the underlying cause of exacerbation allows timely prescription of an optimal treatment plan as well as allows avoiding unnecessary clinical tests and specialist consultations. The aim of this study was to develop a predictive model for early identification of COPD exacerbation by using the clinical notes generated within 24 hours of admission to the hospital. The study cohort included patients with a prior diagnosis of COPD. Four predictive models have been developed, among which the support vector machine showed the best performance based on the resulting 80% F1 score.

Association of Diagnostic Discrepancy with Length of Stay and Mortality in Congestive Heart Failure Patients Admitted to the Emergency Department.

The goal of this study was to analyze diagnostic discrepancies between emergency department (ED) and hospital discharge diagnoses in patients with congestive heart failure admitted to the ED. Using a synthetic dataset from the Department of Veterans Affairs, the patients' primary diagnoses were compared at two levels: diagnostic category and body system. With 12,621 patients and 24,235 admission cases, the study found a 58% mismatch rate at the category level, which was reduced to 30% at the body system level. Diagnostic categories associated with higher levels of mismatch included aplastic anemia, pneumonia, and bacterial infections. In contrast, diagnostic categories associated with lower levels of mismatch included alcohol-related disorders, COVID-19, cardiac dysrhythmias, and gastrointestinal hemorrhage. Further investigation revealed that diagnostic mismatches are associated with longer hospital stays and higher mortality rates. These findings highlight the importance of reducing diagnostic uncertainty, particularly in specific diagnostic categories and body systems, to improve patient care following ED admission.

Assessing Patient Perspectives on Pulmonary Telerehabilitation Using Thematic Analysis of Semi-Structured Qualitative Interviews.

The study was aimed at exploring patients' experiences after the completion of a 12-month pulmonary telerehabilitation (PR) program. Semi-structured qualitative interviews were conducted with 16 COPD patients. The interviews were analyzed using a thematic approach to identify patterns and themes. The patients exhibited high acceptability and satisfaction with the remote PR program and provided valuable input for its improvement. These insights will be used for the implementation of a patient-centered COPD telerehabilitation system.

Immersive Technologies for Depression Care: Scoping Review.

BACKGROUND: Depression significantly impacts quality of life, affecting approximately 280 million people worldwide. However, only 16.5% of those affected receive treatment, indicating a substantial treatment gap. Immersive technologies (IMTs) such as virtual reality (VR) and augmented reality offer new avenues for treating depression by creating immersive environments for therapeutic interventions. Despite their potential, significant gaps exist in the current evidence regarding the design, implementation, and use of IMTs for depression care. OBJECTIVE: We aim to map the available evidence on IMT interventions targeting depression treatment. METHODS: This scoping review followed a methodological framework, and we systematically searched databases for studies on IMTs and depression. The focus was on randomized clinical trials involving adults and using IMTs. The selection and charting process involved multiple reviewers to minimize bias. RESULTS: The search identified 16 peer-reviewed articles, predominantly from Europe (n=10, 63%), with a notable emphasis on Poland (n=9, 56%), which contributed to more than half of the articles. Most of the studies (9/16, 56%) were conducted between 2020 and 2021. Regarding participant demographics, of the 16 articles, 5 (31%) exclusively involved female participants, and 7 (44%) featured participants whose mean or median age was >60 years. Regarding technical aspects, all studies focused on VR, with most using stand-alone VR headsets (14/16, 88%), and interventions typically ranging from 2 to 8 weeks, predominantly in hospital settings (11/16, 69%). Only 2 (13%) of the 16 studies mentioned using a specific VR design framework in planning their interventions. The most frequently used therapeutic approach was Ericksonian psychotherapy, used in 56% (9/16) of the studies. Notably, none of the articles reported using an implementation framework or identified barriers and enablers to implementation. CONCLUSIONS: This scoping review highlights the growing interest in using IMTs, particularly VR, for depression treatment but emphasizes the need for more inclusive and comprehensive research. Future studies should explore varied therapeutic approaches and cost-effectiveness as well as the inclusion of augmented reality to fully realize the potential of IMTs in mental health care.

Machine Learning Approaches for Exercise Exertion Level Classification Using Data from Wearable Physiologic Monitors.

This research aimed to develop a model for real-time prediction of aerobic exercise exertion levels. ECG signals were registered during 16-minute cycling exercises. Perceived ratings of exertion (RPE) were collected each minute from the study participants. Based on the reported RPE, each consecutive minute of the exercise was assigned to the "high exertion" or "low exertion" class. The characteristics of heart rate variability (HRV) in time and frequency domains were used as predictive features. The top ten ranked predictive features were selected using the minimum redundancy maximum relevance (mRMR) algorithm. The support vector machine demonstrated the highest accuracy with an F1 score of 82%.

The Efficacy of Long-Term Hydroxychloroquine Use in the Prevention of COVID-19: A Retrospective Cohort Study.

The use of hydroxychloroquine (HCQ) in the prevention or treatment of COVID-19 remains controversial due to the insufficient supporting evidence and clinical studies indicating that it does not reduce COVID-19 mortality. Its potential protective effects against SARS-CoV-2 are still unclear. Big data resources, such as MarketScan database containing over 30 million insured participants annually, have not been used systematically to assess the association between long-term HCQ use and the risk of COVID-19. This retrospective study aimed to determine the protective effect of HCQ using the MarketScan database. We examined COVID-19 incidence from January to September 2020 among adult patients with systemic lupus erythematosus or rheumatoid arthritis who had received HCQ for at least 10 months in 2019 compared to those who did not. Propensity score matching was used to control for confounding variables and make the HCQ and non-HCQ groups comparable in this study. After matching at the ratio of 1:2, the analytical dataset comprised 13,932 patients who received HCQ for over 10 months and 27,754 HCQ-naïve patients. Multivariate logistic regression showed that long-term HCQ use was associated with a lower likelihood of COVID-19 in patients who had been receiving HCQ for over 10 months (OR=0.78, 95% CI: 0.69-0.88). These findings suggest that long-term HCQ use may provide protection against COVID-19.

Using Big Data to Uncover Association Between Sildenafil Use and Reduced Risk of Alzheimer's Disease.

Alzheimer's disease is a chronic neurodegenerative disease with multiple pathogenesis pathways. Sildenafil, one of the phosphodiesterase-5 inhibitors, was proven to have effective benefits in transgenic Alzheimer's disease mice. The purpose of the study was to investigate the relationship between sildenafil use and the risk of Alzheimer's disease based on the IBM® MarketScan® Database covering over 30 million employees and family members per year. Sildenafil and non-sildenafil-matched cohorts were generated using propensity-score matching with the greedy nearest-neighbor algorithm. The propensity score stratified univariate analysis and the Cox regression model showed that sildenafil use was significantly associated with a 60% risk reduction of developing Alzheimer's disease (HR=0.40; 95%CI:0.38-0.44; P<.0001) compared to the cohort of individuals who did not take sildenafil. Sex-stratified analyses revealed that sildenafil was related to a lower risk of Alzheimer's disease in subgroups of both males and females. Our findings demonstrated a significant association between sildenafil use and a lower risk of Alzheimer's disease.

Combining NLP and Machine Learning for Differential Diagnosis of COPD Exacerbation Using Emergency Room Data.

Chronic Obstructive Pulmonary Disease (COPD) exacerbation exhibits a set of overlapping symptoms with various forms of cardiovascular disease, which makes its early identification challenging. Timely identification of the underlying condition that caused acute admission of COPD patients in the emergency room (ER) may improve patient care and reduce care costs. This study aims to use machine learning combined with natural language processing (NLP) of ER notes to facilitate differential diagnosis in COPD patients admitted to ER. Using unstructured patient information extracted from the notes documented at the very first hours of admission to the hospital, four machine learning models were developed and tested. The random forest model demonstrated the best performance with F1 score of 93%.

Identification of Subphenotypes of Opioid Use Disorder Using Unsupervised Machine Learning.

This paper aimed to detect the latent clusters of patients with opioid use disorder and to identify the risk factors affecting drug misuse using unsupervised machine learning. The cluster with the highest proportion of successful treatment outcomes was characterized by the highest percentage of employment rate at admission and discharge, the highest percentage of patients who also recovered from alcohol and other drug co-use, and the highest proportion of patients who recovered from untreated health issues. Longer participation in opioid treatment programs was associated with the highest proportion of treatment success.

Feasibility of a Virtual Reality App to Promote Pulmonary Rehabilitation.

One of the major barriers to joining pulmonary rehabilitation (PR) programs is a lack of awareness about its benefits, combined with overall skepticism about regular exercise among COPD patients. Empowering COPD patients with foundational knowledge about PR may potentially facilitate their decision to join a PR program. A virtual reality (VR) app may serve as an engaging and interactive means to deliver PR education; however, the feasibility of this approach in COPD patients is unknown. The goal of this project was to assess the feasibility of VR-based PR education in COPD patients. Using mixed methods design, the feasibility of the VR app was assessed by evaluating its usability, patient acceptance, and its impact on patient knowledge about PR. The results of the usability assessment showed high user acceptance of the VR system and the ability to successfully operate the VR appliances. The use of the VR education app resulted in a statistically significant increase in patient understanding of the main concepts of pulmonary rehabilitation. Further development and evaluation of VR-based systems for patient engagement and empowerment are warranted.

Towards an Open-Source Platform to Harmonize and Share Research Results from COVID-19 Interventional and Observational Studies.

As the SARS-CoV-2 virus continues to remain a universal threat on a global scale, a large number of COVID-19 clinical trials and observational studies are being conducted and published. Currently, 9,202 COVID-19 clinical trials have been registered on ClinicalTrials.gov and 293,187 COVID-19 articles were indexed in PubMed. To fully capitalize on the voluminous number of publications reporting COVID-19 interventional and observational studies, their results should be freely accessible via an open-source harmonized shared resource. We introduced ReMeDy (https://remedy.mssm.edu/), an intelligent integrative informatics platform aimed to harmonize and cross-link diverse COVID-19 trial outcomes and observational data. We tested the potential of the platform by uploading 52 COVID-19 clinical trials and 48 COVID-19 observational retrospective studies. ReMeDy was validated based on its capability to store and organize diverse data. The next steps include developing a crowdsourcing functionality coupled with automated outcome extraction using natural language processing.