Predictive Modeling of Stress in the Healthcare Industry During COVID-19: A Novel Approach Using XGBoost, SHAP Values, and Tree Explainer

There was a substantial medicine shortage and an increase in morbidity due to the second wave of the COVID-19 pandemic in India. This pandemic has also had a drastic impact on healthcare professionals' psychological health as they were surrounded by suffering, death, and isolation. Healthcare practitioners in North India were sent a self-administered questionnaire based on the COVID-19 Stress Scale (N = 436) from March to May 2021. With 10-fold cross-validation, extreme gradient boosting (XGBoost) was used to predict the individual stress levels. XGBoost classifier was applied, and classification accuracy was 88%. The results of this research show that approximately 52.6% of healthcare specialists in the dataset exceed the severe psychiatric morbidity standards. Further, to determine which attribute had a significant impact on stress prediction, advanced techniques (SHAP values), and tree explainer were applied. The two most significant stress predictors were found to be medicine shortage and trouble in concentrating.

Evaluation of the model malaria elimination strategy in Mandla district along with its neighbouring districts: a time series analysis from 2008 to 2020.

BACKGROUND: Compared to 2017, India achieved a significant reduction in malaria cases in 2020. Madhya Pradesh (MP) is a tribal dominated state of India with history of high malaria burden in some districts. District Mandla of MP state showed a considerable decline in malaria cases between 2000 and 2013, except in 2007. Subsequently, a resurgence of malaria cases was observed during 2014 and 2015. The Malaria Elimination Demonstration Project (MEDP) was launched in 2017 in Mandla with the goal to achieve zero indigenous malaria cases. This project used: (1) active surveillance and case management using T4 (Track fever, Test fever, Treat patient, and Track patient); (2) vector control using indoor residual sprays and long-lasting insecticidal nets; (3) information education communication and behaviour change communication; and (4) regular monitoring and evaluation with an emphasis on operational and management accountability. This study has investigated malaria prevalence trends from 2008 to 2020, and has predicted trends for the next 5 years for Mandla and its bordering districts. METHODS: The malaria prevalence data of the district Mandla for the period of January 2008 to August 2017 was obtained from District Malaria Office (DMO) Mandla and data for the period of September 2017 to December 2020 was taken from MEDP data repository. Further, the malaria prevalence data for the period of January 2008 to December 2020 was collected from DMOs of the neighbouring districts of Mandla. A univariate time series and forecast analysis was performed using seasonal autoregressive integrated moving average model. FINDINGS: Malaria prevalence in Mandla showed a sharp decline [- 87% (95% CI - 90%, - 84%)] from 2017 to 2020. The malaria forecast for Mandla predicts zero cases in the next 5 years (2021-2025), provided current interventions are sustained. By contrast, the model has forecasted a risk of resurgence of malaria in other districts in MP (Balaghat, Dindori, Jabalpur, Seoni, and Kawardha) that were not the part of MEDP. CONCLUSION: The interventions deployed as part of MEDP have resulted in a sustainable zero indigenous malaria cases in Mandla. Use of similar strategies in neighbouring and other malaria-endemic districts in India could achieve similar results. However, without adding extra cost to the existing intervention, sincere efforts are needed to sustain these interventions and their impact using accountability framework, data transparency, and programme ownership from state to district level.

Predictive Modeling of Stress in the Healthcare Industry during COVID-19: A Novel Approach Using XGBoost, SHAP Values, and Tree Explainer

There was a substantial medicine shortage and an increase in morbidity due to the second wave of the COVID-19 pandemic in India. This pandemic has also had a drastic impact on healthcare professionals' psychological health as they were surrounded by suffering, death, and isolation. Healthcare practitioners in North India were sent a self-administered questionnaire based on the COVID-19 Stress Scale (N = 436) from March to May 2021. With 10-fold cross-validation, extreme gradient boosting (XGBoost) was used to predict the individual stress levels. XGBoost classifier was applied, and classification accuracy was 88%. The results of this research show that approximately 52.6% of healthcare specialists in the dataset exceed the severe psychiatric morbidity standards. Further, to determine which attribute had a significant impact on stress prediction, advanced techniques (SHAP values), and tree explainer were applied. The two most significant stress predictors were found to be medicine shortage and trouble in concentrating. © 2022 IGI Global. All rights reserved.

Healthcare resource utilization and costs associated with COVID-19 patients with preexisting type 2 diabetes and concurrent cardiovascular disease

BACKGROUND: The impact of migr2D) and cardiovascular disease (CVD) are associated with increased morbidity and mortality in COVID-19 (C-19) patients. However, the economic burden associated with these pre-existing comorbidities is not well understood. OBJECTIVE(S): This study aimed to compare the healthcare resource utilization (HCRU) and costs among C-19 patients with pre-existing T2D + CVD, T2D only, or neither. METHOD(S): This retrospective study used administrative claims in the HealthCore Integrated Research Database from US commercial and Medicare Advantage health plans. Patients with C-19 were identified from March 1, 2020-May 31, 2021, and stratified by presence of T2D and CVD. HCRU and costs were identified during follow-up and presented on a per patient per month (PPPM) basis. Propensity score matching and multivariable analyses were performed to adjust for differences between the three groups. RESULT(S): A total of 321,232 C-19 patients were identified (21,651 T2D + CVD, 28,184 T2D only, and 271,397 neither) with a mean follow-up of 5.4 months. C-19 patients with T2D + CVD were significantly older and had a greater comorbidity burden than those with neither. The unadjusted analysis shows that PPPM costs during follow-up were $14,790, $5,717, and $1,891 for T2D + CVD, T2D only, and neither, respectively, with C-19 related costs contributing 78%, 75%, and 64% of the overall costs. The majority of costs occurred during the first month after C-19 infection. After matching, 6,967 patients were identified for each cohort. Hospitalization occurred in 34.2% (T2D + CVD), 26.0% (T2D only), and 21.2% (neither), with a mean length of stay of 9.5, 9.9, and 8.9 days. Emergency room visits were reported in 28.6%, 24.5% and 20.4%, respectively. In-person physician and telehealth visits followed a similar pattern, with the highest number of visits among C-19 patients with T2D + CVD. Multivariable models show that C-19 patients with T2D + CVD were 59% more likely to be hospitalized, incurring 54% greater costs than those with neither. Patients with T2D only were 28% more likely to be hospitalized with 21% greater costs than those with neither. CONCLUSION(S): C-19 patients with pre-existing T2D + CVD had the greatest economic burden even after accounting for baseline differences between groups. The magnitude of increased HCRU and costs suggests that more aggressive triage and management of C-19 patients with both T2D and CVD may favorably impact economic outcomes.

Role of biochemical and inflammatory markers in assessing COVID-19 severity among the Indian population: An observational study

Introduction: Different laboratory parameters get altered in coronavirus disease 2019 (COVID-19);therefore, the changes of these parameters could help recognize the patients with severe disease. This study was conducted to achieve a comprehensive biochemical and inflammatory profile of COVID-19 among the Indian population. Methods: The study consisted of 730 patients admitted to Jaya Arogya Hospital, Gwalior, with COVID-19 from August 2020 to December 2020. The patients were divided into mild disease group (MDG) (n=533) and severe disease group (SDG) (n=197) depending on certain criteria, and their biochemical and inflammatory markers were collected. Data were analyzed using SPSS version 25. Results: Statistically significant rise in blood urea (P=0.011), serum creatinine (P=0.008), serum bilirubin (P=0.012), interleukin 6 (IL-6) (P<0.001), and troponin I (P<0.001) was observed in SDG as compared to MDG. Serum electrolytes (sodium and potassium) and serum protein (total protein and albumin) showed a significant fall in SDG as compared to MDG (P<0.001 for electrolytes and P=0.023 for proteins). The area under the receiver operating characteristic curve (AUROC) showed a high diagnostic value of IL-6. Conclusion: Patients with severe COVID-19 showed a high prevalence of hyperbilirubinemia, hypoproteinemia, electrolyte imbalance, and raised inflammatory markers (IL-6, troponin I, and procalcitonin). Results showed their effectiveness in assessing disease severity and predicting outcomes in patients with COVID-19. © 2022 The Author(s).

MEDIATION OF BIOMARKERS OF INFLAMMATION IN SLEEP-RELATED HYPOXIA AND COVID-19 CLINICAL OUTCOMES

Introduction: Central to the pathophysiology of SARS-CoV-2 is immune dysregulation and systemic inflammation, however, it is yet unknown whether sleep-related hypoxemia-which we have recently noted to be associated with worse COVID-19 clinical outcomes-is mediated by these biomarkers and pathways. Methods: Data from patients who tested positive for SARS-CoV-2 and part of the integrated Cleveland Clinic COVID-19 and sleep laboratory registries from March-November 2020 were included. To assess the mediation effect of biomarkers, the relationship between sleep-related hypoxia measures (% sleep time<90%SaO2,T90) and moderate/severe WHO-7 COVID-19 score (use of supplemental oxygen, non-invasive ventilation, mechanical ventilation/ ECMO or death) was first tested. The mediation effect, or natural indirect effect, of biomarkers of inflammation (C-Reactive Protein (CRP), white blood cell (WBC) count (with a focus on lymphocyte count) and lactate) was then estimated by logistic regression models adjusted for demographics, comorbidities, smoking pack year and site location using PROC CAUSALMED statement in SAS software (version 9.4, Cary, NC). Results: The analytic sample included 446 patients hospitalized due to COVID-19: age:63.3.±13.8 years,51.3% female,39% African American with body mass index(BMI)=36.1±9.3kg/ m2. Thirty-six percent used supplemental oxygen, 4% used highflow or non-invasive ventilation,5% required ECMO or mechanical ventilation and 2% died. Hypoxic measures were associated with moderate/severe WHO-7 COVID-19 outcome: T90 median (>1.8%vs.≤1.8%) (OR=2.04, 95%CI:1.28-3.23,p=0.003), 5% increases in both mean SaO2 (OR=0.43, 95%CI: 0.26-0.70,p=<0.001) and minimum SaO2 (OR=0.84, 95%CI: 0.72-0.99,p=0.03). CRP was associated with mean SaO2 (p=0.040) and minimum SaO2 (p=0.029), likewise mediation analysis showed that there was a significant natural indirect effect of CRP in both hypoxia measures (OR=0.86,95%CI 0.73-0.99,p=0.036;OR=0.95,95%CI 0.90- 1.00,p=0.034 respectively). WBC count, but not lymphocyte count subset, was associated with mean SaO2 (p=0.044), but the natural indirect effect was not significant (p=0.23. Lactate was associated with minimum SaO2 (p=0.044), but the natural indirect effect was not significant (p=0.23). T90 median was not associated with CRP(p=0.13), WBC count(p=0.87) or lactate(p=0.28). Conclusion: CRP appears to represent a relevant mediator of sleep-related hypoxia and WHO-7 clinical outcomes. Further investigation is needed to elucidate if treatment of sleep-related hypoxia downregulates biomarkers of systemic inflammation to modify disease course.

SLEEP DISTURBANCES IN POST-ACUTE SEQUELAE OF COVID-19 (PASC)

Introduction: Sleep difficulties and fatigue are highly prevalent, pervasive symptoms reported in patients with Post-Acute Sequelae of COVID-19 (PASC). As little is known of the predictors and severity of PASC-related sleep disturbance and intersection with fatigue, we leverage systematic data collected from the Cleveland Clinic ReCOVer Clinic for further elucidation Methods: Analysis of data collected from Cleveland Clinic ReCOVer Clinic patients (February-November 2021) who completed the Patient-Reported Outcomes Measurement (PROMIS) Sleep Disturbance and PROMIS Fatigue questionnaires was performed. Data were extracted from the Cleveland Clinic COVID-19 registry and the electronic health record.PROMIS scores are standardized to the general U.S. adult population on a T-scale with mean 50±10. PROMIS sleep disturbance and fatigue T-scores ≥60 indicates at least moderate disturbance and ≥70 indicate severe disturbance. T-test and Chi-square tests were used to examine cross-group differences. Multivariable logistic regression adjusted for age, race, sex, and body mass index(kg/m2) was performed to investigate factors associated with sleep disturbance severity. Results: Out of 1321, 682 patients completed the PROMIS Sleep Disturbance questionnaire with age 49.8±13.6, 75.2% female and 12.3% black race. Average T-scores were 57.7±8.3, 281 (41.2%) patients reported at least moderate sleep disturbance and 50 (7.3%) reported severe sleep disturbances. Average PROMIS Fatigue T-score was 63.0±9.2;68.6% patients reported at least moderate fatigue, 22.6% reported severe fatigue. Patients with moderate-severe compared to normal-to-mild sleep disturbances respectively had higher BMI (32.3±8.7 vs 30.9±7.5, p=0.049), were more likely of black race (40.0±10.0 vs 41.0±15.7,p=0.010), had worse eneral Anxiety Disorder (GAD)-2 questionnaires scores (2.8±2.1 vs 1.6±1.7,p<0.001), Patient Health Questionnaire (PHQ)-2 scores (2.8±2.0 vs 1.6±1.7,p<0.001) and PROMIS fatigue scores (66.7±7.8 vs 60.4±9.1,p<0.001) with no difference in age, sex, or hospitalization due to COVID-19. In the adjusted model, black race was associated with moderate-severe sleep disturbance (OR=3.42, 95%CI:1.64-7.13). Conclusion: The prevalence of moderate to severe sleep disturbances reported by patients presenting for PASC was very high i.e.>40% and associated with obesity, black race and mood symptoms. Notably, after adjustment for demographics, black race conferred a 3-fold higher odds of moderate-severe sleep disturbance emphasizing the need to characterize race-specific determinants and disparities in COVID-19 survivors.

Predictive Modeling of Healthcare Professional's Stress Based on XGBoost Model

In India, the second wave of the COVID-19 pandemic has resulted in a significant shortage of medicines and increased morbidity. COVID-19 has also had a profound influence on the psychological well-being of health professionals, who are surrounded by agony, death, and isolation as a result of the epidemic. The goal of this cross-sectional study is to look into the mental health of Indian healthcare workers during the second wave of the COVID-19 outbreak. From March 2021 to May 2021, a self-administered questionnaire based on the COVID-19 Stress Scale was delivered online to healthcare professionals (N = 836) in north India. An ensemble learning technique - Extreme Gradient Boosting (XGBoost) was applied to predict individual stress levels with 10-fold cross-validation. XGBoost had predicted stress with an average accuracy of 0.8889. According to the findings of this study, around 52.6 percent of healthcare professionals in the sample meet the threshold for severe psychiatric morbidity. In addition, advanced methodologies (SHAP values) were employed to determine which features had a significant impact on stress prediction. Medicine shortages and trouble concentrating were found to be the two most significant CSS predictors. © 2021 IEEE.

Sleep disordered breathing polysomnographic measures and COVID-19 risk of who-7 clinical outcomes in a large health care system

Introduction: There is lack of clarity of sleep disordered breathing (SDB)-including the role of nocturnal hypoxia and confounding influence of obesity-on the clinical course of human coronavirus disease 2019 (COVID-19). We postulate that SDB portends increased risk of adverse COVID-19 clinical outcomes even after accounting for confounding factors. Methods: A retrospective cohort analysis of COVID-19 and sleep laboratory observational registries March-November 2020 within the Cleveland Clinic health system was performed. Ordinal logistic regression assessed the association of SDB indices and World Health Organization (WHO)-7 COVID-19 clinical outcome (hospitalization, use of supplemental oxygen, non-invasive ventilation, mechanical ventilation/ ECMO and death) in an unadjusted model and adjusted for age, sex, race, body mass index(BMI,kg/m2),diabetes mellitus, hypertension, coronary artery disease, heart failure, asthma, chronic obstructive pulmonary disease (COPD), cancer and smoking using SAS software. Results: Of 19,449 (32%) patients positive for SARS-CoV-2,2,290 (6%) had an available sleep study. The analytic sample included 1788 of which 1,484(64%) had an apnea hypopnea index (AHI, 3-4% hypopnea oxygen desaturation)≥5. The median duration from sleep study to COVID test was 5.8 years (IQR:3.3-9.0). Age was 56.5±14.4 years,50.4% female,28% African American with BMI=35.9±8.9kg/m2. Nine percent of patients were hospitalized,10% with supplemental oxygen,6% used non-invasive ventilation,2% required ECMO or mechanical ventilation and 2% died. For every AHI increase of 5, the odds of a higher WHO-7 level increased 2% (OR=1.02,95%CI1.01-1.04,p=0.005),but the association was mitigated in the adjusted model (OR=1.00,95%CI:0.98,1.02,p=0.80). Per 5% increase in time spent with SaO2<90%, the odds of a higher WHO-7 level increased 10% (OR=1.10,95%CI1.06-1.13,p=<0.001) persisting in the adjusted model(OR=1.06,95%CI:1.02-1.10,p=0.002). For every decrease of 5% mean SaO2, the odds of a higher level WHO-7 increased 56% (OR=0.56,95%CI:0.46-0.67,p<0.001) persisting in the adjusted model(OR=0.72,95%CI:0.58-0.89,p=0.003). Conclusion: Even after adjustment for obesity, underlying cardiopulmonary disease and smoking, sleep-related hypoxemia was a potential key pathophysiologic mechanism associated with increased morbidity and mortality in COVID-19. Elucidation of sleep-related hypoxemia as a risk stratification measure, particularly given the silent hypoxia inherent to early COVID-19, is critical for future investigation, as is the role of sleep-related hypoxia reversal as a target to improve COVID-19 outcomes.

Pediatric obstructive sleep apnea (OSA) and COVID-19-related adverse clinical outcomes

Introduction: The relationship of OSA and human coronavirus (COVID-19) in the pediatric population is unknown. We postulate that OSA is associated with SARS-CoV-2 positivity and with adverse COVID-19 outcomes in children. Methods: A retrospective review of 120 consecutive patients (<18 years) with prior polysomnogram (PSG) and COVID-19 testing from the Cleveland Clinic COVID-19 registry was conducted. Using a case control design of SARS-CoV-2 positive and negative pediatric patients, we examined COVID-19 and pre-existing OSA (dichotomized AHI≥1) using logistic (OR,95%CI) regression and as continuous measures: AHI, oxygen(SpO2) nadir, %time SpO2<90%) using linear regression(beta+/-SE). In those positive for SARS-CoV- 2(cases only), we assessed the association of OSA and World Health Organization(WHO) COVID-19 clinical outcome composite score (hospitalization, requiring supplemental oxygen, non-invasive ventilation/ high-flow oxygen, invasive ventilation/ECMO or death) using Wilcoxon rank sum test for ordinal data. Results: Cases (n=36) were 11.8±4.4 years, 61% male, 27.8% black and 88.9% with OSA, while 85.7% of controls (n=84) had OSA. OSA was not associated with increased SARS-CoV-2 positivity: OR=1.33(0.40, 4.45,p=0.64). No significant difference between cases and controls for mean AHI 3.7(1.5,6.0) vs 3.5(1.5,7.1),p=0.91,SpO2 nadir 88.6±5.4 vs 89.1±4.4,p=0.58,%time SpO2<90% 0.05[0.00,1.00) vs 0.10 (0.00,1.00, p=0.65) respectively was noted. WHO-7 COVID- 19 clinical outcome did not meet statistical significance in relation to OSA due to the low event frequency (p=0.49). Of note, those with OSA vs without OSA had a higher WHO-7 outcome score of 2 vs 0 and prevalence of hospitalization: 12.5 vs 0% respectively. Of hospitalized patients, the following was observed: 23% had moderate/severe OSA vs 4.3% mild OSA, 50% required supplemental oxygen and 25% required intubation/invasive ventilation. No deaths or readmissions were reported. High risk conditions included: 75% obesity, 50% asthma, 25% sickle cell disease and 25% hypoplastic left heart. Conclusion: In this first report of which we are aware focused on COVID-19 in pediatric OSA, we use a case control design leveraging COVID-19 and sleep laboratory registries. Albeit not statistically significant, pediatric patients with OSA had a higher percentage of worse clinical outcomes. Larger network studies are needed to clarify whether poorer COVID-19 outcomes may be attributable to OSA or modulated via high risk health conditions.

Evaluating the impact of the coronavirus(COVID-19) pandemic on treatment paradigms in headand neck cancer at a tertiary care hospital

Objective: The study aimed to assess the impact of the coronavirus (COVID-19) pandemic on head and neckoncologic care at a tertiary care hospital. The pandemic caused by the novel severe acute respiratory syndromecoronavirus (SARS-CoV-2) has led to policies designed to limit its spread. Policies including eliminating routineappointments, statewide stay-at-home orders, and restricting surgical cases to emergent surgeries have limitedaccess to care. Therefore, we hypothesized that treatment modifications would be implemented for patients andcancer consultations would decrease compared to historical data. Methods: Information regarding treatment modifications was collected prospectively during interdepartmental tumorconferences from March 18, 2020 to May 20, 2020. Information regarding patient demographics, tumorcharacteristics, and incidence of new cancer consultations was collected via chart review. Treatment modifications were categorized as follows: Elimination of Systemic Therapy, Treatment Delay, Change to Non-SurgicalManagement, or Alteration in Adjuvant Therapy. Rationales for modification were similarly grouped as follows:Operating Room Limitations, Medical Co-Morbidities, COVID-19 positive, Patient concern, or System limitations. Wedetermined the rate of treatment modifications and the frequencies of rationales and modification types. Demographic and tumor characteristics were compared between this population and a retrospectively collectedcohort from 2019. Results: 117 patients were presented during the review period in 2020. There were 69 patients presented acrossthe same time period in 2019. There were no differences in demographic characteristics between the groups. There was no difference between the tumor or nodal stages of the presented cases year over year. During the 2020 timeperiod there were more total case presentations and new cancer cases compared to the 2019 time period. Of the117 cases presented during the study period, there were 10/117 (8.4%) treatment modifications. The most commonreason for modification was limited PPE supply. The most common modification was treatment delay. The secondmost common modification was change from primary surgical management to nonsurgical management. Treatmentmodifications occurred most commonly early in the review period and declined subsequently. Conclusions: Despite the ongoing pandemic and resulting state and institutional restrictions, there was noappreciable reduction in new cancer consultations for head and neck cancer. There were a small number oftreatment modifications, particularly early in the course of our state and institutional response to the virus. However, over the 2-month period examined, patient care for these patients remained largely unaffected. While the restrictionin elective surgical care was implemented across the state of Maryland, the oncologic triage and emergency surgeryprioritization sustained the volume of oncologic practice.