Risk factors for adverse outcomes during mechanical ventilation of 1152 COVID-19 patients: a multicenter machine learning study with highly granular data from the Dutch Data Warehouse

Background The identification of risk factors for adverse outcomes and prolonged intensive care unit (ICU) stay in COVID-19 patients is essential for prognostication, determining treatment intensity, and resource allocation. Previous studies have determined risk factors on admission only, and included a limited number of predictors. Therefore, using data from the highly granular and multicenter Dutch Data Warehouse, we developed machine learning models to identify risk factors for ICU mortality, ventilator-free days and ICU-free days during the course of invasive mechanical ventilation (IMV) in COVID-19 patients. Methods The DDW is a growing electronic health record database of critically ill COVID-19 patients in the Netherlands. All adult ICU patients on IMV were eligible for inclusion. Transfers, patients admitted for less than 24 h, and patients still admitted at time of data extraction were excluded. Predictors were selected based on the literature, and included medication dosage and fluid balance. Multiple algorithms were trained and validated on up to three sets of observations per patient on day 1, 7, and 14 using fivefold nested cross-validation, keeping observations from an individual patient in the same split. Results A total of 1152 patients were included in the model. XGBoost models performed best for all outcomes and were used to calculate predictor importance. Using Shapley additive explanations (SHAP), age was the most important demographic risk factor for the outcomes upon start of IMV and throughout its course. The relative probability of death across age values is visualized in Partial Dependence Plots (PDPs), with an increase starting at 54 years. Besides age, acidaemia, low P/F-ratios and high driving pressures demonstrated a higher probability of death. The PDP for driving pressure showed a relative probability increase starting at 12 cmH2O. Conclusion Age is the most important demographic risk factor of ICU mortality, ICU-free days and ventilator-free days throughout the course of invasive mechanical ventilation in critically ill COVID-19 patients. pH, P/F ratio, and driving pressure should be monitored closely over the course of mechanical ventilation as risk factors predictive of these outcomes.

Abstract 15941: Derivation of Risk-Scores That Stratify Risk of Hospitalization and Death in SARS-CoV-2 Positive Patients

Introduction: It is well-documented that COVID-19 patients with pre-existing cardiovascular-related disorders are at higher risk of a complicated course. It would be valuable to integrate individual risk factors into overall risk scores for hospitalization and death from COVID-19. Methods: The Intermountain Healthcare medical record database was searched for all individuals tested for SARS-CoV-2 infection up to June 8, 2020. Data from test-positive patients (pts) was analyzed to determine the characteristics of pts requiring hospitalization. From these data, 2 risk scores for hospitalization were derived using multi-variable modeling: of only demographic and risk-factor data, or also including concurrent medications. The risk scores were also applied to predict the risk of dying from COVID-19. Results: Of 104,018 people tested at Intermountain Healthcare for SARS-CoV-2, 5505 (5.3%) were positive. Of test-positive pts, 451 (8.2%) were hospitalized, and 37 (0.7%) died. Using a demographic/risk factor only score, 1.4, 7.0, and 36.6% of low-, moderate-, and high-risk groups, respectively, were hospitalized (AUC=0.826). Using demographic risk-factors and medications, 1.4, 5.6, and 40.3% of low-, moderate-, and high-risk patients were hospitalized (AUC=0.854, Table 1). The demographic/risk factor-score was also predictive of the risk of dying, with 0%, 0.9% and 4.5% in low-, moderate-, and high-risk groups dying (AUC=0.918). Adding medications to the risk-factors model further improved the prediction of death with 0.1, 0.04, and 4.9% in the low-, moderate-, and high-risk groups dying (AUC=0.942, Table 2). Conclusions: We demonstrate the derivation of highly predictive risk scores for COVD-19 patients at low, moderate, and high risks of hospitalization or death. Pending appropriate validation in another cohort, application of these risk-scores may allow healthcare systems to risk-stratify COVID-19 patients requiring variable intensity of care.

Abstract TP238: Encephalopathy Only Stroke Codes (EoSC) Do Not Result in rt-PA Treatments

Background: Isolated mental status changes as presenting sign (EoSC+), are not uncommon stroke code triggers. As stroke alerts, they still require the same intensive resources be applied. We previously showed that EoSC+ strokes (EoSC+CVA+) account for 8-9% of EoSC+ codes but only 0.1-0.2% of all codes. Whether these result in thrombolytic treatment (rt-PA), and the characteristics/ risk factor profiles of EoSC+CVA+ patients, have not been reported. Methods: Retrospective analysis of stroke codes from an IRB approved registry, from 2004 to 2018, was performed. EoSC+ definition used was consistent with prior publications (NIHSS>0 for Q1a, 1b, or 1c with remaining elements scored 0). Other definitions were also assessed. Characteristics and risk factors were compared for EoSC+, EoSC+CVA+, and rt-PA (EoSC+ CVA+TPA+) patients. Results: EoSC+ occurred in 59/2982 (1.98%) of all stroke codes. EoSC+CVA+ occurred in 8/59 (13.56%) of EoSC+ codes and 8/2982 (0.27%) of all stroke codes. 6/8 (75%) of EoSC+CVA+ scored NIHSS=1. Hispanic ethnicity (p=0.009), HTN (p=0.02), and history of stroke/TIA (p=0.002) were less common in EoSC+. No demographic/ risk factor differences were noted for [EoSC+CVA+ vs. EoSC+CVA-]. No cases of rt-PA eligibility/ treatment were noted. In EoSC+CVA+ analysis, imaging positive stroke/intracranial hemorrhage was noted on only 3 cases (3/2982=0.10% of all stroke codes) and none were posterior stroke. Conclusions: EoSC+ is not an uncommon reason to activate stroke codes, but rarely results in stroke/TIA (0.27%) or stroke (0.10%), and in our analysis never (0%) resulted in rt-PA. Sub-analysis did not show missed rt-PA or posterior strokes. This adds information for application of limited acute stroke code resources. Though stroke codes must still to be activated, understanding characteristics, and knowing that EoSC+CVA+ patients are unlikely to receive rt-PA, may help triage stroke resources. Further investigation is warranted.