Clinically Applicable Approach for Predicting Future Mechanical Ventilation in Patients with COVID-19 Infection

Abstract

Background: COVID-19 patients requiring mechanical ventilation have a high mortality and resource utilization. The ability to predict which patients may require mechanical ventilation, prior to clinical decompensation, allows increased acuity of care and targeted interventions to potentially mitigate deterioration. Methods: We included COVID-19 positive hospitalized patients in this single-center retrospective observational study. Our primary outcome was mechanical ventilation or death within the subsequent 24-hours prediction window. As clinical decompensation is likely more recognizable, but less modifiable, as the prediction window shrinks, we also assessed 4-hour and 12-hour, and, for more notice, 48-hour prediction window. General classes of model features included demographic information, laboratory results, comorbidities, medication administration, and vital signs. We created a Random Forest model and assessed performance using 10-fold cross validation. The model was then compared to models derived from generalized estimating equations using discrimination.Findings: Ninety-three (23%) of the 398 COVID-19 positive patients meeting our inclusion criteria required mechanical ventilation or died within 14-days of admission. The Random Forest model predicts pending mechanical ventilation with good discrimination (c-statistic = 0.858, 95% confidence interval 0.841 to 0.874), which is comparable to the discrimination of the generalized estimating equation regression. Vitals sign data including SpO 2 /FiO 2 ratio (Z-score = 8.56), respiratory rate (5.97), and heart rate (5.87) had the highest predictive utility in our algorithm. In our highest risk cohort, the number of patients needed to identify (NNI) a single new case of mechanical ventilation was 3.2 and for our second risk quintile, NNI was 5.0. Interpretation: Machine learning techniques can be leveraged to improve the ability to predict which COVID-19 patients are likely to require mechanical ventilation, identifying unrecognized bellwethers and providing insight into the constellation of accompanying signs of respiratory failure in COVID-19.Funding : National Institutes of Health and the Foundation for Anesthesia Education and ResearchFunding: MRM reports grants from the National Institutes of Health (K01-HL141701). NJD reports a grant from the Foundation for Anesthesia Education and Research. All other author declare no funding. Declaration of Interest: CBD is paid consultant for Thrive Earlier Detection. He is also an inventor on various technologies unrelated to the work described in this manuscript. Some of the licenses are or will be associated with equity or royalty payments. The terms of all these arrangements are being managed by Johns Hopkins University in accordance with its conflict of interest policies. All other authors declare no competing interests.Ethical Approval Statement: For this retrospective observational study performed at our academic quaternary care center, we obtained Institutional Review Board approval (Ann Arbor, MI; HUM00052066). As no patient care interventions were made through conducting the study, patient consent was waived. This manuscript follows multidisciplinary guidelines for reporting machine learning predictive models in biomedical research.14 Study outcomes, data collection, and statistical analyses were established a priori and presented at a multidisciplinary peer-review forum on May 20, 2020 prior to data access.