Prediction Models Of Mortality Risk Of Inpatients With COVID-19 Based On Clinical Variables

Part 1: Predicting mortality risk of COVID-19 patients based on neutrophil to lymphocyte ratio on admission Objective To analyze and identify the complete blood count(CBC)parameters that are most associated with the mortality of COVID-19 inpatients,and to develop a prediction model based on a single variable to predict the mortality risk of COVID-19 patients at admission.Methods We retrospectively analyzed 12,862 COVID-19 cases from 21 hospitals in Hubei Province and randomly assigned them to a training set cohort and a validation set cohort.The association between 10 CBC parameters and 60-day all-cause mortality was evaluated by Cox regression and least absolute shrinkage and selection operator(LASSO)Cox regression models in the training set.The CBC parameters that were most associated with mortality was selected.The Cox proportional hazard regression model of the selected parameter was constructed,and the prediction capacity of the parameter on mortality risk was evaluated by the area under the receiver operating characteristic curves(AUROC).The optimal cutoff point to distinguish the risk of death at admission was determined by the biggest Youden index.The prediction capacity of the selected parameter was also validated in the validation set.Results Cox regression and LASSO Cox regression analysis showed that,lymphocyte counts decrease and neutrophil counts increase were the two variables most associated with60-day all-cause mortality in the training set cohort.We further constructed the Cox regression model of the composite variable of these two parameters,i.e.neutrophil to lymphocyte ratio(NLR).NLR had an AUROC of 0.89(95% confidence interval [CI],0.87-0.91),which was higher than that of lymphocyte counts or neutrophil counts alone.Based on the biggest Youden index,NLR=6.11 was identified as the optimal cutoff point to distinguish the risk of death in COVID-19 patients.NLR also showed good predictive performance in the validation set,with an AUROC of 0.86(95% CI 0.84-0.88).Conclusions The present study indicates that NLR is a practical and economical parameter to predict the risk of death in COVID-19 patients at admission.Predicting the risk of COVID-19 mortality based on a single parameter would help first-line clinicians to quickly stratify risk for patients,thus optimizing the allocation of medical resources.Part 2: Development of a risk score using time series data of multiple complete blood count parameters during hospitalization to predict COVID-19 mortalityObjective To develop a sensitive risk score that can dynamically predict the risk of mortality in COVID-19 patients at different time points during hospitalization using repeated measures data of complete blood count(CBC).Methods This is a retrospective cohort study,which included a total of 13,138 COVID-19 inpatients in Hubei Province,China,and Milan,Italy.Among these patients,9,810 patients with ?2 CBC records from Hubei Province were assigned to the training cohort for the development of prediction model.CBC parameters were analyzed as potential risk factors for all-cause mortality of COVID-19 and were selected by the generalized linear mixed model(GLMM).Results Five risk factors were derived from GLMM,i.e.platelet counts,age,white blood cell counts,neutrophil counts,and NLR.A composite score(PAWNN score)was constructed based on the five risk factors using the Cox proportional hazard regression model.The PAWNN score showed good accuracy for predicting mortality in 10-fold cross-validation(AUROCs 0.92–0.93)and subsets with different quartile intervals of follow-up(AUROCs0.89-0.94)and preexisting diseases in the training set.The performance of the PAWNN score was further validated in 2,949 patients with only one CBC record from the Hubei cohort(AUROC 0.97)and 227 patients with only bseline CBC record from the Italian cohort(AUROC 0.80).The latent Markov model(LMM)identified three latent statues of COVID-19 and the PAWNN score was a distinguishing feature of these latent statues.The PAWNN score had good prediction power for transition probabilities between different latent statues.Conclusions The PAWNN score is a simple and accurate risk assessment tool that can predict the mortality for COVID-19 patients at different time points during their entire hospitalization.This tool can assist clinicians in prioritizing management of COVID-19 inpatients,especially for less developed areas with limited medical resources.Part 3: A risk score based on baseline risk factors for predicting mortality in COVID-19 patientsObjective To develop a sensitive and clinically applicable risk assessment tool identifying COVID-19 patients with a high risk of mortality at hospital admission.This model would assist frontline clinicians in optimizing medical treatment with limited resources.Methods 6,415 patients from seven hospitals in Wuhan city were assigned to the training and testing cohorts.A total of 6,351 patients from another three hospitals in Wuhan,2,169 patients from 11 hospitals outside of Wuhan,and 553 patients from Milan,Italy were assigned to three independent validation cohorts.A total of 64 candidate clinical variables at hospital admission were included in variable selection process and analyzed by random forest and LASSO analyses.Results Eight factors,namely,Oxygen saturation,blood Urea nitrogen,Respiratory rate,admission before the date when the national Maximum number of daily new cases was reached,Age,Procalcitonin,C-reactive protein(CRP),and absolute Neutrophil counts,were identified as having significant associations with mortality in COVID-19 patients.A composite score based on these eight risk factors,termed the OURMAPCN score,was developed.The OURMAPCN score predicted the risk of mortality among the COVID-19 patients in the trianing set with a C-statistic of 0.92(95% CI 0.90-0.93).The hazard ratio for all-cause mortality between patients with OURMAPCN score >11 compared with those with scores ?11 was 18.18(95% CI 13.93-23.71;P<0.0001).The predictive accuracy,specificity,and sensitivity of the score were validated in three independent cohorts.Conclusions The OURMAPCN score is a risk assessment tool to predict the mortality risk in COVID-19 patients based on a limited number of baseline risk factors.This tool can assist clinicians in optimizing the clinical management of COVID-19 patients with limited medical resources.