ABSTRACT
Background: Acute kidney injury (AKI) is common in patients with COVID-19 and associated with poor outcomes. Urinary biomarkers have been associated with adverse kidney outcomes in other settings and may provide additional prognostic information in patients with COVID-19. Methods: We evaluated 19 urinary biomarkers of injury, inflammation, and repair in patients hospitalized with COVID-19 at 2 academic medical centers between April and June 2020. We associated biomarkers with a primary composite outcome of KDIGO stage 3 AKI, requirement for dialysis, or death within 60 days of admission. We also compared various kidney biomarker levels in the setting of COVID-19 versus other common AKI settings. Results: Out of 157 patients, 24 (15.3%) experienced the primary outcome. Twofold higher levels of neutrophil gelatinase-associated lipocalin (NGAL) (HR: 1.53;95% CI: 1.33-1.76), monocyte chemoattractant protein (MCP-1) (HR: 1.86;95% CI: 1.48-2.33), and kidney injury molecule-1 (KIM-1) (HR: 2.32;95% CI: 1.69-3.18) were associated with highest risk of the primary outcome. Higher epidermal growth factor (EGF) levels were associated with a lower risk of the primary outcome (HR 0.52;95% CI: 0.40-0.69). Individual biomarkers provided moderate discrimination and biomarker combinations improved discrimination for the primary outcome. Conclusions: Urinary biomarkers are associated with severe kidney complications in patients with COVID-19 and provide valuable information to monitor kidney disease recovery and progression.
ABSTRACT
Background: It is not clear how to best control for comorbidities when examining short-term mortality among individuals with COVID-19. The Charlson and Elixhauser Comorbidity Index were developed to predict 1-year and in-hospital mortality, respectively, and both indices can be operationalized using individual comorbidities or a weighted summary score. We compared the predictive accuracy for these comorbidity scores in predicting in-hospital death among adults hospitalized with COVID-19 from 5 hospitals comprising a health care system in the Mid-Atlantic United States. Methods: We used electronic health record data from adults hospitalized for COVID-19 from March 4-November 6, 2020. We ascertained comorbidities using all available lookback data from January 1, 2018 through COVID-19 hospital admission.We operationalized both comorbidity scores using individual comorbidities - 17 for Charlson and 29 for Elixhauser. We calculated weighted Charlson scores four ways, separately, using weights proposed by Deyo (1992), Schneeweiss (2003), Quan (2011) and Mehta (2016).We calculated the Elixhauser comorbidity score using weights proposed by van Walraven (2009) and Thompson (2015). We used logistic regression to compare the performance of different comorbidity scores in predicting in-hospital death. Nine models were constructed (1 baseline model that included age and sex, 1 for Charlson individual comorbidities, 4 for weighted Charlson scores, 1 for Elixhauser individual comorbidities and 2 for weighted Elixhauser scores). All models included age and sex as covariates.We evaluated the performance of each model using the c-statistic, and compared cstatistics using chi-square statistics, with a p-value < 0.05 considered significant model fit improvement. Secondarily, we compared model fit using Akaike Information Criteria (AIC), where lower values indicate better model fit.We used PROC LOGISTIC in SAS version 9.4. Results: Of 2,815 COVID-19 hospitalized patients, 12% (n=349) died in the hospital. Each comorbidity score performed significantly better (p < 0.001) than age and sex alone (c-statistic 0.775) at predicting COVID-19 related death. Overall, the ranking of the top 4 comorbidity scores were as follows: individual Elixhauser comorbidities (c-statistic 0.822) > Elixhauser-Thompson (c-statistic 0.803) > Elixhauser-van Walraven (c-statistic 0.796) = individual Charlson comorbidities (c-statistic 0.796).Weighted Elixhauser comorbidity scores (c-statistics ranging from 0.796 to 0.803) had significantly better performance than weighted Charlson comorbidity scores (c-statistics ranging from 0.786 to 0.790). Conclusions were similar when using AIC values to assess model fit. Conclusion: The individual comorbidities in the Elixhauser were the most accurate in predicting in-hospital death. If the weighted score needs to be used due to sample size limitations, we found that the Elixhauser-Thompson score was the most accurate in this training set. While statistically significant, the magnitude of predictive accuracy gained by adding covariates to the model for in-hospital mortality were small. Future research should investigate the utility of a customized COVID-19-specific comorbidity score in predicting mortality among adults hospitalized with COVID-19.