Evolution of Clinical Phenotypes of COVID-19 Patients During Intensive Care Treatment: An Unsupervised Machine Learning Analysis (preprint)

Background: Identification of distinct clinical phenotypes in critically ill COVID-19 patients could improve understanding of the disease heterogeneity and enable prognostic and predictive enrichment facilitating more personalized treatment. However, previous attempts did not take into account temporal dynamics of the disease. By including the dimension of time, we aim to gain further insights into the heterogeneity of COVID-19.Methods: We used highly granular data from 3202 adult critically ill COVID patients in the multicenter Dutch Data Warehouse that were admitted to one of 25 Dutch ICUs between February 2020 and March 2021. Parameters including demographics, clinical observations, medications, laboratory values, vital signs, and data from life support devices were selected based on relevance and availability. Twenty-one consecutive datasets were created that each covered 24 hours of ICU data for each day of ICU treatment up until day 21. After aggregation and multiple imputation of the temporal data, clinical phenotypes in each dataset were identified by performing multiple cluster analyses. Clinical phenotypes were identified by aggregating values from all patients per cluster. Both evolution of the clinical phenotypes over time and patient allocation to these clusters over time were tracked.Results: The final patient cohort consisted of 2438 critically ill COVID-19 patients with a registered ICU mortality outcome. Forty-one parameters were chosen for the cluster analysis. On admission, both a mild and a more severe clinical phenotype were found. After day 4, the severe phenotype split into an intermediate and a severe phenotype for 11 consecutive days. Heterogeneity between phenotypes appears to be strongly driven by inflammation and dead space ventilation. During the 21-day period only 8.2% and 4.6% of patients in the initial mild and severe clusters remained assigned to the same phenotype respectively. The clinical phenotype half-life was between 5 and 6 days for the mild and severe phenotypes, and about 3 days for the medium severe phenotype.Conclusions: Patients typically do not remain in the same cluster throughout intensive care treatment. This may have important implications for prognostic or predictive enrichment. Prominent dissimilarities between clinical phenotypes are predominantly driven by inflammation and dead space ventilation.

Joint Modeling of Repeated Measurements of Different Biomarkers Predicts Mortality in COVID-19 Patients in the Intensive Care Unit. (preprint)

Background: Predicting disease severity is important for treatment decisions in patients with COVID-19 in the intensive care unit (ICU). Different biomarkers have been investigated in COVID-19 as predictor of mortality, including C-reactive protein (CRP), procalcitonin (PCT), interleukin-6 (IL-6) and soluble urokinase-type plasminogen activator receptor (suPAR). Using repeated measurements in a prediction model may result in a more accurate risk prediction than the use of single point measurements. The goal of this study is to investigate the predictive value of trends in repeated measurements of CRP, PCT, IL-6 and suPAR on mortality in patients admitted to the ICU with COVID-19. Methods: This was a retrospective single center cohort study. Patients were included if they tested positive on SARS-CoV-2 by PCR test and if IL-6, PCT, suPAR was measured during any of the ICU admission days. There were no exclusion criteria for this study. We used joint models to predict ICU-mortality. This analysis was done using the framework of joint models for longitudinal and survival data. The reported hazard ratios express the relative change in the risk of death resulting from a doubling or 20% increase of the biomarker’s value in a day compared to no change in the same period. Results: A total of 107 patients were included, of which 26 died during ICU admission. Adjusted for sex and age, a doubling in the next day in either levels of PCT, IL-6 and suPAR was significantly predictive of in-hospital mortality with and an HR of 1.523 (1.012 – 6.540), 75.25 (1.116 – 6247) and 24.45 (1.696 – 1057) respectively. With a 20% increase in biomarker value in a subsequent day, the HR of PCT, IL-6 and suPAR were 1.117 (1.03 – 1.639), 3.116 (1.029 – 9.963) and 2.319 (1.149 – 6.243) respectively. Conclusion: Joint models for the analysis of repeated measurements of PCT, suPAR and IL-6 are a useful method for predicting mortality in COVID-19 patients in the ICU. Patients with an increasing trend of biomarker levels in consecutive days are at increased risk for mortality.

A pragmatic approach to estimating average treatment effects from EHR data: the effect of prone positioning on mechanically ventilated COVID-19 patients (preprint)

Despite the recent progress in the field of causal inference, to date there is no agreed upon methodology to glean treatment effect estimation from observational data. The consequence on clinical practice is that, when lacking results from a randomized trial, medical personnel is left without guidance on what seems to be effective in a real-world scenario. This article proposes a pragmatic methodology to obtain preliminary but robust estimation of treatment effect from observational studies, to provide front-line clinicians with a degree of confidence in their treatment strategy. Our study design is applied to an open problem, the estimation of treatment effect of the proning maneuver on COVID-19 Intensive Care patients.