Analytical Approaches to Quantify Illness Severity

Continuous ECG data from bedside monitors, vital signs, laboratory values, and clinical assessment findings in the electronic health record can be analyzed in real-time to identify patients at rising risk of deterioration. This continuous analytic monitoring has been employed to predict future clinical deterioration. In this dissertation, we investigated how the risk scores from continuous prediction algorithms can be used as a proxy for illness severity.
Using risk scores as a proxy for illness severity, we characterized illness trajectories over time. Specifically, we examined the illness trajectory of PICU patients immediately following a sepsis diagnosis using Markov chain modeling. We analyzed 18,666 illness state transitions over 157 pediatric intensive care unit admissions in the three days following cultures for suspected sepsis. We found the population-based transition matrix based on the sepsis illness severity scores in the hours following a sepsis diagnosis can describe a sepsis illness trajectory. We used Shannon entropy to quantify the differences in transition matrices stratified by clinical characteristics. We found a different structure of dynamic transitions based on ventilator use but not age group. Stochastic modeling of transitions in sepsis illness severity scores can be useful in describing the variation in transitions made by patient and clinical characteristics.
We also proposed a method to characterize individual illness dynamics in patients who experienced sepsis in the pediatric intensive care unit. We then examined the association between individual entropy scores and a composite variable of negative outcomes. In a cohort of 164 intensive care unit admissions where at least one sepsis event occurred, for each admission, we calculated transition probabilities to characterize movement among illness states. We calculated Shannon entropy based on these transition probabilities. We considered entropy as a measure of illness state dynamics. Using hierarchical clustering based on entropy, we identified high- and low-risk phenotypes. Compared to the low-risk phenotype, the high-risk phenotype was defined by the highest entropy values and had the most ill patients as defined by negative outcomes and multiple sepsis events. We found that characterizing illness dynamics using a measure of entropy offers additional information in conjunction with static assessments of illness severity. A stochastic approach to characterizing the entropy of an illness trajectory provides a novel way of assessing the complexity of a course of illness.
To summarize, we found that continuous analytic monitoring can be used in ways that extend beyond early warning of clinical deterioration. Risk scores produced by well-calibrated predictive models can be used as physiological markers of illness severity. We can model the course of illness states through which patients progress following sepsis diagnosis using risk scores as measures of illness severity. Using a measure of entropy to quantify illness dynamics over the critical illness course may add additional information to our understanding of the illness trajectory. Understanding the dynamics associated with a child’s critical illness trajectory provides a novel analytic lens with the potential for further clinical and research applications.