Risk Prediction Model Forecasts Intensive Care Unit Admission, Survival

Researchers have developed a risk score to predict intensive care unit (ICU) admission and ICU survival among community-dwelling older adults, according to a study published in Health Science Reports.

The research team noted that existing predictive models to identify risk of critical illness typically rely on clinical data from emergency departments to assist in patient triage upon hospital admission. However, these models often cannot flag outpatients who may be admitted to the ICU with a critical illness in the future.

This presents a significant challenge as aging populations increasingly utilize critical care services, facilitating the need for models to predict future ICU admission, survivorship, and mortality. The research team indicated that many middle-age patients are equally at risk for poor health outcomes as their older counterparts, making a population-based approach to risk favorable.

“Our Risk Prediction Score tool is designed to be used by health systems and researchers so they can engage with adults in a certain population – perhaps those with certain specific health issues or those living in a specific geographic area – who may be at higher risk of ICU admission and higher or lower risk of ICU survival,” explained Sikandar Khan, DO, MS, an assistant professor of Medicine in the Division of Pulmonary and Critical Care Medicine at Indiana University and medical director of the Indiana University Health Intensive Care Unit (ICU) Survivor Center, in a press release.

“The good news is that if researchers are able to identify populations likely to become future ICU patients using the Prediction Risk Score, they may be able to enroll patients in these populations in studies earlier, and health systems may be able to develop new programs and new models of care for at-risk populations to improve the outcomes of individual patients in and after the ICU,” Khan continued.

The model was trained on data from 48,127 patients who were 50 years and older with at least one primary care visit between January 1, 2017, and December 31, 2017. Using electronic health record data, the research team identified multiple variables that impact ICU survivorship.

Patients in the cohort were divided into three groups – “alive without ICU admission,” “ICU survivors,” and “death” – using ICU admission and mortality within two years following an initial primary care visit. During the study period, 92.2 percent of patients were alive without ICU admission, 6.2 percent were admitted to the ICU at least once and survived, and 1.6 percent died.

The highest risk score weights for ICU mortality were greater deciles of age over 50 years; diagnoses of chronic obstructive pulmonary disorder (COPD) or chronic heart failure; and laboratory abnormalities in alkaline phosphatase, hematocrit, and albumin.

Overall, the model achieved high performance, discriminating between patients who died versus those who remained alive without ICU admission with an area under the receiver operating characteristics (AUC) curve of 0.858, and between ICU survivors versus those alive without ICU admission with an AUC of 0.765.

While these results are promising, the research team cautioned that more research is needed before the model can be deployed in clinical settings.

“Currently we have imprecise tools to identify which groups of patients will become severely ill. This study presents the first step in creating infrastructure for further research by us and others to identify and follow cohorts of patients who may become critically ill and ultimately to improve their outcomes,” Khan stated.