A UCLA research team has created the Comorbid Operative Risk Evaluation (CORE) score to better take into account the role chronic diseases play in a patient’s risk of mortality after surgery, allowing surgeons to adapt to patients’ pre-existing conditions and more easily determine mortality risk.
For almost 40 years, researchers have used two tools, the Charlson Comorbidity Index (CCI) and the Elixhauser Comorbidity Index (ECI), to measure the impact of existing health conditions on patient outcomes. These tools use ICD codes entered by medical professionals and billers to account for the patient’s illness. However, these tools were not designed for patients undergoing surgery and often address chronic diseases that are not relevant to surgical populations. They often capture data from medical billing records and lack nuanced information about pre-existing health conditions.
A total of 699,155 patients were used to develop the model, of which 139,831 (20%) formed the test cohort. Researchers questioned adults undergoing 62 operations in 14 specialties from the 2019 National Inpatient Sample (NIS) using International Classification of Diseases, Tenth Revision (ICD-10) codes. They classified ICD-10 codes for chronic diseases into Refined Clinical Classifications Software (CCSR) groups. They used logistic regression on CCSR with non-zero feature importance on four machine learning algorithms predicting in-hospital mortality and used the outcome.
coefficients to calculate the Comorbid Operational Risk Evaluation (CORE) score based on a previously validated methodology. The final score ranges from zero, which represents the lowest risk, to 100, which represents the highest risk.
Health services and outcomes research using retrospective databases continues to represent an increasing proportion of surgical research. Researchers who highlight quality problems and disparities mean well. However, without the right tools, it may be unclear whether poor outcomes are independent of preexisting conditions.
“The advent of new software and statistical methodologies has allowed researchers to exploit large databases to answer questions about health care quality, disparities and outcomes,” said Dr. Nikhil Chervu, resident physician in the Department of UCLA Surgery and lead author of the study. “However, these databases often capture data from medical billing records and lack nuanced information about pre-existing health conditions. Without addressing differences in patients’ chronic illnesses, population comparisons can fail. Incorporating this score in additional research will further validate its use and help improve the analysis of surgical outcomes using large databases.”