SENSITIVITY AND SPECIFICITY AFFECTED BY COVARIATES (FACTORS) - ONE WAY TO ANALYZE
SENSITIVITY AND SPECIFICITY AFFECTED BY COVARIATES (FACTORS) - ONE WAY TO ANALYZE

Summary/Abstract: Taking into account the high frequency of elderly patients affected by colorectal cancer and the increased prevalence of comorbidities, there have been created different classification systems for the prediction of postoperative results which depends on the number and type of comorbidities. In order to be useful in clinical practice, a system has to be easily useable, to allow comparisons amongst patients, and to offer a good estimation of risk based on the preoperative score, for a correct evaluation of patients who will probably best benefit from the surgery. The Charlson comorbidity index (CCI) is widely used in the surgical fields. It comprises of a weighted scoring system based on 19 comorbidities. The absolute obtained number can be used for comparison and evaluation of the risk. The Elixhauser method is a more recent model which utilises the risk adjustment using 31 conditions. The comorbidities' evaluation index for adults (ACE-27) was developed by modifying the Kaplan-Feinstein comorbidity index (KFI). The modifications have been made following discussions with experts regarding clinical aspects and a review of literature. Only a few studies had compared the results about comorbidities index in estimating mortality after a colorectal cancer surgery. Recent studies compared Charlson comorbidity index and the Elixhauser method or ACE-27. Their aim was to investigate each comorbidity's impact and their prognostic ability in colorectal cancer regarding the survival causes. This study compares the punctual obtained estimations in accordance with the number of comorbidities and the adjusted indexes for the well-known preoperative risk factors, for a complete evaluation of the preoperative risk. We have evaluated the prediction capacity by ROC analysis and choose the cutoff for maximum accuracy. The results were not acceptable and the need of a proper score is mandatory. More than this by logistic regression (multiple regression) we found that sensitivity and specificity depends on other covariate like age and TNM stage. Therefore we have improved the prediction of death by a score that take into account beside Charlson score other covariates.

• Details
• Contents