Mortality prediction model for community-acquired pneumonia patients based on machine learning

Machine Learning

Below is a summary of “”.A machine learning model for mortality prediction in community-acquired pneumonia patients: a study of development and validation,” was published in the January 2023 issue. chest To Sironis,other.

Machine learning (ML) and other forms of artificial intelligence are being explored as viable ways to improve the predictive power of existing clinical tools such as prognostic scores. However, more studies are needed evaluating the performance of his ML technique in improving the predictive ability of existing scores for community-acquired pneumonia (CAP). This study aimed to implement and validate a causal probability network (CPN) model for predicting mortality in CAP patients. This study included a retrospective derived validation analysis at his two hospitals affiliated with a Spanish university.

This study compared the predictive ability of the Sequential Organ Failure Assessment (SOFA), Pneumonia Severity Index (PSI), Brief Sequential Organ Failure Assessment (qSOFA), and CURB-65 criteria (confusion, urea, respiratory rate). , BP, age 65 years) into a CPN adapted to the CAP (SeF-ML) originally developed to predict sepsis mortality. It is important to note that the SeF model is proprietary. His DeLong method of correlating receiver operating characteristic curves was used to assess differences between curves.

There were 4,531 patients in the derived cohort and 1,034 patients in the validation cohort. The AUCs for 30-day mortality prediction using SeF-ML, CURB-65, SOFA, PSI and qSOFA were 0.801, 0.759, 0.671, 0.799 and 0.642 respectively in the derived cohorts. The SeF-AUC ML was 0.826 in the validation study, matching the AUC of the training data (0.801) (P. =.51). SeF-AUC ML was 0.764, significantly higher than CURB-65 (0.764, P. =.03) and (0.729, P. =.005). However, the PSI (0.830; P. =.92) or SOFA (0.771; P. =.14). Using structured health data, SeF-ML shows potential to enhance patient mortality prediction in her CAP. More studies using external validation are needed to enhance generalizability.


Source link

Leave a Reply

Your email address will not be published. Required fields are marked *