Development and validation of a prediction model for tuberculous peritoneal effusion

The study aimed to develop and validate a nomogram for diagnosing tuberculous peritoneal effusion (TPE). A total of 351 patients were included, with 245 in the training group and 106 in the validation group. Seven independent predictors were identified: age, fever, ascites adenosine deaminase, ascites carcinoembryonic antigen, ascites total protein, serum carcinoembryonic antigen, and serum creatinine. A diagnostic nomogram was developed based on these seven variables. The area under the curve (AUC) of the nomogram was 0.974 for the training group and 0.955 for the validation group. Calibration curves demonstrated good agreement, and decision curve analysis showed the model provides the greatest net benefit within the threshold probability range. The nomogram demonstrated superior diagnostic performance in distinguishing TPE from non-TPE cases compared to any single indicator.