Abstract
BACKGROUND AND AIMS: Spontaneous bacterial peritonitis (SBP) represents a significant complication in the decompensated phase of cirrhosis. The challenges in treating SBP and the associated mortality rates are markedly elevated in elderly individuals. Timely detection and intervention for SBP are imperative. We aimed to develop a predictive tool for the occurrence of SBP in elderly individuals with decompensated cirrhosis (DC). METHODS: Elderly patients diagnosed with DC were enrolled from Chengdu Fifth People's Hospital in China, spanning from January 1, 2015, to September 31, 2023. Among the patients, 337 were assigned to the training cohort, while 145 were designated to the validation cohort. A multivariate logistic regression analysis was performed to identify significant predictors and to develop a nomogram for predicting the occurrence of SBP. To evaluate the model's discrimination and calibration, a bootstrap method with 1000 resamples was utilized. RESULTS: Findings from the multivariate logistic regression analysis indicated that constipation (odds ratio [OR] 2.09, 95% confidence interval [CI] 1.25-3.49, P=0.005), ascites (OR 2.84, 95% CI 1.64-4.92, P<0.001), Child-Pugh-Turcotte (CPT) score (OR 4.80, 95% CI 1.69-13.60, P=0.003), and high sensitivity C-reactive protein (hs-CRP) (OR 2.96, 95% CI 1.54-5.45, P=0.001) were significant independent predictors for the occurrence of SBP in elderly individuals with DC. The generated nomogram showed an area under the curve of 0.779 for the training cohort and 0.817 for the validation cohort. The nomogram's calibration curve nearly matched the perfect diagonal line, and decision curve analysis showed an improved net benefit for the model. Subsequent validation further corroborated the reliability of the predictive nomogram. CONCLUSION: In conclusion, the nomogram, incorporating variables such as constipation, ascites, CPT score, and hs-CRP, effectively predicted the occurrence of SBP in elderly patients with DC, underscoring its substantial clinical applicability.