Factors and clinical prediction score for complication development after cellulitis diagnosis in adult patients

Abstract

Background Cellulitis is defined as a bacterial infection of the skin and subcutaneous tissue that can cause multiple complications, such as sepsis and necrotizing fasciitis. In extreme cases, it may lead to multiorgan failure and death. We sought to analyze the clinical factors that contribute to the development of complicated disease, including demographics, clinical presentation, initial vital signs, and laboratory studies. Methods Our study is a retrospective cohort study carried out in a university-based tertiary care hospital in Bangkok, Thailand. Adult patients who presented with cellulitis from January 1, 2018, to December 31, 2022, were evaluated for eligibility and inclusion in this study. All related variables for both outcomes, bacteremia and necrotizing fasciitis, were gathered from electronic medical records and analyzed using multivariable logistic regression analysis. Results Of the 1,560 visits to this hospital, 47 cases reported at least one complication, with bacteremia noted in 27 visits (1.73%) and necrotizing fasciitis in 20 visits (1.27%). From the multivariable logistic regression analysis, six variables emerge as predictors of cellulitis complications. These are: Age ≥ 65 years, Body Mass Index ≥ 30 kg/m2, diabetes mellitus, body temperature ≥ 38 °C, systolic blood pressure ≤ 100 mmHg, and involvement of lower extremities. The predictive score was developed from these factors and was named the Ramathibodi Necrotizing Fasciitis/Bacteremia (RAMA-NFB) Prediction Score. Our predictive score has an accuracy of 82.93% (95% CI, 0.77–0.89). Patients in the high-risk group (RAMA NFB score > 6) have a likelihood ratio of 8.75 (95% CI, 4.41–18.12; p < 0.001) times to develop complications of cellulitis. Conclusion In our study, the RAMA-NFB Prediction Score predicts complications of necrotizing fasciitis and bacteremia in adult patients who present with cellulitis. External validation of this predictive score is still needed for further practical application.