Cut of point classifier.

Background

Severe community-acquired pneumonia related treatment failure is persistence of features of severe pneumonia after initiation of antimicrobial therapy or a worsening clinical condition within 48–72 hours of the commencement of the antibiotics. Even though it is the most devastating public health problem in Ethiopia, there is no study to derivate and validate a model to predict treatment failure. To do this, nomogram was used to estimate the probability of treatment failure for each individual child and to classify their risk of treatment failure.

Objective

to develop and validate the model to predict treatment failure among under five children with severe community-acquired pneumonia in Debre Tabor comprehensive specialized hospital.

Method

A secondary analysis of the previously collected prospective follow-up study was used for further analysis among 590 under-5 children hospitalized with severe community-acquired pneumonia. The STATA version 17 software was used for analysis. Descriptive analysis was summarized by frequency and percentage. A multivariable binary logistic regression was also conducted, and the model performance was evaluated using the receiver operating characteristics curve with its area under the curve and calibration curve. Internal validation of the model was assessed using the bootstrap technique. The decision curve analysis was also used to evaluate the usefulness of the nomogram.

Results

The incidence of treatment failure among severe community-acquired pneumonia children was 28.1% (95% CI: 24.7%–30.8%). The previous history of severe community-acquired pneumonia, abnormal pulse rate, chest indrowing, anemia, HIV status, and plural effusion remained for the final model. The area under the curve for the original model and validated model was 0.7719 (95%CI: 0.729, 0.815) and 0.7714 (95% CI: 0.728–0.82), respectively. The decision curve analysis showed that the nomogram had a better net benefit across the threshold probability.

Conclusion

The incidence of treatment failure among children with severe community-acquired pneumonia was high in Debre Tabor comprehensive hospital. The previous history of severe community-acquired pneumonia, abnormal pulse rate, chest indrowing, anemia, HIV status, and plural effusion were the significant factors to develop the predictive model. The model had good discriminatory performance and internally valid. Similarly, the model has a good calibration ability with an insignificant loss of accuracy from the original. The models can have the potential to improve treatment outcomes in the clinical settings. But needs external validation before use.