Influence of Comorbidities on Treatment Costs and Development of a Cost Prediction Model for Pneumonia Patients in a Community Hospital

Abstract

Background and objectives: Pneumonia is a major cause of hospitalization, and comorbidities are an important component in determining medical costs under the diagnosis-Related Groups (DRGs) reimbursement system. However,  empirical  evidence remains  limited  regarding  the  structural  role  of  comorbidities  in  influencing  inpatient medical  costs  among  pneumonia  patients,  as  well  as  the  potential  for  developing medical  cost  prediction models in the context of community hospitals. This study aimed to analyze the factors influencing inpatient treatment costs among pneumonia patients and to compare prediction models for medical costs in the community hospital setting.

Methods: This retrospective analytical cross-sectional study used data from 874 hospitalized pneumonia patients obtained from the hospital information system of a community hospital in southern Thailand during 2020–2024. Associations  between variables  and  medical  costs  were  examined  using  multiple  linear  regression.  Three prediction models were developed and compared: Linear Regression (LR), Support Vector Machine (SVM), and Neural Network (NNET).

Results: The average treatment cost was 25,222 Thai baht. The multiple linear regression model explained approximately 50.0% of the variance in treatment costs (R² = 0.50). Chronic respiratory comorbidities, particularly chronic  obstructive  pulmonary  disease  and asthma,  were  significantly  associated  with  treatment  costs  as  modifiers of baseline cost levels. In contrast, bacterial or viral infection status was the principal driver of costs within a single episode of care. Among the three prediction approaches, the linear regression model demonstrated the best predictive performance and showed substantially greater consistency than both SVM and NNET.

Conclusions: Comorbidities exert a contextual influence on treatment costs among pneumonia patients by increasing the complexity of care rather than serving as the primary determinant of cost. Linear models appear to be the most appropriate approach for cost prediction in community hospitals. These findings may provide empirical evidence to support further development of the DRGs system in Thailand.