Diagnosis-code co-occurrence patterns and their associations with hospitalization costs among patients with lung cancer: a province-wide retrospective study in Sichuan, China

Background Hospitalized patients with lung cancer often have complex diagnosis records, yet conventional comorbidity measures provide limited information on which specific diagnosis-code combinations are associated with differences in hospitalization costs. We examined diagnosis-code co-occurrence patterns and their associations with hospitalization costs in patients with lung cancer. Methods We analyzed de-identified records from the province-wide hospital discharge abstract database of Sichuan Province, China, from January 2016 to December 2024. Diagnoses were standardized to three-character categories of the International Classification of Diseases, 10th Revision (ICD-10). Elastic-net regression was used to select core diagnosis-code pairs associated with hospitalization cost based on each patient’s latest eligible admission. A weighted co-occurrence network was then constructed from the selected pairs. Their adjusted cost associations were estimated across all eligible admissions using patient-clustered generalized estimating equations. Pairs with significant positive associations were further classified according to patient-level co-occurrence frequency and the magnitude of their cost associations. Results The study included 133,977 patients and 199,740 hospitalizations. Elastic-net regression selected 88 core diagnosis-code pairs, which formed a network of 49 nodes and five communities. After false discovery rate correction, 69 pairs were significantly associated with hospitalization costs, including 59 positive and 10 negative associations. Adjusted cost ratios for positively associated pairs ranged from 1.022 to 1.659. Of the 59 positively associated pairs, 9 were classified in the higher-frequency/larger-cost-association region, 21 in the lower-frequency/larger-cost-association region, 21 in the higher-frequency/smaller-cost-association region, and 8 in the lower-frequency/smaller-cost-association region. Conclusion Diagnosis-code pairs differed substantially in both how frequently they occurred and their associations with hospitalization costs. Joint consideration of these two dimensions identified recurrent patterns with distinct frequency–cost profiles. These findings show how routinely collected discharge data can be used to distinguish cost-relevant patterns of diagnostic complexity in hospitalized patients with lung cancer.