Think sepsis, write sepsis, code sepsis - patient characteristics associated with sepsis (under-)coding in administrative health data.

PURPOSE: Sepsis is a leading cause of morbidity and mortality, yet its documentation and coding in administrative health data remain unreliable. Accurate coding is essential for epidemiological surveillance, quality assurance, and reimbursement. This study aims to identify patient characteristics associated with under-diagnosis and under-coding of sepsis in German inpatient administrative health data (IAHD).

METHODS: This secondary analysis of the multicenter OPTIMISE study included 10,334 hospital cases from ten German hospitals (2015-2017). Sepsis cases were identified via structured chart review and compared to ICD-coded diagnoses. Logistic regression and classification tree analyses were used to determine predictors of under-diagnosis and under-coding, including ICU admission, organ dysfunction, and infection source.

RESULTS: Among 1,310 cases fulfilling severe sepsis-1 criteria, only 30.7% were correctly coded. The strongest predictor for coding accuracy was explicit mention of sepsis in the medical chart (OR 19.58). ICU treatment, organ dysfunction severity, and mechanical ventilation were also associated with higher coding rates, while pneumonia as the infection source was linked to a lower probability of sepsis being named and coded.

CONCLUSION: Sepsis coding in administrative data is frequently inaccurate. Explicit naming of sepsis and severity markers strongly influence correct coding. As Germany introduces mandatory sepsis quality assurance in 2026, targeted interventions - including enhanced clinician documentation and electronic coding support - are essential to improve coding reliability and patient care.