BACKGROUND: The quality of registry data is important for epidemiological research. The Swedish inpatient registry (IPR) is a national database with mandatory registration of all hospitalisations since 1987, and since 2007, the medical procedure codes which can be used for grading severity of infectious diseases. However, the completeness of procedure code registration has rarely been studied.

OBJECTIVES: To determine the quality and completeness of procedure codes for ICU admission, mechanical ventilation and extra-corporeal membrane oxygenation (ECMO) in the Swedish IPR utilising the Swedish Intensive Care Registry (SIR) as the gold standard. DESIGN A Swedish nationwide observational study.

SETTING: Covid-19 patients in Sweden who required intensive care in Sweden between March 2020 and August 2022. PATIENTS Covid-19 patients with a laboratory-verified SARS-CoV-2 infection who required ICU admission (n=8992), mechanical ventilation (n=5262) or ECMO (n=29).

MAIN OUTCOME MEASURES: The sensitivity and/or positive predictive values of procedure code registration for ICU, mechanical ventilation, ECMO and Covid-19 diagnosis code registration in the IPR were evaluated using SIR as the reference. Factors associated with low reporting were explored and the dates of ICU admission registration compared between IPR and SIR.

RESULTS: For Covid-19 patients registered in SIR as needing intensive care, mechanical ventilation or ECMO, the completeness of procedure codes in the IPR was 39.7, 78.2 and 100%, respectively. Of the 39.7% with an ICU code in the IPR, the ICU date in the IPR corresponding to the actual ICU admission date was 52.3%. The completeness of ICU registration in the IPR varied from 0.6 to 96.9% between healthcare regions

CONCLUSIONS: Procedure codes for intensive care in the Swedish IPR showed low sensitivity and varied greatly between healthcare regions. This negatively influences their usability for epidemiological research and calls for updated guidelines on coding.