Administrative coding for alpha-1 antitrypsin deficiency including the pi*ZZ phenotype is accurate in Sweden

We aimed to validate the International Classification of Diseases (ICD)-10 codes for alpha-1 antitrypsin deficiency (AATD) (E880A = asymptomatic AATD; E880B = symptomatic AATD) in a large hospital reporting to the Swedish National Patient Register and to ascertain their relation to the protease inhibitor (Pi)*ZZ-phenotype. We randomly selected 150 adults and 50 children who visited Karolinska University Hospital (Stockholm, Sweden) between 2014 and 2024 with coding for E880A or E880B (1:1). Positive predictive values (PPVs) of AATD ICD-10 codes were calculated for correctly assigned codes and the Pi*ZZ-phenotype using medical charts as gold standard. Information on smoking status, lung disease, liver disease, and living area, were also retrieved. The PPV of AATD ICD-10 codes (E880A + E880B) in adults was 99% (95%CI = 95–100%; n = 148/150). The PPV for the Pi*ZZ-phenotype was only 59% (95%CI = 50–67; n = 83/141) but increased to 79% (95%CI = 67–88%; n = 50/63) when only considering outpatients with E880B coding. Of adult participants, 13% had liver disease, 51% had lung disease, and 50% were ever-smokers. In children, the PPV of E880A + E880B was 100% (95%CI = 91–100%; n = 50/50) for any AATD diagnosis and was 88% for the Pi*ZZ-phenotype (95%CI = 75–95%; n = 43/49). Liver or lung disease occurred in 6% of children. Results were consistent across several sensitivity analyses. In a tertiary care setting, the validity of ICD-10 codes for AATD is excellent. The PPV of these codes for delineating the Pi*ZZ-phenotype is high in children but requires an algorithm in adults with coding for E880B in outpatients.