Validation of Hemolytic Anemia Codes in the French Hospital Database

Introduction: Hemolytic anemia is a group of rare diseases. National databases are useful data sources to assess the epidemiology and the management of such diseases. The French national health database covers the entire French population (66 million inhabitants). It links sociodemographic data, out-hospital data including dispensing data and hospital data. As a consequence, this database is very useful for population-based studies on rare diseases. For instance, it has been used for epidemiological and pharmacoepidemiological studies on immune thrombocytopenia. Diseases are identified using the national hospital database (named PMSI) that contains discharge diagnoses for all hospital stays in all private and public hospitals in France. This study was aimed at assessing the positive predictive value (PPV) of hemolytic anemia diagnoses in the French hospital database.

Methods: In the PMSI database, every hospital stay contains one primary diagnosis, and possibly one related and several associated diagnoses. They are coded using the International Classification of Diseases, 10^th revision (ICD-10) by the physician in charge of the patient or trained nurses from the medical chart. We selected all hospital stays at Toulouse University Hospital, South of France (2860 beds) with a diagnosis of hemolytic anemia (D55.0-D59.9 ICD-10 codes) between January 2017 and December 2017. Medical charts and biological data were reviewed. Hemolytic anemia was defined by anemia with high reticulocyte count (>150 x 10⁹/L), plus at least two of the three following signs of hemolysis: low serum haptoglobin level, hyperbilirubinemia and elevated serum lactate dehydrogenase level. PPVs and their 95% confidence intervals (CI) were calculated by categories of hemolytic anemia: autoimmune hemolytic anemia (AIHA; D59.1 ICD-10 code), enzyme deficiency (D55.0-D55.9 ICD-10 codes), hereditary spherocytosis (D58.0 ICD-10 code) and hemoglobinopathy (D56.0-D57.9 and D58.2 ICD-10 codes). AIHA was defined by positive direct antiglobulin test (DAT); enzyme deficiency was defined by low enzyme dosage and negative DAT; hereditary spherocytosis was defined by negative DAT and either positive flow cytometry osmotic fragility test either positive eosin-5'-maleimide binding test; hemoglobinopathy was defined by compatible blood count and positive hemoglobin electrophoresis.

Results: During the study period, 54 patients had at least one hospital stay with a discharge diagnosis of AIHA, 12 with enzyme deficiency, 10 with hereditary spherocytosis, 92 with thalassemia and 285 with sickle cell disease. We further excluded 13 patients due to missing data, precluding disease classification. AIHA was confirmed in 49/53 patients; the PPV was 92.5% (95% CI: 85.3%-99.6%). Enzyme deficiency was confirmed in 8/12 patients (including G6PD deficiency: 7/9) and hereditary spherocytosis in 10/10 patients. Thalassemia was confirmed in 72/83 patients; the PPV was 86.7% (95% CI: 79.5%-94.0%); however, the code of thalassemia type was not adequately coded in most cases (PPV<30%). Sickle cell disease was confirmed in 279/284 patients; the PPV was 98.2% (95% CI: 96.7%-99.8%).

Conclusions: Overall, this study confirms high PPV values for hemolytic anemia discharge diagnoses recorded in the French hospital database, allowing epidemiological studies using this source of data.