Reliability of recurrent pregnancy loss diagnosis coding in the Swedish National Patient Register: a validation study

Background: The Swedish National Patient Registry (NPR) is a nationwide registry that is used extensively for epidemiological research. Using the NPR, we recently found a recurrent pregnancy loss (RPL) incidence of 650/100,000 (0.65%) pregnant women in Sweden. It is of great importance that the quality of the coding is good and reliable in order to use NPR data for research. To specifically study RPL in Sweden, a general validation of this diagnosis in the NPR is needed. Objective: To validate the diagnosis of RPL, defined as ≥3 consecutive miscarriages before 22 gestational weeks, recorded in the NPR and assess how registered miscarriages were verified clinically (ultrasound or urine/serum hCG) by reviewing the medical records. Materials and methods: In a cohort of 6,852 women diagnosed with RPL in Sweden, during 2003-2012, a total of 238 complete medical records from 38 hospitals were reviewed. A power calculation estimated that 228 medical records had to be reviewed for a positive predictive value (PPV) of 85% (95% CI) with a power of 90%. The ICD-10 diagnoses used for RPL were N96.9 and O26.2. Results: The diagnosis of RPL was confirmed in 202 out of 238 medical records resulting in a PPV of 85% (95% CI 78-89%) out of which 59% were verified with ultrasound whereas 35% were verified only by urine/serum hCG. Conclusion: The Swedish NPR is a valuable tool for epidemiological research. We found a high PPV of RPL in the NPR, supporting the use of these data for future research.