Background: The organization of obstetric care in France brings all women in contact with the hospital system. Thus, hospital discharge data from the Program of Medicalization of the Information System (PMSI) constitute a potentially valuable source of information, particularly regarding rare events such as severe maternal morbidity. These data cover a large population but their quality has not been assessed in that field. Our objectives were to study the processes of production and the validity of PMSI data related to severe maternal morbidity.
Methods: The study was conducted in four French tertiary teaching hospitals (Caen, Cochin [AP-HP, Paris], Grenoble and Lille). First, the organization of each step of the medical information process -production, formatting, verification and processing- was detailed in each center with a standardized form. Second, the validation study was based on the comparison of data related to severe maternal morbid events in the PMSI from these centers for 2006 and 2007, with the content of medical records which constituted the gold standard. Indicators of sensitivities and positive predictive values of PMSI were calculated.
Results: The processes of PMSI data production showed major differences between the four centers. In hospital discharge data, diagnoses (eclampsia and pulmonary embolism) had a high proportion of false-positives (68%). Inversely, procedures (four procedures for management of severe haemorrhage) had less than 1% of false-positives, but a low sensitivity with 37% false-negatives which could be corrected in 95%. Regarding intensive care provision, all indicators of hospital data quality were very high. In addition, the validity of hospital data in centers 1 and 2 was higher for all events.
Conclusion: The heterogeneity of the process of PMSI data production is associated with a variable quality of these data. Intensive care provision can be used in the PMSI, as well as procedures after correction. For diagnoses, the quality of the PMSI data is better in centers having both computerized medical records and steps for verification of medical information.
Copyright © 2012 Elsevier Masson SAS. All rights reserved.