A Pierron1, M Revert2, K Goueslard3, A Vuagnat4, J Cottenet1, E Benzenine1, J Fresson5, C Quantin6. 1. Service de biostatistique et d'informatique médicale (DIM), CHRU de Dijon, 21000 Dijon, France. 2. École de sages-femmes Saint-Antoine, hôpital Saint-Antoine, AP-HP, 184, rue du Faubourg-Saint-Antoine, 75012 Paris, France; Unité de recherche EA7285, risques cliniques et sécurité en santé des femmes et en santé périnatale, université Versailles St-Quentin, 2, avenue de la Source-de-la-Bièvre, 78180 Montigny-le-Bretonneux, France. 3. École de sages-femmes, service de biostatistique et d'informatique médicale (DIM), CHRU de Dijon, 21000 Dijon, France. 4. Ministère des Affaires sociales et de la Santé, direction de la recherche, des études, de l'évaluation et des statistiques, 14, avenue Duquesne, 75350 Paris, France. 5. Département d'information médicale, maternité régionale, CHU de Nancy, 54000 Nancy, France. 6. Service de biostatistique et d'informatique médicale (DIM), CHRU de Dijon, 21000 Dijon, France; Inserm, U866, université de Bourgogne, 21000 Dijon, France. Electronic address: catherine.quantin@chu-dijon.fr.
Abstract
BACKGROUND: In order to assess public health policies for the perinatal period, routinely produced indicators are needed for the whole population. In France, these indicators are used to compare the national public health policy with those of other European countries. French administrative and medical data (PMSI) are straightforward and reliable and may be a valuable source of information for research. This study aimed to measure the quality of PMSI data from three university health centers for core indicators in perinatal health. METHOD: PMSI data were compared with medical files in 2012 from 300 live births after 22 weeks of amenorrhea, drawn at random from University Hospitals in Dijon, Paris and Nancy. The variables were chosen based on the Europeristat Project's core and recommended indicators, as well as those of the French National Perinatal survey conducted in 2010. The information gathered blindly from the medical files was compared with the PMSI data positive predictive value (PPV) and the sensitivity was used to assess data quality. RESULTS: Data on maternal age, parity and mode of delivery as well as the rates of premature births were superimposable for the two sources. The PPV for epidural injection was 96.2% and 94.3% for perineal tears. Overall, maternal morbidity was underdocumented in the PMSI, so the PPV was 100.0% for pre-existing diabetes, 88.9% for gestational diabetes and 100.0% for high blood pressure with a rate of 9.0% in PMSI and 6.3% in the medical files. The PPV for bleeding during labor was 89.5%. CONCLUSION: To conclude, PMSI data are apparently becoming more and more reliable for two reasons: on one hand, the importance of these data for budgetary promotion in hospitals; on the other, the increasing use of this information for statistical and epidemiological purposes.
BACKGROUND: In order to assess public health policies for the perinatal period, routinely produced indicators are needed for the whole population. In France, these indicators are used to compare the national public health policy with those of other European countries. French administrative and medical data (PMSI) are straightforward and reliable and may be a valuable source of information for research. This study aimed to measure the quality of PMSI data from three university health centers for core indicators in perinatal health. METHOD:PMSI data were compared with medical files in 2012 from 300 live births after 22 weeks of amenorrhea, drawn at random from University Hospitals in Dijon, Paris and Nancy. The variables were chosen based on the Europeristat Project's core and recommended indicators, as well as those of the French National Perinatal survey conducted in 2010. The information gathered blindly from the medical files was compared with the PMSI data positive predictive value (PPV) and the sensitivity was used to assess data quality. RESULTS: Data on maternal age, parity and mode of delivery as well as the rates of premature births were superimposable for the two sources. The PPV for epidural injection was 96.2% and 94.3% for perineal tears. Overall, maternal morbidity was underdocumented in the PMSI, so the PPV was 100.0% for pre-existing diabetes, 88.9% for gestational diabetes and 100.0% for high blood pressure with a rate of 9.0% in PMSI and 6.3% in the medical files. The PPV for bleeding during labor was 89.5%. CONCLUSION: To conclude, PMSI data are apparently becoming more and more reliable for two reasons: on one hand, the importance of these data for budgetary promotion in hospitals; on the other, the increasing use of this information for statistical and epidemiological purposes.