• I Fournel, M Schwarzinger, E Benzenine, C Binquet, C Hill, and C Quantin.
• Département d'information médicale, CHU de Dijon, 1, boulevard Jeanne-d'Arc, 21000 Dijon cedex, France.
• Rev Epidemiol Sante. 2007 Oct 1; 55 (5): 365-73.

BackgroundA subject's vital status is essential for epidemiological studies. This information may be obtained for large numbers of patients with different methods, but these are often expensive. This study was aimed at assessing the performance of patient vital status determination using a record linkage method between hospital data and national mortality data once the information was made anonymous in compliance with French legislation.MethodsAll patients hospitalised in the Gustave-Roussy Institute, a cancer center in Villejuif France, were eligible for inclusion if they lived in France (mainland or Overseas Departments). The study cohort included patients admitted for the first time for malignant or suspected malignant-tumor during the period 1998-2000. Nominal data from the Gustave Roussy Institute hospital files as well as from the French National Institute of Statistics and Economic Studies (INSEE) mortality databases were then anonymised using irreversible hash coding. Once anonymised, the Gustave Roussy Institute and INSEE mortality databases were linked using the Jaro probabilistic method. Record linkage involved the following variables: birth name, first given name and birth date, along with the INSEE code of birth place.Results10,089 patients were included. The linkage record results were very satisfactory for all the patients included; the percentage of those properly classified was 97.2%, sensitivity was 94.8% and specificity 99.5%. The performance of the probabilistic record linkage method on anonymised data was very satisfactory (sensitivity 96.8% and specificity 99.8%) for determining vital status for patients born in France, suffering from cancer and in-patients at the Institute Gustave-Roussy. Results were inferior for patients born abroad (sensitivity 82.8% and specificity 97.7%) but the method achievements may be enhanced by additional manual validation steps.ConclusionProbabilistic linkage on data rendered anonymous enables to obtain information on vital status for a great number of subjects at low cost, in compliance with French legislation.

18 keywords...

hide…