Article Text

PDF

204 Searching for signals of potentially new disease-exposure associations: Interest of screening work-related diseases surveillance databases with Data Mining approaches?
  1. V B Bonneterre1,
  2. Bicout2,
  3. Agius3,
  4. De Gaudemaris2
  1. 1CHU Grenoble (Teaching Hospital), Grenoble Cedex 9, France
  2. 2UJF-Grenoble 1/CNRS/TIMC-IMAG UMR 5525 (EPSP team), Grenoble, France
  3. 3Centre for Occupational and Environmental Health, The University of Manchester, Manchester, United Kingdom

Abstract

Objective We defined potential new work-related diseases as either new couples “disease x agent”, or new triads “disease x agent x occupational setting” (“old friends in new places”) with, at least, a reasonably strong causality presumption. Before such cases have been reported a high number of times and raise clinicians’ concern, they will have been encountered only a few times, sometimes by different physicians. Some cases might have been captured by surveillance networks. Thus, it is for interest for these networks to develop tools trying to highlight pro-actively these kinds of early potential signals within their growing databases (rather than waiting to be asked, after clinician concern was expressed, how many cases they have been recorded the previous years). The objective of this communication, is to show and illustrate to what extent, a Data Mining approach could help identifying such cases of interest for vigilance purposes.

Methods Databases from the French National Surveillance Scheme on occupational diseases Surveillance and Prevention (rnv3p), and from the UK team of occupational diseases surveillance (THOR) have been explored (both schemes being part of a wider Modernet consortium, whose networking is currently funded by EU-COST program).

Analyses of the existing couples and triads and identification of those generating a signal with disproportionality measures used in pharmacovigilance (ex PRR: Proportional Reporting Ratio).

Results New couples and triads already derived from these analyses conducted on rnv3p and THOR schemes will be presented.

Conclusion These methods stand at the frontier of conventional epidemiological surveillance of work-related diseases, and might be beneficial for vigilance in highlighting similar cases, in order to investigate them as early as possible. The higher the “background noise” in the database, the more efficient they are to highlight disproportionalities.

Statistics from Altmetric.com

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.