Results

In all, 535 physicians (191 SGPs and 344 non-SGPs) accepted voluntary participation—that is, 10% of the physicians to whom participation was offered. During the study period, 262 279 medical encounters were reported via 18 178 connections to the host computer. The hours of connection ranged from 5 am to 2 am. The duration of connections to the host server averaged 139 (174) seconds (mean (SD)). The participation time curve was bell shaped. Median participation time was 65% of the surveillance period (calculated by the Kaplan-Meier survival method). Attrition from the network was not found to be associated with the physician's work load in terms of the total number of patients seen daily, or with his proximity to an event's venue within a radius of 30 kilometres of a World Cup city (see reference 1for details). SGPs reported the same mean number of patients seen daily as non-SGPs—that is, 18 (7.66) and 17.4 (7.5) patients, respectively (not significant, Wilcoxon test). They participated for a longer period than non-SGPs, for median periods of 45 and 38 days respectively (p = 0.0001, Wilcoxon test). The level of medical activity reflected by the items surveyed remained stable throughout the WFC, both in the cities that hosted matches and those that did not (data not shown).

Discussion

For monitoring health issues in the general community during mass gatherings, a nationwide real time electronic active surveillance system based on sentinel physicians proved very feasible. It also proved acceptable in terms of participation time. The surveillance system was sensitive, because it would have captured one case of each disorder under surveillance with a 95% probability if at least 100 cases had been seen in general practice, according to the Poisson distribution, and because about 3% of the GPs were represented in the WFC cities. We may assume that our sample was fairly representative of French GPs, for four reasons: (1) the participating SGPs did not differ from all SGPs (data not shown) and the characteristics of the SGPs are comparable to those of all French Gps5; (2) non-SGPs were recruited by random digit dial procedures; (3) non-SGPs did not differ from SGPs as regards the declared incidences of the surveyed items; and (4) the overall male/female ratio of the GPs participating in this study was 4, compared with 3.8 for all French GPs. Real time electronic surveillance ensured appropriate timeliness. This system could therefore serve as a model for any ongoing monitoring of general community health during international special events, or for any other short-term intensive nationwide health surveillance.

During the study period, using the same model, other surveillance systems were implemented in emergency hospital departments, private emergency community services, community health centres, and the medical centre of the Paris airports.1 Like the sentinel GPs, none of these sources detected any significant health events.

This initiative may be generalisable to countries without a sentinel surveillance system. Most of the physicians participating in the present surveillance programme had never been involved in a surveillance system and were not experienced in data communication. All the electronic devices used were especially created for this particular programme, and were distinct from existing computer developments. We used hardware equipment that was light (a single desktop PC), and software based on free access internet technologies. Internet access is now increasingly widespread. The expenses and resources necessary for such an operation were not excessive (marginal costs of the programme were evaluated at about US$20 000).