Commentary: How good are the epidemiological data?BMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7069.1382 (Published 30 November 1996) Cite this as: BMJ 1996;313:1382
- Norman Noah, professor of epidemiology and public healtha
This surveillance system was provoked into existence by the virtual non-functioning of the existing notification system in France. In its 12 years, 330 000 separate episodes of communicable disease from a selected list of eight rubrics have been reported. This gives an average of one diagnosis a week from each of about 500 general practitioners, a surprisingly low rate considering the conditions that are reported. These are an interesting but curious mix, ranging from specific (measles, mumps, chickenpox, and HIV tests) through broad (male urethritis, acute viral hepatitis) to diffuse (acute diarrhoeas, flu-like illness).
The strength of the system is undoubtedly in its technical sophistication. The use of paper is (presumably) minimal or non-existent. Thus the statistics, provided the sentinel general practitioners report promptly, are right up to date, while the analysis and feedback are not only virtually instantaneous but also make use of modern statistical methods such as “kriging.” Ease of access to the system and feedback are enviable.
The system nevertheless remains a general practitioner based sentinel system and is not a substitute for the existing notification system in France. Diseases for which local or individual public health action need to be taken—such as rabies, meningococcal meningitis, or food poisoning—cannot adequately be covered by this or indeed any other general practitioner sentinel system. Measles and mumps—two diseases for which it has been most useful—will decrease in incidence with increasing vaccine coverage and become no longer viable for sentinel reporting. Moreover, even when the incidences of these infections were high, the number of reported cases to the sentinel system was fairly low—between 401 and 1558 cases annually between 1985 and 1990 for measles.1 Since the general practitioners' list sizes and the age and sex distributions of their patients were not known, cases were often reported as numbers per general practitioner and it is not clear how age specific incidence rates are calculated.
Including acute diarrhoeal disease in the system without laboratory backup may be of limited value unless there is an acute and overwhelming epidemic, which is uncommon and unlikely nowadays. Flu-like illness without appropriate laboratory backup will act as an effective early warning system for influenza, but several other infections masquerade under the heading of “flu-like” and in recent years have undoubtedly accounted for more illnesses than influenza itself. It is a pity that much valuable epidemiological information on this type of illness will have been lost because the organism or organisms responsible were not known. The time and place maps of flu-like illness produced by the system2 are beautiful but can be something of a mystery without this information.
The HIV tests give a positivity rate. This may be difficult to interpret in a sentinel system if the distribution of HIV infection and AIDS varies considerably by geographical area, especially when the system has a high turnover rate of general practitioners.3 What would also need to be known is the pattern of patients' access to their general practitioners, and not only for HIV.
The system is strong on collection, analysis, and feedback, three cornerstones of successful surveillance. The public will, however, mainly depend on the interpretation of the data if the system is to be of use to them. With as many as 25 000 different maps and 10 000 graphs to choose from, even the average epidemiologist may find the system a daunting one and the dedicated researcher is the person most likely to want access. I suspect that the “weekly updated electronic report, written by epidemiologists in simple words” will be all that most of the public will ever want to see.