• routine statistics

The World Health Organisation’s annual report of the year 2000¹ illustrated at least two features of health information systems at present. One is that the public is eager to compare, evaluate, and discuss health topics using “information products” such as quantitative figures, rankings, and the like. By providing information on health results achieved in each country and on the levels of resources invested, the WHO was able to trigger off a debate of unprecedented vigour.

However, this report also produced a vivid example on the frailty of the data on which most health information is based. As a matter of fact, the WHO study is based on scanty data: a reasonable system of vital statistics is still missing in large areas of the world, and data on the activities in health services are rudimentary in many countries, even in the richest parts of the world. Even a novel concept, like the one brilliantly developed in the WHO report, cannot overcome basic failures. Expectedly, but sadly, a large part of the discussion on this WHO report has been about the (poor) information system rather than on the (poor) performance of health systems.

Part of this problem is investigated by Unal and colleagues in this issue.² They explored systematically the existing data sources currently used in the UK to monitor and evaluate coronary heart diseases (CHD) and related aspects of prevention and care. With a standardised tool, the authors review the sources related to persons and to treatments, as well as data sources on classic risk factors for cardiovascular diseases. This study is refreshing for readers embarked in the daily utilisation of routine statistics—that is, for most of those people working in public health. It provides substantial information for UK researchers and practitioners: they now have access to a relevant evaluation of the databases available in the field of CHD.

It is also refreshing for the non-UK reader: it reminds everybody that we have to take care of the health information system if we want to develop public health strategies. Indeed, the situation is distressing in many countries, and in most areas of public health. Because of the recent and huge development of electronic tools (from data storage to data transmission and the ability to perform complex statistical analysis), there is the vague belief that the health information is performing increasingly well. This reveals a common confusion, just as technical feats of the television tend to mask the intellectual poverty of (most) programmes.

Unal et al typically show that a large amount of data related to CHD is available, even if many data sources must be looked for outside the health system. Although the United Kingdom might enjoy a richer set of data sources than others, a similar profusion is observed in most developed countries. However, the authors identify substantial problems, which most practitioners and researchers will easily recognise. One is related to the difficulty to analyse trends because of the irresolute character of many surveys, and also because of the continuous changes in definition of items, formulation of questions, etc.

These changes are clearly inappropriate when studying slow secular changes in the pattern of risk factors. The problem is partly related to the fact that, for many data, the data collection is defined outside the department of health: data on physical activity for example might be collected by the statistical unit of the department of transport, or data on housing by the social statistics. A practical proposal, thus, is to set up coordinating mechanisms within the official statistics in order to be sure that the needs of public health will be appropriately represented, both in terms of data collection and of requirement for analysis. Such a machinery does not exist in many countries I know of.

A further problem is the paucity of research in the field of information system, which corresponds to a lack of interest in the academic public health. Except for the management of health services (where information is a flourishing market), public health has not produced recently many new ideas on routine statistics. Consider for example the registration of cause of death, a remarkable epidemiological tool implemented and developed during the past century in the emerging modern states.³ Using death as an official event with compulsory registration, the cause of the death was registered and analysed as a surrogate for disease. This has provided enormous contribution to the field of public health, from the monitoring of the population health to the evaluation of the impact of public health interventions.

However, causes of death are notoriously inadequate to capture the health status of an aged or very aged population: doctors have to choose a “causal” disease among two or three serious conditions, among 10 known pathologies. One consequence is that causes of death tend to be ignored when this population is analysed despite the fact that people aged more than 60 provide 80% of the deaths in developed countries. A more sensible approach would be to develop novel approaches for the use of death in health statistics: an option would be to measure prevalent diseases at death.

Another area where research is lacking is the systematic analysis of the impact of information system on decision in public health. As pointed out by Unal et al, the health system requires “good evidence, particularly as resources are limited”. However, it is difficult to find good systematic reviews showing how the information provided by routine statistics has been conclusive to start, to modify or to stop a public health strategy. Of special interest for systematic research is the way public health professionals interpret observed trends from routine statistics in term of proof or decisive argument (see for example the debate about the impact of the control of cardiovascular risk factors⁴). Unlike clinical medicine, where the development of evidence based guidelines triggered off several research lines to analyse the impact of information on doctors’ activities, public health has been slow to develop an interest in this field.

Addressing these three points, and several others, requires an active involvement of each practitioner and researcher. Routine statistics are an important resource for public health, but are yet underresearched and underfunded: as stated by Unal and colleagues, the 2 billion spent annually by the NHS to take care of CHD patients does not have a reasonable counterpart in terms of investment in the information system. As a matter of fact, routine statistics have been a prerequisite for the development of the modern public health. The future of public health also depends on the professional willingness to maintain these resources.