Reproduced from THE LANCET, November 6, 1971, pp 1024 - 1026
MEDICINE - the study, prevention, and treatment of ill-health - is one of the largest industries there is. In 1968 - 69, the United Kingdom spent 4.69% of her gross national product on health care, while the United States spent 6.71%.1 Despite the size of this industry we have as yet no rational basis upon which to organise and distribute our resources. This must in part be attributed to the absence of any method of measuring efficacy: we have no method of measuring a patient's health before and after treatment. If we had, we might be able to allocate limited treatment resources to areas where they would be most beneficial and research to areas where it is most required, and we could also apply cost-effectiveness calculations to much medical work.
This paper reports the development and assessment of a health index which measures health and allows severity of disease, efficacy of treatment, and cost to be compared.
Medical care includes frequent evaluation of the patient's health, but the measurement is practical or relative and is expressed in such terms as "fit for light duty" or "better than yesterday". Such assessments serve the patient and his doctor fairly well but they do not lend themselves to numerical representation. Indeed, such representation is impossible without first having a suitable definition of health upon which to base the score. The World Health Organisation 2 defines health as a state of physical, mental, and social well-being, and this definition was used as a basis for some early trials. However, it soon became apparent that although it was a reasonable definition, it did not encourage precise scoring. What one observer considered "mental" another considered "social", and what one observer felt was minor social ill-health was major to another.
After this experience, we examined the activities and aspects which together comprise the pattern of daily life; many of these appeared to overlap or encompass each other, and a cumbersome and inefficient method of scoring would have been the result. However, by elimination of some items, and grouping of others, the list was reduced to ten which were comprehensive but without obvious redundancy:
(3) Physical suffering
(4) Mental suffering
(7) Dependency on others
(10) Sexual activity
These ten factors were used as the basis for scoring a patient's health, each factor being considered in the same way: "Is there any impairment due to the patient's ill-health?". This approach corresponds to a definition of health as: "No impairment in the ten aspects of life listed." Although not an elegant definition, it seemed likely to lend itself to unambiguous scoring.
The scoring system selected was to allocate for each factor 1, 1/2, or 0, according to whether the patient was normal, impaired, or incapacitated. The total was then divided by ten to yield the health index. The score for a normal person was thus 1, ill health causing a reduced score and extreme ill-health scoring 0. This method of scoring health was evaluated.
The ten questions were printed as a list (see figure), with space for recording details of patient and observer, and room for several scores to be made. The interviewer was provided with instructions (see legend to figure) on the interpretation and scoring of some of the questions.
Twenty observers made 180 assessments on twenty seven patients. The observers included consultants, registrars, housmen, and students, and the patients were volunteers with a variety of complaints who were chosen when an opportunity presented for several observers to see the patient. The scores were made quickly and easily, much of the information being already obtained from a normal consultation. The scores in this series ranged from 0.25 scored by a man with severe asthma to 1.0 scored by a man awaiting admission for excision of a tongue papilloma.
There is no accepted alternative measurement against which the accuracy of the scoring system can be gauged, but when the patients were ranked according to score, there did not seem to be any striking anomalies (see table).
DIAGNOSES OF PATIENTS PLACED IN RANK ORDER
ACCORDING TO HEALTH INDEX SCORED.
|Benign Tongue Papilloma
Infertility for investigation
Pregnancy for termination
Carpal tunnel syndrome
|Carcinoma of stomach
Lumbar pain for investigation
|Gout in a patient with coarctation
and alcoholic neuropathy
The variation between observers was slight, with scores usually differing by no more than 0.05. This reproducibility was assessed in this series in the following way. All the scores obtained on any patient at one session were paired in an arrangement based on the alphabetical order of the names of the observers, any unpaired score being discarded. This yielded 72 pairs of observations for which the correlation coefficient was 0.883, with a standard error of 0.118.
Many patients were seen more than once. For example, a woman with asthma was treated for five days; one observer judged her score to have risen from 0.4 to 0.55, compared to 0.45 to 0.70 for a second observer. In such cases the change in score was calculated and comparison was again made on an arrangement based on the alphabetical order of observers, any unpaired score being discarded. This yielded 25 pairs of observations. On all but one occasion the observers agreed on the direction of change, and there was good agreement on the magnitude of the change. The correlation coefficient was 0.884 with a standard error of 0.196.
The development of this health index was stimulated by consideration of the statistics in current use. For example, the mean bed occupancy has been widely used in the United Kingdom. This provides a very inadequate representation of efficiency and tends to encourage such practices as keeping a patient in hospital until the next operating-list. To run a factory on "mean warehouse occupancy" would be commercial suicide, for it would encourage overcrowding, stagnation, and buildings of inadequate size.
Other measurements, although useful, have only limited applicability. For example, the benefit of influenza vaccination may be inferred from the percentage of factory staff off work. However, to extend this measurement to the general assessment of health services would be inappropriate; apparent "efficiency" might result from various measures ranging from fining absentees to compulsory sterilisation of the female labour force. A satisfactory measure of efficiency must be applicable to a wide range of problems and should produce no anomalies in the society in which it is to be used.
The health index described here may be used as a weighting factor giving a value in health terms to any period of time. The product is conveniently expressed in "health years". Thus, one year in full health is one health-year, one year with a health index of 0.5 is 1/2 health year, one month at full health is 1/12 health-year, and one month with a health index of 0.5 is 1/24 health year. Benefit from medical treatment can thus be expressed as health years or parts of a health year gained to the patient as a result of the measures under consideration. In the same way the impact of a disease can be expressed as health-years lost to an individual or to the community. Such knowledge could be used in allocating resources available for treatment and research.
If the health index is combined with satisfactory measurements of cost it should be possible to apply cost-benefit analysis to health. Hitherto the appropriate information about cost has not been available, but the massive collection of data now being undertaken in the Hospital Activity Analysis project could provide a basis for satisfactory costing.
1 Office of health Economics Information Sheet no. 9. London, 1970.
2 World Health Organisation. Constitution of the World Health Organisation: annex 1. World Health Organisation, 1958.
|Copyright © 2004
February 4th 2005