Socioeconomic scoring in an urban area of a developing country*

BRIEF COMMUNICATIONS Socioeconomic scoring in an urban area of a developing country * D. J. POLE 1 & A. C. IKEME 2 Abstract A group of variables that might define " life-style" was selected and a questionnaire based on these was distributed to the families of 2350 schoolchildren in Accra, Ghana to determine the presence or absence of these variables. Analysis of the replies revealed a set of answers that formed a related group. A scoring system was developed in which rarer attributes were weighted, and the community was thus graded by socioeconomic status on a ten-point scale. Principles for developing questionnaires applicable to other types of community were defined. Social classification has been a most productive adjunct to the study of diseases of industrialized communities in developed countries. Clearer social groupings in such countries permit classification based on easily obtainable variables (e.g., occupation or educational level). In developing countries, however, although certain conditions such as endomyocardial fibrosis (1, 2) and low serum chol- esterol levels (3) have been reported as being more common in persons of low socioeconomic status, social classification has often been based on the impressions of clinicians. A more systematic approach to grading social status in developing countries is therefore needed. As part of a survey of blood pressure in African schoolchildren, a questionnaire was used to define the socioeconomic determinants appropriate to families in one developing country. This paper will describe how the questionnaire was used to generate a scoring system for social status in order to illus- trate general principles that would be applicable to other developing countries. * From the WHO Cardiovascular Research Team, Department of Medicine, University of Ghana, P.O. Box 4236, Accra, Ghana. 1 Epidemiologist. 2 Cardiologist. Method The parents of 2350 five- to twelve-year-old schoolchildren in Accra, Ghana were given a ques- tionnaire to be completed and returned to school the following day. The response rate was over 99 %. The following questions were asked: 1. How many rooms are there in the dwelling? 2. Is the kitchen shared? 3. Is the toilet shared? 4. What type of vehicle (if any) is owned? 5. Is a television or radio set owned? 6. What level of education has the father attained? 7. How many children are there in the family? The schools selected were from different parts of the Accra metropolitan area, covered a wide social range, and included both state and fee-paying schools. The answers were coded, punched on to magnetic tape, and analysed by means of the computer programme SPSS (4). Following the analysis, certain questions were rejected and a scoring system was developed to provide a numerical value for each family for use in subsequent studies. Results Analysis of responses. An appropriate response was received to each question in about 97 % of cases (Table 1). The consistency of the questions was checked by determining how the responses to each question correlated with each other. If the different questions were measuring the same par- ameter, the responses would all follow the required trend. In the case of house size, this regular trend was not observed with respect to most of the other questions. Only 54% of large " houses " (> 7 rooms) had their own kitchens, compared with 62%0 and 73% of small and medium-sized "houses". This unexpected relationship was also observed between house size on the one hand and toilet, car, and TV ownership, education, and family size on the other. This indicated either that " house size " was not 3461 -476- BULL. WORLD HEALTH ORGAN., Vol. 53, 1976 BRIEF COMMUNICATIONS rable 1. Results from the questionnaire Attribute Percentage a Rooms 1-2 12 3-6 40 7+ 46 Kitchen shared 48 separate 50 Toilet shared 48 separate 50 Vehicle car 26 bicycle 6 none 66 riadio TV & radio 22 radio only 50 neither 28 Father's education university 14 technical & secondary 25 primary 32 none 25 Number of children 1-2 11 3-6 53 7+ 36 a Inappropriate results were excluded from the analysis. related to the other life-style determinants or that the question itself had not been properly framed. For this reason the answers to this question were excluded from the ultimate score. The answers to other questions showed a cor- relation but not a complete overlap. For example, if the kitchen was shared, the toilet was also likely to be shared; car ownership was indicative of television ownership; and the father's education was indicative of both car and television ownership (Table 2). These correlations were significant (P < 0.01) and were in the expected direction, i.e., that of increasing material possessions. These questions were therefore considered satisfactory in that were they sufficiently related to be measuring the same parameter without being so closely related as to be repetitive. "Family size " however, showed smaller and less significant correlations to the other question. In framing this question, we had assumed that wealthier Table 2. Correlations between answers to the questionnaire Own Car TV/ Education No. oftoilet radio children Own kitchen 0.49 0.22 0.17 0.18 0.04 NSa Own toilet 0.27 0.20 0.19 0.04 NSa Car 0.54 0.51 0.12 TV/radio 0.49 0.10 Education 0.15 a NS = not significant. families would have fewer children. However, the response suggested that a more complex relation- ship was involved and the answers to this question were excluded from the final score. However, the existence of a weak correlation with other life- style indices suggests that the question could be further refined. Examination of the replies led to a breakdown of the families that responded to the questionnaire into definite groups. For example, the very low level of bicycle ownership, which was probably due to the existence of good public transport, suggested that this item was not a significant measure of material status in this community and those who owned bicycles were therefore grouped with those who owned no vehicle. Development ofscore. The next step was to develop a scoring system based on the acceptable answers. From the premise that each question concerned a desirable item, it followed that the rarer items were proportionally harder to obtain. A formula was therefore evolved to give suitable weighting to each item. The score was the whole number that, when multiplied by the number of people possessing the desirable item, gave a total close to half the population. For example, since 1154 families (49.5%) had a separate toilet the score for that item was 1, but since only 586 (25.1 %) had a father with secondary education the score for that item was 2. The scores developed in this way are shown in Table 3. The breakdown of the scores obtained can be seen in Table 4. A convenient breakdown into ten groups was produced. The median score was 3.2 and the mean was 3.8 ± 2.5. The distribution is skewed to the right (skewness = + 0.79). 477 BRIEF COMMUNICATIONS Table 3. Scoring based on answers Score 0 1 2 3 Kitchen/toilet share share one share neither Radio/TV neither own one own both Vehicle none own car Father's none primary secondary/ university education technical Discussion The concept of " social class " in industrialized countries is based on a relationship involving educational status, material possessions, occupation, and other empirical variables, which tend to group people into different life-styles. These life-styles correlate with many health factors as widely different as the frequency of rheumatic heart disease and mortality from car accidents (5). In developing countries, the association of many cardiovascular diseases, such as endomyocardial fibrosis (1) or puerperal cardiomyopathies (6), with socioeconomic status has been suggested. These observations, and the probable influence of changing life styles on the patterns of disease in developing countries (7), demand that methods of assessing social status be accurately refined. The principles of assessment Table 4. Breakdown of scores on a ten-point scale (2333 families) Value Relative frequency 0 4.2% 1 10.7% 2 20.9% 3 20.9% 4 13.9% 5 8.0% 6 3.4% 7 3.2% 8 4.8% 9 10.0% Total 100.0 % developed in one community should be applicable to another, even though different empirical variables might be measured. The aim of this analysis, therefore, has been to show how a set of questions relating to life style indices may be used to determine socioeconomic status. The validity of this technique depends considerably on providing a set of questions that are relevant to that community and that can be shown to measure much the same thing. Require- ments for such questions can therefore be proposed as follows: 1. The set of questions should be properly framed and should be appropriate for the community under study. 2. Each question should divide the population into groups of a reasonable size, if necessary by a suitable combination of the answers. Rare or very common attributes should be rejected since these cannot add significantly to the value of the infor- mation obtained. 3. Answers should correlate partially but not completely with each other. If the correlation is weak, the question should be further refined or rejected as not being a valid measure of the relation- ship being examined. If too strong, the questions are measuring almost identical parameters and one of them is redundant. 4. Where there is a range of responses, the ques- tion is to be considered satisfactory only if it correlates with other responses over its entire range. 5. A scoring system should give weight to rarer attributes on the assumption that each attribute is equally desired. Socioeconomic scoring does not obviate the need to relate disease to occupational groups where this is possible. In urban communities of developing countries however, where occupational groupings are ill-defined and do not expose subjects to dis- tinctive risks, scoring of material attributes presents a rational method of socioeconomic classification for examining etiological hypotheses. ACKNOWLEDGEMENTS Computer analysis of the results was performed by Mr E. Grant, WHO, Geneva. Helpful criticism was received from Dr P. H. N. Wood, ARC Epidemiology Research Unit, Manchester, England and from Dr J. S. Stromberg, WHO, Geneva. The study is part of a WHO supported research project. 478 BRIEF COMMUNICATIONS 479 REFERENCES 1. ABRAHAMS, D. G. Endomyocardial fibrosis of the right ventricle. Quarterly journal of medicine, 31: 1 (1962). 2. SHAPER, A. G. & COLES, R. M. The tribal distribution of endomyocardial fibrosis in Uganda. British heart journal, 27: 121 (1965). 3. EDOZIEN, T. C. Biochemical normals in Nigerians: chemical composition of the blood of adults. West African medical journal, 9: 204 (1960). 4. NIE, N. H. ET AL. Statistical package for the social sciences. New York, McGraw-Hill, 1970. 5. MORRIS, J. N. Uses of epidemiology. Edinburgh & London, Livingstone, 1967, p. 66. 6. STUART, K. L. Cardiomyopathy of pregnancy and the puerperium. Quarterly journal of medicine, 37: 463 (1968). 7. BERTRAND, E. & BAUDIN, L. Coronaropathies et deve- loppement economique. Nouvelle presse medicale, 3: 285 (1974).