ORIGINALES
Clustering of behavioural risk factors and their association
with subjective health
Iñaki Galána / Fernando Rodríguez-Artalejob / Aurelio Tobíasa / Lucía Díez-Gañána,b /
Ana Gandarillasa / Belén Zorrillaa
aServicio de Epidemiología. Instituto de Salud Pública. Consejería de Sanidad y Consumo de la Comunidad
de Madrid. Madrid. España. bDepartamento de Medicina Preventiva y Salud Pública.
Facultad de Medicina. Universidad Autónoma de Madrid. Madrid. España.
| (Agregación de factores de riesgo ligados al comportamiento y su relación con la salud subjetiva) |
| Abstract ]]>
Objectives: To describe the clustering of behavioural risk factors in the adult population of the Autonomous Community of Madrid (Spain), and to evaluate the association between the level of aggregation of such factors and suboptimal subjective health. | Resumen |
The Spanish version of this manuscript can be downloaded (PDF format) from the web (www.doyma.es/gs).
Correspondence: Iñaki Galán.
Servicio de Epidemiología. Instituto de Salud Pública.
Consejería de Sanidad y Consumo de la Comunidad de Madrid.
Julián Camarillo, 4 B. 28037 Madrid. España.
E-mail: iñaki.galan@saludmadrid.org
Received: 14 de septiembre de 2004.
Acepted: 10 de enero de 2005.
]]> Introduction
Several behavioural risk factors such as smoking, excessive alcohol consumption, inactivity and an unbalanced diet are responsible for most of the burden of disease in developed societies, expressed in terms of general mortality1, or premature mortality and disability2.
The simultaneous occurrence of several factors in the same individual has been associated with a greater risk of general mortality, and more specifically with mortality from cancer, heart disease and stroke3-6. Furthermore, the accumulation of several factors increases the risk of suboptimal perceived health7, although most of this effect might be due to the health disorders they induce8. It has also been shown that the clustering of classic risk factors (low physical activity, unbalanced diet, smoking, and excessive alcohol consumption) is associated with an atherogenesis lipid high and blood pressure profile9.
Although lifestyle is treated as a one-dimensional structure, an approach employing diverse methodological options has demonstrated their multidimensionality10-14. This means that completely healthy or unhealthy patterns of behaviour are infrequent: most people show various combinations of healthy and unhealthy habits. For example, the relationships between smoking and alcohol consumption15, between smoking and diet16, and between physical activity and other factors are well known17. A wider-range of combinations in which a higher than expected frequency of 3- and 4-factor clustering has been observed has also been evaluated18-20.
Risk-factor clustering analysis can contribute towards designing improved public health interventions21. In particular, it can be used to identify lifestyle-related risk factors which lead to other unhealthy habits. Furthermore, it can improve the efficiency of interventions by directing them at the sectors of the population who exhibit the highest aggregation of risk factors. This approach may also be used to stimulate research into the underlying influences responsible for the observed risk-factor clusters. Nevertheless, earlier studies have shown that the prevalence of multiple behavioural patterns differs between socio-demographic groups and regions22,23. This study therefore focuses its attention on describing the composition and aggregation pattern of the main behaviour-related risk factors for the adult population of the Community of Madrid. In addition, it evaluates the degree of clustering of these factors with respect to suboptimal subjective health.
Methods
Data source and study population
The information source used was the Non-communicable Disease Risk Factor Surveillance System (SIVFRENT), which was based on continuous telephone surveys on health behaviour and preventive practices among the non-institutionalised population aged 18-64 years, living in the Community of Madrid. The study sample was selected from a telephone directory listing homes with landline telephone: in Madrid, this currently covers 94.8% of homes24. The interview was carried out using a CATI (Computer Assisted Telephone Interviewing) system25. The questionnaire consisted of a central core of questions which have remained unchanged since 1995, the year in which the survey was first conducted. The methods of this system have been described in detail elsewhere26. For this study, data analysis focussed on 16,043 interviews carried out from 1996 through 2003.
Study variables
The behavioural factors analysed were: smoking, alcohol consumption, physical activity at leisure time and food habits. State of health was assessed as self-rated health during the previous twelve months. The following socio-demographic variables were also considered: age, educational level and social class.
]]> Smokers were defined as people who had smoked more than 100 cigarettes in their lives and who still smoked at the time of completing the questionnaire. Risk-drinkers were defined as men who consumed a daily average of ≥ 50 ml of pure alcohol and women who consumed ≥ 30 ml per day, or men who consumed ≥ 80 ml and women who consumed ≥ 60 ml over a short period of time, such as during an afternoon or a night («binge drinking»). Estimation of average daily consumption was based on recall of the type, frequency and quantity of consumption of different alcoholic drinks during the previous week. Allocation of «binge drinking» pattern was based on recalled consumption of 8 units of pure alcohol («drinks») in men and 6 in women over a short period of time in the course of the previous 30 days. Leisure time inactivity was defined as not undertaking activities involving at least moderate-intensity activity for 30 minutes at a time at least 3 times a week. To estimate free-time physical activity, metabolic equivalents (METs)27 were calculated from the frequency and duration of sporting activities during the previous 2 weeks. The CDC (Centers for Disease Control and Prevention) recommendation of carrying out at least moderate-intensity activities was used: these were defined as activities whose assigned METs27 were at least three times greater than those associated with resting28. Finally, an unbalanced diet was considered as consumption of less than 2 servings of fruit, juice or vegetables in the previous 24 hours.State of health was assessed as perceived health over the previous twelve months: the categories were very good, good, fair, bad and very bad, with the categories fair, bad and very bad being considered as indicators of suboptimal health. Finally, the following socio-demographic variables were considered: age in 9 groups (18-24 years old and subsequent 5-year groupings up to the age of 64); education: higher (university studies), medium-high (second degree secondary studies), medium-low (first degree secondary studies), and low (primary studies or lower); social class29: class I (professionals and management positions in companies with 10 or more employees), class II (management positions in companies with fewer than 10 employees and intermediate professions), class III (qualified non-manual workers), class IVa (skilled manual workers), class IVb (semi-skilled manual workers), class V (unskilled manual workers).
Analysis
All the possible risk factor combinations were studied, estimating each factor's prevalence and comparing observed and expected proportions. The expected probability was calculated assuming the independence of the different factors and multiplying the individual prevalence of each factor. The observed/expected ratios measured the direction and degree of behavioural clustering, and their 95% confidence interval was calculated assuming a Poisson distribution, as described by Breslow and Day30.
To identify population subgroups with the greatest probability of factor clustering, a logistic regression model was built adjusting for age, educational level, social class, and the year of the interview. Similarly, a logistic regression model was used to summarize the relationship between the number of risk factors present and the frequency of suboptimal subjective health, adjusting for age, educational level, social class, body mass index (weight in kg/square of the height in m2), and year of interview. The study years included in this analysis were 2000-2003, as subjective health was recorded from 2000 on. Analyses were done for each sex separately.
Statistical analysis was performed with the Stata v.7.0 (StataCorp, College Station, 2001).
Results
The average response rate for the period 1996-2003, measured as the number of completed interviews, divided by the number of complete and incomplete interviews plus the number of interviews not performed (including negative responses and non-contacts)31, was 66.1%. Response rates ranged from 61.7% in 1999 to 69.5% in 1996.
Table 1 shows the socio-demographic characteristics of the study sample and the frequency of each factor presented both individually and by cluster. In total, 9.5% of men and 8.3% of women showed no risk factors, while 69.0% of men and 77.8% of women had only one or two factors. High levels of aggregation, with the accumulation of 3 and 4 factors, were respectively present in 17.2% and 4.4% of men, and 12.2% and 1.6% of women.
Table 3 shows the relationship between the presence of a specific risk factor and the aggregation of the remaining behaviours. The individual factor most associated with this clustering was tobacco smoking; in fact, as compared to non-smokers, men and women who smoke had, respectively, odds ratios (OR) = 3.72 (IC 95%, 2.98-4.66) and 3.15 (IC 95%, 2.25-4.42) for having the other 3 risk factors. In second place comes high-risk alcohol consumption, followed by an unbalanced diet. The factor with the lowest tendency for clustering was leisure time inactivity. Except for tobacco smoking, where the association was greatest in men, the relationship was very similar for both sexes.
The presence of 3 or 4 risk factors occurred almost as twice as often in men as in women (table 4). The aggregation of 3 or 4 factors was also more frequent in the younger age groups (18 and 34 year olds in men and 18-24 year olds in women). In men, the frequency of clustering decreased with age after the age of 34. A similar pattern was shown for women, with the frequency of clustering decreasing from the 25-29 year old age group, with subsequent reductions being more pronounced than in men. The frequency of factor clustering in men also increased with a decreasing educational level. This gradient was not observed in women, although in comparison with women with university studies the probability of aggregation was always greater in groups with lower educational level. With regard to social class based on occupation, men exhibited greater accumulation of factors in the manual classes (IVa, IVb and V) in comparison with men in class I, although this was only statistically significant in category IVa. For women, there was no clearly observable pattern, although those of class IVb showed an OR of 1.39 (IC 95%, 1.04-1.82) with respect to members of the highest social class (table 4).
]]> Finally, the frequency of suboptimal health increased with the accumulation of behavioural factors (table 5). As compared to people with none of the risk factors studied, those with only one risk factor showed an OR for suboptimal subjective health of 1.90 (IC 95%, 1.24-2.93) in men, and 1.44 (IC 95%, 1.00-2.08) in women. In people with 3 or 4 factors these OR increased to 2.49 (IC 95%, 1.59-3.90) and 1.96 (IC 95%, 1.29-2.97) for men and women, respectively.
Discussion
The results of this study suggest that an important percentage of the population, about 20%, shows 3 or 4 important risk factors simultaneously: smoking, high-risk drinking, leisure time inactivity and having an unbalanced diet. These factors cluster on a multidimensional structural base, with tobacco smoking being the factor most closely related with the accumulation of other factors. The existence of high levels of aggregation was more common in men, in younger age groups and in the case of lower educational level, and was associated with a suboptimal subjective health. These results are consistent with those observed in previous studies19,20,23,32.
The frequency and distribution of the indicators studied, both individually and as a cluster, depends on the definition employed. In this work, the definition of tobacco smoking was the same as that regularly used in other health surveys33. The definition for risk-drinking was partly established in relation to average daily intakes in line with criteria proposed by the Programme for Preventive Activities and Health Promotion (PAPPS) of the Sociedad Española de Medicina de Familia y Comunitaria (Spanish Society for Family and Community Medicine)34, and also took into consideration «binge drinking», whose relationship with an increase in mortality is now well-known and documented35. The definition of leisure time inactivity was also elaborated according to the recommendations of the PAPPS34. Finally, insufficient consumption of fruit and vegetables, as an indicator of an unbalanced diet, was limited to the consumption of less than 2 rations per day. This frequency is situated in the lower quartile of quintile, and is a reference category used to calculate the risk of cardiovascular diseases and cancer36,37.
A limited number of people (about 9%) have a very healthy profile, having none of the indicated risk habits, and another minority (3%) has a very unhealthy profile, with all of the positive risk factors being present. These data are coherent with the absence of a one-dimensional structure10,13, according to which there should be 2 majority groups within the population; one with completely healthy habits and the other with unhealthy habits. In our case most people exhibit 1 or 2 risk factors, although the proportion of people with three or four factors is also high (close to 20%), but it is distributed with different frequencies for different combinations of aggregation, according to the multidimensional concept of these behavioural habits11,14. Our results are very similar to those reported in the studies of Schuit et al.19 for Germany and Laaksonen et al20 for Finland, in which the same risk factors were investigated. Of the 4 indicators studied, tobacco smoking is the one that presents the greatest probability of clustering with other risk factors. This is followed by excessive alcohol consumption and an unbalanced diet, while inactivity exhibits a much weaker relationship. This important role for tobacco in clustering has been described by Prättälä et al32 as the «gateway» to other risk factors, and Burke et al18 and Laaksonen et al23 have reached similar conclusions. Moreover, the weakest association -that of inactivity with the other risk factors- is also in line with observations based on other studies17,38.
The simultaneous existence of several unhealthy habits is more common in men than in women, and in younger people as opposed to older people. This age-related distribution probably reflects the higher survival rate of subjects who have maintained healthier habits and lifestyles, since -as many studies have shown- the presence of these risk factors is responsible for a significant incidence of premature mortality1-6. This situation could also be due to improvements in diet and the abandoning of addictive habits such as smoking or excessive alcohol consumption39 by older subjects. As well as being associated with abandoning unhealthy habits and/or differential survival, the more pronounced age-related differences associated with women could express a certain cohort effect in the adoption of risk factors23. This has, for example, occurred in our geographical area in the case of tobacco smoking40.
People with lower socio-economic status generally exhibit less healthy behaviour41. From comparisons among different indicators, it seems that education rather than income or occupation is the factor most consistently associated with different behavioural habits42. In our study, a greater aggregation of unhealthy behaviour was observed in people with low educational levels, while the relationship with occupation appeared less pronounced when the 2 variables were modelled simultaneously. This relationship with educational level has also been described by other authors considering similar risk factors.9,19,23,33. Our data showed this association as being greater for men than for women, with the difference being greater than that ob served by Laaksonen et al23. These results can be explained by a differential degree of incorporation of women into unhealthy and particularly addictive habits. For example, it is well documented that in the early stages of the development of epidemic tobacco addiction, the people who start smoking first belong to higher socio-economic levels, while in later stages, the greatest frequency of consumption occurs in the lower social class categories43. This effect can be clearly seen in our region, where - until recently - tobacco smoking was most prevalent in women from the highest socio-economic groups and the same was true of alcohol consumption. However, in recent years a change in this pattern has been observed, with a tendency towards similar frequency in all strata. In men, however, all of the risk indicators are most frequent in the lowest socio-economic categories44.
Subjective health is considered a valid indicator of the state of health and is an important independent predictor of morbidity and mortality45,46. Many investigations have detected an association between various individually assessed risk factors and a worse state of health8,46-50. This situation is also repeated with factor clustering9,19; as in our study, previous works show that as simultaneous risk factors accumulate, state of health worsens. This relationship could reflect the effects of both physical health problems and the functional limitations arising from these risk factors, because the physical symptom component tends to be related to perceived health51. When the model considers chronic diseases related to these factors, such as diabetes or known obstructive respiratory diseases (data not shown), this relationship is less marked. Indeed, it would be expected to diminish even further if other health problems were taken into account. This could be interpreted as the potential effect on perceived health being measured by the existence of chronic health problems, which would act as an intermediate step between the risk factors and the subjective state of health8. Even so, it is also likely that there is another direct relationship with these unhealthy habits that is independent of the existence of other health problems49.
]]> The 4 indicators have been aggregated with each receiving a similar weighting. Several authors have criticised the construction of these additive indices in which each factor is given equal treatment52,53, despite the fact that their contribution to the development of chronic health problems is different. Nevertheless, such indices have been successfully employed to explain the risk of morbidity and mortality3,4,54. Furthermore, as this is a cross-sectional study, it is not possible to make causal inferences on the relationships detected. For example, a person with health problems is likely to modify his/her behaviour by giving up smoking or excessive alcohol consumption, or by making health-favouring changes in diet or physical activity. This change in lifestyle would subsequently lead to this person being placed in a different category in the current classification, with little or no consequent factor clustering. Because recent ex-smokers49 and ex-drinkers50 have a worse state of health, the magnitude of the observed relationship would tend to decrease.In conclusion, a high percentage of the population, almost one in five people, simultaneously exhibits 3 or 4 of the following risk factors: smoking, high-risk drinking, leisure time inactivity and an unbalanced diet. These factors cluster in a multidimensional fashion, with smoking being the risk factor with the highest frequency of clustering. Clustering varies among socio-demographic strata, being most common in men, in younger age groups and in people with low educational level. The accumulation of factors is associated with suboptimal perceived health.
The tendency for these risk factors to cluster, the description of the pattern of combinations, and the identification of population groups with high clustering frequencies may have important implications for the design of population health promotion strategies, and also for the elaboration of preventive strategies for primary health care, largely based on the detection of individual risk factors.
Acknowledgwent
The translation of the manuscript was done by Malcom Hayes.
References
1. Mokdad AH, Marks JS, Stroup DF, Gerberding JL. Actual causes of death in the United States, 2000. JAMA. 2004;10:1238-45. [ Links ] 2. Murray CJL, López AD. Global mortality, disability, and the contribution of risk factors: global burden of disease study. Lancet. 1997;349:1436-42. [ Links ] 3. Meng L, Maskarinec G, Lee J, Kolonel LN. Lifestyle factors and chronic diseases: application of a composite risk index. Prev Med. 1999;29:296-304. [ Links ] 4. Yusuf HR, Giles WH, Croft JB, Anda RF, Casper ML. Impact of multiple risk factor profiles on determining cardiovascular disease risk. Prev Med. 1998;27:1-9. [ Links ] 5. Luoto R, Prättälä, Uutela A, Puska P. Impact of unhealthy behaviors on cardiovascular mortality in Finland, 1978-1993. Prev Med. 1998;27:93-100. [ Links ] 6. Chyou PH, Burchfiel CM, Yano K, Sharps DS, Rodríguez BL, Curb JD, et al. Obesity, alcohol consumption, smoking, and mortality. Ann Epidemiol. 1997;7:311-7. [ Links ] 7. Johansson S, Sundquist J. Change in lifestyle factors and their influence on health status and all-cause mortality. Int J Epidemiol. 1999;28:1073-80. [ Links ] 8. Manderbacka K, Lundberg O, Martikainnen P. Do risk factors behaviours contribute to self-ratings of health? Soc Sci Med. 1999;48:1713-20. [ Links ] 9. Raitakari OT, Leino M, Räikkönen K, Kimmo VK, Taimela S, Räsänen L, et al. Clustering of risk habits in young adults. The cardiovascular risk in young Finnish Study. Am J Epidemiol. 1995;142:36-44. [ Links ] 10. Kronenfeld JJ, Goodyear N, Pate R, Blair A, Howe H, Parker G, et al. The interrelationships among preventive health habits. Health Educ Res. 1988;3:317-23. [ Links ] 11. Krick JP, Sobal J. Relationships between health protective behaviors. J Community Health. 1990;15:19-34. [ Links ] 12. Hulshof KFA, Wedel M, Löwik MRH, Kok FJ, Kistemaker C, Hermus RJJ, et al. Clustering of dietary variables and other lifestyle factors (Dutch Nutritional Surveillance System). J Epidemiol Community Health. 1992;46:417-24. [ Links ] 13. Mulder M, Ranchor AV, Sanderman R, Bouma J, Van den Heuvel WJA. The stability of lifestyle behaviour. Int J Epidemiol. 1998;27:199-207. [ Links ] 14. Berrigan D, Dodd K, Troiano RP, Krebs-Smith SM, Ballard R. Patterns of health behavior in US adults. Prev Med. 2003;36: 615-23. [ Links ] 15. Sobell LC, Sobell MB, Kozlowski LT, Toneatto T. Alcohol or tobacco research versus alcohol and tobacco research. Br J Addict. 1990;85:263-9. [ Links ] 16. Dallongeville J, Marécaux N, Fruchart JC, Amouyel P. Cigarette smoking is associated with unhealthy patterns of nutrient intake: a meta-analysis. J Nutr. 1998;128:1450-7. [ Links ] 17. Johnson NA, Boyle CA, Heller RF. Leisure-time physical activity and other health behaviours: are they related? Aust J Public Health. 1995;19:69-75. [ Links ] 18. Burke V, Milligan RAK, Beilin LJ, Dunbar D, Spencer M, Balde E, et al. Clustering of health-related behaviors among 18-year-old Australians. Prev Med. 1997;26:724-33. [ Links ] 19. Schuit AJ, Van Loon AJM, Tijhuis M, Ocké MC. Clustering of lifestyle risk factors in a general adult population. Prev Med. 2002;35:219-24. [ Links ] 20. Laaksonen M, Prättälä R, Karisto A. Patterns of unhealthy behaviour in Finland. Eur J Public Health. 2001;11:294-300. [ Links ] 21. Sobal J, Revicki D, DeForge BR. Patterns of interrelationships among health-promotion behaviors. Prev Med. 1992;8: 351-9. [ Links ] 22. Liang W, Shediac-Rizkallah C, Celentano DD, Rohde C. A population-based study of age and gender differences in patterns of health-related behaviors. Am J Prev Med. 1999;17:8-17. [ Links ] 23. Laaksonen M, Prättälä R, Lahelma E. Sociodemographic determinants of multiple unhealthy behaviours. Scand J Public Health. 2003;31:37-43. [ Links ] 24. Encuesta a hogares españoles sobre equipamiento, uso y valoración de servicios de telecomunicaciones, audiovisual e Internet. Informe preliminar (diciembre de 2002). Comisión del mercado de las telecomunicaciones e Instituto Nacional de Estadística (accessed 11 Aug 2004). Available in: http://www.aece.org/docs/EstudioINE+CMT.2002.pdf. [ Links ] 25. Nicholls II WL. Computer-assisted telephone interviewing: a general introduction. In: Groves RM, Biemer PP, Lyberg LE, Massey JT, Nicholls II WL, Waksberg J editors. Telephone survey methodology. New York: John Wiley & Sons Inc.; 1988. p. 377-85. [ Links ] 26. Galán I, Rodríguez-Artalejo F, Zorrilla B. Reproducibilidad de un cuestionario telefónico sobre factores de riesgo asociados al comportamiento y prácticas preventivas. Gac Sanit. 2004;18:118-24. [ Links ] 27. Ainsworth BE, Haskell WL, Whitt MC, Irwin ML, Swartz AM, Strath SJ, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc. 2000;32 Suppl 9:498-504. [ Links ] 28. Pate R, Pratt M, Blair SN, Haskell WL, Macera CA, Bouchard C, et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA. 1995;273: 402-7. [ Links ] 29. Álvarez C, Alonso J, Domingo A, Regidor E. La medición de la clase social en ciencias de la salud. Informe de un grupo de trabajo de la Sociedad Española de Epidemología. Sociedad Española de Epidemiología. Barcelona: SG editores; 1995. [ Links ] 30. Breslow NE, Day NE. Statistical methods in cancer research. The design and analysis of cohort studies. Lyon: IARC Scientific Publications n.o 82; 1987. [ Links ] 31. The American Association for Public Opinion Research. 2004. Standard definitions: Final dispositions of case codes and outcome rates for surveys. 3rd ed. Lenexa, Kansas: AAPOR (accessed 11 Aug 2004). Avaliable in: http://www.aapor. org/pdfs/standarddefs2004.pdf [ Links ] 32. Prättälä R, Karisto A, Berg M-A. Consistency and variation in unhealthy behaviour among Finnish men, 1982-1990. Soc Sci Med. 1994;39:115-22. [ Links ] 33. Husten C, Jackson K, Lee C. Cigarette smoking among adults: United States, 2002. Morb Mortal Wkly Rep. 2004;53:427-31. [ Links ] 34. Brotons C, Ciurana R, Iglesias M, editores. Manual de prevención en Atención Primaria. Programa de Actividades Preventivas y de Promoción de la Salud (PAPPS). Barcelona: Sociedad Española de Medicina de Familia y Comunitaria; 2003. [ Links ] 35. Murray RP, Connett JE, Tyas SL, Bond R, Ekuma O, Silversides CK, et al. Alcohol volume, drinking pattern, and cardiovascular disease morbidity and mortality: is there a U-shaped function? Am J Epidemiol. 2002;155:242-8. [ Links ] 36. Bazzano LA, He J, Ogden LG, Loria CM, Vupputuri S, Myers L, et al. Fruit and vegetable intake and risk of cardiovascular disease in US adults: the first National Health and Nutrition Examination Survey Epidemiologic Follow-up Study. Am J Clin Nutr. 2002;76:93-9. [ Links ] 37. Van't Veer P, Jansen MC, Klerk M, Kok FJ. Fruits and vegetables in the prevention of cancer and cardiovascular disease. Public Health Nutr. 2000;3:103-7. [ Links ] 38. Johnson MF, Nichols JF, Sallis JF, Calfas KJ, Hovell MF. Interrelationships between physical activity and other health behaviors among university women and men. Prev Med. 1998; 27:536-44. [ Links ] 39. Cockerham WC. Lifestyles, social class, demographic characteristics, and health behavior. In: Gochman D, editors. Handbook of health behavior research. I Personal and social determinants. New York: Plenum Press, 1997. p. 253-65. [ Links ] 40. Fernández E, Schiaffino A, García M, Saltó E, Villalbí JR, Borrás JM. Prevalencia del consumo de tabaco en España entre 1945 y 1995. Reconstrucción a partir de las Encuestas Nacionales de Salud. Med Clin (Barc). 2003;120:14-6. [ Links ] 41. Pomerlau J, Pederson LL, Ostbye T, Speechley M, Speechley KN. Health behaviours and socio-economic status in Ontario, Canada. Eur J Epidemiol. 1997;13:613-22. [ Links ] 42. Winkleby MA, Jatulis DE, Franck E, Frank E, Fortmann SP. Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. Am J Public Health. 1992;82:816-20. [ Links ] 43. López A, Collishaw NE, Piha T. A descriptive model of the cigarette epidemic in developed countries. Tobacco Control. 1994;3:242-7. [ Links ] 44. Galán I. Desigualdades sociales en hábitos de salud en la Comunidad de Madrid. En: Desigualdades sociales en salud: situación en España en los últimos años del siglo xx. Murcia: Universidad de Alicante, 2002. p. 91-116. [ Links ] 45. Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav. 1997;38:21-37. [ Links ] 46. Manor O, Matthews S, Power C. Self-rated health and limiting longstanding illness: inter-relationships with morbidity in early adulthood. Int J Epidemiol. 2001;30:600-7. [ Links ] 47. Ross CE, Bird CE. Sex stratification and health lifestyle: consequences for men's and women's perceived health. J Health Soc Behav. 1994;35:161-78. [ Links ] 48. Parkatti T, Deeg DJ, Bosscher RJ, Launer LL. Physical activity and self-rated health among 55- to 89-year-old Dutch people. J Aging Health. 1998;10:311-26. [ Links ] 49. Guallar-Castillón P, Rodríguez Artalejo F, Díez Gañán L, Banegas Banegas JR, Lafuente Urdinguio P, Del Rey Calero J. Consumo de tabaco y salud subjetiva en España. Med Clin (Barc). 2001;116:451-3. [ Links ] 50. Poikolainen K, Vartiainen E, Korhonen HJ. Alcohol intake and subjective health. Am J Epidemiol. 1996;144:346-50. [ Links ] 51. Bailis DS, Segall A, Chipperfield JG. Two views of self-rated general health status. Soc Sci Med. 2003;56:203-17. [ Links ] 52. Dean K, Salem N. Detecting measurement confounding in epidemiological research: construct validity in scaling risk behaviours: based on a population sample in Minnesota, USA. J Epidemiol Community Health. 1998;52:195-9. [ Links ] 53. Slater CH, Linder SH. A reassessment of the additive scoring of health practices. Med Care. 1988;26:1216-27. [ Links ] 54. Segovia J, Bartlett RF, Edwards AC. Health status and health practices -Alameda and Beyond. Int J Epidemiol. 1991; 20:259-63. [ Links ] ]]>