Chapter 24
Distribution of risk factors by poverty Tony Blakely, Simon Hales, Charlotte Kieft, Nick Wilson and Alistair Woodward
Summary Socioeconomic position is an important distal risk determinant for many health outcomes. While it was not possible (owing to limitations of data and other factors) to directly map socioeconomic position to the burden of disease, it was considered possible to map some risk factors by absolute poverty, which is one measure of socioeconomic position. The proportions of the population living on US$ 2 per day were estimated for each of the 14 subregions1 using World Bank estimates of poverty by country. The counterfactual scenario was no absolute poverty in the world (no one living on US$ 2/day
Reference prevalencea
EMR-B
Reference prevalencea
EMR-D
Reference prevalencea
EUR-B
Reference prevalencea
Subregion
Underweight (low weightfor-age)
15.4
32.4 25.9 19.4 0.0
5.2
72.7 32.0 11.2 0.0
13.7
35.1 29.8 26.0 0.0
10.1
Unimproved water and/or sanitation
—
— — — —
—
— — — —
—
— — — —
47.3
Non-marital sex (men)
—
— — — —
—
— — — —
—
— — — —
4.1
Non-marital sex (women)
—
— — — —
—
— — — —
—
— — — —
50.7
Condom use (men)
—
— — — —
—
— — — —
—
— — — —
60.1
Condom use (women)
39.9
9.3 9.6 8.9 0.0
21.9
66.2 45.3 15.0 0.0
—
— — — —
12.2
Indoor air pollution
36.7
–5.6 –8.2 –8.2 0.0
17.6
11.5 7.2 2.2 0.0
—
— — — —
31.9
Tobacco use
64.9
–13.6 –17.3 –16.6 0.0
—
— — — —
—
— — — —
65.9
Alcohol use
38.9
–9.1 –6.5 –1.8 0.0
17.3
–5.2 –3.2 –1.1 0.0
—
— — —
31.7
Body weight (women)
Table 24.49 Absolute percentage differences in risk factor prevalence by subregion by three-level poverty variable: reference group >US$ 2 per day prevalence (continued)
2028 Comparative Quantification of Health Risks
28.0 18.6 7.0 0.0
26.1
1.1 5.5 4.0 0.0
13.9
US$ 2/day
>US$ 2/day
US$ 2/day
>US$ 2/day
SEAR-D
Reference prevalencea
WPR-B
Reference prevalencea
47.1
31.0 16.1 2.3 0.0
16.0
64.7 50.0 17.0 0.0
22.1
—
— — — —
—
— — — —
—
— — — —
—
— — — —
—
— — — —
—
— — — —
15.4
No data.
15.5
>US$ 2/day
Reference prevalencea
21.2 15.4 9.2 0.0
—
–9.7 –8.6 –7.6 0.0
The actual percentage prevalence of the risk factor in the >US$ 2 per day reference category.
34.9 19.0 10.8 0.0
US$ 2/day
SEAR-B
1.9
— — — —
a
2.0
>US$ 2/day
Reference prevalencea
20.2 14.0 8.2 0.0
—
2.8 2.2 1.7 0.0
US$ 2/day
EUR-C
—
— — — —
—
— — — —
—
— — — —
—
— — — —
—
— — — —
—
— — — —
—
— — — —
22.6
–0.9 –0.4 –0.5 0.0
64.3
28.5 25.6 18.4 0.0
29.2
70.8 54.1 22.5 0.0
—
— — — —
20.7
6.3 8.3 6.6 0.0
34.6
1.4 –0.3 –0.7 0.0
—
— — — —
—
— — — —
34.4
2.7 1.0 –0.9 0.0
62.0
–10.7 –7.6 –4.4 0.0
—
— — — —
—
— — — —
89.1
–21.1 –15.8 –13.7 0.0
20.1
1.7 –0.5 3.0 0.0
0.2
–0.1 –0.1 0.0 0.0
—
— — — —
—
— — — —
Tony Blakely et al. 2029
17.3 7.5 0.0
22.6
19.2 13.6 0.0
25.8
4.8 3.3 0.0
4.5
16.0 9.5 0.0
9.6
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
AFR-D
Reference prevalencea
AFR-E
Reference prevalencea
AMR-B
Reference prevalencea
AMR-D
Reference prevalencea
Subregion
Underweight (low weightfor-age)
20.2
68.5 50.1 0.0
15.5
79.7 59.0 0.0
46.2
41.7 32.6 0.0
16.4
53.1 17.4 0.0
Unimproved water and/or sanitation
45.7
–14.1 –8.0 0.0
45.7
–11.9 –9.4 0.0
37.5
–6.2 3.7 0.0
49.2
–18.9 –6.8 0.0
Non-marital sex (men)
4.1
–1.2 –0.3 0.0
18.2
–2.1 –4.1 0.0
17.1
–5.4 –3.8 0.0
20.4
–10.4 –2.6 0.0
Non-marital sex (women)
46.1
–26.7 –18.4 0.0
57.6
–18.8 –15.8 0.0
44.4
–27.0 –15.4 0.0
44.7
–19.0 –4.7 0.0
Condom use (men)
47.3
–32.0 –26.3 0.0
37.4
–24.8 –20.7 0.0
24.7
–21.4 –16.1 0.0
26.5
–17.0 –7.6 0.0
Condom use (women)
43.0
57.0 100.0 0.0
17.8
62.1 49.7 0.0
80.5
19.5 49.6 0.0
70.6
5.1 4.7 0.0
Indoor air pollution
29.2
–9.6 –6.2 0.0
30.6
0.9 3.9 0.0
24.6
–6.0 –5.3 0.0
14.2
2.4 1.5 0.0
Tobacco use
64.1
–9.0 –8.4 0.0
70.0
–34.8 –26.6 0.0
50.5
–77.3 –80.2 0.0
41.6
–5.9 –4.9 0.0
Alcohol use
31.1
–5.0 –3.8 0.0
27.4
–5.1 –2.9 0.0
33.1
–18.4 –13.8 0.0
20.1
–10.1 –4.7 0.0
Body weight (women)
Table 24.50 Absolute difference in risk factor prevalence by subregion by two-level poverty variable: reference group >US$ 1 per day prevalence
2030 Comparative Quantification of Health Risks
7.2 7.0 0.0
8.0
5.5 4.6 0.0
24.0
5.0 4.8 0.0
7.4
2.7 2.6 0.0
2.5
26.1 23.1 0.0
24.1
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
US$ 1/day
>US$ 1/day
EMR-B
Reference prevalencea
EMR-D
Reference prevalencea
EUR-B
Reference prevalencea
EUR-C
Reference prevalencea
SEAR-B
Reference prevalencea
29.1
14.1 13.7 0.0
4.7
22.2 21.7 0.0
19.5
23.1 22.5 0.0
23.3
54.1 41.7 0.0
15.9
32.0 30.9 0.0
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
13.4
–8.1 –8.0 0.0
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
—
— — —
22.5
–0.8 –0.7 0.0
—
— — —
—
— — —
—
— — —
64.2
35.8 70.7 0.0
22.5
4.7 7.5 0.0
41.3
8.3 8.6 0.0
50.0
27.1 23.7 0.0
—
— — —
—
— — —
34.7
2.5 2.2 0.0
35.5
–4.3 –4.8 0.0
22.0
5.6 3.6 0.0
—
— — —
—
— — —
85.5
–18.0 –16.0 0.0
62.2
–10.6 –11.5 0.0
—
— — —
—
— — —
continued
—
— — —
—
— — —
37.9
–8.4 –8.0 0.0
15.5
–3.5 –3.4 0.0
—
— — —
Tony Blakely et al. 2031
46.9
—
— — — —
— — —
—
The actual percentage prevalence of the risk factor in the >US$ 2 per day reference category.
16.2
>US$ 1/day
Reference prevalencea
62.6 61.1 0.0
—
— — —
Non-marital sex (women)
No data.
–1.2 –0.2 0.0
US$ 1/day
WPR-B
53.1
— — —
Non-marital sex (men)
a
39.9
>US$ 1/day
Reference prevalencea
27.7 27.6 0.0
Unimproved water and/or sanitation
—
14.2 10.8 0.0
US$ 1/day
SEAR-D
Subregion
Underweight (low weightfor-age)
—
— — —
—
— — —
Condom use (men)
75.0
— — —
—
— — —
Condom use (women)
34.5
17.4 18.1 0.0
71.4
28.6 30.4 0.0
Indoor air pollution
58.9
1.5 0.1 0.0
—
— — —
Tobacco use
20.2
–7.4 –6.9 0.0
—
— — —
Alcohol use
0.2 0.6 0.0
0.2
–0.1 –0.1 0.0
Body weight (women)
Table 24.50 Absolute difference in risk factor prevalence by subregion by two-level poverty variable: reference group >US$ 1 per day prevalence (continued)
2032 Comparative Quantification of Health Risks
Tony Blakely et al.
2033
underweight among children living on >US$ 2 per day in AFR-D was 17.9%. Accordingly the prevalence of underweight among children living on 65 years. Also, there was a very strong association with higher levels of education and smoking among younger Spanish women (in marked contrast to the data for men) (1987 data). More recent data from Spain suggest that the gap in smoking cessation rates by educational level has become wider over time (i.e. cessation rates are higher for those with higher levels of education) (Fernandez et al. 2001). This pattern was also apparent for women with higher educational levels. Similarly, other data from Spain suggest that the quitting rate is higher among women aged 25–44 years with non-manual occupations compared to those with manual occupations (Regidor et al. 2001). A review of smoking among adolescents identified studies in five developed countries (Iceland, Japan, New Zealand, Sweden and the United Kingdom) in which higher personal income was associated with higher levels of smoking (Tyas and Pederson 1998). Smoking not associated with SES One review identified two studies that reported non-significant effects of parental education on adolescent smoking in Canada and the United States (Tyas and Pederson 1998). However, these studies examined maternal education only. In the United Kingdom there was little occupational class gradient in the proportion of children who had ever smoked (although some gradient existed for higher average consumption of cigarettes with lower occupational class) (Acheson 1998). It has been suggested that smoking in adolescence is an indicator of occupational
2038
Comparative Quantification of Health Risks
class of destination rather than occupational class of origin (Glendinning et al. 1994). DISCUSSION
The relationship In general in developing countries, lower SES was associated with higher prevalence of smoking. The association was generally stronger for men. There was also a much lower smoking prevalence among women in most developing countries. Nevertheless, some studies in the developing world actually show the opposite pattern, with high-SES adults and adolescents having higher smoking rates, while other studies show no association. This overall pattern contrasts with that for other cardiovascular disease risk factors in developing countries (cholesterol level, obesity, blood pressure and to some extent physical inactivity). This might reflect the low cost of tobacco products in many developing countries and the influence of advertising and the mass media in promoting their use. Nevertheless, there is a need for more historical data on smoking in developing countries to better interpret this pattern. In developed countries the gradient of increased smoking with lower SES is even more predominant. There are still some exceptions, however, such as for adolescent smoking in some settings and for Spanish women. Furthermore, the SES gradient for tobacco-related disease and overall mortality is likely to be partly explained by the SES gradient for tobacco use. In both developing and developed countries there is some evidence that the type of tobacco product smoked by low-SES adults may also be more hazardous (e.g. beedis in India) than that consumed by high-SES adults. Similarly, the number of cigarettes consumed daily may be higher among low-SES adults. Possible mechanism Several factors are likely to be involved in the relatively increased prevalence of smoking by lower-SES groups. • Cultural effects relating to health values and risk perception may be relevant. Similarly, neighbourhood effects may be important, such as the influence of disadvantaged neighbourhoods on smoking in men (though not women) described in the United States by Ross (2000). • Parental influences on adolescent smoking may possibly be more important in low-SES populations (e.g. in countering the benefits of education about the risks of smoking). • Access to cheap tobacco in tobacco-producing rural areas of the developing world may be relevant.
Tony Blakely et al.
2039
• Lower educational attainment may lead to poorer knowledge of the hazards of smoking and of how to obtain support for quitting. • The relative lack of restrictions on smoking at the workplace for many blue-collar workers could reduce incentives to quit (compared to those in higher occupational classes). • Lower rates of smoking cessation in low-SES groups might be related to a greater burden of psychosocial stress and the effect of others smoking. Advertising by the tobacco industry is probably an important determinant of smoking trends (Stellman and Resnicow 1997), particularly the aggressive marketing in developing countries (Connolly 1992; Mackay 1992). 4.2
Hazardous alcohol use
LITERATURE ON DEVELOPING COUNTRIES
Hazardous alcohol use associated with low SES In China, one cohort study of urban male workers reported that high levels of daily alcohol consumption were significantly more prevalent among men with low educational attainment (Siegrist et al. 1990). Similarly, a Brazilian study of urban residents (using a logistic regression model) found that both heavy drinking and alcohol dependence were associated with low educational attainment and low income (Moreira et al. 1996). Another study in a Brazilian city found that low social class was significantly associated with excessive alcohol consumption in both men and women (Duncan et al. 1993). A study in a Caribbean island reported that men and women of low SES (based on educational attainment) drank more alcohol per week (Grol et al. 1997b). However, high-SES women were more likely to be regular drinkers than low-SES women. Of the two African studies identified, one in Ethiopia found that problem drinking (based on the CAGE screening instrument) was associated with a lower level of educational attainment among residents of Addis Ababa (Kebede and Alem 1999). Similarly, in a Nigerian study, alcohol consumption was higher in those with low SES (Bunker et al. 1992). In Brazil, the results of one study suggested that higher rates of oesophageal cancer among lower-SES groups are likely to be attributable to the higher use of sugar-cane spirit, black tobacco and mate in these groups (Bouchardy et al. 1993). In contrast, data from Colombia indicate that the incidence of alcohol-related cancers tended to show positive social class gradients (Cuello et al. 1982).
2040
Comparative Quantification of Health Risks
Hazardous alcohol use associated with high SES In an Ethiopian study, problem drinking (based on the CAGE screening instrument) was associated with higher income and education among men or women in a rural district (Alem et al. 1999) (in contrast to another Ethiopian study described above). In Thailand, a study reported that hazardous alcohol use (based on AUDIT scores) was associated with higher income among those seeking emergency treatment (Lapham et al. 1998). By contrast, the study found that those with a university degree were significantly less likely to have positive AUDIT scores. A study in the Republic of Korea found a trend towards greater alcohol consumption with increasing years of education among women, but no such trend among men (Chung et al. 1992). Alcohol intake among high-SES males explained a small but statistically significant part of the relationship between hypertension and high SES in villagers in North India (Singh et al. 1997b). This suggests that at least some members of this population were drinking fairly large quantities of alcohol. No association with SES A study in a metropolitan area in Brazil found no statistically significant differences in alcohol use associated with SES, although there was a trend towards higher rates of drinking among low-SES men (Martins et al. 1995). LITERATURE ON DEVELOPED COUNTRIES
Hazardous alcohol use associated with low SES A major review of alcohol drinking, social class and cancer suggested a very likely role for alcohol drinking in the observed negative social class gradients for alcohol-related cancers in men in France, Italy and New Zealand (Moller and Tonnesen 1997). Evidence that was less strong, but still suggestive of such a role, was also reported for men in Denmark, Switzerland and the United Kingdom. This review identified studies showing associations between hazardous alcohol use and low SES in Denmark (for average number of drinks per week in men and women); Finland (for rates of alcohol intoxication and rates of hospital admission for acute alcohol-related conditions for men and women); France (for prevalence of “heavy drinkers” in men); Sweden (alcoholism and high alcohol consumption in young people); Switzerland (for alcoholism); the United Kingdom (for the proportion of heavy drinkers in manual vs non-manual occupations among men); and the United States (for heavy consumption—five or more drinks on one occasion in men and women). The major review by Colhoun et al. (1998) on hypertension and SES reported that higher alcohol consumption by low-SES men explains part, though not all, of the association between SES and blood pressure in men in developed countries. This collectively considered the findings of
Tony Blakely et al.
2041
studies in Australia, Israel, the Netherlands, Norway, the United Kingdom (two studies) and the United States (two studies). Similarly, the INTERSALT study, involving 52 centres in 18 developed and 12 developing countries, found that adjusting for alcohol (along with BMI, smoking, and sodium and potassium excretion) halved the association between low SES and higher blood pressure in men so that it was no longer statistically significant (Stamler et al. 1992). A combined analysis of population samples from France, Italy, Spain and Switzerland (Pequignot et al. 1988) found that daily alcohol consumption was higher in male manual workers than in male professionals, whereas in women there was no significant variation by occupational group. More recent data from the United Kingdom indicated that 10% of low-SES men (social classes IV and V) were dependent on alcohol compared to 5% of high-SES men (social classes I and II) (Acheson 1998). Similarly, in the United States, heavy alcohol use was reported to be higher among those with poorer education among men and women in virtually all ethnic groups (Pamuk et al. 1998). Data from surveys in 48 American states showed that hazardous consumption of alcohol was significantly more frequent in households with below-median income and in those with a lower educational level (for five or more drinks on one occasion at least once a week during the previous year) (Midanik and Clark 1994). For weekly drinking, however, the pattern was the opposite, being higher in those with above-median income and a higher level of education. In a national survey in New Zealand based on AUDIT scores, lower social class was associated with an increased risk of a hazardous pattern of alcohol use (Howden-Chapman and Tobias 2000). There was some suggestion that lower educational attainment was associated with a more hazardous pattern of alcohol use, but this was not statistically significant. Earlier New Zealand work indicated that the pattern of drinking “high quantities but less often” was more common in the lower social classes, along with an increased prevalence of high daily consumption (Casswell and Gordon 1984). Hazardous alcohol use associated with high SES The review by Moller and Tonnesen (1997) reported a French study indicating that high-SES women were more likely to be regular consumers of alcohol, and that in Sweden high alcohol consumption was associated with high SES among older people. The review by Colhoun et al. (1998) on hypertension and SES reported that higher alcohol consumption among high-SES women had been identified in studies in Australia and England (two studies) and in the INTERSALT study, where some of 52 centres collected data on alcohol. A study in Wales revealed a tendency towards a higher prevalence of consumption of 22 drinks or more per week in the higher social classes (Farrow et al. 1988). A study of three regions in the United Kingdom
2042
Comparative Quantification of Health Risks
reported that weekly alcohol consumption increased with household income in both men and women (Crawford 1988), but that manual workers drank more alcohol than non-manual workers. In the United States, a combined analysis of 10 surveys showed some positive gradient for the frequency of alcohol drinking with SES (Knupfer 1989). However, there was a weakly inverse gradient for “frequent drunkenness”. Among employees in Minneapolis-St Paul, the frequency of alcohol drinking was associated with high SES in women but not in men (Jeffery et al. 1991). A national survey in New Zealand (Statistics New Zealand/Ministry of Health 1993) reported a higher alcohol intake with greater educational attainment, in contrast to the other results from New Zealand detailed above. No association with SES The review by Colhoun et al. (1998) reported on two British studies that showed no SES gradient with alcohol consumption. Similarly, there was no association with social class among a sample of attendees at a London health centre (King 1986). In New Zealand, a child cohort study found that family SES did not have a marked impact on drinking behaviour, but that those from low-SES families in this cohort drank more at age 15 years (Connolly 1992). The international review of alcohol drinking, social class and cancer by Moller and Tonnesen (1997) found no evidence for such an association for Japan, except that women in the highest social classes had slightly elevated mortality from liver cirrhosis. There was a similar finding for Finland and for Sweden, except that pharyngeal cancer in Swedish women was more prevalent in low-SES women. Nevertheless, these two countries do show some evidence for SES gradients in hazardous alcohol use (as detailed above). DISCUSSION
The relationship The general pattern in developed countries is for lower-SES men to have a more hazardous pattern of alcohol consumption in terms of high or excessive intake of alcohol. This pattern is also evident in the distribution of alcohol-related disease, particularly alcohol-related cancers. The pattern for women can resemble that for men, although there are also many studies indicating excessive alcohol use among high-SES women. Different studies in the same countries sometimes show contrasting patterns that may reflect temporal trends, regional variation and different methodological approaches. The pattern is also mixed in developing countries, although in this review more studies were identified that indicated that lower SES was associated with a more hazardous pattern of alcohol consumption. This pattern is similar to that found for tobacco use, but it does contrast with
Tony Blakely et al.
2043
the SES gradient seen for other disease risk factors where high SES is associated with increased risk (e.g. for cholesterol level, obesity, blood pressure and to some extent physical inactivity). Possible mechanisms Some of the factors likely to be involved in the excessive use of alcohol by lower-SES groups include: • cultural influences, including the presence of greater numbers of alcohol users in certain areas and the social acceptability of heavy drinking; • the existence of psychological stressors associated with poverty and inequality from which alcohol users may seek to escape; • access to cheap alcohol, especially in rural areas of the developing world and where alcohol is not taxed; and • the effect of lower educational attainment on poorer knowledge of the hazards of excessive alcohol use. High levels of alcohol use among higher-SES groups (especially women) may reflect in particular cultural patterns and the role of additional income in settings where alcohol is relatively expensive. The alcohol industry may also target advertising at those with the highest incomes. It is also plausible that alcohol is used as a coping mechanism by high-SES women who have heavy demands on their time. Some of the studies in this review also reported higher levels of abstinence among lower-SES groups. This behavioural pattern may have health benefits in those aged
