Sot. Sci. Med. Vol. 34. No. 5, pp. 523-532. 1992 Printed in Great Britain
0277-9536/92$5.00+ 0.00 Pcrgamon Press
plc
COMPETING EXPLANATIONS FOR ASSOCIATIONS BETWEEN MARITAL STATUS AND HEALTH SALLY WYKE and GRAEMEFORD MRC Medical Sociology Unit, 6 Lilybank Gardens, Glasgow G12 8QQ, U.K. Abstract-This paper is based on baseline data from a survey of 1042 fifty-five year olds living in the Central Clydeside Conurbation, who constitute the eldest cohort of the ‘West of Scotland Twenty-07 Study’-a longitudinal study of health and everyday life. The relationship between marital status and a number of measures of health and illness is explored. The paper examines which of four ‘social causation’ explanations-that married people have better health because they have more material resources, less stress, indulge in less risky health behaviour and have more social support-can actually account for the observed patterning. It finds that more risky health behaviour (measured by smoking and drinking), and ‘objective’ levels of social support, cannot account for very much of the effect of marital status on health measures; but that material resources, stress and perceived quality of social support could do so. However, elucidation of the direction of the relationships between these explanations and health measures, and indeed of the effect of health ‘selection’ into and out of marriage must await future sweeps of this longitudinal study. Key words-marital
status, health, social causation
INTRODUCTION
For over a century, abundant evidence has existed linking people’s life expectancy with their marital status. William Farr, Registrar General for England and Wales in the mid-nineteenth century, examined age specific death rates for single, married and widowed women and men in France in 1853. He concluded from the figures that: Marriage is a healthy state. The single individual is more likely to be wrecked on his voyage than the lives joined together in matrimony (quoted in Refs [I, 21).
Subsequent research has echoed Farr’s findings: at any given age, married people have lower death rates than unmarried people; among the unmarried, divorced people have the highest rates, while single (never married) people either have intermediate rates or rates closer to married people [2-91. Information about marital status and morbidity is more scarce, though data from the British General Household Survey [I, 3, IO] and the American National Health Interview Survey [l 1] suggest that self-reported morbidity rates mirror mortality rates, being highest among separated or divorced people. Similar findings are consistently reported for psychiatric morbidity and measures of psychological wellbeing [12-141. Review papers have hypothesized various explanations for this patterning [I 1,3, 11. two models have been suggested: the ‘health selection’ and the ‘social causation’ model. In the-foT6
High M>21)
P < 0.05
68
32
25
33
I6
I4
I2
6
18
25
27
24
30
70
26
I4
I%
34
8
4
I8
26
24
28
32
68
18
32
I8
23
9
I2
26
I2
29
21
P < 0.001
P < 0.05
ns
SALLY WYKE and GRAEME FORD
528
Table 2. Associations between marital status, ‘felt’ distress, and soaal support Females
Mean values for:
M&s
Married (N = 383)
No longer married (N = 144)
NCVU marned (N = 381
2.2 4.5 3.8 5.6
2.8 3.9 3.8 4.6
2.6 3.6 3.1 4.1
‘Felt’ distress/eustress Available support Actual support Quality support
support and quality/intimacy of support. It shows that married people reported lower levels of distress then either the divorced and widowed or the never married, and that they consistently reported highest levels on all three measures of social support. However, it was the never married who reported lowest levels of social support, with the widowed and divorced reporting intermediate levels on most of the measures. Were health measures marital status?
signiJcantly
associated
with
All analyses of associations between health and marital status were performed separately for women and men. We took no a priori decisions about which marital status categories would fare worse on the health measures. Instead, multiple comparison tests (Sheffe [35,36] and Newman-Keuls [36, 37]), which correct the significance values for the number of possible comparisons, were used to identify significant differences between marital status categories. The results of these analyses can be seen in Table 3. The brackets underneath the mean scores on each health measure show which categories were significantly different from which other, whilst the significance levels refer to overall significance (F-test, analysis of variance). Neither systolic blood pressure nor standardised respiratory function showed any significant association with marital status for women. However, no longer married women reported significantly more limiting chronic conditions, worse self-rated
Sik?
Married Cl4 = 393)
No longer married fN = 50)
Never married (N = 341
2.4 5.0 3.6 5.9
3.0 4.4 3.2 4. I
3.0 3.0 3.0 3.8
0.001 0.01 0.E I
Sin 0.00 I 0.001 O.&
health and worse mental health compared to married women. The differences between no longer married and single women, or between single and married women were not significant. Although the difference between marital status categories in the number of symptoms reported in the last month did reach overall significance, the multiple comparison tests showed that no marital status category reported significantly more symptoms than any other. No longer married men had significantly higher systolic blood pressure than married men. It was possible that this association was confounded by a relationship between obesity and marital status, since obesity is correlated with both blood pressure and marital status. We explored this by examining the association between body mass index and marital status, but found no significant differences in body mass index between marital status groups. There was little difference between marital status categories in standardised respiratory function. Although the difference in the number of limiting chronic conditions reported by each category did reach overall significance for men, the multiple comparison tests showed that no one category reported significantly more than any other. The number of symptoms reported in the last month was not significantly different between marital status categories, but no longer married men reported significantly worse self-rated health compared to married men, and both no longer married and single men reported significantly worse mental health compared to married men.
Table 3. Associations between health variables and marital status Females All (530) Systolic blood pressure (mmHg) Standardised rap function (FLY,) Number of limiting chronic conditions Number of symptoms in last month Self assessedhealth in last year GHQ score (mental health)
M (359)
(&
Males
NLM (134)
Overall significance*
All (453)
(3?8)
(351)
NLM (44)
Overall significance*
136.2
136.0
141.8
135.2
ns
138.0
137.3 1
135.0
147.0
0.05
-1.1
-1.1
-1.1
- I.3
“S
-
-1.1
-0.9
- I.2
“s
I.1
I
I.2
I.1
1.2
1.4
0.05
0.9
0.9
1.3
I.2
0.05
5.7
5.4
6.2
6.4
0.05
4.3
4.2
4.9
4.8
“s
2.4
2.4
2.3
2.6
0.01
2.4
2.3 d7
2.6
11.1
IO.5
12.4
0.01
10.6
10.2
12.3
Il.4
0.05 12.3
0.01
a
M = Married. S = Single. NLM = No Longer Married. *F-test oneway analysis of variance. L Indicates categories stgnificantly ddkrent from each other (Shefk and Newman-Keuls multiple comparison tests).
Marital status and health Which explanations can account for these associations and which cannot?
included in these tables to facilitate clear presentation of the other results. As already stated, Table 4 compares no longer married with married women. It shows that car ownership, felt distress/eustress and quality of social support consistently explain quite large proportions of the differences in means between married and no longer married women on each health measure (e.g. for number of conditions causing limitation, the proportions are respectively 64%, 74% and 52%); marital status no longer has any significant effect after control by any of these three explanatory variables. An exception to this pattern is quality of social support, which explains only 21% of the difference in means for self-rated health in the last 12 months, and marital status retains a significant main effect after control. On the other hand, smoking, drinking and available or actual social support accounted for much smaller proportions of the differences in means between married and no longer married women on each health measure (e.g. on number of conditions causing limitation, the proportions are l%, 1% and 0% respectively); marital status still had a significant main effect on the health measures after control by these variables. Table 5 compares no longer married with married men. It can be seen that none of the explanatory variables could explain the differences between married and no longer married men in mean systolic blood pressure. However, for the self-reported measures a similar pattern emerged to that in the women: car ownership, felt distress/eustress and quality of social support accounted for quite large proportions in the differences in means, and marital status did not have a significant main effect on health after control (apart from car ownership on the relationship between marital status and GHQ); but smoking, drinking and actual social support explained much smaller proportions, and marital status retained its significant main effect on health. Finally, Table 6 considers the proportion of the difference in mean GHQ score between married and single men that can be explained by each of the explanatory variables. The proportions are somewhat smaller than when married were compared to no longer married men or women, although the explanatory variables nearly fall into the same two groups: felt distress/eustress and quality of social support account for the largest proportions of the original
The main purpose of this paper is to explore which of the four ‘social causation’ explanations can explain the statistically significant associations between marital status and some health measures, and which can not. The four explanations were measured with seven explanatory variables (see methods section), four of which were continuous (perceived distress/eustress, structural/available support, actual/companionship support and quality/intimacy of support) and three of which were categorical (car ownership, smoking and drinking behaviour). The analysis presented here uses analysis of covariance (for the continuous explanatory variables) or hierarchical analysis of variance (for the categorical explanatory variables) to identify those explanations which do remove the main effect of marital status on health and those which do not. It goes on to compare the proportion of the original effect of marital status on the health measures that each explanatory variable can account for. The analysis does not try to explain all of the variance in the health measures-to do that would require use of many more of the potential explanatory variables available in this wide-ranging data set. The results of these analyses are shown in Tables 4 to 6. The tables compare only marital status categories which were significantly different from one another according to the multiple comparison tests described in the previous section. The figures in these tables represent the proportions of the differences in mean health measures between marital status categories that are explained by each of the explanatory variables separately. Asterisks denote differences which remain statistically significant after controlling for the relevant explanatory variable. Thus for women (Table 4), car ownership explains 64% of the difference in mean number of limiting conditions between being married and no longer married. The remaining differences between marital status categories is no longer statistically significant. In contrast, smoking behaviour explains only 1% of the difference in means, the remaining difference, as was the initial difference, being statistically significant. Since the results for these analyses are almost identical for available and for actual social supportthey have the same effect on the relationship between marital status and health measures-the former is not Table
4. The
effect
of
being
married
proportion
Health
variables
Number
of conditions
causing
score
*Marital **Marital
to no longer
is explained
being
married
by the following
on
WOMEN’S
health:
what
variables
Car ownership
distres?!eustress
Smoking
Drinking
0.64
0.74
0.01
0.01.
0.00-
0.52
0.84
0.80
0.25.
0.10**
o.oo**
0.21’
0.21.
o.oo-
o.oo**
0.55
‘Felt’
l
Actual
Quality
support
support
limitation
Self rated health in last 12 months GHQ
in comparison of the cl%3
529
0.50
status has statistically status
has statistically
significant significant
0.96 effect AFTER effect AFTER
control control
P < 0.05. P < 0.01.
530
SALLY WYKE and GRAEME FORD Table
5. The
effect of being married
compared
to no longer
the effect is explained Car Health
variables
Systolic
blood
Self-rated
pressure
distrcss/custress
0.04**
in
on MEN’S
health:
0.51
I 0.68
0.21.
0.47
what
proportion
of
variables
‘Felt’
ownership
health
being married
by the following
Smoking
Drinking
0.0’.
0.01
0.2.
0.00**
l*
Actual
Quality
support
support
o.oo**
0.02**
0.00**
0.38
I2 months GHQ
score
/ No
proportions
*Marital
status
**Marital
were calculated has statistically
status
due to interactions
significant
has statistically
eff&t
significant
AFTER
DISCUSSION
Information about marital status is routinely collected in health surveys, although not so routinely analyzed or reported. This paper has explored associations between both direct measures and selfreported measures of health and marital status in a sample of fifty-five year olds living in the West of Scotland. Standardized respiratory function did not show any significant association with marital status for women or men, and systolic blood pressure showed significant association only for men (no longer married men had significantly higher systolic blood pressure than married men, but not single men). However, almost all self-reported measures of health and illness did show significant associations with marital status. Significant differences between marital status categories were similar to those found in mortality data. No-longer-married men and women fared worst on almost every health measure compared to married men and women. Single women reported similar levels of ill health to married women, but the differences between them and no longer married women did not reach statistical significance. However, single men reported worse health compared to married men, and this difference reached statistical significance for mental health. The paper went on to explore which of four ‘social causation’ explanations-that married people have more material resources, fewer risky health behaviours, less stress, and more social support than non-married people-could best account for these associations. Since only baseline, cross-sectional data are currently available, the other explanation, that married people are ‘selected’ because of their better health status, could not be explored in the present paper. The four explanations were measured 6. The
effect of being married
compared
GHO
variables Score
*Marital **Marital
status status
Car
ownershiti 0.14. has statistically has statistically
significant significant
0.03** variables
P < 0.01.
on MEN’S
health:
0.45
proportion
of the
Actual
Quality
Smokine
Drinkine
sulmort
s”ooOrt
o.oo**
o.oo**
o.oo**
0.66
P < 0.05. control P < 0.01.
control
What
variables
‘Felt‘ distress;eustress
effect AFTER
status.
with seven explanatory variables: car ownership; felt distress/eustress; smoking levels; drinking levels; available/structural social support; actual/ companionship support; and quality/intimacy of support. We explored associations between these variables and found, as expected, that married people had highest levels of car ownership, lowest levels of felt distress, lowest levels of smoking and highest levels of all three measures of social support. However, clear patterns of drinking behaviour were more difficult to discern. In order to ascertain which explanatory variables could, and which could not, account for the relationship between marital status and health, the proportion of the original effect of marital status on health measures accounted for by each explanatory variable was examined. We found that none of the explanatory variables could account for very much of the effect of being married compared to no longer married on systolic blood pressure. Exactly what it is that causes higher blood pressure in no-longer married men is intriguing-it certainly is not smoking, drinking, available material resources for healthy living, or lack of social support. This needs further exploration. However, much of the effect of being married compared to no longer married on self-reported measures of health disappeared after control by either car ownership, perceived distress/eustress or quality/intimacy of social support. On the whole, nolonger married people, and single men, were indeed less likely to have a car (and presumably fewer resources of any kind), were more likely to feel distressed and were less likely to feel they had a high quality/intimacy of social support, and these things each accounted for quite large proportions of the original association between marital status and health measures. The direction of the relationships between marital status, car ownership, distress/eustress, quality of support and the health measures cannot be ascertained with the present cross-sectional data. Income is known to drop with separation and divorce: low income could cause poor health through
bv the followine
effect AFTER
0.63
I
and marital
P < 0.05.
control
to single (never married)
effect is exolained
Health
explanatory
control
effect AFTER
effect of marital status on GHQ, but this time car ownership as well as smoking, drinking and actual social support explain little of the difference between the means.
Table
0.05**
between
531
Marital status and health
lack of access to the necessities for ‘healthy living’, or through the stressful effect of worry over inadequate income; but equally illness could cause a drop in income. The same is true with the other explanatory variables: both distress/eustress and quality of support obviously have psychological dimensions. No longer married people could either have high levels of distress or feel that their support is poor (that is they are generally fed up!), and this could cause either more illness or cause them to fell ill and report more illness. However, an equally plausible scenario is that being (or feeling) ill causes more distress. Some of these relationships will become clearer as longitudinal data become available. The principal finding of this paper is that material resources (measured by car ownership), felt distress/eustress and quality/intimacy of support could individually account for the effect of being married compared to no longer married on health measures; whereas smoking, drinking and the more ‘objective’ measures of social support-available/ structural support and actual/companionship support-could not. Thus our data support the hypotheses suggested in the literature that marital status is associated with health through higher levels of material resources and lower levels of stress amongst married people, and that the reported quality/intimacy of their social support may also be an important factor. However, they are less supportive of the suggestions that marital status is associated with health through higher levels of risky health behaviours or through higher levels of available or actual social support amongst married people. However, health behaviours are not just smoking and drinking and it is, of course, possible that other types of health promoting behaviour not considered here, such as leading an orderly lifestyle, eating regular meals and getting regular sleep could contribute to the better health experienced by married people. Acknowledgemenrs-The
study on which this paper was based is funded by the U.K. Medical Research Council. We would like to thank the respondents and interviewers for participating in this study. We are very grateful to Russell Ecob for statistical advice, and would like to thank Kate Hunt, Sally Macintyre and Steve Platt and two anonymous referees for helpful comments on an earlier version of this paper. REFERENCES
I. Macintyre S. Marriage is good for your health, or is it? Proc. ROY. Phil. Sot. Glas. New Series Number Three. 1986.
_
2. Koskenvuo M., Kaprio J., Kesaniemi A. and Sarna S. Differences in mortality from ischemic heart disease by marital status and social class. J. Chron. Dis. 33.95-106, 1980. 3. Morgan M. Marital status, health, illness and service use. Sot. Sci. Med. 14, 633-643, 1980. 4. Fox A. J. and Goldblatt P. Longitudinal Study: Martaliry 1971-5. HMSO, London, 1982. 5. Riessman C. K. and Gerstel N. Marital dissolution and health: do males or females have greater risk? Sot. Sci. Med. 20, 627435, 1985.
6. Koskenvuo M., Kaprio J., Lonnqvist J. and Sama S. Social factors and the gender difference in mortality. Sot. Sci Med. 23, 605-609, 1986. 7. Kotler P. and Wingard D. L. The effect of occupational, marital and parental roles on mortality: the Alameda County study. Am. J. Pub1 Hlth. 79, 607612, 1989. 8. Rosengren A., Wedel H. and Wilhemsen L. Marital status and mortality in middle-aged Swedish men. Am. J. Epidemiol. 129, 54-&l. 1989. 9. Trovato F. and Lauris G. Marital status and mortality in Canada. J. Marr. Fam. 51, 907-922, 1989. 10. Arber S. Class, paid employment and family roles: making sense of structural disadvantage, gender and health status. Sot. Sci. Med. 32. 425-436, 1991. II. Verbrugge, L. M. Marital status and health. J. Marr. Fam. 41, 267-285,
1979.
12. Warheit G. J., Holzer C. E., Bell R. R. A. and Arey S. A. Sex, marital status and mental health: a reappraisal. Sot. Force. 55, 459-470, 1976. 13. Roberts C. R., Roberts R. E. and Stevenson J. M. Women, work, social support and psychiatric morbidity. Sot. Psychiaf. 17, 167-173, 1982. 14. Gove W. R., Hughes M. and Briggs S. C. Does marriage have positive effects on the psychological well-being of the individual? J. Hlth Sot. Behau. 24, 122-131, 1983. 15. Brown J. S. and Giesy B. Marital status of persons with spinal cord injury. Sot. Sci. Med. 23, 313-322, 1986.
16. Bieliauskas L. A. Srress and its Relationship to Health and Illness. Westview Press, Boulder, Colorado, 1982. 17. Pearlin I. L. The sociological study of stress. J. Hlfh Sot. Behav. 30, 241-256,
1989.
18. McCubbin H., Joy C. B., Cauble A. E., Comeau J. K., Patterson J. M. and Needle R. H. Family Stress and Coping: a decade review. J. Marr. Fam. 42, 855-871, 1980. 19. Berkman L. F. and Breslow L. Health and Ways of Living: The Alameda Counry Study. Oxford University Press, New York, 1983. 20. Umberson D. Family status and health behaviours: Social control as a dimension of social integration. J. Hlrh Sot. Behav. 28, 306319,
1987.
21. Broadhead W. E., Kaplan B. H., James S. A., Wagner E. H., Schoenbach V. J.. Grimson R.. Hevden S.. Tibblin G. and Gehlbach S. H. The epidemiologic evidence for a relationship between social support and health. Am. J. Epidemiol. 117, 521-537, 1983. 22. Berkman L. F. and Syme S. L. Social networks, host resistance and mortality: A nine-year follow-up study of Alameda County residents. Am. J. Epidemiol. 109, 186-204,
1979.
23. Blazer D. G. Social support and mortality in an elderly community population. Am. J. Epidemiol. 115,684-694, 1982. 24. House J. S., Robbins C. and Metzner H. L. The association of social relationships and activities with mortality: Prospective evidence from the Tecumseh Community Health Study. Am. J. Epidemiol. 116, 123-140, 1982.
25. Aneshensel C. S. and Frerichs R. R. Stress, support and depression: A longitudinal causal model. J. Commun. Psychol. 10, 363-376, 1982. 26. Cohen S. and Wills T. A. Stress, Social Support and the Buffering Hypothesis. Psycho/. BUN. 98.310-357, 1985. 27. Anson 0. Marital status and women’s health revisited: the importance of a proximate adult. J. Marr. Fam. 51, 185-194, 1989. 28. Macintyre S. Wesr of Scotland Twenty-07Study: Background and Rafionale. MRC Medical Sociology Unit Working Paper No. 7, Glasgow, 1987. 29. Black R. Instead of the 1986 Census: The potential contribution of enhanced registers (with Discussion). J. Roy. Stat. Sot. A, 148, 287-316, 1985.
532
SALLY WYKE and GRAEMEFORD
30. Ecob R. An investigation of weighting schemes for compensation for non-response in the Twenty-07 baseline surveys. MRC Medical Sociology Unit Working Paper No. 20, Glasgow, 1990. 31. Knudson R. J., Slatin R. C., Lebowitz M. D. and Burrows B. The maximal expiratory flow-volume curve. Am. Rev. Resp. Dis. 713, 587, 1976. 32. Goldberg D. Manual of the General Health Questionnaire. NFER-Nelson, Windsor, 1978. 33. Andrews F. M. and Withey S. B. Social Indicators of
34. 35. 36.
37.
Well-Being: American Perceptions of Ltfe Quality. Plenum Press, New York, 1976. Marsh C. Exploring Data. Polity Press, Cambridge, 1988. SheBe H. The Analysis of Variance. Wiley, New York, 1959. Armitage P. and Berry G. Statistical Methodc in Medical Research 2nd edn. Blackwell Scientific Publication, Oxford, 1987. Miller R. G. Jr Simultaneous Statistical Inference 2nd edn. Springer, New York, 1981.
0277-9536/92$5.00+ 0.00 Pcrgamon Press
plc
COMPETING EXPLANATIONS FOR ASSOCIATIONS BETWEEN MARITAL STATUS AND HEALTH SALLY WYKE and GRAEMEFORD MRC Medical Sociology Unit, 6 Lilybank Gardens, Glasgow G12 8QQ, U.K. Abstract-This paper is based on baseline data from a survey of 1042 fifty-five year olds living in the Central Clydeside Conurbation, who constitute the eldest cohort of the ‘West of Scotland Twenty-07 Study’-a longitudinal study of health and everyday life. The relationship between marital status and a number of measures of health and illness is explored. The paper examines which of four ‘social causation’ explanations-that married people have better health because they have more material resources, less stress, indulge in less risky health behaviour and have more social support-can actually account for the observed patterning. It finds that more risky health behaviour (measured by smoking and drinking), and ‘objective’ levels of social support, cannot account for very much of the effect of marital status on health measures; but that material resources, stress and perceived quality of social support could do so. However, elucidation of the direction of the relationships between these explanations and health measures, and indeed of the effect of health ‘selection’ into and out of marriage must await future sweeps of this longitudinal study. Key words-marital
status, health, social causation
INTRODUCTION
For over a century, abundant evidence has existed linking people’s life expectancy with their marital status. William Farr, Registrar General for England and Wales in the mid-nineteenth century, examined age specific death rates for single, married and widowed women and men in France in 1853. He concluded from the figures that: Marriage is a healthy state. The single individual is more likely to be wrecked on his voyage than the lives joined together in matrimony (quoted in Refs [I, 21).
Subsequent research has echoed Farr’s findings: at any given age, married people have lower death rates than unmarried people; among the unmarried, divorced people have the highest rates, while single (never married) people either have intermediate rates or rates closer to married people [2-91. Information about marital status and morbidity is more scarce, though data from the British General Household Survey [I, 3, IO] and the American National Health Interview Survey [l 1] suggest that self-reported morbidity rates mirror mortality rates, being highest among separated or divorced people. Similar findings are consistently reported for psychiatric morbidity and measures of psychological wellbeing [12-141. Review papers have hypothesized various explanations for this patterning [I 1,3, 11. two models have been suggested: the ‘health selection’ and the ‘social causation’ model. In the-foT6
High M>21)
P < 0.05
68
32
25
33
I6
I4
I2
6
18
25
27
24
30
70
26
I4
I%
34
8
4
I8
26
24
28
32
68
18
32
I8
23
9
I2
26
I2
29
21
P < 0.001
P < 0.05
ns
SALLY WYKE and GRAEME FORD
528
Table 2. Associations between marital status, ‘felt’ distress, and soaal support Females
Mean values for:
M&s
Married (N = 383)
No longer married (N = 144)
NCVU marned (N = 381
2.2 4.5 3.8 5.6
2.8 3.9 3.8 4.6
2.6 3.6 3.1 4.1
‘Felt’ distress/eustress Available support Actual support Quality support
support and quality/intimacy of support. It shows that married people reported lower levels of distress then either the divorced and widowed or the never married, and that they consistently reported highest levels on all three measures of social support. However, it was the never married who reported lowest levels of social support, with the widowed and divorced reporting intermediate levels on most of the measures. Were health measures marital status?
signiJcantly
associated
with
All analyses of associations between health and marital status were performed separately for women and men. We took no a priori decisions about which marital status categories would fare worse on the health measures. Instead, multiple comparison tests (Sheffe [35,36] and Newman-Keuls [36, 37]), which correct the significance values for the number of possible comparisons, were used to identify significant differences between marital status categories. The results of these analyses can be seen in Table 3. The brackets underneath the mean scores on each health measure show which categories were significantly different from which other, whilst the significance levels refer to overall significance (F-test, analysis of variance). Neither systolic blood pressure nor standardised respiratory function showed any significant association with marital status for women. However, no longer married women reported significantly more limiting chronic conditions, worse self-rated
Sik?
Married Cl4 = 393)
No longer married fN = 50)
Never married (N = 341
2.4 5.0 3.6 5.9
3.0 4.4 3.2 4. I
3.0 3.0 3.0 3.8
0.001 0.01 0.E I
Sin 0.00 I 0.001 O.&
health and worse mental health compared to married women. The differences between no longer married and single women, or between single and married women were not significant. Although the difference between marital status categories in the number of symptoms reported in the last month did reach overall significance, the multiple comparison tests showed that no marital status category reported significantly more symptoms than any other. No longer married men had significantly higher systolic blood pressure than married men. It was possible that this association was confounded by a relationship between obesity and marital status, since obesity is correlated with both blood pressure and marital status. We explored this by examining the association between body mass index and marital status, but found no significant differences in body mass index between marital status groups. There was little difference between marital status categories in standardised respiratory function. Although the difference in the number of limiting chronic conditions reported by each category did reach overall significance for men, the multiple comparison tests showed that no one category reported significantly more than any other. The number of symptoms reported in the last month was not significantly different between marital status categories, but no longer married men reported significantly worse self-rated health compared to married men, and both no longer married and single men reported significantly worse mental health compared to married men.
Table 3. Associations between health variables and marital status Females All (530) Systolic blood pressure (mmHg) Standardised rap function (FLY,) Number of limiting chronic conditions Number of symptoms in last month Self assessedhealth in last year GHQ score (mental health)
M (359)
(&
Males
NLM (134)
Overall significance*
All (453)
(3?8)
(351)
NLM (44)
Overall significance*
136.2
136.0
141.8
135.2
ns
138.0
137.3 1
135.0
147.0
0.05
-1.1
-1.1
-1.1
- I.3
“S
-
-1.1
-0.9
- I.2
“s
I.1
I
I.2
I.1
1.2
1.4
0.05
0.9
0.9
1.3
I.2
0.05
5.7
5.4
6.2
6.4
0.05
4.3
4.2
4.9
4.8
“s
2.4
2.4
2.3
2.6
0.01
2.4
2.3 d7
2.6
11.1
IO.5
12.4
0.01
10.6
10.2
12.3
Il.4
0.05 12.3
0.01
a
M = Married. S = Single. NLM = No Longer Married. *F-test oneway analysis of variance. L Indicates categories stgnificantly ddkrent from each other (Shefk and Newman-Keuls multiple comparison tests).
Marital status and health Which explanations can account for these associations and which cannot?
included in these tables to facilitate clear presentation of the other results. As already stated, Table 4 compares no longer married with married women. It shows that car ownership, felt distress/eustress and quality of social support consistently explain quite large proportions of the differences in means between married and no longer married women on each health measure (e.g. for number of conditions causing limitation, the proportions are respectively 64%, 74% and 52%); marital status no longer has any significant effect after control by any of these three explanatory variables. An exception to this pattern is quality of social support, which explains only 21% of the difference in means for self-rated health in the last 12 months, and marital status retains a significant main effect after control. On the other hand, smoking, drinking and available or actual social support accounted for much smaller proportions of the differences in means between married and no longer married women on each health measure (e.g. on number of conditions causing limitation, the proportions are l%, 1% and 0% respectively); marital status still had a significant main effect on the health measures after control by these variables. Table 5 compares no longer married with married men. It can be seen that none of the explanatory variables could explain the differences between married and no longer married men in mean systolic blood pressure. However, for the self-reported measures a similar pattern emerged to that in the women: car ownership, felt distress/eustress and quality of social support accounted for quite large proportions in the differences in means, and marital status did not have a significant main effect on health after control (apart from car ownership on the relationship between marital status and GHQ); but smoking, drinking and actual social support explained much smaller proportions, and marital status retained its significant main effect on health. Finally, Table 6 considers the proportion of the difference in mean GHQ score between married and single men that can be explained by each of the explanatory variables. The proportions are somewhat smaller than when married were compared to no longer married men or women, although the explanatory variables nearly fall into the same two groups: felt distress/eustress and quality of social support account for the largest proportions of the original
The main purpose of this paper is to explore which of the four ‘social causation’ explanations can explain the statistically significant associations between marital status and some health measures, and which can not. The four explanations were measured with seven explanatory variables (see methods section), four of which were continuous (perceived distress/eustress, structural/available support, actual/companionship support and quality/intimacy of support) and three of which were categorical (car ownership, smoking and drinking behaviour). The analysis presented here uses analysis of covariance (for the continuous explanatory variables) or hierarchical analysis of variance (for the categorical explanatory variables) to identify those explanations which do remove the main effect of marital status on health and those which do not. It goes on to compare the proportion of the original effect of marital status on the health measures that each explanatory variable can account for. The analysis does not try to explain all of the variance in the health measures-to do that would require use of many more of the potential explanatory variables available in this wide-ranging data set. The results of these analyses are shown in Tables 4 to 6. The tables compare only marital status categories which were significantly different from one another according to the multiple comparison tests described in the previous section. The figures in these tables represent the proportions of the differences in mean health measures between marital status categories that are explained by each of the explanatory variables separately. Asterisks denote differences which remain statistically significant after controlling for the relevant explanatory variable. Thus for women (Table 4), car ownership explains 64% of the difference in mean number of limiting conditions between being married and no longer married. The remaining differences between marital status categories is no longer statistically significant. In contrast, smoking behaviour explains only 1% of the difference in means, the remaining difference, as was the initial difference, being statistically significant. Since the results for these analyses are almost identical for available and for actual social supportthey have the same effect on the relationship between marital status and health measures-the former is not Table
4. The
effect
of
being
married
proportion
Health
variables
Number
of conditions
causing
score
*Marital **Marital
to no longer
is explained
being
married
by the following
on
WOMEN’S
health:
what
variables
Car ownership
distres?!eustress
Smoking
Drinking
0.64
0.74
0.01
0.01.
0.00-
0.52
0.84
0.80
0.25.
0.10**
o.oo**
0.21’
0.21.
o.oo-
o.oo**
0.55
‘Felt’
l
Actual
Quality
support
support
limitation
Self rated health in last 12 months GHQ
in comparison of the cl%3
529
0.50
status has statistically status
has statistically
significant significant
0.96 effect AFTER effect AFTER
control control
P < 0.05. P < 0.01.
530
SALLY WYKE and GRAEME FORD Table
5. The
effect of being married
compared
to no longer
the effect is explained Car Health
variables
Systolic
blood
Self-rated
pressure
distrcss/custress
0.04**
in
on MEN’S
health:
0.51
I 0.68
0.21.
0.47
what
proportion
of
variables
‘Felt’
ownership
health
being married
by the following
Smoking
Drinking
0.0’.
0.01
0.2.
0.00**
l*
Actual
Quality
support
support
o.oo**
0.02**
0.00**
0.38
I2 months GHQ
score
/ No
proportions
*Marital
status
**Marital
were calculated has statistically
status
due to interactions
significant
has statistically
eff&t
significant
AFTER
DISCUSSION
Information about marital status is routinely collected in health surveys, although not so routinely analyzed or reported. This paper has explored associations between both direct measures and selfreported measures of health and marital status in a sample of fifty-five year olds living in the West of Scotland. Standardized respiratory function did not show any significant association with marital status for women or men, and systolic blood pressure showed significant association only for men (no longer married men had significantly higher systolic blood pressure than married men, but not single men). However, almost all self-reported measures of health and illness did show significant associations with marital status. Significant differences between marital status categories were similar to those found in mortality data. No-longer-married men and women fared worst on almost every health measure compared to married men and women. Single women reported similar levels of ill health to married women, but the differences between them and no longer married women did not reach statistical significance. However, single men reported worse health compared to married men, and this difference reached statistical significance for mental health. The paper went on to explore which of four ‘social causation’ explanations-that married people have more material resources, fewer risky health behaviours, less stress, and more social support than non-married people-could best account for these associations. Since only baseline, cross-sectional data are currently available, the other explanation, that married people are ‘selected’ because of their better health status, could not be explored in the present paper. The four explanations were measured 6. The
effect of being married
compared
GHO
variables Score
*Marital **Marital
status status
Car
ownershiti 0.14. has statistically has statistically
significant significant
0.03** variables
P < 0.01.
on MEN’S
health:
0.45
proportion
of the
Actual
Quality
Smokine
Drinkine
sulmort
s”ooOrt
o.oo**
o.oo**
o.oo**
0.66
P < 0.05. control P < 0.01.
control
What
variables
‘Felt‘ distress;eustress
effect AFTER
status.
with seven explanatory variables: car ownership; felt distress/eustress; smoking levels; drinking levels; available/structural social support; actual/ companionship support; and quality/intimacy of support. We explored associations between these variables and found, as expected, that married people had highest levels of car ownership, lowest levels of felt distress, lowest levels of smoking and highest levels of all three measures of social support. However, clear patterns of drinking behaviour were more difficult to discern. In order to ascertain which explanatory variables could, and which could not, account for the relationship between marital status and health, the proportion of the original effect of marital status on health measures accounted for by each explanatory variable was examined. We found that none of the explanatory variables could account for very much of the effect of being married compared to no longer married on systolic blood pressure. Exactly what it is that causes higher blood pressure in no-longer married men is intriguing-it certainly is not smoking, drinking, available material resources for healthy living, or lack of social support. This needs further exploration. However, much of the effect of being married compared to no longer married on self-reported measures of health disappeared after control by either car ownership, perceived distress/eustress or quality/intimacy of social support. On the whole, nolonger married people, and single men, were indeed less likely to have a car (and presumably fewer resources of any kind), were more likely to feel distressed and were less likely to feel they had a high quality/intimacy of social support, and these things each accounted for quite large proportions of the original association between marital status and health measures. The direction of the relationships between marital status, car ownership, distress/eustress, quality of support and the health measures cannot be ascertained with the present cross-sectional data. Income is known to drop with separation and divorce: low income could cause poor health through
bv the followine
effect AFTER
0.63
I
and marital
P < 0.05.
control
to single (never married)
effect is exolained
Health
explanatory
control
effect AFTER
effect of marital status on GHQ, but this time car ownership as well as smoking, drinking and actual social support explain little of the difference between the means.
Table
0.05**
between
531
Marital status and health
lack of access to the necessities for ‘healthy living’, or through the stressful effect of worry over inadequate income; but equally illness could cause a drop in income. The same is true with the other explanatory variables: both distress/eustress and quality of support obviously have psychological dimensions. No longer married people could either have high levels of distress or feel that their support is poor (that is they are generally fed up!), and this could cause either more illness or cause them to fell ill and report more illness. However, an equally plausible scenario is that being (or feeling) ill causes more distress. Some of these relationships will become clearer as longitudinal data become available. The principal finding of this paper is that material resources (measured by car ownership), felt distress/eustress and quality/intimacy of support could individually account for the effect of being married compared to no longer married on health measures; whereas smoking, drinking and the more ‘objective’ measures of social support-available/ structural support and actual/companionship support-could not. Thus our data support the hypotheses suggested in the literature that marital status is associated with health through higher levels of material resources and lower levels of stress amongst married people, and that the reported quality/intimacy of their social support may also be an important factor. However, they are less supportive of the suggestions that marital status is associated with health through higher levels of risky health behaviours or through higher levels of available or actual social support amongst married people. However, health behaviours are not just smoking and drinking and it is, of course, possible that other types of health promoting behaviour not considered here, such as leading an orderly lifestyle, eating regular meals and getting regular sleep could contribute to the better health experienced by married people. Acknowledgemenrs-The
study on which this paper was based is funded by the U.K. Medical Research Council. We would like to thank the respondents and interviewers for participating in this study. We are very grateful to Russell Ecob for statistical advice, and would like to thank Kate Hunt, Sally Macintyre and Steve Platt and two anonymous referees for helpful comments on an earlier version of this paper. REFERENCES
I. Macintyre S. Marriage is good for your health, or is it? Proc. ROY. Phil. Sot. Glas. New Series Number Three. 1986.
_
2. Koskenvuo M., Kaprio J., Kesaniemi A. and Sarna S. Differences in mortality from ischemic heart disease by marital status and social class. J. Chron. Dis. 33.95-106, 1980. 3. Morgan M. Marital status, health, illness and service use. Sot. Sci. Med. 14, 633-643, 1980. 4. Fox A. J. and Goldblatt P. Longitudinal Study: Martaliry 1971-5. HMSO, London, 1982. 5. Riessman C. K. and Gerstel N. Marital dissolution and health: do males or females have greater risk? Sot. Sci. Med. 20, 627435, 1985.
6. Koskenvuo M., Kaprio J., Lonnqvist J. and Sama S. Social factors and the gender difference in mortality. Sot. Sci Med. 23, 605-609, 1986. 7. Kotler P. and Wingard D. L. The effect of occupational, marital and parental roles on mortality: the Alameda County study. Am. J. Pub1 Hlth. 79, 607612, 1989. 8. Rosengren A., Wedel H. and Wilhemsen L. Marital status and mortality in middle-aged Swedish men. Am. J. Epidemiol. 129, 54-&l. 1989. 9. Trovato F. and Lauris G. Marital status and mortality in Canada. J. Marr. Fam. 51, 907-922, 1989. 10. Arber S. Class, paid employment and family roles: making sense of structural disadvantage, gender and health status. Sot. Sci. Med. 32. 425-436, 1991. II. Verbrugge, L. M. Marital status and health. J. Marr. Fam. 41, 267-285,
1979.
12. Warheit G. J., Holzer C. E., Bell R. R. A. and Arey S. A. Sex, marital status and mental health: a reappraisal. Sot. Force. 55, 459-470, 1976. 13. Roberts C. R., Roberts R. E. and Stevenson J. M. Women, work, social support and psychiatric morbidity. Sot. Psychiaf. 17, 167-173, 1982. 14. Gove W. R., Hughes M. and Briggs S. C. Does marriage have positive effects on the psychological well-being of the individual? J. Hlth Sot. Behau. 24, 122-131, 1983. 15. Brown J. S. and Giesy B. Marital status of persons with spinal cord injury. Sot. Sci. Med. 23, 313-322, 1986.
16. Bieliauskas L. A. Srress and its Relationship to Health and Illness. Westview Press, Boulder, Colorado, 1982. 17. Pearlin I. L. The sociological study of stress. J. Hlfh Sot. Behav. 30, 241-256,
1989.
18. McCubbin H., Joy C. B., Cauble A. E., Comeau J. K., Patterson J. M. and Needle R. H. Family Stress and Coping: a decade review. J. Marr. Fam. 42, 855-871, 1980. 19. Berkman L. F. and Breslow L. Health and Ways of Living: The Alameda Counry Study. Oxford University Press, New York, 1983. 20. Umberson D. Family status and health behaviours: Social control as a dimension of social integration. J. Hlrh Sot. Behav. 28, 306319,
1987.
21. Broadhead W. E., Kaplan B. H., James S. A., Wagner E. H., Schoenbach V. J.. Grimson R.. Hevden S.. Tibblin G. and Gehlbach S. H. The epidemiologic evidence for a relationship between social support and health. Am. J. Epidemiol. 117, 521-537, 1983. 22. Berkman L. F. and Syme S. L. Social networks, host resistance and mortality: A nine-year follow-up study of Alameda County residents. Am. J. Epidemiol. 109, 186-204,
1979.
23. Blazer D. G. Social support and mortality in an elderly community population. Am. J. Epidemiol. 115,684-694, 1982. 24. House J. S., Robbins C. and Metzner H. L. The association of social relationships and activities with mortality: Prospective evidence from the Tecumseh Community Health Study. Am. J. Epidemiol. 116, 123-140, 1982.
25. Aneshensel C. S. and Frerichs R. R. Stress, support and depression: A longitudinal causal model. J. Commun. Psychol. 10, 363-376, 1982. 26. Cohen S. and Wills T. A. Stress, Social Support and the Buffering Hypothesis. Psycho/. BUN. 98.310-357, 1985. 27. Anson 0. Marital status and women’s health revisited: the importance of a proximate adult. J. Marr. Fam. 51, 185-194, 1989. 28. Macintyre S. Wesr of Scotland Twenty-07Study: Background and Rafionale. MRC Medical Sociology Unit Working Paper No. 7, Glasgow, 1987. 29. Black R. Instead of the 1986 Census: The potential contribution of enhanced registers (with Discussion). J. Roy. Stat. Sot. A, 148, 287-316, 1985.
532
SALLY WYKE and GRAEMEFORD
30. Ecob R. An investigation of weighting schemes for compensation for non-response in the Twenty-07 baseline surveys. MRC Medical Sociology Unit Working Paper No. 20, Glasgow, 1990. 31. Knudson R. J., Slatin R. C., Lebowitz M. D. and Burrows B. The maximal expiratory flow-volume curve. Am. Rev. Resp. Dis. 713, 587, 1976. 32. Goldberg D. Manual of the General Health Questionnaire. NFER-Nelson, Windsor, 1978. 33. Andrews F. M. and Withey S. B. Social Indicators of
34. 35. 36.
37.
Well-Being: American Perceptions of Ltfe Quality. Plenum Press, New York, 1976. Marsh C. Exploring Data. Polity Press, Cambridge, 1988. SheBe H. The Analysis of Variance. Wiley, New York, 1959. Armitage P. and Berry G. Statistical Methodc in Medical Research 2nd edn. Blackwell Scientific Publication, Oxford, 1987. Miller R. G. Jr Simultaneous Statistical Inference 2nd edn. Springer, New York, 1981.
