(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
Aging Clinical and Experimental Research
Association of health behavior and social role with total mortality among Japanese elders in Okinawa, Japan Aya Goto1, Seiji Yasumura1, Yuko Nishise2, and Seizo Sakihara3 1Department of Public Health, Fukushima Medical University School of Medicine, Fukushima, 2Department of Public Health, Yamagata University School of Medicine, Yamagata, 3Department of Social Welfare, Okinawa International University, Okinawa, Japan
ABSTRACT. Background and aims: Longitudinal studies on the relationship of life-style factors and social networking with mortality have been only slightly conducted in a comprehensive manner among the Japanese elderly. This study aims at examining associations of health behavior and social role with total mortality among Japanese elders in Okinawa. Methods: All residents aged 65 years or older were recruited in 1987 from the village of Ohgimi, and 724 subjects were followed for 12 years. A comprehensive health assessment questionnaire was used, and the resulting data were analyzed by the Cox proportional hazards model. Results: The number of deaths during the study period was 139 (55.4%) for men and 166 (35.1%) for women. The standardized mortality rates of malignancy, cerebrovascular disease, heart disease and pneumonia were higher for men. In the final model of multivariate analysis for women, less than 6 hours of sleep [Hazard ratio (HR)= 2.62] and lower social role (HR=1.88) significantly increased the risk of mortality. As for men, neither health behavioral items nor social role showed a significant association with mortality. Conclusions: Our study results suggest that having sufficient hours of sleep and enhancing social role could greatly help elderly women in Okinawa to maintain good health. Moreover, our study provides direction for future research to explore the difference in adequate health promotion measures between women and men. (Aging Clin Exp Res 2003; 15: 443-450) ©2003,
Editrice Kurtis
INTRODUCTION The elderly in Japan are growing at a faster rate than ever. The proportion of the elderly aged 65 years or old-
er increased from 10.3% in 1985 to 17.4% in 2000 (1). The life expectancies at birth for women and men were 83.2 years and 76.7 years in 1995, and the data for Okinawa, one of the “longevity prefectures”, were respectively 85.1 years and 77.2 years in the same year (1). The Japanese government has been engaged in the promotion of national health through the First-Phase Measures for National Health Promotion since 1978 and the Second-Phase Measures since 1988 (2). Major activities of these measures were the establishment of physical check-up and health guidance systems and the development of health promotion infrastructures. In order to face the increasing needs of welfare and medical services for the elderly, the new health plan, Healthy Japan 21, was launched in 2000. The plan focuses more on supporting individual health promotion through multiple social networking channels, and aims at preventing deaths and at extending the period people can live without becoming ill or bedridden. However, longitudinal studies on the relationship of life-style factors and social networking with total mortality have been only slightly conducted in a comprehensive manner among the Japanese elderly. Using 12-year longitudinal data from a communitybased cohort of women and men aged 65 years and older in Okinawa, we examined the association of health behavior and social interaction with total mortality. Aging is a multidimensional progressive process (3), and the importance of comprehensive geriatric assessment in epidemiological studies has been recognized (4). We thus made adjustments by means of various variables such as sociodemographic factors, functional status, and biological factors. The first goals of the health strategies in many countries are to prolong people’s lives (5-7), and we chose total mortality as an objective outcome indicator. It
Key words: Cohort studies, elderly, health behavior, Japan, longevity, mortality, sleep patterns, social role. Correspondence: A. Goto, MD, MPH, PhD, Department of Public Health, Fukushima Medical University School of Medicine, Hikarigaoka 1, Fukushima, 960-1295, Japan. E-mail:
[email protected] Received July 17, 2002; accepted in revised form June 25, 2003.
Aging Clin Exp Res, Vol. 15, No. 6 443
A. Goto, S. Yasumura, Y. Nishise, et al.
is a general health status indicator, which is not just an indication of the rates of deaths per year, but also of general levels of ill health and disease in a community. METHODS All residents aged 65 years or older were recruited in 1987 from the village of Ohgimi in Okinawa, Japan. The baseline survey and health check-up were conducted in the same year, and those who participated in both were followed for 12 years. The methods used in the present study have been reported previously (8, 9). In brief, the cohort was selected from the residence registry, using birthdays of residents as of January 1, 1987. The village of Ohgimi was selected because of less migration among the elderly and a high participation rate in the annual health check-up. Nine hundred and four subjects were selected from the residence registry, but 87 were excluded from the study: 14 had died before the interview or health check-up, 48 had been admitted to hospital, and 27 had moved out of the village. A total of 724 subjects participated in both the health check-up and the baseline survey, and the overall participation rate was 88.6% (724/817). Check-ups were conducted for 6 days in April 1987, and the baseline survey for 8 days in July of the same year. Personal interviews and check-ups were conducted at several community centers in Ohgimi, except for 35 health check-ups and 157 interviews which were carried out at home. The survey team consisted of epidemiologists, psychologists, physicians, public health workers, and university students from the Tokyo Metropolitan Institute of Gerontology, University of Rhyukyu, a local public clinic, and a village office. The questionnaire used in the baseline survey was based on the health status assessment questionnaire of the elderly, developed by the Tokyo Metropolitan Institute of Gerontology and previously used with another elderly cohort in Japan (10). It consisted of 31 items, including basic sociodemographic information, subjective health, past medical history, history of falls and fractures, reproductive history, health behavior, psychological function, functional status, family structure, education, occupation, and economic status. Tests included in the health check-up were blood pressure, anthropometric measurements, electrocardiographs, and blood tests. Outcomes (dead/alive) of people in the study cohort after the 12-year follow-up was obtained from the Ohgimi Village office, and data on the major causes of death from the Japanese Vital Statistics of the Ministry of Health and Welfare. Two data were matched using sex and date of births. Age-adjusted mortality rates of both women and men for major causes of death were calculated using the population structure in 1985, which is commonly applied for calculating the standardized mortality rate in Japan.
444 Aging Clin Exp Res, Vol. 15, No. 6
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
Statistical analysis All data were entered into a computer and analyzed using the statistical software STATA version 6 for Macintosh (STATA Corporation, College Station, Texas). Variables included in the analysis are listed in Table 1. They were chosen from factors that were previously reported to relate with the total mortality of elderly cohorts (10-15). Social roles were measured using the Tokyo Metropolitan Institute of Gerontology (TMIG) Index of Competence (16). This scale consisted of the following four questions (“yes” was scored as 1 and “no” as 0): “Do you visit your friends’ houses?”, “Do you give advice to your family or friends?”, “Can you visit your friends in hospital?”, and “Do you talk to younger
Table 1 - Characteristics of the elderly cohort recruited from the village of Ohgimi in Okinawa, Japan.
Items
Men Women (N=251) (N=473) % %
Sociodemographic items Age (75 or over) Spouse (not present) Education (up to elementary school) Current working status (not working) Health behavior and social activity Hours of sleep (hours) less than 6 6-7 more than 7 Exercise (seldom) Smoking (ex- or current smoker) Drinking (everyday) Social rolea (score<2) Medical history and subjective health Cerebrovascular disease (yes) Hypertension (yes) Fracture (yes) Subjective health (fair or poor) Functional status ADLb (low) Hearing (impaired) Vision (impaired) BADLc Walking (dependent) Eating (dependent) Going to toilet (dependent) Bathing (dependent) Dressing (dependent) Biological items Body Mass Index (<22 kg/m2) Hemoglobin (men: <13 g/dL, women: <12 g/dL) Serum albumin (<3.9 g/dL) Total cholesterol (>219 mg/dL) Creatinine (men: >1.2 mg/dL, women: >1.0 mg/dL) Systolic blood pressure (>139 mmHg) Diastolic blood pressure (>89 mmHg) Electrocardiograph (abnormal)
37.1 13.8 88.7 55.2
48.2 63.6 92.9 69.4
** **
7.2 24.5 68.4 69.3 82.0 31.8 8.5
7.3 34.7 58.0 68.3 20.1 1.6 8.4
4.4 34.6 12.9 21.1
1.6 35.8 13.7 29.1
2.5 17.9 5.8
2.0 9.7 7.8
14.6 2.1 0.4 3.3 3.4
18.4 4.2 2.2 3.6 3.3
43.8
34.0
*
8.1 11.4 24.5
9.8 6.0 40.8
* **
21.6 84.1 10.4 14.8
13.7 81.6 15.7 11.0
** *
** ** * * **
**
*p<0.05, **p<0.01. Chi-square test was used. aSocial role was measured using the Tokyo Metropolitan Institute of Gerontology Index of Competence. bActivities of daily living. cBasic activities of daily living.
Predictors of mortality among Japanese elders
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
Table 2 - Twelve-year mortality of the elderly cohort recruited from the village of Ohgimi in Okinawa, Japan. Men
Age group 65-69 70-74 75-79 80Total
Total N 76 82 40 53 251
Women
Death N % 25 41 24 49 139
32.9 50.0 60.0 92.5 55.4
Total N
N
Death %
152 93 90 138 473
23 23 31 89 166
15.1 24.7 34.4 64.5 35.1
people?”. Continuous variables were categorized into dichotomies. Total cholesterol and Body Mass Index (BMI) are known to have a U-shaped relationship with total mortality, but they did not show such a relationship in this cohort and were categorized into dichotomies. Categorical variables were also re-classified into dichotomies, except for hours of sleep. Hours of sleep had a U-shaped relation with mortality in this cohort, and were categorized into a trichotomy: shorter than 6, 6-7, and longer than 7. Electrocardiographs were evaluated using the Minnesota Code, and codes IV and V were classified as abnormal. Survival curves for both sexes were drawn using the Kaplan-Meier method, and statistical differences between the two were calculated using the log-rank test. The Cox proportional hazards model was applied to examine factors associated with total mortality stratified by sex. Survival time from April 1987 to November 1999 was counted in months. The maximum survival time for subjects who were alive by November 1999 was 151 months. The independent variables of our interest were hours of sleep, exercise, smoking, drinking, and social role. Potential con-
founders were grouped into four and added sequentially to the model. Sociodemographic items included age, presence of spouse, education, and working status. Past history of cerebrovascular disease, hypertension and fracture, and subjective health were grouped into one. Functional status consisted of activities of daily living (ADL), hearing, vision, and basic activities of daily living (BADL). Biological items were body mass index, hemoglobin, serum albumin, total cholesterol, creatinine, blood pressure, and electrocardiograph abnormality. Differences in basic characteristics between sexes were analyzed using the Chi-square test. For analysis of factors associated with mortality, bivariate analysis adjusting for age was first conducted for each controlling factor. Factors that showed a significant association with total mortality (p<0.05) were put into multivariate analysis sequentially. The group of sociodemographic items was introduced first, then medical history and subjective health, functional status, and lastly, biological items. For women, only three models were examined, since the second group of medical history and subjective health did not show significance in the bivariate analysis. The proportionality of the model was tested using the Grambsch and Therneau method (17). RESULTS The characteristics of the 724 subjects are listed in Table 1. The proportions of women and men were 34.7 and 65.3%, respectively. Median age for men was 73 years (range 65 to 96 years), and that for women was 74 years (range 65 to 97 years). The proportions of those who did not live with their spouses and worked at the time of survey were much higher for women. Among health behavior, the proportions of smokers (men: 82.0%, women: 20.1%) and everyday drinkers (men: 31.8%, women: 1.6%) were much higher for men. More women
Table 3 - Leading causes of deaths and cause-specific standardized mortality rate of the elderly cohort recruited from the village of Ohgimi in Okinawa, Japan. Men
Causes of death Malignant neoplasm (Respiratory system) Cerebrovascular disease Pneumonia Heart disease Accident Other Unknown Total aStandardized
Women
Men
Women
Standardized mortality rate (per 100,000)a
N
%
N
%
39 (13) 14 14 13 2 35 22 139
28.1 (8.6) 10.1 10.1 9.4 1.4 25.2 15.8
28 (3) 12 20 19 5 40 42 166
16.9 (1.8) 7.2 12.0 11.4 3.0 24.1 25.3
1021.4 319.5 332.8 255.9 290.0 47.2 3156.6
441.8 25.5 63.9 104.4 154.6 44.7 1496.6
mortality rate was calculated using the population structure in 1985.
Aging Clin Exp Res, Vol. 15, No. 6 445
A. Goto, S. Yasumura, Y. Nishise, et al.
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
Survival rate 1 Women 0.9
Men
0.8
0.7 0
50 100 Analysis time (months)
150
Figure 1 - Sex-specific 12-year survival curves of the elderly cohort recruited from the village of Ohgimi in Okinawa, Japan.
slept 6-7 hours. While a higher proportion of men had suffered from cerebrovascular disease and had some hearing impairment, subjective health was lower for women. As for the biological items, more men had lower BMI, lower albumin and higher creatinine, and more women had higher total cholesterol. The total number of deaths during the 12-year study period was 139 (55.4%) for men and 166 (35.1%) for women (Table 2). Mortality was higher for men and for
the older age groups. Table 3 shows the main causes of death. Of 305 deaths, causes were unknown for 22 men and 42 women. The three leading causes of death, in rank order, were malignant neoplasm, cerebrovascular diseases and pneumonia for men, and malignant neoplasm, pneumonia and heart disease for women. The standardized mortality rate for cancer of the respiratory system and cerebrovascular disease was 12.5 and 5.2 times higher for men than for women, respectively. Figure 1 shows the sex-specific survival curves, with a significant difference between them (p<0.001). In the bivariate analysis of women, the items showing significant association with total mortality numbered 7: ADL, BADL (eating, bathing, dressing), BMI, serum albumin, and electrocardiograph abnormality. As for men, 9 controlling variables showed a significant association: history of cerebrovascular disease and hypertension, ADL, BADL (walking, bathing, dressing), serum albumin, diastolic blood pressure, and electrocardiograph abnormality. The results of sequential multivariate analysis are shown in Tables 4 and 5. The hazard ratio for less than 6 hours of sleep and social role remained significant for women in Models 1 to 3. More than 7 hours of sleep was significant in Models 1 and 2, but the significance diminished as biological items were introduced in Model 3. The hazard ratio in the final model was 2.62 (95% CI=1.36-5.07) for less than 6 hours of sleep and 1.88 (95% CI=1.13-3.13) for social role. As for men, nei-
Table 4 - Association of health behaviors and social activity with total mortality among elderly women recruited from the village of Ohgimi in Okinawa, Japan (Cox proportional hazards model). Model 1 Items Health behavior and social activity Hours of sleep less than 6 more than 7 Exercise (seldom) Smoking (ex- or current smoker) Drinking (everyday) Social rolea (score<2) Sociodemographic items Age Functional status ADLb (low) BADLc Eating (dependent) Bathing (dependent) Dressing (dependent) Biological items Body Mass Index (<22 kg/m2) Serum albumin (<3.9 g/dL) Electrocardiograph (abnormal)
Model 2
Model 3
HR
95% CI
HR
95% CI
HR
95% CI
2.65** 1.60* 1.04 1.25 0.51 1.92**
1.42-4.95 1.06-2.42 0.70-1.53 0.87-1.79 0.16-1.64 1.22-3.01
2.29* 1.57* 1.03 1.20 0.59 1.70*
1.18-4.44 1.03-2.38 0.70-1.52 0.83-1.74 0.18-1.90 1.03-2.82
2.62** 1.40 0.97 1.19 0.46 1.88*
1.36-5.07 0.91-2.15 0.66-1.44 0.81-1.74 0.14-1.49 1.13-3.13
1.10**
1.07-1.13
1.10**
1.07-1.12
1.09**
1.07-1.12
0.90
0.27-2.97
0.54
0.15-1.98
1.74 1.23 1.45
0.90-3.36 0.45-3.30 0.52-4.05
1.61 1.30 1.51
0.81-3.20 0.46-3.63 0.52-4.42
1.41* 2.15** 1.84**
1.00-1.99 1.22-3.76 1.18-2.87
*p<0.05, **p<0.01. aSocial role was measured using the Tokyo Metropolitan Institute of Gerontology Index of Competence. cBasic activities of daily living.
446 Aging Clin Exp Res, Vol. 15, No. 6
bActivities
of daily living.
Predictors of mortality among Japanese elders
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
Table 5 - Association of health behaviors and social activity with total mortality among elderly men recruited from the village of Ohgimi in Okinawa, Japan (Cox proportional hazards model). Model 1 Items Health behavior and social activity Hours of sleep less than 6 more than 7 Exercise (seldom) Smoking (ex- or current smoker) Drinking (everyday) Social rolea (score<2) Sociodemographic items Age Medical history Cerebrovascular disease (yes) Hypertension (yes) Functional status ADLb (low) BADLc Walking (dependent) Bathing (dependent) Dressing (dependent) Biological items Serum albumin (<3.9 g/dL) Diastolic blood pressure (>89 mmHg) Electrocardiograph (abnormal)
Model 2
Model 3
Model 4
HR
95% CI
HR
95% CI
HR
95% CI
HR
95% CI
1.42 1.62 0.98 1.08 1.42 1.79
0.61-3.27 0.99-2.66 0.65-1.48 0.65-1.79 0.95-2.13 1.00-3.21
1.29 1.61 1.01 1.16 1.50 1.40
0.53-3.15 0.97-2.68 0.66-1.55 0.68-1.97 0.99-2.27 0.75-2.61
1.39 1.55 0.93 1.18 1.53 1.02
0.57-3.40 0.93-2.59 0.60-1.45 0.67-2.10 1.00-2.34 0.51-2.05
1.29 1.54 1.07 0.95 1.43 1.00
0.50-3.34 0.92-2.58 0.68-1.67 0.54-1.69 0.91-2.25 0.45-2.24
1.13**
1.09-1.16
1.13**
1.09-1.16
1.13**
1.09-1.17
1.13**
1.09-1.17
2.27* 1.49
1.07-4.79 0.99-2.24
1.90 1.43
0.81-4.45 0.94-2.19
1.41 1.23
0.54-3.68 0.78-1.93
0.79
0.12-5.22
1.99
0.25-15.73
1.18 5.95 0.94
0.66-2.11 0.83-42.41 0.24-3.74
1.15 1.31 1.06
0.62-2.14 0.13-13.02 0.23-4.84
2.08* 3.14** 1.88*
1.13-3.85 1.75-5.64 1.08-3.28
*p<0.05, **p<0.01. aSocial role was measured using the Tokyo Metropolitan Institute of Gerontology Index of Competence; bActivities of daily living; cBasic activities of daily living.
DISCUSSION The primary advantage of this study was that we sampled all elderly residents aged 65 years or older in one longevity region in Japan, and that we followed them for
ther health behavioral items nor social role showed significance in Models 1 to 4. Only biological items, albumin, diastolic blood pressure, and electrocardiograph abnormality, remained significant in the final model for men.
Table 6 - Ten-year mortality of Japanese and European elderly people aged 70-75 yearsa. Men Total Country Japan Belgium Denmark France France Italy Netherlands Portugal Spain Switzerland Switzerland Switzerland Poland aData
Town
N
Outcome confirmed N
Ohgimi Hamme Roskilde Haguenau Romans Padua Culemborg Vila Franca de Xira Betanzos Yverdon Burgdorf Bellinzona Marki
87 126 101 109 142 97 114 111 88 123 30 30 19
85 126 101 109 137 79 114 111 81 122 30 30 19
Women Death N
%
30 70 54 58 74 38 67 51 39 58 14 14 12
35.3 55.6 53.5 53.2 54.0 48.1 58.8 45.9 48.1 47.5 46.7 46.7 63.2
Total N
Outcome confirmed N
Death N
%
111 105 101 110 137 93 124 111 119 126 30 30 23
111 105 100 109 133 76 124 111 103 125 30 29 22
19 30 36 26 31 17 40 37 27 31 7 11 8
17.1 28.6 36.0 23.9 23.3 22.4 32.3 33.3 26.2 24.8 23.3 37.9 36.4
from European towns are based on the SENECA Study (18).
Aging Clin Exp Res, Vol. 15, No. 6 447
A. Goto, S. Yasumura, Y. Nishise, et al.
over a decade. We applied a comprehensive geriatric health assessment by using a questionnaire including social, economic, health behavioral, and functional factors as well as biological factors. When compared with European cohorts, the mortality rate of our cohort was lower for both women and men. Table 6 shows the difference in 10-year mortality among 70-75 year-old cohorts in Japan and European countries. The data for Europe were based on the Survey in Europe on Nutrition in the Elderly (SENECA) (18). The SENECA was used for comparison because it was a multicenter study following elderly cohorts from 1988 to 1999, and is similar to that of our study. Currently, Japan is one of the longevity countries. However, health behavior is changing, and it is questionable whether Japan will continue to lead other countries in life expectancy. The proportion of smokers among young women in Japan has increased from 8.6% in 1986 to 11.5% in 2000 (1). Although the proportion of smokers among men has decreased from 59.7 to 47.4% during the same period, it is still much higher than in other developed countries. Hours of sleep, which was shown to have a positive effect on mortality in our study, has been slightly shortening in recent years (19). We also found that social interactions had a positive influence on survival, but the proportion of elders living alone has increased steadily from 9.3% in 1985 to 13.2% in 1998 (2). Moreover, one recent report in Japan revealed that the prevalence of frailty, a precursor of bedridden status, among non-institutionalized people aged 65 years or older was as high as 6%, and nearly 60% of these patients rarely had contacts with neighbors (20). Together with the rapid aging of the population and the change in health behavior, the need for health promotion aiming at successful aging is becoming greater than ever in Japan. It is known that the sex difference is a strong predictor of mortality (21, 22), and Japan is reported as one of the developed countries in which the sex mortality ratio (the ratio of male to female death rates) among older adults increased during the 1980s (23). In the present study, the total mortality among men was significantly higher than that among women after 12 years of follow-up. The most important contributor to the difference in mortality between women and men is the major cause of death (24). A recent study in Japan reported that 60-75% of sex difference in life expectancy was due to the difference in mortality from diseases of the circulatory system, malignancy (particularly that of trachea, bronchus and lung) and other diseases of the respiratory system (25). Other studies revealed that the major contributors to the sex mortality ratio were cardiovascular disease, malignancy, cerebrovascular disease, accidents, pneumonia, suicide, and liver diseases (26, 27). Correspondingly, we found that the standardized mortality rates of malignancy, cere-
448 Aging Clin Exp Res, Vol. 15, No. 6
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
brovascular disease, heart disease and pneumonia were much higher for men. Various factors have been proposed to account for the sex difference in mortality: biological variations in blood pressure and plasma glucose level, hormonal balance, genetic factors, health behavior such as smoking and drinking, social activities, psychological factors, and interactions between genetic and environmental factors (21, 28, 29). We found a large difference in smoking and drinking behavior between elderly women and men in the village of Ohgimi, and this may have caused the difference in mortality rates due to major causes of death. Although these two factors did not show any significant association with total mortality in both sexes in the present study, cause-specific analysis is currently being conducted and this will enable us to explore further the difference in predictors of mortality between sexes. Among health behaviors, sleep for less than 6 hours remained significant in all models for women. Cohort studies in the United States such as the Cancer Prevention Study II and the Alameda County Study, have reported a significant relationship between hours of sleep and mortality (30, 31). In Japan, one prospective study in a remote area in Aichi found that mortality was lower for women who had sufficient sleep every day (32), and another study in Gifu reported that both longer and shorter sleep, compared with 7-8 hours of sleep, increased total mortality in men (33). Whether the strength of association between hours of sleep and total mortality differs between women and men is inconclusive. Epidemiological studies tend to show that women are more likely to report sleep difficulties in old age, but electroencephalogram studies report greater deterioration in sleep architecture among men (34). This discrepancy may reflect a gender difference in the willingness to disclose symptoms, and the association of sleep duration and mortality among men may have been underestimated in our study. Sleep disturbances are often secondary to medical illnesses, including heart failure, depression, peptic ulcer, and respiratory diseases (35). The fact that the introduction of biological items diminished the significance of more than 7 hours of sleep may indicate that long hours of sleep reflect underlying medical illness. Moreover, habitual sleep patterns may independently affect morbidity and mortality related to stroke, heart disease, and other diseases (36). Careful assessment of sleeping habits among the elderly should be undertaken to detect underlying diseases and to give them adequate advice on sleeping patterns. It was also unique for women that their social roles were constantly associated with mortality in all models, whereas they did not show significance at all for men. The result was even more interesting, considering no significant difference in the distribution of social role scores between women and men. Social network ties have been reported to influence mortality (15, 37-39), and this influ-
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
ence seems to be higher for women. The Alameda County Study reported that fewer social contacts revealed a higher relative risk of mortality for women than for men (28). One report described that the reasons for the long life expectancy in Okinawa, despite relatively low socioeconomic status, are healthier life-styles and active social networking (40). People in Okinawa place a strong emphasis on family ties, and family activities are frequent. It is women who play an important role on these occasions in Japan, and the score on the social role of the TMIG index is higher among Japanese women in general (16). Thus, social roles reflecting the degree of involvement in social networking may be a strong predictor of mortality among women in our study. Widowhood is a normal life condition among many elderly women, because they tend to marry older men and their life span is longer than that of men. The proportion of women who did not live with their spouses was nearly five times higher than that among men in our study. In addition, more elderly Japanese women wish to expand their networks with friends and younger generations than men do (41). Responding to elderly women’s demands and considering the prevailing state of widowhood and the stronger relationship between social network and mortality among them, programs focusing on bringing the elderly into contact with their peers and neighbors could improve social support and strengthen social roles. There are four limitations to interpreting the results of this study. First of all, although the baseline questionnaire contained a comprehensive set of variables, we cannot deny that some important factors like dietary habits and income were omitted. While none of the health behavior or social roles had a significant association with total mortality among men, biological items such as serum albumin, blood pressure and electrocardiograph abnormality had a significant impact on their mortality. Inclusion of additional factors in the multivariate analysis model might have shown different results. As for women, some dependent variables of interest remained significant in the multivariate analysis, but this may also have been due to incomplete control. Secondly, the causes of a large proportion of deaths were unknown in our study. Data recorded in the Ministry of Health and Welfare files were limited to deaths occurring in the village of Ohgimi. Therefore, we could not confirm the causes of death for people who had moved out of the village during the follow-up period. Older people tend to be institutionalized outside the village, and most of the unknown cases (15 for men, 34 for women) were in the group of 85-year olds or older. Thirdly, the social role of the TMIG index was the only indicator of social interaction included in our study. In Japan, elderly women tend to have more frequent contacts with family members and friends than elderly men
Predictors of mortality among Japanese elders
(42). Instead, a higher proportion of elderly men hold group memberships and take the lead in local clubs compared with elderly women (43). The social role of the TMIG index may be better suited to assess women’s social interactions rather than men’s. Indicators regarding group memberships should be included in future research to examine their association, especially with mortality in men. Lastly, as pointed out previously, our cohort was from one of the longevity prefectures, so that the obtained mortality rate should not be generalized to other populations in Japan. Moreover, people in Okinawa have a somewhat different life-style from those in other regions of Japan. They have better sleeping patterns, and the proportions of smokers and everyday drinkers are lower than in other regions (44, 45). Thus, we cannot rule out the possibility that the associations of health behavioral variables with total mortality are underestimated in the present study. Our study results suggest that having sufficient hours of sleep and enhancing social networks could help elderly women in Okinawa to maintain good health. As for elderly men, biological items were the stronger predictors of their mortality. The Healthy Japan 21 covers nine areas: 1) nutrition, 2) physical activities, 3) mental health, 4) tobacco, 5) alcohol consumption, 6) dental health, 7) diabetes, 8) cardiovascular disease and 9) cancer. The interest of the elderly in staying healthy tends to focus on three of these targets: proper nutrition, physical activities, and adequate rest (2). In addition to these three targets, our study suggests that the enhancement of social roles deserves more attention. Staying socially active during daytime hours and taking adequate rest at night may help elderly women in Okinawa to increase their length of life. Moreover, our study provides direction for future research to explore the difference between inadequate health promotion measures in women and men. ACKNOWLEDGEMENTS This work was supported by the Ministry of Health and Welfare, Health Science Research Grants, as part of the project “Comprehensive Research on Aging and Health: A Longitudinal Study on Social Environments and Longevity in Okinawa” (H10-Choju-020).
REFERENCES 1. Health and Welfare Statistics Association. Trends in the nation’s health. Tokyo: Health and Welfare Statistics Association, 2002 (in Japanese). 2. Ministry of Health and Welfare. Annual Report on Health and Welfare 1999-2000; Seeking for new image of the elderly as social aging proceeds into the 21st century. Tokyo: Japan International Corporation of Welfare Services, 2001. 3. World Health Organization. Growing older - staying well, aging and physical activity in everyday life. Geneva: World Health Organization, 1998. 4. WHO Scientific Group on the Epidemiology of Aging. The uses of epidemiology in the study of the elderly. WHO Technical Report Series 706. Geneva: World Health Organization, 1984.
Aging Clin Exp Res, Vol. 15, No. 6 449
A. Goto, S. Yasumura, Y. Nishise, et al.
5. WHO Regional Office for Europe. WHO Regional Publications, European Series, No. 97: The European health report 2002. Copenhagen: WHO Regional Office for Europe, 2002. 6. U.S. Department of Health and Human Services. Healthy people 2010. McLean, VA: International Medical Publishing, Inc., 2001. 7. Great Britain Department of Health. Saving lives: our healthier nation. London: The Stationery Office, 1999. 8. Tokyo Metropolitan Institute of Gerontology. Health survey among elderly in Ohgimi, Okinawa. Tokyo: Tokyo Metropolitan Institute of Gerontology, 1988 (in Japanese). 9. Matsuzaki T. Epidemiological study on factors associated with longevity among Japanese. Kosei no Sihyo 1989; 36: 3-11 (in Japanese). 10. Shibata H, Haga H, Nagai H, et al. Predictors of all-cause mortality between ages 70 and 80: the Koganei Study. Arch Gerontol Geriatr 1992; 14: 283-97. 11. Ho SC. Health and social predictors of mortality in an elderly Chinese cohort. Am J Epidemiol 1991; 133: 907-21. 12. Fried LP, Kronmal RA, Newman AB, et al. Risk factors for 5-year mortality in older adults: the Cardiovascular Health Study. JAMA 1998; 279: 585-92. 13. Ruigomez A, Alonso J, Anto JM. Relationship of health behaviors to five-year mortality in an elderly cohort. Age Ageing 1995; 24: 113-19. 14. Iwamoto H, Kiyohara Y, Kato I, et al. Mortality from major causes of death and its risk factors in the elderly: 26-year followup study in Hisayama. Japan. Jpn J Geriatr 1994; 31: 671-6 (in Japanese). 15. Seeman TE, Berkman LF, Kohout F, Lacroix A, Glynn R, Blazer D. Intercommunity variation in the association between social ties and mortality in the elderly: a comparative analysis of three communities. Ann Epidemiol 1993; 3: 325-35. 16. Koyano W, Shibata H, Haga H, Suyama Y, Nakazato K. Measurement of competence in the elderly living at home: development of an index of competence. Jpn J Pub Health 1987; 34: 109-14 (in Japanese). 17. Stata Corporation. Stata reference manual, Rel. 6, Vol. 3. College Station, TX: Stata Press, 1999. 18. Amorim-Cruz JA, Haveman-Nies A, Schlettwein-Gsell D, De Henauw S. Gender, cohort and geographical differences in 10year mortality in elderly people living in 12 European towns. J Nutr Health Aging 2002; 6: 269-74. 19. Ministry of Public Management, Home Affairs, Posts and Telecommunications. Statistics Bureau and Statistics Center. Basic statistics of life-style. Tokyo: Statistics Bureau and Statistics Center, 2002 (in Japanese). 20. Imuta H, Yasumura S, Abe H, Fukao A. The prevalence and psychosocial characteristics of the frail elderly in Japan: a communitybased study. Aging Clin Exp Res 2001; 13: 443-53. 21. Wingard DL, Suarez L, Barrett-Connor E. The sex differential in mortality from all causes and ischemic heart disease. Am J Epidemiol 1983; 117: 165-72. 22. Janghorbani M, Hedley AJ, Jones RB, et al. Gender differential in all-cause and cardiovascular disease mortality. Int J Epidemiol 1993; 22: 1056-63. 23. Waldron I. Recent trends in sex mortality ratios for adults in developed countries. Soc Sci Med 1993; 36: 451-62. 24. Johansson S. Longevity in women. Cardiovasc Clin 1989; 19: 3-16. 25. Qureshi MH, Katoh T, Iibuchi Y. Sex differential in life ex-
450 Aging Clin Exp Res, Vol. 15, No. 6
(Aging 15: 443-450, 2003),©2003, Editrice Kurtis
26.
27.
28.
29. 30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41. 42. 43. 44.
45.
pectancy in Japan and Scotland: age and causes of death. Acta Med Okayama 1995; 49: 97-106. Nishida S, Watahiki N. Sex differential in life expectancy at birth in Japan: (1) The sex differential in recent years. Minzoku Eisei 1996; 62: 127-38 (in Japanese). Watahiki N, Nishida S. The sex differential in life expectancy at birth in recent decades in Japan. Bull Natl Inst Public Health 1996; 45: 139-49 (in Japanese). Wingard DL. The sex differential in mortality rates: demographic and behavioral factors. Am J Epidemiol 1982; 115: 205-16. Waldron I. Sex differences in human mortality: the role of genetic factors. Soc Sci Med 1983; 17: 321-33. Kripke DF, Garfinkel L, Wingard DL, Klauber MR, Marler MR. Mortality associated with sleep duration and insomnia. Arch Gen Psychiatry 2002: 59; 131-6. Wingard DL, Berkman LF, Brand RJ. A multivariate analysis of health-related practices: a nine-year mortality follow-up of the Alameda County Study. Am J Epidemiol 1982; 116: 765-75. Takezaki T, Tajima K, Yoshida M, et al. Risk of death by health habit index from a cohort study among the residents of a rural area in Aichi, Japan. Jpn J Pub Health 1999; 46: 904-14 (in Japanese). Kojima M, Wakai K, Kawamura T, et al. Sleep patterns and total mortality: a 12-year follow-up study in Japan. J Epidemiol 2000; 10: 87-93. Tallis R, Fillit H, Brocklehurst JC. Brocklehurst’s Textbook of Geriatric Medicine and Gerontology, 5th ed. Edinburgh: Churchill Livingstone, 1998. Raiha I, Seppala M, Impivaara O, et al. Chronic illness and subjective quality of sleep in the elderly. Aging Clin Exp Res 1994; 6: 91-6. Qureshi AI, Giles WH, Croft JB, Bliwise DL. Habitual sleep patterns and risk for stroke and coronary heart disease: a 10-year follow-up from NHANES I. Neurology 1997; 48: 904-11. Hanson BS, Isacsson S-O, Janzon L, et al. Social network and social support influence mortality in elderly men. The prospective population study of “men born in 1914,” Malmö, Sweden. Am J Epidemiol 1989; 130: 100-11. Seeman TE, Kaplan GA, Knudsen L, et al. Social network ties and mortality among the elderly in the Alameda County study. Am J Epidemiol 1987; 126: 714-23. Steinbach U. Social networks, institutionalization, and mortality among elderly people in the United States. J Gerontol 1992; 47: S183-90. Cockerham WC, Hattori H, Yamori Y. The social gradient in life expectancy: the contrary case of Okinawa in Japan. Soc Sci Med 2000; 51: 115-22. Inoue T, Ehara Y. Women’s Data Book. Tokyo: Yuhikaku, 1998 (in Japanese). Sugisawa H. Social integration and mortality in Japanese elderly. Jpn J Pub Health 1994; 41: 131-9 (in Japanese). Tamakoshi A, Aoki R, Ohno Y, et al. Social activities in the elderly. Jpn J Pub Health 1995; 42: 888-96 (in Japanese). Konishi M, Kondou H, Okada K. Health status, life habits, and social background among the JPHC study participants at baseline survey. J Epidemiol 2001; 11 (Suppl): S57-74. Sobue T, Yamamoto S, Watanabe S. Smoking and drinking habits among the JPHC study participants at baseline survey. J Epidemiol 2001; 11 (Suppl): S44-56.