In September 2001, we targeted a questionnaire survey to all elderly individuals (aged 65 years and over) who lived at home in suburban Tokyo, Japan. Of 16,462 eligible elderly individuals, 13,195 people (80.2% as a respondent percentage) gave informed consent to participate in the study and returned the self-administered questionnaire by mail. In September 2004, we sent via mail a second questionnaire (the same as used previously) to the respondents, and 8558 participants responded (505 cases had moved, 914 were deceased and 3218 did not respond). The inclusion criterion was all elderly dwelling in their home aged 65 years or more. We followed up all of the participants until August 31, 2007, and ultimately obtained individual data, deceased or living, for a total of 8162 participants through the municipal resident’s registry.

The city used as the study setting had developed partly as a commuter town to accommodate increased workers and their families in the metropolitan Tokyo area between the period of the 1970s and 1990s, which included a recent era of high Japanese economic growth. The majority of dwellers were middle class. The total population of the city was approximately 140,000 as of 2001 with 9.6 per cent of the population aged 65 years or older. This number was nearly half of the national average in 2000 (17.3%).

The current study aimed to examine the causal relationships among health and life conditions, socioeconomic status, healthy dietary and lifestyle habits, and three health-related dimensions.

Health and life conditions in 2004 were dependent endogenous latent variables, including the number of survival days between September 1, 2004 and August 1, 2007 and the degree of long-term health care. The degree of health care was used in order to detect elderly health conditions in September 2004 using a public assessment tool devised by the Japanese Ministry of Health, Labor and Welfare. This tool is comprised of six levels ranging from the lightest degree (requiring mild support) to the most severe degree (requiring comprehensive care). In our analysis, a respondent that had received no care scored zero, while a respondent scored 1 if assigned at the lightest degree and scored 6 if assigned at the most severe degree.

Socioeconomic status in 2001 was assessed via educational attainment and annual income. Levels of educational attainment were categorized into four groups; graduation from junior high school, graduation from high school, those achieving a higher educational level than junior college and those who did not want to respond. Annual income levels fell into four categories; less than one million Japanese yen (equivalent to less than US $13,000), less than three million yen, less than seven million yen and more than seven million yen in 2001.

Healthy dietary and lifestyle habits in 2004 were assessed by the dietary health score and the healthy lifestyle score, respectively. The dietary health score was determined through the following three steps. First, food items consumed were assessed by the question; “Do you eat food items such as meat, soy products, eggs, bluebacked fish (sardine, mackerel and herring), dairy products, fruits, vegetables, salt-cured products, oily foods, strongly flavored foods, snacks and breakfast?” If the respondent consumed a particular food item, we then asked them to choose only one answer describing the frequency with which they consumed the food item as follows: eat everyday, eat 5 to 6 days per week, eat 3 to 4 days per week, eat 1 or 2 days per week or never eat. The frequency of item consumption per day was also measured (ranging from once to four times per day). Second, as being significantly and positively associated with number of survival days of the respondents examined at the end of August 2007, healthy dietary habits were defined. Our analyses resulted in a selection of ten healthy dietary habits as follows: consumption of meat, eggs, foods with eggs, and blue-backed fish 1 to 4 days a week, consumption of soy foods, dairy products, fruits and salt-cured products more than 3 days a week, consumption of vegetables more than 5 days a week and having breakfast everyday. Finally, a dietary health score was calculated as a score pertaining to the consumption of the ten identified healthy dietary food categories (one point for each category), ranging from 0 to 10, with a higher score indicating healthier dietary habits. Healthy lifestyle habits were assessed in terms of alcohol consumption, smoking, exercise habits, sleeping habits and body mass index (BMI). Such habit items were regarded as being healthy lifestyle habits when significantly associated with the number of survival days of the respondents as detected at the end of August 2007. As a result of our analyses, the habits significantly associated with number of survival days were alcohol consumption, never smoking (even in the past), less than nine hours of sleep per night, exercising more than once a week and having a BMI of more than 20. We scored these healthy lifestyle habits in a range from 0 to 5, with a higher score representing a healthier lifestyle.

The three health-related dimensions examined in our study consisted of physical, mental and social health components. Physical health parameters included the basic activities of daily living (BADL) and the instrumental activities of daily living (IADL) as well as the number of diagnosed diseases from which the respondents had suffered from at the time of the 2001 survey. The BADL score was assessed by three questions: “Can you go to the toilet by yourself?”, “Can you take a bath by yourself?” and “Can you take a walk outside?” [26]. Individuals obtained one point if they could perform these functions, and overall scores ranged from 0 to 3 with higher scores indicating a greater level of basic activity competence. The IADL score was measured via five questions related to instrumental activity: “Can you buy daily necessities by yourself?”, “Can you cook daily meals by yourself?”, “Can you deposit and withdraw money in a bank account?”, “Can you complete documents related to insurance and pensions?” and “Can you read books and newspapers?” [27]. As the IADL was scored in a similar fashion to the BADL, total scores ranged from 0 to 5 with higher scores indicating greater instrumental activity. The prevalence of comorbidity was determined by instructing the individuals to “Choose the diagnosed diseases with which you now are treated”. We discovered four diseases (cerebrovascular disease, cardiovascular disease, hepatic disease and diabetes mellitus) that were significantly and negatively associated with the number of survival days of the respondents between September 2004 and August 2007, and thus the extent of comorbidity was defined as the number of diseases with which the respondent was diagnosed among these four.

Mental health was determined via self-rated health, self-rated health compared with that of the previous year and daily life satisfaction at the time of the 2001 survey. Self-rated health was quantified via a four-point Likert scale using questions pertaining to the respondent’s perceived overall health: “In general, would you say your health is: very poor, poor, fair or excellent?” [28,29]. The score ranged from 1 to 4, with a higher score indicating higher perceived health. If unknown, we assigned a score of 0. Self-rated health compared to the previous year was determined on a three-point Likert scale with the question “How is your current health compared to your health at this time last year?” and the answer was chosen among the following responses: poorer, may or may not be better and better. In this respective order, scores of 1 to 3 were assigned and we assigned a score of 0 if the answer was unknown. Daily life satisfaction was detected by the question “Are you satisfied with your current daily life?” with answers ranging from not satisfied, may or may not be satisfied and satisfied. In this respective order, scores ranged from 1 to 3, and we assigned a score of zero if the answer was unknown.

Social health was measured by variables including the frequency of going outside, communication with the neighborhood and hobby-related activities. The frequency of going outside was examined by the question “How often do you go outside including around your neighborhood?” and answers ranged from less than once a month, more than once a month and more than 3 to 4 times a week. Scores ranged from 1 to 3 in this respecttive order. If the answer was unknown, a score of zero was assigned. Communication with the neighborhood was assessed by asking “How often do you have communication with your friends or neighbors?” and scores ranged from seldom, once a month, 3 to 4 times a week and almost everyday, using a scale ranging from 1 to 4 in respective order and zero if unknown. We measured hobby-related activities by inquiring “Do you participate in hobby-related activities?” with two possible answers (yes and no). Scores were assigned as 1 (no hobby), 2 (presence of a hobby) and 0 (unknown).

Respondents aged 80 years and older and those associated with more than a middle degree of long-term care at the 2001 survey were excluded in the analyses due to an increased and indispensable deviation in their measurement variables. All data obtained were evaluated by score or degree to examine differences between males and females using the Kendall’s tau rank correlation coefficient. Exploratory factor analysis was used to fit all of the observed variables to corresponding latent variables. This analysis was performed by the maximum likelihood procedure and a Promax oblique rotation, using IBM SPSS Statistics 21 for Windows (IBM Inc., New York, USA).

We used structural equation modeling (SEM) to examine presumptive underlying causal relationships between the latent variables by Amos ver.19 for Windows (IBM Inc.). The analyses were performed using a crosslagged effects variation model with longitudinal data followed over seven years. Four latent variables were fitted with socioeconomic status and three health-related dimensions as causes using data obtained in the 2001 survey, healthy dietary and lifestyle habits as intermediate causes using data from the 2004 survey, and health and life conditions as outcomes by using data at the 2004 survey and data between 2004 and 2007. All observed variables corresponding to the three health-related dimensions were collected at the questionnaire survey in September 2001. Estimation of the best-fitting model was carried out by the method of maximum likelihood of SEM. The optimization algorithm was implemented with no-missing-data parameters. The direct, indirect and total standardized effects of different latent variables on the endogenous health and life condition variable were measured by gender. The models employed indices criteria for assessing model fitness. Goodness-of-fit was approved by CMIN, NFI, IFI and RMSEA. Results were regarded as statistically significant if the p value was less than 0.05.

Of 8162 eligible participants, included in the analysis were 7066 individuals (male, 3409; female, 3657) aged more than 80 years and with either no long-term care or the mildest degree of long-term care utilized in September 2001. The average ages (±standard deviation) of males and females were 70.0 ± 3.9 and 70.5 ± 4.2 years, respectively. Table 1 shows the frequencies of measured variables by gender. The mean number of survival days between September 1, 2004 and August 31, 2007 was fewer in males than in females. 2.3 per cent of male and 3.0 per cent of female participants had degenerated to either the middle or severest degrees of long-term care in 2004 compared to 2001. Compared with males, females displayed significantly lower educational attendance as well as annual income, which were deemed a gender disparity (Kendall’s tau rank correlation coefficients were −0.376, p < 0.001 and −0.236, p < 0.001, respecttively). Regarding the dietary health score, the majority of the participants were distributed below seven points, while in the healthy lifestyle category, most participants scored above two points. The results indicated that many people enjoyed healthy lifestyle habits, although a significant number of the elderly had lived on an unhealthy diet. More than 80 per cent of the participants had been autonomously active with high BADL and IADL scores. There were some people who had physical inconveniences and one in five participants suffered from at least one disease. In terms of self-rated health, the respondents who answered that they were in poor health approached 20 per cent. More than twice as many female respondents than male respondents reported being in poorer health compared to the previous year. Two-thirds of respondents revealed satisfaction with their daily life among both males and females. Although half of the respondents across both sexes enjoyed hobby-related activities, nearly 10 per cent remained in their homes, and more than half of males as well as one-third of females had little communication with their own neighbors. We also observed that some elderly people were likely to experience poor human relationships.

Table 2 presents factor loadings for a hypothesized model consisting of the latent variables obtained through exploratory factor analysis. Factor 1 indicated high loadings particularly for self-rated health variables, and displayed a high confidence coefficient. Self-rated health and daily life satisfaction, with the exception of the number of comorbid conditions, were termed “mental health factors” related to the three health-related dimensions. The number of comorbid conditions was a negative factor by itself and deemed to considerably associate with physical condition since the comorbidities tended to include more common diseases such as hypertension, diabetes and cardiovascular disease apart from mental disorders among the target population. Along with the number of comorbid conditions, BADL and IADL, excluding the frequency of going outside, informed the labeling of factor 2 as “physical health”. We emphasized