• Open Access

Underweight/overweight and the risk of long-term care: Follow-up study using data of the Japanese long-term care insurance system

Authors

  • Akiko Honda,

    1. Division of Health Promotion, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan
    2. Niigata Prefectural Government, Niigata, Japan
    Search for more papers by this author
  • Naohito Tanabe,

    Corresponding author
    1. Division of Health Promotion, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan
    • Correspondence: Professor Naohito Tanabe MD PhD, Department of Health and Nutrition, Faculty of Human Life Studies, University of Niigata Prefecture, 471 Ebigase, Higashi-ku, Niigata 950-8680, Japan. Email: tanabena@gmail.com

    Search for more papers by this author
  • Nao Seki,

    1. School of Health Sciences, Faculty of Medicine, Niigata University, Niigata, Japan
    Search for more papers by this author
  • Yoshiko Ogawa,

    1. Division of Public Health, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan
    2. Niigata Prefectural Government, Niigata, Japan
    Search for more papers by this author
  • Hiroshi Suzuki

    1. Division of Public Health, Niigata University Graduate School of Medical and Dental Sciences, Niigata, Japan
    Search for more papers by this author

Abstract

Aim

Both being underweight and overweight can lead to reduced activity of daily living, which subsequently can require long-term care. The aim of the present study was to clarify the association between underweight/overweight and the subsequent risk of long-term care introduction.

Methods

We tracked the data of long-term care insurance for 1580 men and women aged ≥65 years who had participated in the official population-based health check-up program in 2001 in Tsunan town and Sekikawa village, Japan. The health check-up data and medical expenditure data for the fiscal year 2001 were used as baseline data. Participants were classified into underweight (body mass index (BMI) <18.5 kg/m2), normal range (BMI = 18.5 to <25.0 kg/m2) and overweight (BMI ≥25.0 kg/m2); the normal range was used as a reference category in Cox proportional hazards models.

Results

During the average 5.8 years of follow up, 156 participants were identified to start using long-term care services. Among the young-old elderly (65–74 years-of-age), underweight was significantly associated with the risk of long-term care introduction (multivariable-adjusted HR 4.26, 95% CI 1.69–10.72), whereas overweight was not (multivariable-adjusted HR 1.45, 95% CI 0.69–3.06). Neither underweight nor overweight were significantly associated with long-term care introduction among the old-old elderly (≥75 years-of-age).

Conclusions

Underweight could be a good predictor of long-term care introduction in the young-old elderly. We should pay attention to underweight in the elderly, as it might be a manifestation of some physical or mental problems related to future long-term care introduction. Geriatr Gerontol Int 2014; 14: 328–335.

Introduction

The aging population is a serious social issue in Japan. The increasing burden of the long-term care of older adults with decreased activity of daily living (ADL) has been becoming a nationwide problem. To cope with these problems, the Japanese government established the long-term care insurance system in April 2000. It was reorganized into a system that placed emphasis on the preventive long-term care of the elderly in 2006. In this system, one facet of preventive long-term care[1] is “improvement of nutrition”,[2] and body mass index (BMI) <18.5 is one of the screening criteria. Meanwhile, the major reasons for long-term care are cerebrovascular disease, followed by dementia, debilitation as a result of aging, articular diseases and fractures.[3] Underweight is a risk factor for dementia[4, 5] and fractures,[6] which are major reasons for long-term care in Japan. In contrast, overweight can also lead to the need for long-term care, as it is a risk factor for cerebrovascular diseases,[7] as well as some articular diseases.[8, 9] Furthermore, in Japan, countermeasures against metabolic syndrome, which is closely associated with being overweight, are regarded as priorities in public health strategies, and their effects on preventing long-term care is also expected.[10-13] However, the literature has not sufficiently discussed whether overweight or underweight is a better predictor of the need for long-term care.

Accordingly, we examined the association between underweight/overweight and the subsequent risk of long-term care introduction in Japanese older adults using data from health check-ups and long-term care insurance.

Methods

Participants

The study areas were Tsunan town and Sekikawa village, both in the Niigata Prefecture, Japan. According to the 2000 national census, the population aged ≥65 years was 4062 in Tsunan and 2309 in Sekikawa. Participants were selected from 2378 people aged ≥65 years who had participated in the official population-based health check-up program in 2001. Initially, 2190 people were selected for whom information on medical and long-term care expenditures for the entire 12 months of the fiscal year (FY) 2001 (April to March of the following year) was available from the municipal government office. From this group, people who had been hospitalized or who had received long-term care in FY2001 and those with a history of cerebrovascular, heart, liver, or renal diseases were excluded. The remaining 1580 people (627 men and 953 women) comprised the participants of the present study.

Baseline cardiovascular risk factor assessments

Baseline health check-ups based on the standardized procedures of the Niigata Prefecture were carried out from April to November of 2001 in Tsunan, and from June to July of 2001 in Sekikawa. Public health nurses interviewed the participants regarding their medical history, smoking status and alcohol-drinking habits; they also measured their height, weight and blood pressure, and took blood samples. BMI was calculated on the basis of the measured height and weight. Blood pressure was measured with a mercury sphygmomanometer, and was classified according to the 1999 World Health Organization and International Society of Hypertension criteria (1999 WHO/ISH criteria).[14] The levels of serum total cholesterol, serum high-density lipoprotein (HDL) cholesterol and plasma glucose were measured from the blood samples. Plasma glucose levels, which are strongly affected by fasting, were categorized in quartiles for each of the fasting participants (i.e. >10 h after eating) and postprandial participants, and integrated categories were used in the analysis. The quartile categories were ≤4.78, 4.83–5.06, 5.11–5.44 and 5.50 + mmol/L for fasting values, and ≤5.39, 5.44–6.11, 6.17–7.06, and 7.11 + mmol/L for postprandial values.

Baseline medical expenditures

Until the medical insurance system in Japan was partially reformed in April 2008, in principle, all Japanese citizens held one of two types of health insurance. One was insurance for employees and their dependents; the other was the National Health Insurance (NHI) for all other citizens, such as farmers, the self-employed, people living on pensions, the unemployed and their dependents. In 2001, the medical-expenditure data for residents aged ≥70 years, regardless of their insurance type, and medical-expenditure data for residents in the NHI system who were aged <70 years were managed by the municipality government office. The baseline medical expenditure was calculated using this database; the monthly total medical expenditure for each participant was calculated for the entire 12 months of FY2001 (monthly medical expenditure). As people who had been hospitalized during this period were excluded from the present study, the baseline expenditure did not include expenditures for hospitalized medical care. In the present study, medical expenditure data are shown in Japanese yen.

Follow-up for long-term care

All people aged ≥40 years in Japan also belong to the long-term care insurance system. Long-term care expenditures covered by that system could be assessed by reviewing the municipal database. The insurance data were tracked through this database up to September 2007. The follow up for the present study terminated when the participants began to use long-term care services (long-term care introduction). When it was certified that the participant required long-term care, the participant was considered to have reached this end-point. The certified grade and long-term care expenditure of each certified participant in the month they were certified were also retrieved from this database. During the 9218 person-years of follow up, 5.8 years per person in average, 156 participants began to use long-term care services.

All data were electronically provided by Tsunan town and Sekikawa village in an anonymous format under written permission from the head of each municipality. For data linkage, the staff of the municipalities gave unique identifiers specific to each participant. Researchers of the university analyzed the anonymous data, and the linkage between the baseline health check-up data, baseline medical expenditure data and long-term care expenditure data was carried out using these identifiers. The study protocol was reviewed and approved by the ethics committee of the Niigata University School of Medicine.

Statistical analysis

Participants were divided into two age-class categories: the young-old elderly (65–74 years-of-age) and the old-old elderly (≥75 years-of-age), and all analyses were carried out within each age category. Participants were classified into three groups according to their BMI based on the criteria of World Health Organization: <18.5 kg/m2 (underweight), 18.5 to <25.0 kg/m2 (normal range) and ≥25.0 kg/m2 (overweight). To examine the crude statistical differences in the baseline characteristics between these three BMI groups, one-way analysis of variance was used for continuous variables and the χ2-test was used for categorical variables. As the distributions of medical expenditure and long-term care expenditure were skewed with heavy right tails, log-transformed values were used in the statistical analysis.

The hazard ratios (HR) and the 95% confidence intervals (CI) of underweight and overweight with respect to long-term care introduction were calculated by sex- and age-adjusted Cox proportional hazards models for both sex-combined analyses and by the age-adjusted model for sex-specific analyses, treating normal-range weight as a reference category. Multivariable adjusted HR and 95% CI were also calculated, and further adjusted for hypertension (positive/negative), hyperglycemia (positive/negative), dyslipidemia (positive/negative), smoking status (current/former/never), alcohol drinking habit (habitual/occasional/none) and monthly medical expenditure (continuous) using propensity scores. Hypertension was considered positive when any of treated hypertension, high systolic blood pressure (≥140 mmHg) or high diastolic blood pressure (≥90 mmHg) was present; hyperglycemia was considered positive when any of a history of diabetes mellitus or high plasma glucose (in the highest quartile category) was present; and dyslipidemia was considered positive when any of a history of dyslipidemia, high serum total cholesterol (≥5.68 mmol/L) or low serum HDL cholesterol (<1.03 mmol/L) was present. Multinomial logistic regression analysis was applied in the calculation of propensity scores using these confounders as independent variables and BMI group as a dependent variable, with the “normal range” BMI category as a reference.[15] spss 11.0J for Windows (SPSS Japan, Tokyo, Japan) was used in the analyses, and two-tailed P-values of <0.05 were considered to be statistically significant.

Results

On the basis of the baseline BMI, 53 participants, 675 participants and 272 participants among the young-old elderly, and 71 participants, 399 participants, and 110 participants among the old-old elderly were classified as underweight, normal range and overweight, respectively (Table 1). At baseline, the proportion of men and the prevalence of current/former smokers and mean serum HDL cholesterol in both age groups, the prevalence of alcohol drinkers in the young-old elderly, and mean age in the old-old elderly tended to be higher in the lower BMI groups; whereas the prevalence of treated hypertension and hyperlipidemia, mean values of blood pressure and serum total cholesterol in both age groups, and mean monthly-medical-expenditure in the young-old elderly tended to be higher in the higher BMI groups.

Table 1. Baseline characteristics of participants by age and body mass index
 Body mass index (kg/m2)P-value
–<18.518.5 to <25.0≥25.0
(Underweight)(Normal range)(Overweight)
  1. Data are presented as the mean ± SD or n (%). Drink alcoholic beverages almost every day. The quartile categories were ≤4.78, 4.83–5.06, 5.11–5.44 and 5.50 + mmol/L for fasting values, and ≤5.39, 5.44–6.11, 6.17–7.06, and 7.11 + mmol/L for postprandial values.
Young-old elderly (aged 65–74 years)    
No. participants53 (100.0)675 (100.0)272 (100.0) 
Sex, n (men)24 (45.3)290 (43.0)77 (28.3)<0.001
Age (years)70.2 ± 2.670.1 ± 2.869.9 ± 2.90.763
Treated hypertension9 (17.0)189 (28.0)114 (41.9)<0.001
History of diabetes mellitus1 (1.9)18 (2.7)15 (5.5)0.075
History of hyperlipidemia1 (1.9)58 (8.6)35 (12.9)0.020
Current/former smokers13 (24.5)159 (23.6)25 (9.2)<0.001
Alcohol drinkers20 (37.7)247 (36.6)73 (26.8)0.014
Systolic blood pressure (mmHg)130.9 ± 18.6133.5 ± 17.6139.8 ± 18.2<0.001
Diastolic blood pressure (mmHg)76.7 ± 11.775.2 ± 10.178.1 ± 10.6<0.001
Serum total cholesterol (mmol/L)4.81 ± 0.745.15 ± 0.805.27 ± 0.77<0.001
Serum HDL-cholesterol (mmol/L)1.75 ± 0.531.51 ± 0.381.37 ± 0.34<0.001
Plasma glucose (quartile)   0.078
Q122 (41.5)179 (26.5)63 (23.2) 
Q27 (13.2)165 (24.4)79 (29.0) 
Q310 (18.9)168 (24.9)66 (24.3) 
Q414 (26.4)163 (24.1)64 (23.5) 
Monthly medical expenditure in FY 2001 (Yen/month, in common logarithmic value)
 2.16 ± 2.422.84 ± 2.133.08 ± 2.060.014
Old-old elderly (aged ≥75 years)    
No. participants71 (100.0)399 (100.0)110 (100.0) 
Sex, n (men)38 (53.5)169 (42.4)29 (26.4)<0.001
Age (years)79.6 ± 3.978.6 ± 3.478.1 ± 3.30.023
Treated hypertension18 (25.4)132 (33.1)56 (50.9)<0.001
History of diabetes mellitus2 (2.8)14 (3.5)8 (7.3)0.179
History of hyperlipidemia0 (0.0)46 (11.5)11 (10.0)0.011
Current/former smokers27 (38.0)76 (19.0)7 (6.4)<0.001
Alcohol drinkers23 (32.4)130 (32.6)29 (26.4)0.452
Systolic blood pressure (mmHg)134.3 ± 17.8134.4 ± 19.0142.9 ± 19.7<0.001
Diastolic blood pressure (mmHg)72.8 ± 10.774.1 ± 10.277.3 ± 11.70.008
Serum total cholesterol (mmol/L)4.80 ± 0.865.07 ± 0.865.38 ± 0.82<0.001
Serum HDL-cholesterol (mmol/L)1.69 ± 0.371.54 ± 0.381.41 ± 0.34<0.001
Plasma glucose (quartile)   0.152
Q123 (32.4)96 (24.1)23 (20.9) 
Q214 (19.7)107 (26.8)21 (19.1) 
Q312 (16.9)95 (23.8)33 (30.0) 
Q422 (31.0)101 (25.3)32 (29.1) 
Monthly medical expenditure in FY 2001 (Yen/month, in common logarithmic value)
 1.98 ± 2.451.84 ± 2.521.82 ± 2.610.903

During the follow up, long-term care introduction was identified for 38 young-old participants and 124 old-old participants (Table 2). The certified grade of long-term care and long-term care expenditure in the month when they were certified for long-term care insurance were compared between BMI groups. Among the young-old elderly, the proportion of the “support required level” grade was the lowest, and the proportion of “care level 5” was the highest in the underweight group (P = 0.040). In addition, among the young-old elderly, the long-term care expenditure was the highest in the underweight group (P = 0.007).

Table 2. Certified grade of long-term care and long-term care expenditure in the month when participants were certified for long-term care insurance
 Body mass index (kg/m2)P-value
–<18.518.5 to <25.0≥25.0
(Underweight)(Normal range)(Overweight)
  1. Data are n (%) or mean ± SD. The higher the levels, the more severe the disability status.
Young-old elderly, aged 65–74 years (n = 38)    
Certified grade   0.040
Support required level1 (16.7)7 (35.0)3 (25.0) 
Care level 13 (50.0)3 (15.0)4 (33.3) 
Care level 20 (0.0)5 (25.0)3 (25.0) 
Care level 30 (0.0)2 (10.0)2 (16.7) 
Care level 40 (0.0)3 (15.0)(0.0) 
Care level 52 (33.3)(0.0)(0.0) 
Long-term care expenditure (Yen/month, in common logarithmic value)
 5.0 ± 0.44.5 ± 0.34.4 ± 0.30.007
Old-old elderly, aged 75 years (n = 124)    
Certified grade   0.064
Support required level7 (43.8)26 (31.7)11 (42.3) 
Care level 12 (12.5)34 (41.5)5 (19.2) 
Care level 25 (31.3)14 (17.1)4 (15.4) 
Care level 31 (6.3)5 (6.1)1 (3.8) 
Care level 40 (0.0)2 (2.4)4 (15.4) 
Care level 51 (6.3)1 (1.2)1 (3.8) 
Long-term care expenditure (yen/month, in common logarithmic value)
 4.5 ± 0.34.6 ± 0.34.6 ± 0.30.550

Among the young-old elderly, the long-term care introduction rate was the highest in the underweight group, whereas it was the lowest in the normal-range group for both men and women (Table 3). For men and women combined, underweight was significantly associated with the risk of long-term care introduction (multivariable-adjusted HR 4.26, 95% CI 1.69–10.72), whereas overweight was not (multivariable-adjusted HR 1.45, 95% CI 0.69–3.06). These relationships were similarly observed for men and women. Among the old-old elderly, neither underweight nor overweight was significantly associated with long-term care introduction.

Table 3. Cox proportional hazard ratios for 5-year long-term care introduction associated with underweight and overweight
 Body mass index (kg/m2)
–<18.518.5 to <25.0>25.0
(Underweight)(Normal range)(Overweight)
  1. Multivariable model adjusted for sex (combined analyses of both sexes), age, hypertension, hyperglycemia, dyslipidemia, smoking status, alcohol drinking habit and monthly medical expenditure in the fiscal year 2001 using propensity scores. *P < 0.05, **P < 0.01 compared with the “normal range” category. CI, confidence interval; HR, hazard ratio.
Young-old elderly (aged 65–74 years)   
Men and women, no. at risk (person-years)53 (308.1)675 (4112.7)272 (1652.6)
No. long-term care introduction (/1000 person-years)6 (19.48)20 (4.86)12 (7.26)
Sex- and age-adjusted HR (95% CI)4.03 (1.62–10.03)**1.001.60 (0.78–3.31)
Multivariable, HR (95% CI)4.26 (1.69–10.72)**1.001.45 (0.69–3.06)
Men, n at risk (person-years)24 (132.8)290 (1740.9)77 (455.3)
No. long-term care introduction (/1000 person-years)3 (22.60)11 (6.32)4 (8.78)
Age-adjusted HR (95% CI)3.58 (1.00–12.86)1.001.43 (0.45–4.50)
Multivariable, HR (95% CI)3.78 (1.05–13.54)*1.001.14 (0.35–3.75)
Women, n at risk (person-years)29 (175.3)385 (2371.8)195 (1197.3)
No. long-term care introduction (/1000 person-years)3 (17.11)9 (3.79)8 (6.68)
Age-adjusted HR (95% CI)4.52 (1.22–16.71)*1.001.76 (0.68–4.57)
Multivariable, HR (95% CI)4.53 (1.16–17.66)*1.001.72 (0.65–4.56)
Old-old elderly (aged ≥75 years)   
Men and women, n at risk (person-years)71 (354.5)399 (2179.8)110 (611.0)
No. long-term care introduction (/1000 person-years)16 (45.13)82 (37.62)26 (42.55)
Sex- and age-adjusted HR (95% CI)1.05 (0.61–1.81)1.001.15 (0.74–1.80)
Multivariable, HR (95% CI)1.01 (0.58–1.78)1.001.08 (0.69–1.70)
Men, n at risk (person-years)38 (185.7)169 (917.8)29 (161.6)
No. long-term care introduction (/1000 person-years)6 (32.32)34 (37.05)7 (43.32)
Age-adjusted HR (95% CI)0.74 (0.31–1.76)1.001.26 (0.56–2.84)
Multivariable, HR (95% CI)0.68 (0.27–1.71)1.001.17 (0.48–2.82)
Women, n at risk (person-years)33 (168.8)230 (1262.1)81 (449.4)
No. long-term care introduction (/1000 person-years)10 (59.23)48 (38.03)19 (42.28)
Age-adjusted HR (95% CI)1.38 (0.69–2.73)1.001.13 (0.66–1.92)
Multivariable, HR (95% CI)1.37 (0.68–2.75)1.001.08 (0.63–1.84)

When the participants were stratified by median baseline medical expenditure, a significantly elevated risk of long-term care introduction associated with underweight was observed only in the stratum of low baseline medical expenditure among the young-old elderly (multivariable-adjusted HR 4.90, 95% CI 1.46–16.49; Table 4).

Table 4. Cox proportional hazard ratios for 5-year long-term care introduction associated with underweight and overweight stratified by baseline medical expenditure
 Body mass index (kg/m2)
<18.518.5 to <25.0>25.0
(Underweight)(Normal range)(Overweight)
  1. The stratification was made on the basis of the medial of the baseline medical expenditure: median = 6347.2 Yen/month. Multivariable model adjusted for sex (combined analyses of both sexes), age, hypertension, hyperglycemia, dyslipidemia, smoking status, alcohol drinking habit and monthly medical expenditure in the fiscal year 2001 using propensity scores. *P < 0.05, **P < 0.01 compared with the “normal range” category. HR, hazard ratio.
Young-old elderly (aged 65–74 years)   
Low baseline medical expenditure, n at risk (person-years)30 (176.5)310 (1883.5)105 (636.2)
No. long-term care introduction (/1000 person-years)4 (22.66)9 (4.78)3 (4.72)
Sex- and age-adjusted HR (95% CI)4.91 (1.50–16.00)**1.001.09 (0.29–4.09)
Multivariable, HR (95% CI)4.90 (1.46–16.49)*1.000.92 (0.24–3.52)
High baseline medical expenditure, n at risk (person-years)23 (131.6)365 (2229.2)167 (1016.4)
No. of long-term care introduction (/1000 person-year)2 (15.20)11 (4.93)9 (8.85)
Sex- and age-adjusted HR (95% CI)2.97 (0.66–13.46)1.001.89 (0.78–4.61)
Multivariable, HR (95%CI)3.21 (0.71–14.55)1.001.83 (0.73–4.61)
Old-old elderly (aged ≥75 years)   
Low baseline medical expenditure, n at risk (person-years)43 (219.0)242 (1344.6)60 (335.9)
No. long-term care introduction (/1000 person-years)11 (50.23)43 (31.98)11 (32.75)
Sex- and age-adjusted HR (95% CI)1.29 (0.65–2.55)1.001.12 (0.57–2.18)
Multivariable, HR (95% CI)1.31 (0.64–2.69)1.000.97 (0.49–1.91)
High baseline medical expenditure, n at risk (person-years)28 (135.5)157 (835.3)50 (275.1)
No. long-term care introduction (/1000 person-years)5 (36.90)39 (46.69)15 (54.53)
Sex- and age-adjusted HR (95% CI)0.74 (0.29–1.88)1.001.10 (0.60–2.03)
Multivariable, HR (95% CI)0.71 (0.28–1.82)1.001.16 (0.63–2.13)

Discussion

Regarding the young-old elderly, underweight was associated with an elevated risk of long-term care introduction, whereas overweight was not. Neither underweight nor overweight was significantly associated with long-term care introduction among the old-old elderly.

Several cohort studies have reported that underweight is a predictive factor of decline in physical functioning.[5, 16-18] Regarding the Asian populations, Ho et al. reported a significant association between underweight (BMI <20 kg/m2) and a decline in physical functioning in the elderly aged 70 years or older living in Hong Kong.[16] In addition, BMI <18.5 or 20 kg/m2 is an international indicator of undernutrition to predict a decline in physical and mental functioning,[19] and also BMI <18.5 is a screening criterion in Japan to identify candidates who require intervention to improve nutrition.[2] Such a decline in physical functioning and undernutrition might increase the risk of long-term care introduction. Furthermore, regarding the underweight young-old elderly, the long-term care expenditure in the month of certification for long-term care insurance was more expensive than those of greater weight. This result may indicate that decreased ADL was more serious for underweight people than for those of greater weight, although we cannot conclude this due to the very small number of certified participants in this BMI category among the young-old elderly. The possible reasons why the risk of underweight was not clear in the old-old elderly in this study are: (i) people who were not underweight might have a manifested a decline in physical functioning, such as senility; (ii) the influence of other diseases of which underweight is not a risk, such as cardiovascular diseases, might have become stronger along with aging; and (iii) some high-risk underweight participants might have already dropped out from participating in a health check-up program because of some physical or mental problems before they had reached the old-old generation; that is, the risk of long-term care introduction in the underweight old-old participants of a health check-up program might not represent the risk in the underweight old-old elderly living in the community.

In the present study, both underweight men and women among the young-old elderly showed an elevated risk of long-term care introduction, whereas Hirai et al. reported that this association was seen only in men.[20] Unlike the present study, their study distinguished neither the young-old from the old-old, nor the normal-range weight from the overweight. Therefore, the difference in the methodology could explain the difference in the results.

As a reason for the positive association between being underweight and the risk of long-term care introduction, one may point out the possibility that patients with chronic diseases that cause individuals to be underweight are more likely to be included in the underweight participants at baseline. However, in the present study, baseline medical expenditure in the underweight group was the lowest between the three BMI categories. Furthermore, the risk of long-term care introduction associated with being underweight was high in the young-old participants with low baseline medical expenditure. Therefore, it seems unlikely that pre-existing chronic diseases that cause individuals to be underweight confounded the results. One may also point out the possibility that people with low economic status, dementia or depression might not be interested in health-seeking behaviors, despite having chronic diseases, which could influence the apparently low medical expenditure in the underweight elderly. However, we believe that the probability of such bias is not so high, because the study participants were the participants of a health check-up program; most of them should have participated in the program in order to maintain or improve their health status.

Obesity is a risk factor for cardiovascular diseases and can increase the risk of long-term care introduction through the onset of these diseases. The most common cause of long-term care in Japan is cerebrovascular diseases, and their most important risk factor is high blood pressure. According to the present study results at baseline, similar to previous reports,[11-13] blood pressure and the prevalence of treated hypertension tended to increase with BMI. Furthermore, overweight participants showed a slightly elevated risk of long-term care introduction among the young-old elderly, although it was not statistically significant. According to some reports, the association between long-term care introduction and risk factors for lifestyle related diseases is relatively weak in the elderly.[21-24] Therefore, the number of participants in the present study could have been insufficient to detect an association between overweight and the risk of long-term care introduction. To clarify this association, a larger scale cohort study is necessary.

As this cohort study used data administratively accumulated in the past, there were some limitations. First, although the data of health check-ups were used as baseline data, we did not obtain treatment statuses except for hypertension. Accordingly, as an index of the severity of treated diseases at baseline, we had to use medical expenditure data. However, we were still unable to identify the diseases that required medical care from the expenditure database. Second, data regarding the diseases that caused the need for long-term care were not obtained. Therefore, we had no choice but to guess regarding the cause of long-term care introduction. To resolve these problems, it is necessary to carry out prospective cohort studies with planned data collection, instead of using existing data. Third, some elderly people with physical impairment or dementia might be cared for by their relatives without using long-term care insurance services. Additional registration systems will be necessary in future studies in order to include such subjects in assessing the risk of long-term care introduction.

The present study has shown that underweight could be a good predictor of long-term care introduction in the young-old elderly who received health check-ups. In Japan, overweight individuals are targeted to prevent lifestyle-related diseases. However, we should also pay attention to underweight in the elderly, as it might be a manifestation of some physical or mental problems that lead to future long-term care introduction.

Acknowledgments

We are grateful for the cooperation of Tsunan town and Sekikawa village during the data collection. This study was supported in part by research grants from the Ministry of Health, Labour, and Welfare in the fiscal years 2006 and 2007.

Disclosure statement

The authors declare no conflict of interest.

Ancillary