A cross-sectional study was carried out in Taiwan. Data were obtained from the Mei Jau (MJ) Health Management Institution, which is a private health evaluation provider with multiple health screening centers in Taiwan and Asia. The multiple centers of the MJ Health Management Institution in Taiwan located in Taipei, Taoyuan, Taichung and Kaohsiung cities provide comprehensive health management services for their members in Taiwan.

This study included data of 482,641 Taiwanese subjects between 2001 and 2010, extracted from the MJ Health Management Institution database. The sample size was 316191 after excluding subjects with multiple entries, and included 19,215 participants aged 65 years or above. Of the 19,215 participants, 6,872 participants were further excluded due to missing data. After screening for exclusion of chronic diseases such as cardiovascular disease, cerebrovascular disease, diabetes mellitus or renal failure, and all types of cancer (n = 4,396), 7,947 adults aged 65 years or above were finally recruited for analyses (Fig. 1).

This study used the data extracted from the MJ Health Management Institution database. All participants read, understood and signed informed consent forms before data collection procedures by the MJ Health Management Institution, and the protocol was approved by the Taiwan National Health Research Ethics. The detailed information concerning the data collection procedure can be found at the Taiwan MJ Health Screening Centers. The MJ Health Management Institution delinked the data, and the data were anonymous without any information regarding the identity of participants. Additionally, this study was approved by the Taipei Medical University-Joint Institutional Review Board (TMU-JIRB No N201612035).

Socio-demographic characteristics, lifestyle and health status were collected using a self-reporting questionnaire, or was recorded face-to-face by the nurse prior to health screening. Socio-demographic characteristics included age, gender, occupation, education level (uneducated with illiteracy or educated with elementary education or above), marital status and annual family income. Lifestyle factors, including cigarette smoking (no or yes), alcohol drinking (no or yes), betel nut chewing (no or yes) and daily sleeping duration (< 6 hours, 6 to 8 hours or ≥ 9 hours) were recorded. Health status such as morbidity, medication use and swallowing difficulty (dysphagia) was also analyzed in this study.

The nutritional status of the elderly was defined by using anthropometric indicators, biochemical data and dietary intake scores. In terms of anthropometric indicators, height (cm) and weight (kg) of older adults were obtained from medical records, and body mass index (BMI) was computed by dividing the body weight in kilograms by the square of body height in meters. Underweight was defined as BMI < 18.5 kg/m² as per the guidelines of the World Health Organization (WHO) [26] and the Ministry of Health and Welfare, Taiwan.

Biochemical data included serum albumin and blood hemoglobin (Hb) as protein status and anemia indicators, respectively. Blood samples were obtained during the health screening at the Taiwan MJ Health Screening Centers, and analysis of the collected blood samples was performed at the central laboratory of the MJ Health Management Institution. Serum albumin level < 3.5 g/dL was defined as protein malnutrition. The WHO definition of anemia is blood Hb level < 13 g/dL in men and < 12 g/dL in non-pregnant women [27].

Dietary intake in the past month was assessed using a food frequency questionnaire (FFQ). Fourteen individual food groups were analyzed in this study: rice, bread, instant noodles, root crops, legumes/beans, light-colored vegetables, dark and green leafy vegetables, fruits, milk, other dairy products, eggs, meat, seafood and organ meat. The dietary intake score ranged from 1 to 5, and was assigned for the consumption frequency of each food group, from the lowest to the highest. Consumption frequencies of the various foods were assessed as per their respective criteria, as follows: vegetables, < 0.5 bowl/day, 0.5–1 bowl/day, 1–1.5 bowls/day, 1.5–2 bowls/day and ≥ 2 bowls/day (a bowl equal to the size of 11 cm in diameter); milk, < 1 glass/week, 1–3 glasses/week, 4–6 glasses/week, 1 glass/day and ≥ 2 glasses/day (1 glass is equivalent to 240 mL); fruits and rice, < 1 serving/day, 1–2 servings/day, 2–3 servings/day, 3–4 servings/day and ≥ 4 servings/day; other food groups, < 1 serving/week, 1–3 servings/week, 4–6 servings/week, 1 serving/day and ≥ 2 servings/day. The dietary intake score was calculated according to dietary habits (vegetarians vs non-vegetarians). Vegetarians included 6 categories of food groups: grains (rice, bread and instant noodles), root crops, legumes/beans, vegetables (light-colored vegetables and dark and green leafy vegetables), fruits, and dairy products (milk and other dairy products). Non-vegetarians included 7 categories of food groups: 6 categories included for vegetarians, plus animal products (eggs, meat, seafood and organ meat). Dietary intake score of all categories of food groups was summed up. The minimum and maximum dietary intake scores for grains, root crops, legumes/beans, vegetables, fruits, dairy products and animal products are 3–15, 1–5, 1–5, 2–10, 1–5, 2–10 and 4–20, respectively. Dietary intake scores for vegetarians and non-vegetarians ranged from 10 to 50 and from 14 to 70, respectively. Inadequate dietary intake was defined as a dietary intake score below the median of 30 for vegetarians, and 42 for non-vegetarians. The median was used as a cutoff to define inadequate or adequate dietary intake as reported in previous studies, by using healthy eating index or scores of healthy eating indices [282930].

Data analysis was performed using the IBM SPSS Statistics 21.0 (SPSS Inc., Chicago, IL, USA). Normality test was done by the Shapiro-Wilk test to determine the distribution of outcome variables. The data were normally distributed, and the mean, standard deviation and proportions were used in the descriptive analysis of the variables. Chi-square test for categorical data and Pearson's correlation were performed to assess the relationship between variables. Potential risk factors for malnutrition in older adults were determined by multiple logistic regression analysis. Data are considered statistically significant at an alpha level of P < 0.05.

The socio-demographic characteristics, lifestyle and health status of the participants are presented in Table 1. The mean age of the participants was 70.1 (SD = 4.5) years, ranging from 65 to 92 years. Majority of the participants were female (56.1%), retired or unemployed (91.7%), educated (80.9%), married (70.3%), and having annual family income less than 400,000 NTD (13,200 USD, 56.1%). Considering the lifestyle factors, most participants were non-smokers (76.0%), non-drinkers (57.9%), non-betel nut chewers (70.1%), and having daily sleep duration of 6-8 hours (50.6%). Around half the participants took 2 to 4 different medicines (48.0%) and were afflicted with 2 to 4 diseases (50.5%). Only 4.8% of the participants experienced dysphagia.

The distribution of education, smoking, alcohol drinking, betel nut chewing, daily sleep duration, medication use, and morbidity were significantly different between vegetarians and non-vegetarians (P < 0.001). Compared to non-vegetarians, a higher proportion of vegetarians were educated (92.9% vs 77.6%), smokers (27.3% vs 23.1%), drinkers (76.8% vs 32.8%) and betel nut chewers (54.7% vs 23.3%), had daily sleep duration of less than 6 hours (85.8% vs 39.6%), used multiple medications (80.2% vs 39.4%) and experienced comorbidity (53.3% vs 49.8%).

Among the elderly subjects, the prevalence of underweight, protein malnutrition, anemia and inadequate dietary intake was 20.2%, 6.6%, 10.5% and 52.5%, respectively (Table 2). Anemia was prevalent in 70.1% female subjects, and inadequate dietary intake included 77.4% non-vegetarians.

The correlation coefficients between variables and nutritional status are presented in Table 3. Age shows a negative correlation with body weight (r = −0.19, P = 0.02), BMI (r = −0.41, P < 0.001), albumin level (r = −0.93, P < 0.001) and hemoglobin level (r = −0.30, P = 0.008). Except for no significant association between BMI and dietary intake score, significantly positive correlations were observed among all nutritional status variables such as body weight, BMI, albumin level, hemoglobin level and dietary intake score at P < 0.001.

Factors associated with inadequate nutritional status were analyzed by odds ratios and are presented in Table 4. Male gender (OR, 1.27; 95% CI, 1.16–1.39), retired (OR, 1.10; 95% CI, 1.02–1.39), unemployed (OR, 1.19; 95% CI, 1.05–1.41), uneducated (OR, 1.68; 95% CI, 1.19–1.90), smoking (OR, 1.19; 95% CI, 1.06–1.32), vegetarian diet (OR, 1.19; 95% CI, 1.01–1.62), slept between 6 and 8 hours a night (OR, 1.16; 95% CI, 1.06–1.30), or used multiple medications (OR, 1.34; 95% CI, 1.02–1.85) were associated with the risk of being underweight. However, those who chewed betel nut (OR, 0.83; 95% CI, 0.43–0.98) were less likely to be underweight compared to those who did not chew betel nut.

The odds of protein malnutrition were significantly higher for drinkers (OR, 3.28; 95% CI, 1.11–9.70), vegetarians (OR, 3.13; 95% CI, 1.03–9.48), or those who had comorbidity (OR, 2.64; 95% CI, 1.01–6.85). The risk of being anemic was higher in older adults aged > 70 years (OR, 1.62; 95% CI, 1.39–1.90), uneducated (OR, 1.20; 95% CI, 1.02–1.44), unmarried (OR, 1.19; 95% CI, 1.01–1.40), drank (OR, 1.15; 95% CI, 1.04–1.38), or had comorbidity (OR, 1.30; 95% CI ,1.11–1.76). However, males (OR, 0.47; 95% CI, 0.40–0.55) were less likely to be anemic compared to females.

Inadequate dietary intake was more prevalent in adults aged > 70 years (OR, 1.20; 95% CI, 1.07–1.98), uneducated (OR, 1.89; 95% CI , 1.06–2.12), smoked (OR, 1.50; 95% CI, 1.10–2.05), drank alcohol (OR, 1.25; 95% CI , 1.01–1.79), consumed vegetarian diet (OR, 1.31; 95% CI , 1.18–2.04), were in low income group (OR, 1.12; 95% CI, 1.07–1.94), or had dysphagia (OR, 1.34; 95% CI, 1.15–2.56).