We carried out a cross sectional population based study which enrolled 1200 community dwelling elderly individuals aged 65 years and above, living in rural areas of seven out of the eight Mohafazat in Lebanon.

The sample was selected through multistage cluster sampling. The sample size has been established according to the prevalence of home-living elderly people, either suffering from malnutrition or considered at risk of malnutrition (average 40%). These results are based on a review of 48 studies including either healthy or frail elderly individuals 19 and a retrospective study of pooled data 20 based on the Mini-Nutritional-Assessment (MNA) 21 . Although some of the study samples were of convenience nature and mainly focused on European population, we considered these results as the most appropriate for our sample calculation, as representative data from developing countries are lacking. A minimum of 1024 subjects was necessary to establish a 95% confidence interval, with an accuracy of +/− 3%, taking clustering into account. Due to possible missing values in several items, a final sample size of 1200 elderly was chosen.

Lebanon is divided into eight Mohafazat (governorates); each of them consists of several districts (Caza) forming a total of 25 Caza. As our study included only rural elderly subjects, the Caza of Beirut (urban area) was excluded. In each of the remaining 24 Caza (stratum), two villages were randomly selected from the list of villages provided by the Central Agency of Statistics in Lebanon, except for two Caza where only one large village was selected giving a total of 46 villages 22 . Within each village, a random sample of 25 elderly individuals was drawn from the small villages and 50 from the larger villages, based on the list of households provided by the municipality or other local authority. A replacement list was prepared in case of absence or refusal of participation. The inclusion criteria were: to be at least 65 years old, to live at home in rural areas, to be free from terminal illness and not tube fed. The study received the approval of the ethics committee of St Josephs University of Beirut.

The study was based on a comprehensive multi-component questionnaire, administrated by trained interviewers, including the assessment tools as described below. The questionnaire was translated back and forth from French to Arabic by two persons fluent in both languages. People were questioned after oral consent at their place of residence. Written consent was not considered necessary because it was an observational study. Also, participants remained anonymous and individual results were kept confidential. A pilot study including 100 individuals was performed previously in order to pretest the feasibility of the questionnaire. According to the results, some minor changes were made. The interview duration of 40–50 minutes was considered acceptable. If the participant was unable to answer, the help of a family member was required. The main survey was conducted over a period of 12 months between April 2011 and April 2012.

The variables recorded included demographic characteristics (age, gender, marital status, village of residence) and living conditions (living alone or with others). Information about the financial situation was recorded by two questions: one question regarding the personal monthly income which, according to the national minimum wages, was categorized into < 300.000 LL (~200 USD), 300.000 – 600.000 LL and > 600.000 LL (~400 USD). According to a survey conducted by the Central Administration of Statistics of Lebanon in 2004, about 8,7% of Lebanese adults live below the extreme poverty line corresponding to an annual income of 2400 USD (200 USD monthly) per household 23 . The second question asked about financial dependency from children, with answers ranging between “totally dependent, partially dependent or independent”. Information about educational level was obtained using the following categories: illiteracy, less or at least 8 years schooling (primary school), less or at least brevet (middle school), less or at least high school graduation and university level. Regarding the main occupation, individuals were questioned about the longest occupation held, which was categorized into: farmer, employee or manager, self-employed, without work (including housework); they were also asked if they were still working. The question about health insurance was categorized into three groups: without insurance, private insurance, national social security fund (NSSF) and others.

Weight was taken in light indoor clothes without shoes by electronic digital scale to the nearest 0.1 kg, whereas height was measured in standing position to the nearest 0.1 cm. Body Mass Index (BMI) was computed as [weight (kg)/height (m²)]. Nutritional status was assessed by the MNA in its Arabic version 24 . The MNA, developed by Guigoz et al. 25 , is the most established, best validated and widespread nutritional assessment tool used among geriatric population 26 . It has been translated into more than 20 languages and is cited in nearly 200 publications. The MNA includes 18 questions regarding anthropometric, general, dietetic, and subjective assessment. Based on the total score, subjects were classified into three categories: malnourished (<17), at nutritional risk (17 ≤ score < 24), adequate nutritional status (≥ 24) 25 .

Health related characteristics were assessed by self related health status (SRH) based on a 5 item scale. This measure has shown to be a reliable indicator for overall health status 27 in developed countries but also in most Arabic countries 28 . Comorbidities were recorded by asking participants if they suffered from chronic physician-diagnosed conditions such as hypertension, diabetes, etc. Drug intake was estimated by the number of drugs taken daily on a regular basis as prescribed by a physician and checked with packages shown to the interviewer. Furthermore, the participants were questioned about chronic pain (yes/no), defined as feeling pain for at least 3 months, insomnia (no or occasionally/often or always) as well as recent hospitalization (<1 year). Frailty was assessed by the SOF (Study of Osteoporotic Fractures) frailty Index 29 , which included 3 items: involuntary weight loss for one year, inability to rise from a chair without using arms and reduced energy level for at least 3 days during the past week. Based on the original SOF frailty index, frailty status was defined as robust (0 component), pre-frail (1 component), and frail (2 or more components). Oral health assessment included three dichotomous questions about chewing problems, total or partial loss of dentition and wearing dental prosthesis. Finally, tobacco use was estimated by asking about current smoking, whereas the level of physical exercise was assessed by estimating the frequency and duration of exercise (excluding housework) by “a minimum of 30 min daily”, “occasionally” or “never”.

Functional status was investigated through the Activities of Daily Living (ADL), a 5-item scale commonly used in comprehensive geriatric assessment evaluating the basic activities like bathing, toileting, clothing, walking and eating by his/her own 30 . This scale was validated in its Arabic version by Nasser & Doumit 31 in a sample of Lebanese elderly living in nursing homes. Continence was not considered in this scale because difficulties in bladder or bowel control reflect an abnormality in a particular physical system and should therefore be considered as impairment rather than a disability 32 . According to several authors 33 34 , we defined three main groups: “not disabled” was defined as independent in all 5 ADL, “moderately disabled” as dependent in one or two items, “severely disabled” as dependent in three or more items. The Instrumental ADL (IADL) included four activities: telephone use, use of transportation, responsibility for drugs intake and budget management. This 4 item IADL scale has been shown to be associated with cognitive impairment in community dwelling elderly subjects 35 . Individuals were considered as fully independent (coded 0) if they could perform the IADL item without any help, otherwise they were considered partially dependent (coded 1). The final score, ranging from 0 to 4, was computed by summing the number of IADL dependencies for these four items. Subjects without telephone (n = 288) as well as subjects free of drug intake (n = 152) were not considered in the analysis. Balance disorder was assessed by the “one-leg balance” test, a useful screening tool of fall risk among elderly individuals 36 . Subjects were asked to stay on one leg without using their arms. Moreover, participants were asked if they experienced one or more falls during the past year.

Mental status was assessed by the 5 item Geriatric Depression Scale (GDS-5), a dichotomized 5- item scale (score ranges from 0–5) allowing to detect depressive disorder in elderly people 37 . Presence of depressive symptoms was defined as a score of two or above 37 . The 5-item WHO Well Being Index 38 was used to assess the mood of our study sample, as this instrument had previously been validated in the Arabic version by Sibai et al. 38 among Lebanese elderly. The WHO-5 Arabic version allows the detection of depression among Lebanese elderly at a cut-off point less than 13 38 . Cognitive status was assessed by the Mini-mental-state (MMS) examination 39 , the most commonly used screening tool for cognitive impairment worldwide. In order to take into account the high level of illiteracy, we constructed a modified version adapted to illiterate subjects (MMS 2), whereas the original MMS (MMS 1) translated in Arabic language was used for literate elderly. In the MMS 2 form, question 28 including a written order (“Close your eyes”) was replaced by the same order but given verbally (the interviewer asks the subject to close the eyes). Furthermore, regarding question 29, where the patient is required to write a sentence, the participant was asked to construct a sentence orally including a subject, a verb and an object. As no cut-off points were defined in Lebanese elderly, the results were divided into quartiles. To investigate the social network we used the Lubben Social Network Scale 6 (LSNS 6) as described by Lubben et al. 40 . This tool is an abbreviated version of the original LSNS scale 41 , which was especially developed for elderly populations and has been shown to be associated with a wide range of health indicators. The LSNS 6 is based on 3 questions assessing the family network, as follows: “How many relatives do you see or hear from at least once a month? How many relatives do you feel close to such that you could call on them for help? How many relatives do you feel at ease with that you can talk to about private matters?” These same questions are repeated by replacing the word “relatives” with the word “friends”. The answers were as follows: none (coded 0), one (coded 1), two (coded 2) three or four (coded 3), five to eight (coded 4), nine or more (coded 5). The total score is the sum of the 6 items, ranging from 0 to 30. According to the author 40 , at a score below 12, the person is considered as at risk for social isolation. Subjective loneliness was assessed by the modified version of the Jong- Gierveld Loneliness Scale as described by Wilson et al. 42 . This 5 item scale included the following: “I experience a general sense of emptiness,” “I miss having people around,” “I feel like I don’t have enough friends,” “I often feel abandoned,” and “I miss having a really good friend”. As Wilson 42 , we used a dichotomous scale where “yes” was scored for 1 and “no” for 0. Following the authors’ instruction, higher values indicated more loneliness.

The Statistical Package for Social Sciences (SPSS) version 19.0 was used to enter and analyze data. Cluster effect was taken into account when computing confidence intervals, according to Rumeau-Rouquette et al. 43 . Percentages were used to present nominal variables, while means and standard deviations were applied for presenting continuous variables. The Chi Square test was used for cross tabulation of qualitative variables in bivariate analysis, while the Student T test was used to compare the means between genders.

Table 1 presents the socio-demographic characteristics of the study participants. The sample included 591 men (49.2%) and 609 women (50.8%). The mean age was 75.3 (SD = 7.1) years and similar in both genders. Nearly 60% of the participants were Christians, 27.7% were Muslims and 13.7% Druze. About 10% of the elderly individuals lived alone; within these, elderly women were three times more likely to live alone than men. Most of the men (84.8%), but only 44.2% of women were still married. Women were two times more likely to be illiterate than men and to suffer from worse financial status. About 46.8% of the participants had a monthly income of less than 300.000 LL and two thirds were partially or totally dependent on their children. More than 40% of the study sample did not have any health insurance. Regarding the current work-status, nearly 30% of men were still working.

¹Number of respondents to this question; ² NSSF (national social security fund) and others.

Table 2 shows the results of the nutritional assessment and the health status of the participants. The mean BMI of our study sample was 27.5 (SD = 5.4) and was significantly higher in women than in men. Obesity was more prevalent in the female population compared to men. According to the MNA cut off points, 8.0% (95%CI 4.9% - 11.1%) of the individuals were considered as malnourished whereas 29.1% (95%CI 24.0% - 34.2%) were at risk for malnutrition. Poor nutritional status, defined as malnutrition (MN) and risk of malnutrition was significantly more frequent in females (MN: 9.1% [95%CI 4.5% - 13.7%], risk of MN: 35.3% [95%CI 28.2% - 42.6%]) than in males (MN: 6.9% [95%CI 2.8% - 11.0%], risk of MN: 22.9% [95%CI 16.1% - 29.7%]), (p < 0.001). Regarding the SRH scale, the original 5 item responses were categorized into good (good or very good), average and poor (bad or very bad). About 20% of the individuals reported poor health, whereas nearly 40% described their health as good. Women were two times more likely to consider their health as poor compared to men (p < 0.001). Comorbidity was particularly high in this population with 54.7% reporting more than three chronic diseases. The most prevalent disease was hypertension, followed by diabetes, dyslipidemia and cardiovascular disease (Figure 1). Women showed higher prevalence of the most chronic, age-related diseases (p < 0.001). Women also reported a particularly high drug intake (70.7% took more than 3 drugs daily), high frequency of chronic pain and insomnia. Considering frailty, 44.6% of elderly females, but only 28.1% of males, were considered as frail. Regarding oral health status, about two thirds of the participants, especially females, were partially or totally edentulous.

¹Number of respondents to this question.

Functional status of the participants is shown in Table 3. Almost 76% of the participants declared being able to perform the basic ADL. Males showed a higher level of functional ability (ADL and IADL items) compared to women. For example, regarding the IADL, males reported a 3 times higher level of independency, whereas women were two times more likely to be dependent for at least 3 IADL compared to men. Furthermore, female participants reported 1.7 times more falls than their male counterparts. In return, 61.1% of males were able to perform the “one leg balance” test but only 37.3% of females (p < 0.001). Table 3 also reports the psycho-cognitive status and social network of the study participants. Based on the 5 item GDS score, it appears that women were 1.5 times more likely to express depressive symptoms compared to men (49.6% vs 31.9%). Similar results were obtained based on the WHO-5-A score. Regarding cognitive impairment, two groups have to be considered: among the literate group (MMS1), 76.8% of the participants reached a score ≥ 24 (upper quartile), whereas in the illiterate group (MMS2) the majority of the population (68.9%) met a score of less than 24. In both groups, cognitive dysfunction was highly prevalent among women; when compared to men, women of the literate group were at least 2 times more likely to suffer from cognitive impairment compared to males. Risk of social isolation was commonly observed; in fact, nearly half of the individuals had to be considered as socially isolated (Lubben score <12). Regarding subjective loneliness, the original total score (as described by Wilson) ranging from 0 to 5, was categorized into 0–1, 2–3 and 4–5; we found that women were 1.5 times more likely to experience emotional loneliness than men (p < 0.001).

¹Number of respondents to this question; ²IADL = Instrumental activities of daily living: missing values due to participants without telephone or with no drug intake; ³GDS score > =2: depression; ⁴WHO-5A- score <13: depressive disorders; ⁵MMS 1: original version; ⁶MMS 2 = adapted for illiteracy; ⁷Lubbenscore < 12: risk for social isolation; ⁸Wilson score: higher score = emotional loneliness.

When comparing sample characteristics based on cognitive status, we found important differences between the two groups (cf Table 4). Mean age was significantly higher in cognitively impaired elderly individuals (m = 76.4 y, DS = 7.2 y) compared to those with normal cognitive status (m = 73.8 y, DS = 6.6 y). In addition, significantly higher frequencies of cognitive dysfunction were found in following categories: being female (74.2%), widowed (73.9%), illiterate (86.2%), having a monthly income < 300.000 LL (71.6%), being without health insurance (61.5%). Moreover, individuals with lower occupational status had significantly lower cognitive functioning than employees, managers or those who were self-employed. Regarding nutritional status, a very high proportion of malnourished subjects (93.3%) were suffering from cognitive decline, whereas among those who had normal nutritional status, only 42.7% were found to be cognitively impaired (p < 0.001). When considering the health related variables, both mean number of chronic diseases and mean drug intake was significantly higher in cognitively impaired elderly. Furthermore cognitive dysfunction was significantly more frequent in following categories: frail elderly (74.9%), fallers (72.6%), subjects reporting hospitalization during last year (65.4%), and those reporting oral health problems. Finally, in those with the highest ADL disability (dependent for ≥ 2 ADL) and IADL disability (dependent for 4 IADL), nearly 90% showed cognitive decline (p < 0.001).

¹The total number of respondents may not add up to 1189 due to missing values; ² MMS 1 (original version) and MMS 2 (adapted for illiteracy) were computed; ³ GDS score > =2: depressive symptoms.