Depression, malnutrition, and health-related quality of life among Nepali older patients

This study, abbreviated as NepEldQOL I, supplements our previous study, NepEldQOL II [14]; together these studies provide the most comprehensive portrayal of the well-being of Nepal’s older population to date. The current study was conducted in January–April of 2017 in the outpatient department (OPD) of Nepal Medical College and Teaching Hospital (NMCTH) in Kathmandu, Nepal. According to hospital administration data, NMCTH had a total of 138,684 outpatient visits in 2016–2017; of these, 16,567 visits were among patients aged 60 years and above.

The required sample size of 289 was calculated by using StatCalc in Epi Info 7 based on a 24% prevalence of malnutrition among Nepali older adults [6], 5% alpha or Type I error and a 5% margin of error. Surveyors were graduate students in medicine and public health who were provided with a one-day extensive orientation on the study tools, sampling strategy, and data collection techniques. Each day, participants were selected by systematic random sampling from the daily “first-come first served” OPD sign-in lists, since NMCTH, like most health service facilities in Nepal, uses only a walk-in appointment system for all new or revisiting patients. The daily OPD record (updated continuously throughout the day) served as the population sample reference frame for each day of data collection, which took place during normal business hours from January to mid-April, when the required sample size was achieved. Surveyors identified patients ages 60 and above from the OPD list, and subsequently approached every third patient to screen for eligibility. Eligible participants were 60 years or older and present in the OPD one on the data collection days. Those too frail physically or mentally to respond, and/or with hearing or speech impairment were excluded. Consent to participate was requested from all eligible patients. If the approached patient was deemed ineligible or if s/he refused to participate, the next eligible patient from the OPD list was approached. Of the 297 patients approached, eight eligible patients refused to participate; the remainder granted consent.

Individual face-to-face interviews were conducted with patients in the waiting area of the OPD. The survey tools were translated from English to Nepali by the second author and then verified by back-translation to English by another author, following the protocol for translations [15]. Any disagreement on translation was resolved by adjudication by a third author and mutual consensus between the three authors.

Nutritional status was assessed using the Short Form Mini Nutritional Assessment (MNA), validated previously among Nepali older adults [6]. For the measurement of BMI, the surveyors measured each participant’s height with a mechanical stadiometer (Prestige HM 007) and weight with a digital weighing scale (SECA GMBH & Co Model: 874). BMI was calculated as weight in kg/(height in m)² which was categorized as per instructions in the MNA guide [16]. The cumulative MNA score ranges from 0 to 14. A score of less than 8 indicates malnourishment, a score between 8 and 11 points indicates that the subject is at risk of malnutrition and a score of 12 or higher indicates a normal nutritional status. Details on the MNA tool are provided elsewhere [16]; briefly, the MNA short form consists of six items: decline in food intake, involuntary weight loss, mobility, psychological stress, neuropsychological problems and body mass index (BMI). While the World Health Organization defines malnutrition as “deficiencies, excesses, or imbalances in a person’s intake of energy and/or nutrients” [17], we use the words “nutritional status” and “malnourished” throughout the manuscript to reflect only deficient nutritional status in order to be consistent with the MNA tool. The Cronbach’s Alpha, which is the measurement of scale reliability, for the MNA scale in the current study was 0.59.

The Geriatric Depression Scale Short Form (GDS) was used to measure depression [18]. The GDS has been described in detail elsewhere [18]; briefly, it is a 15-item instrument with responses in “Yes/No” format. Of the 15 items, 10 indicate the presence of depression when answered positively while the other five are indicative of depression when answered negatively (reverse coded for cumulation). A cumulative score of 5 or more suggests depression [19]. The validity and reliability of GDS to measure depression among community-dwelling Nepali older adults has been established by a previous study [20]. The Cronbach’s Alpha for the GDS was 0.81 in the current study.

HRQOL was assessed using the European quality of life tool (EQ-5D) [21]. The Nepali versions of the EQ-5D tools have been validated in a previous study [22]. The EQ-5D allows participants to classify their health status in five different dimensions (i.e., mobility, self-care, usual activities, pain/discomfort and anxiety/depression) and within a three-level response (no problems, moderate problems, and severe problems) [21]. The five dimensions of the ED-5D are then converted into a single index value, called EQ-5D index, by using the “EQ-5D-3L crosswalk index value calculator” [23], using United Kingdom (UK) weights as the reference. The EQ-5D index ranges from 0 to 1, where 0 indicates severely ill, and 1 indicates perfect health. Perfect health is represented by no problems across all five dimensions; severely ill corresponds to severe problems on all five dimensions of EQ-5D.

Additionally, EQ-5D has a vertically calibrated scale called the EuroQol visual analytic scale (EQVAS), which allows participants to rate their overall health on a scale ranging from 0 to 100, where 0 and 100 signify the worst and the best imaginable health state level, respectively. Participants rated their overall health in EQVAS at the level they felt best described their health on the study day. For this study, the Cronbach’s Alpha of the EQ-5D scale was 0.79.

Sociodemographic variables were self-reported and included age, gender, ethnicity, religion, marital status, educational status, occupation, monthly family income, family structure, smoking, and alcohol use. For ethnicity, the Nepal Health Management Information System’s ‘caste/ethnic groupings’ were used [24]. Due to sparsity in certain categories, related categories were combined to form three categories: Upper caste, Janjatis, and Dalit/other minorities. Education status was categorized into three groups: illiterate; informal (no formal schooling, some literacy); and formal education (any years of formal schooling). Occupation indicated the primary occupation of the participant in the past or the current occupation if currently employed (n = 27). Adequate rest was defined as sleeping for more than six hours daily. Information on primary caretakers was collected through open-ended questioning and recoded into two categories: son and daughter-in-law as the first; others as the second.

Statistical analyses were performed in IBM SPSS v22. Numerical variables are expressed as a mean and standard deviation (SD); categorical variables as frequency and percentage. Comparisons of means between the groups were made by independent t-tests or analysis of variance, while frequency distributions were evaluated by Pearson’s chi-square (χ²) test. The Spearman correlation coefficient (ϱ) was calculated to estimate the correlation of EQVAS with total MNA and GDS scores. The factors associated with nutritional status, depression, and HRQOL were assessed by three separate linear regression models, each adjusted for age and sex, using the total score of MNA, GDS, and EQVAS as the dependent variable, respectively. To account for non-normally distributed outcome variables and a relatively small sample size, bootstrap models with 5000 replications were used for calculating stable estimates of correlates and their bias accelerated and corrected (BCa) 95% confidence intervals.

Two different mediating models were developed for the depression-nutrition-HRQOL triad (Fig. 1): the first uses total MNA score, or nutritional status, as the mediator (Fig. 1a) while the second uses total GDS score, or depression status, as the mediator (Fig. 1b). For moderation analyses (Fig. 2), an interaction between depression and nutritional score was added to a base regression model with depression and nutritional score as independent factors; EQVAS was used as the outcome. The PROCESS macro for SPSS was used for the mediation-moderation analyses. The mediation analyses were first run without any covariates (Model 1); then adjusted for age and sex (Model 2); then further adjusted for ethnicity, marital status, smoking, alcohol use, educational status, perception of negligence/hatred, and perceived health status compared to peers (Model 3). For all statistical tests, two-tailed p-values< 0.05 were considered statistically significant.

Detailed characteristics of the study participants can be found in Table 1. A total of 170 males and 119 females participated. The mean age of participants was 68.5 years, with a range from 60 to 90 years old. Most were Upper caste (46%) or Janjatis (46%), Hindu (76%), married (80%), illiterate (39%), and lived in a joint family (70.6%). The primary past occupation among males was farming (39%); among females it was household chores (48%). Many participants were reluctant to reveal their monthly family income; among respondents (n = 167), the mean monthly family income was $205. Only a small proportion (10%) were actively involved in earnings; over half of the participants were taken care by their son and daughter-in-law. The mean (±SD) BMI, EQVAS, MNA, and GDS scores of the participants were 24.9 ± 3.5, 65.2 ± 16.8, 10.9 ± 2.5, and 5.9 ± 3.8, respectively (Table 1).

Appendix Table 5 provides the detailed characteristics of participants based on nutritional status categories as defined by MNA. The mean MNA score was 11.0 and ranged from 2 to 14. Only about half of the participants had adequate nutritional status; 10% were malnourished, and 38% were at risk of malnutrition. Comorbidity between depression and malnutrition was prevalent among 6.9% of the participants (Appendix Table 5). In the regression analysis adjusted for age and sex (Table 2), age (β = − 0.07; BCa 95%CI = − 0.12, − 0.01), male gender (β = − 0.65; BCa 95%CI = − 1.25, − 0.07), depression score (β = − 0.18; BCa 95%CI = − 0.26, − 0.10), perception of worsened health (β = − 1.04; BCa 95%CI = − 1.73, − 0.33) and perception of ignorance/hatred due to old age (β = − 1.92; BCa 95%CI = − 2.73, − 1.09) were inversely associated with the nutritional score from the MNA. Likewise, having formal education (β = 0.81; BCa 95%CI = 0.18, 1.47) as well as higher quality of life scores on both the EQ-5D index (β = 2.98; BCa 95%CI = 1.51, 4.30) and the EQVAS (β = 0.06; BCa 95%CI = 0.05, 0.08) were associated with a higher nutritional score on the MNA.

Appendix Table 6 provides the detailed characteristics of participants based on depression status as defined by GDS. More than half (57%) of the participants met the criteria of depression (GDS score ≥ 5). In the regression analysis adjusted for age and sex (Table 2), higher age (β = 0.10; BCa 95%CI = 0.03, 0.17), perception of worsened health (β = 1.68; BCa 95%CI = 0.74, 2.60) and perception of ignorance/hatred due to old age (β = 1.58; BCa 95%CI = 0.32, 2.82) were associated with a higher depression score. Compared to illiterate individuals, those having an informal education (β = 1.06; BCa 95%CI = 0.16, 1.96) scored higher on the depression scale whereas those having a formal education (β = − 1.43; BCa 95%CI = − 2.43, − 0.40) scored lower. A higher depression score on the GDS was associated with a lower score on the nutrition scale (β = − 0.41; BCa 95%CI = − 0.59, − 0.23) and lower quality of life scores: EQ-5D index (β = − 2.49; BCa 95%CI = − 4.61, − 0.48) and EQVAS (β = − 0.06; BCa 95%CI = − 0.09, − 0.03).

The responses of participants in the five dimensions of EQ-5D are provided in Appendix Table 7. The mean EQVAS score and the EQ-5D index were 65.2 and 0.8 respectively; scores were significantly lower among participants meeting the criteria for malnutrition or depression. Thirty-six different health statuses were represented in the EQ-5D (Appendix Table 7).

EQVAS scores were positively correlated with MNA scores (ϱ =0.44, p < 0.001) and negatively correlated with GDS scores (ϱ = − 0.28, p < 0.001). In the regression analysis adjusted for age and sex using EQVAS as the outcome (Table 2), a positive association was observed between the EQVAS QOL score and being married (β = 6.57; BCa 95%CI = 1.75, 11.37), having a formal education (β = 5.54; BCa 95%CI = 0.92, 9.99), working currently (β = 11.60; BCa 95%CI = 4.57, 18.07), better perceived health status (β = 11.06; BCa 95%CI = 7.34, 15.03), and higher MNA score (β = 2.87; BCa 95%CI = 2.12, 3.62). Likewise, an inverse association was observed between the EQVAS QOL score and perception of worsen health status (β = − 18.20; BCa 95%CI = − 22.34, − 13.94), perception of being ignored/hated for old age (β = − 8.63; BCa 95%CI = − 13.61, − 3.80), and the depression score (β = − 1.23; BCa 95%CI = − 1.72, − 0.72).

We present a pioneer study that quantifies the HRQOL among urban Nepali older patients, filling gaps and advancing knowledge about the prevalence of and factors contributing to depression and malnutrition among Nepali older patients. Moreover, we explore the relationship between depression, nutritional status, and HRQOL, three important aspects of aging, in the Nepalese context. To our knowledge, the mediation and moderation effects of depression and nutrition with the outcome of HRQOL among the older adults have not been previously explored in any context. HRQOL, looking at health from an individual’s perspective, is truly multi-faceted as seen in this study. Simple measures to detect and treat depression among the older adults to improve their overall HRQOL should also examine nutritional wellbeing. Further prospective studies are needed to identify the direction of the relationship between depression and nutrition.

Nonetheless, this study is subject to some limitations, including a relatively small sample size. Due to our cross-sectional study design, no inferences can be made regarding the causal relationships between nutritional status, depression, and HRQOL. Future studies should determine if nutritional risk is associated with QOL over time among older adults. This study recruited participants from outpatient clinics in an urban setting; the nutritional status, depression, and HRQOL of the general population in an urban area and/or those in a rural setting may be different, thus limiting the generalizability of our findings. Exclusion of older patients who were too frail physically or mentally to respond may have resulted in a selection bias that underestimated or biased our findings towards the null. In the current study, the internal consistency of GDS and EQ-5D was high but that of the MNA scale was relatively low (Cronbach’s α = 0.59); omission of any MNA component score did not substantially increase the alpha value (data not shown). Given that the MNA has already been validated in various settings [37] as well as among Nepali older adults [6], we nonetheless believe it to be a valid tool to assess nutritional status among the older adults. The use of UK’s general population weights as reference values in the calculation of the EQ-5D index is not ideal; however, no such reference weights exist for Nepal and the same technique was used in the original study validating the Nepali version of EQ-5D [22] as well as another study from Nepal [38]. We defined adequate rest as sleeping for more than six hours daily; however, the National Sleep Foundation recommends 7–8 h of sleep for older adults [39]; thus we may have overestimated the prevalence of adequate rest. Moreover, sleep hours were self-reported. Increasing age is associated with multi-morbidity that may limit functional capacity and reduce the HRQOL [40]; presence of comorbidities were not assessed in this study. Future studies should look at the possible mediating and bi-directional relationship of multiple morbidities, nutritional status and HRQOL. Lastly, the possibility of residual confounding due to unmeasured covariates cannot be ruled out.