The Simplified Nutritional Appetite Questionnaire (SNAQ) as a Screening Tool for Risk of Malnutrition: Optimal Cutoff, Factor Structure, and Validation in Healthy Community-Dwelling Older Adults

Malnutrition is an independent marker of adverse outcomes in older adults. While the Simplified Nutritional Appetite Questionnaire (SNAQ) for anorexia has been validated as a nutritional screening tool, its optimal cutoff and validity in healthy older adults is unclear. This study aims to determine the optimal cutoff for SNAQ in healthy community-dwelling older adults, and to examine its factor structure and validity. We studied 230 community-dwelling older adults (mean age 67.2 years) who were nonfrail (defined by Fatigue, Resistance, Ambulation, Illnesses & Loss (FRAIL) criteria). When compared against the risk of malnutrition using the Mini Nutritional Assessment (MNA), the optimal cutoff for SNAQ was ≤15 (area under receiver operating characteristic (ROC) curve: 0.706, sensitivity: 69.2%, specificity: 61.3%). Using exploratory factor analysis, we found a two-factor structure (Factor 1: Appetite Perception; Factor 2: Satiety and Intake) which accounted for 61.5% variance. SNAQ showed good convergent, discriminant and concurrent validity. In logistic regression adjusted for age, gender, education and MNA, SNAQ ≤15 was significantly associated with social frailty, unlike SNAQ ≤4 (odds ratio (OR) 1.99, p = 0.025 vs. OR 1.05, p = 0.890). Our study validates a higher cutoff of ≤15 to increase sensitivity of SNAQ for anorexia detection as a marker of malnutrition risk in healthy community-dwelling older adults, and explicates a novel two-factor structure which warrants further research.


Introduction
Malnutrition is increasingly recognized as an important and independent marker of adverse outcomes in older adults, including higher chronic disease burden, frailty and mortality [1][2][3]. The process of malnutrition and involuntary weight loss can be driven by anorexia, inadequate dietary intake, sarcopenia, cachexia, or a combination of these factors [4]. Considerable overlap exists between these processes, especially sarcopenia and cachexia in frail older adults with comorbidities and chronic Table 1. Summary of Simplified Nutritional Appetite Questionnaire (SNAQ) Studies [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25]. However, gaps remain in our understanding of the diagnostic performance and optimal cutoff of SNAQ in robust older adults in the community where the prevalence of malnutrition is lower. The good diagnostic performance of SNAQ and the recommended cutoff of ≤14 was derived from community studies which were primarily based on older adults who were either frail with comorbidities [18,20,22] or who were younger in age [17,23]. There also exists uncertainty in the psychometric properties of SNAQ, such that its factor structure and validity in more robust populations cannot be assumed. Earlier studies highlighted that the item on food intake correlates poorly with total SNAQ score [21], and that reliability of SNAQ increases if this item was omitted [15]. In addition, the one-factor solution of SNAQ was derived either from studies in older adults with health-seeking behavior [14] or specialized populations [17,21,23].

Reference
The aim of this study is thus to determine an optimal cutoff for SNAQ in screening for malnutrition in healthy community-dwelling older adults. This study also seeks to examine the factor structure of SNAQ, as well as assess its psychometric properties including validity and reliability.

Study Population
The "Longitudinal Assessment of Biomarkers for characterization of early Sarcopenia and Osteosarcopenic Obesity in predicting frailty and functional decline in community-dwelling Asian older adults Study" (GeriLABS 2) is a prospective cohort study involving cognitively intact and functionally independent community-dwelling adults aged 50 years and older in Singapore. We recruited 230 participants from December 2017 to March 2019. Inclusion criteria were as follows: (i) aged 50 to 99 years at study enrolment, (ii) community-dwelling, (iii) independent in both basic and instrumental activities of daily living (ADLs), and (iv) generally healthy as defined by a score of <3 on the FRAIL criteria. The FRAIL scale comprises 5 components: Fatigue, Resistance, Ambulation, Illnesses & Loss of Weight with a total score of 0-5 points and represents frail (3)(4)(5), pre-frail (1-2) and robust (0) health status [26]. Participants were excluded if they had cognitive impairment (prior diagnosis of dementia or modified Chinese version of Mini-Mental State Examination (CMMSE) score ≤21) [27], unable to walk 8 m independently, or were living in a long-term residential care facility. This study reports cross-sectional data from the point of recruitment into the study.

Clinical Assessment
We collected demographic data and information on cardiovascular and bone health. Anthropometric measurements (standing height and body weight to calculate Body Mass Index (BMI), calf circumference, mid-arm circumference and waist circumference) were collected. Cognition was assessed via the CMMSE [27]. Mood was assessed via the Geriatric Depression Scale (GDS), with a locally-validated cutoff score of ≥4 to distinguish presence of depressive symptoms [28]. Functional status was assessed using the Barthel ADL index [29] and Lawton and Brody's instrumental ADL index [30], while activity level was evaluated via the Frenchay Activities Index (FAI) [31] and International Physical Activity Questionnaire (IPAQ) [32]. We also measured life-space mobility using the Life-Space Assessment (LSA), which comprises five space-levels corresponding to activities outside the bedroom, home, neighborhood, town, and beyond, respectively [33]. Using cohort quintile cutoffs, we defined low physical activity as FAI ≤29 and IPAQ <2826 METS respectively, and low life-space mobility as LSA <76 [34].
Physical frailty was assessed via the modified Fried phenotypic criteria, with a score of 0 denoting nonfrail, 1-2 denoting prefrail, and 3 and above denoting physical frailty [35]. Details of the operationalization of the modified Fried phenotypic criteria have been previously described [36]. Social frailty was assessed via the locally-validated eight-item Social Frailty Scale (SFS-8), which measures the three domains of social resources, social activities and financial resources, and social need fulfilment (score range: 0-8 points). A score of 0-1 denotes social nonfrailty, 2-3 denotes social prefrailty, and a score of 4 and above denotes social frailty [34].
Physical function was assessed via the Short Physical Performance Battery (SPPB) [37], maximal hand grip strength using a hydraulic hand dynamometer, usual gait speed on the three-meter walk test, and the five-time chair stand test. The SPPB score of <11 denoted poorer quality of life and also corresponded to the quintile cutoff in our cohort [38]. Cutoffs for maximal hand grip strength (<28 kg for males, <18 kg for females), gait speed (<1.0 m/s) and five-time chair stand test (≥12 s) were based upon The Asian Working Group for Sarcopenia (AWGS) 2019 guidelines [39].

Nutritional Assessment
Nutritional risk was assessed via the Simplified Nutritional Appetite Questionnaire (SNAQ), a four-item tool comprising items 1, 2, 4 and 6 of the CNAQ [11]. These items assess appetite, satiety, taste of food and number of meals per day respectively. The SNAQ was developed as a self-assessment screening tool that is quick and easy to administer without the need for trained assessors or laboratory measurements. The total score ranges from 4 to 20. Prior validation studies suggest a cutoff of ≤14 to predict malnutrition and involuntary weight loss (Table 1).
We compared SNAQ against the Mini Nutritional Assessment (MNA), which comprises both screening and nutritional assessment items. The MNA has been validated in various settings with high reliability, sensitivity and specificity [40,41]. A cutoff score of <24 on the MNA-long form indicates risk of malnutrition [42]. Laboratory markers of 25-hydroxy Vitamin D and serum albumin levels were also collected.

Statistical Analysis
We performed statistical analyses using IBM SPSS Statistics version 23.0 (IBM Corporation, Armonk, NY, USA). All statistical tests were two-tailed, with p < 0.05 considered statistically significant. Continuous variables were expressed as means (standard error) or as medians (interquartile range). Categorical variables were expressed as counts and percentages.
To derive the optimal cutoff for SNAQ, we performed receiver operating characteristic curve (ROC) analysis against MNA-long form <24 as the reference standard. We calculated the optimal point between sensitivity and specificity using Youden index. Diagnostic performance was ascertained via area under ROC curve.
Internal consistency of the scale was assessed using Cronbach's alpha. To ascertain the factor structure of SNAQ, we conducted exploratory factor analysis (EFA) using the Kaiser-Meyer-Olkin (KMO) statistic as a measure of sampling adequacy and the Bartlett's test of sphericity to ascertain necessity to perform a factor analysis. We conducted principal component analysis with varimax rotation to ascertain the underlying factor structure. The number of factors to be retained was determined by parallel analysis, which was less likely to over-estimate the number of factors [43]. Items with factorial loadings <0.4 were eliminated.
To assess the construct validity of SNAQ, we analyzed convergent (i.e., MNA, calf circumference, GDS) and discriminant (i.e., waist circumference, CMMSE, level of education) validity for SNAQ total and factor scores via correlational analysis using Spearman's rho. For concurrent validity, we ascertained known-groups validity by examining differences in mean SNAQ scores against depressive symptoms (GDS <4 vs. ≥4) via independent samples t-test, as well as social frailty (nonfrail, prefrail and frail) and physical frailty (nonfrail, prefrail and frail) via one-way analysis of variance with Bonferroni correction for post-hoc comparisons. Predictive validity of SNAQ was assessed via logistic regression, with adjustments for age, gender, education and MNA score, for the outcomes of life-space mobility, social frailty, SPPB, handgrip strength and five-time chair stand. We compared the suggested SNAQ cutoff of ≤14 against our derived cutoff of ≤15 in this study.

Ethics Approval and Consent
The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Domain Specific Review Board of the National Healthcare Group (DSRB Ref: 2017/00850). Written consent was obtained from all participants prior to study participation.

Baseline Characteristics
We studied 230 participants with mean age 67.2 ± 7.4 years and mean education of 10.8 ± 4.4 years ( Table 2). A total of 72.6% were female and 92.2% were of Chinese ethnicity. The high cognitive scores (CMMSE, mean ± SD: 26.1 ± 1.7), ADL indices (basic and instructional ADLs, mean score 100 and 23 respectively) and activity levels (FAI mean ± SD: 32.2 ± 5.2 and IPAQ 5023.4 ± 2402.7 metabolic equivalents (METS) per week) attested to the relatively robust health of participants. Based on the modified Fried frailty phenotype, only two (0.9%) participants were classified as physically frail, with the majority (41.3% and 57.8% respectively) classified as prefrail and robust. Based on the SFS-8, 17 (7.4%) participants were classified as socially frail, with 28.8% socially prefrail and 63.8% socially nonfrail. The median GDS score was 1 (interquartile range, IQR = 0-2).    The prevalence of risk of malnutrition in this study was 5.7% based on the MNA cutoff of <24. The risk of malnutrition identified by SNAQ based on cutoff scores of ≤14 and ≤15 was 18.3% and 40.4% respectively. In comparing participants identified by SNAQ to be at risk of malnutrition (i.e., SNAQ total score ≤15, n = 93) against those who were not at risk (n = 137), the differences in age, calf circumference, GDS, MNA, life space (levels 3, 5 and total score), social frailty, SPPB, hand grip strength, repeated chair stand test were significant (all p < 0.05).

Factor Structure and Reliability of SNAQ
Factor analysis was appropriate as the KMO statistic was 0.530, and the Bartlett's test of sphericity was 32.553 (p < 0.001). The optimal number of factors recommended by parallel analysis was two. The two-factor structure of the SNAQ accounted for 61.5% of the total variance. Factor 1 comprised two items representing appetite perception and accounted for 34.9% of variance. Factor 2 comprised two items

Factor Structure and Reliability of SNAQ
Factor analysis was appropriate as the KMO statistic was 0.530, and the Bartlett's test of sphericity was 32.553 (p < 0.001). The optimal number of factors recommended by parallel analysis was two.
The two-factor structure of the SNAQ accounted for 61.5% of the total variance. Factor 1 comprised two items representing appetite perception and accounted for 34.9% of variance. Factor 2 comprised two items representing satiety and intake, and accounted for 26.6% of variance ( Table 3). The Cronbach's alpha of SNAQ was 0.333. There was a slight increase in Cronbach's alpha if item 4 was deleted.

Convergent and Discriminant Validity
SNAQ Factors 1 and 2 correlated strongly with the total SNAQ score (r = 0.767 and 0.689 respectively, p < 0.05). The correlation between Factors 1 and 2 was poor (r = 0.119, p > 0.05) ( Table 4). In terms of convergent and discriminant validity, the strength of correlations was low to moderate in this sample of healthy older adults. SNAQ total and factor scores correlated with MNA, calf circumference and GDS, with stronger correlations for SNAQ total and Factor 1 scores (range of r: 0.151 to 0.238, p < 0.05) compared to Factor 2 (range of r: 0.049 to 0.161, p <0.05 only for correlation with MNA). The correlations for waist circumference, CMMSE and education were weak (range of r: 0.005 to 0.147, p > 0.05 except for Factor 1 correlation with CMMSE), thus corroborating the discriminant validity of SNAQ.  (Table 5). This attests to the concurrent validity of the SNAQ. For the factor scores, there was a similar trend observed for Factor 1, with a significant decreasing trend in Factor 1 scores with increasing social frailty (7.91 vs. 7.70 vs. 6.94, p = 0.001). A similar trend was observed for Factor 2 with physical frailty and depressive symptoms (p > 0.05), but not for social frailty (8.05 vs. 7.86 vs. 8.24, p > 0.05).

Predictive Validity and Outcome Associations
In logistic regression analysis adjusted for age, gender, education and MNA, both SNAQ cutoffs of ≤14 and ≤15 were associated with life-space mobility (odds ratio, OR 2.29 vs. 2.06, p = 0.041 for both), SPPB (OR 3.93, p = 0.003 vs. OR 3.00, p = 0.010) and five-time chair stand test (OR 2.45, p = 0.029 vs. OR 2.18, p = 0.028) ( Table 6). In contrast, only the ≤15 cutoff was significantly associated with social frailty (OR 1.99, p = 0.025 vs. OR 1.05, p = 0.890 for SNAQ ≤14). For the outcome of hand grip strength, the ≤15 cutoff had higher odds ratios albeit not statistically significant.

Discussion
To our knowledge, this is the first study to ascertain the optimal cutoff, factor structure and validity of SNAQ in healthy, cognitively-intact and functionally-independent community-dwelling older adults. Our findings suggest an optimal SNAQ cutoff of ≤15 to screen for risk of malnutrition that is higher than the currently recommended ≤14. In healthy older persons, the higher SNAQ cutoff lowers the threshold for detection of anorexia, thus improving its diagnostic performance as a screening tool by increasing sensitivity to rule out false negative cases of anorexia for further evaluation. In one study involving hospitalized older adults and their spouses, raising the SNAQ cutoff from ≤14 to ≤15 similarly increased sensitivity from 70.8% to 79.2%, albeit also at the expense of specificity [13]. A SNAQ cutoff of ≤15 could thus present an opportunity for early case detection of malnutrition risk in relatively healthy individuals, and consequently early nutrition assessment, education and intervention to prevent adverse health outcomes and allow robust older adults to remain independent in the community [44][45][46].
Our study also found a two-factor structure of the SNAQ, which differs from prior validation studies that suggest a unifactorial model of 'Appetite' [14,17,21,23]. The factors identified in our study (i.e., Factor 1: Appetite Perception; Factor 2: Satiety and Intake) accounted for a higher proportion of the total variance (61.5%) compared with earlier studies (33.7% to 54.0%), and corresponded to the premise that the primary forward indicators of malnutrition risk in healthy older adults are likely to be driven more by anorexia and inadequate dietary intake, than by muscle or disease-related processes such as sarcopenia and cachexia [4]. Between the two factors, convergent and concurrent validity of the SNAQ appear to be driven more by Factor 1 than Factor 2, which is in keeping with prior studies that have identified 'Appetite' as the main outcome measure in the SNAQ. This suggests that while a distinct factor on its own, satiety and intake (i.e., Factor 2) may be less discriminatory in a relatively robust population, and may also be subjected to differences in culture, socioeconomic status or varying perceptions of meal quantity and frequency that are considered to be norms [47][48][49]. The contribution of Factor 2 may also explain the relatively-low reliability of SNAQ in our study (Cronbach's alpha 0.333); in healthy older adults, the brief 4-item SNAQ may not be sufficiently reliable in screening for the risk of malnutrition, and there may be a need to consider supplementing with other items that examine determinants such as the social facilitation of eating and economic determinants of nutrition [50][51][52].
In line with this, we found that SNAQ was significantly associated with social frailty, life-space mobility and muscle function. Notably, a higher SNAQ cutoff of ≤15 showed a stronger association with social frailty compared to the traditional cutoff of ≤14. These observations remained significant even after adjustment for MNA, suggesting that anorexia (as measured by SNAQ) predicts risk of malnutrition above and beyond the effect of MNA in healthy older adults. This finding also reinforces the relationship between appetite and social influences, and echoes increasing evidence that social eating norms play a role in the development of healthy eating behavior and maintenance of adequate nutrition, especially in later life [53,54]. While many factors contribute to poor appetite (e.g., disease, drugs, depression, socioeconomic factors, impaired masticatory function) [55], in robust older adults without comorbid diseases, the predominant precursor of malnutrition risk may be related to the concept of anorexia of aging instead [56]. Anorexia of the aging involves physiological processes such as age-related neurohormonal changes (e.g., decline in central orexigenic neuropeptide activity, increase in cholecystokinin levels), decline in taste and smell, and reduced antral stretch that collectively lead to a notable decline in food intake with age, and represents an entity that exists discrete from pathological causes of anorexia [9,[57][58][59]. Anorexia can lead to malnutrition and weight loss, which in turn increases the risk of frailty, disability and mortality [60][61][62]. This highlights the importance of screening for anorexia as a forward indicator of risk of malnutrition, even in healthy older adults; a good example is the incorporation of SNAQ as part of the Rapid Geriatric Assessment to screen for malnutrition or anorexia as a geriatric syndrome alongside physical frailty, sarcopenia and cognitive impairment [63].
Some limitations in this study are worth highlighting. First, in terms of the predictive validity of SNAQ, the cross-sectional analysis limits definitive conclusions about causality, and reverse causality to account for the observed associations cannot be excluded. Next, our study comprised predominantly Chinese participants who were cognitively-intact, independent and nonfrail. These findings may not be generalizable to other non-Chinese Asian settings with more frail older adults, or communities with culturally-diverse eating behaviors which may limit interpretation of Factor 2 (i.e., satiety and intake) in the two-factor SNAQ structure. Finally, the low prevalence of risk of malnutrition (5.7%) in our study, which is consistent with the more robust health status of study participants, may have decreased the reliability of SNAQ compared to prior validation studies. We recommend more studies in different populations to examine the influence of socio-cultural characteristics on factor structure and reliability of SNAQ as a screening tool for risk of malnutrition in robust older adults.

Conclusions
In conclusion, our study found a higher optimal cutoff of ≤15 for SNAQ as a screening tool for risk of malnutrition in healthy community-dwelling older adults. Compared with the recommended cutoff of ≤14, the higher cutoff improves diagnostic performance by increasing sensitivity of SNAQ for anorexia detection to facilitate earlier detection of malnutrition risk in healthy older adults for timely nutritional assessment and intervention. Our study adds to the growing body of evidence regarding the psychometric properties of SNAQ, by explicating a two-factor structure comprising 'Appetite Perception' and 'Satiety and Intake' and affirming the validity of SNAQ in healthy older adults. Importantly, we also demonstrated the association of anorexia with other important health outcomes such as social frailty, life-space mobility and muscle function. Our findings set the stage for further longitudinal studies among relatively robust community-dwelling older adults to corroborate the predictive validity for adverse outcomes associated with malnutrition risk, and to further delineate the sociocultural and pathophysiological mechanisms that underpin the relationship between anorexia and risk of malnutrition.