Introduction

The World Health Organization (WHO) has dedicated 2021–2030 towards the goal of “healthy aging in older adults” (World Health Organization [WHO], 2020). An area for action is improving integrated care, including through primary care to prevent functional disability and frailty. WHO has called for longitudinal research to support the understanding of healthy aging, as well as policy development and evaluation of interventions (WHO, 2020). Older adults who are frail experience a decline in function and physiological systems, and do not fully recover when a stressor occurs, such as an acute illness (Pilotto et al., 2020; Laur et al., 2017). Screening and assessment are encouraged to identify frailty early and provide interventions (Lee et al., 2018; Pilotto et al., 2020). Food intake (Huang et al., 2021; Khor et al., 2021), anorexia of aging (Merchant et al., 2021), and malnutrition (Rodríguez-Mañas et al., 2021; Laur et al., 2017) are associated with frailty, and screening for nutrition risk and malnutrition is recommended for those who are potentially frail (Lee et al., 2018; Pilotto et al., 2020).

Nutrition(al) risk, although a nebulous term (Bales, 2001), can be used to describe the presence of risk factors and determinants such as food insecurity, poor appetite, and low food intake which could lead to malnutrition (Keller, 2007; Teitelbaum et al., 2005). Malnutrition is demonstrated by changes in body composition (e.g., muscle mass) and body mass that result in functional deficits (e.g., immunity, strength, tolerance, cognition) associated with a nutritional intake that does not meet the body’s needs; this could be due to inadequate food and fluid intake, high metabolic demand, or inability of the body to use the nutrients and energy provided (Cederholm et al., 2017). Although nutrition risk precedes malnutrition in community-living older adults, it is a relevant target for intervention as nutrition risk has been associated with increased use of healthcare resources, and specifically hospital use (Martínez-Reig et al., 2018; Ramage-Morin et al., 2017; Hamirudin et al., 2016). Determinants of food intake (e.g., dysphagia, eating challenges, financial resources), food intake that is inadequate or of poor quality, as well as weight change and other anthropometric indicators should be included on nutrition risk screening tools for community-living older adults as they are early markers of potential malnutrition (Keller, 2007). SCREEN-8 (previously known as SCREEN-IIAB) is a valid and reliable measure of nutrition risk in community-living older adults (Keller et al., 2005). Screening with this tool has the potential to identify nutrition risk before participants lose capacity and become pre-frail or frail. About one third of community-living older Canadians have nutrition risk as per SCREEN-8 (Ramage-Morin et al., 2017).

Nutrition risk and malnutrition more specifically have been described as a “silent threat” that develops in the community (Hamirudin et al., 2016 p. 9) but is often not recognized until a healthcare event such as hospitalization occurs (Allard et al., 2016). Despite the recognition of the importance of malnutrition to frailty (Pilotto et al., 2020), prevention of inadequate food intake and malnutrition in the community is limited (Merchant et al., 2021). To promote healthy aging and reduce malnutrition and frailty, it is recommended that nutrition risk in older adults be identified through screening programs in primary care and community services (Keller et al., 2021; Hamirudin et al., 2016). Understanding individual determinants of nutrition risk is needed to identify further paths for intervention and to mitigate malnutrition in older adults (Volkert et al., 2019).

Examination of factors associated with nutrition risk in community-living older adults is relatively rare (Hamirudin et al., 2016; Volkert et al., 2019). Most research to date has been based on small samples and cross-sectional data with few factors modelled, limiting our understanding of determinants that are most relevant to target for intervention (Hamirudin et al., 2016). The Determinants of Malnutrition in Aged Persons (DoMAP) is a consensus-based model that identifies diverse sociodemographic, function, and health factors specific to older adults that are associated with malnutrition (Volkert et al., 2019). This model was used to guide our longitudinal analysis to identify determinants of nutrition risk in older adults (≥ 65 years) in the Canadian Longitudinal Study on Aging (CLSA). Figure 1 includes the indirect and direct factors that may contribute to malnutrition adapted from the DoMAP model (Volkert et al., 2019). Analyses were stratified by baseline participant nutrition risk, to understand the determinants of risk and the determinants that perpetuate continued nutrition risk, potentially leading to malnutrition and frailty. This study aimed to identify (1) what proportion of community-dwelling older adults experienced a change in nutrition risk over a 3-year period and if the change differed by baseline risk stratification, and (2) what health, function, sociodemographic, and healthcare service use variables are associated with 3-year change in nutrition risk for older persons stratified by baseline risk.

Fig. 1
figure 1

Indirect and direct factors from the Determinants of Malnutrition in Aged Persons (DoMAP) Model available in the Canadian Longitudinal Study on Aging Dataset that may contribute to malnutrition via the low food intake mechanism. Figure adapted from Volkert et al. (2019). Bolded items in red are assessed by the SCREEN-8 tool

Full size image

Methods

Sample

The CLSA is the largest and most comprehensive Canadian prospective cohort study assessing longitudinal data regarding aging. Greater description of the CLSA has been published previously (Raina et al., 2009). Every 3 years, participants complete data collection for a minimum of 20 years, or until death. Only data from participants in the comprehensive cohort (baseline n = 30,097) were used in this analysis. Data collection occurred through questionnaires, physical examination, and biological samples. Participants were recruited from areas within a 25–50 km radius of 11 major academic centres in Canada, representing 4 regions: Pacific Coast, the Prairies, Central Canada, and the Atlantic Region. The CLSA included participants with a minimum age of 45 at baseline. Participants were excluded if they were residing in the three territories, some remote regions, federal First Nations reserves, and other First Nations settlements in the provinces, were full-time members of the Canadian Armed Forces, and/or were living in institutions (e.g., long-term care homes) at baseline; older adults living in settings where minimal care is provided (e.g., retirement homes) were included. Participants had to be able to respond in English or French and have no cognitive impairment at baseline. Informed consent was obtained from all participants. Baseline data collection occurred during 2012–2015 and first follow-up during 2015–2018. For our analysis, we included participants who were  ≥65 at baseline (n = 12,646) and excluded participants who reported current abdominal or nasogastric tube feeding, resulting in an initial eligible sample of 12,630. Ethics review was completed by the University of Waterloo (ORE#42598).

Variables used

Nutrition risk was assessed using the valid and reliable SCREEN-8 tool (Keller et al., 2005; note rebranded in 2019 to SCREEN-14 (original SCREEN-II) and SCREEN-8 (original SCREEN-IIAB see www.olderadultnutritionscreening.com)). SCREEN-8 asks eight questions regarding weight change, appetite, eating challenges, and fluid and fruit/vegetable intake. The minimum and maximum scores are 0 and 48, respectively. A score of  <38 indicates high nutrition risk, a value previously shown to predict hospitalization and mortality (Ramage-Morin et al., 2017). A positive change when comparing follow-up to baseline indicates a score improvement, while a negative change indicates a decline from baseline.

Variables were selected based on the DoMAP model (Volkert et al., 2019), and included health, social determinants of health, healthcare use, and demographic variables. Both time-independent (i.e., variable collected only at baseline or follow-up) and time-dependent variables (i.e., variable was collected at both baseline and follow-up) were used. Variable selection and categorization were dictated by both clinical/practical relevance and data collection protocols to ensure sufficient distribution within variable groupings to prevent challenges that arise with low cell counts within a variable category. All variables were self-reported by participants, except for body mass index which was measured by trained research staff. A summary of items collected at baseline and follow-up in the CLSA and subsequent categorization are in Table 1.

Table 1 Questionnaire items, data collection methods and variable categorization used in descriptive and multivariable statistics
Full size table

Analysis

Statistical analyses were performed using SAS Studio® Release 3.81. Descriptive statistics (proportions, mean, SD) were determined. The sample of participants who were excluded due to missing data and those included in the final analytical sample were compared on key demographic variables to determine potential challenges with generalizability of findings using the Rao-Scott likelihood ratio chi-square test. Multivariable linear regression was used to determine associations between all operationalized variables based on the DoMAP model variables available in the CLSA and change in SCREEN-8 score. Survey methodologies were accounted using the SURVEYREG procedure, and analytic and geographic weights were used. Listwise deletion for all variables was used in all analyses, resulting in a final sample of 5031. Statistical significance was determined by p < 0.05.

Results

Descriptive statistics

Descriptive statistics for baseline and follow-up variables are in Table 2. Key descriptive statistics stratified by baseline nutrition risk status are in Table 3; descriptive statistics stratified by nutrition risk status for all variables are in Supplementary Table 1. Analytic sample participants and those excluded from analyses due to missing data were compared; those excluded were more likely to be female, older, less educated, have a lower income, be widowed, live alone, and be at nutrition risk (see Supplementary Table 2). Statistical significance between risk groups was found for many of the personal characteristics of participants. At baseline, 71.0% of participants were not at high nutrition risk, while 64.4% were not at high risk at follow-up. Of the 5031 participants, 53.6% remained not at nutrition risk at both baseline and follow-up, and 18.3% remained at nutrition risk at both timepoints. Meanwhile, 28.0% of participants experienced a change in nutrition risk status, with 17.3% experiencing a decline (i.e., not at risk at baseline, at risk at follow-up) and 10.7% an improvement (i.e., at risk at baseline, not at risk at follow-up). Mean change in SCREEN-8 score was  −1.03 (SD 5.39; median =  − 1.00). Among the 3570 participants not at high nutrition risk at baseline, change in SCREEN-8 score was  −2.20 (SD 4.63; median =  − 2.00), and among the 1461 participants at high nutrition risk at baseline, change in SCREEN-8 score was +1.82 (SD 6.02; median = 2.00).

Table 2 Descriptive statistics for baseline and follow-up variables (n = 5031)
Full size table
Table 3 Select participant descriptive statistics for all participants and stratified by baseline nutrition risk status
Full size table

Multivariable regression

Model effects are reported in Table 4 and the odds ratios for only significant model effects are included in Table 5. The full model is included in Supplementary Table 3.

Table 4 Model effects testing the association factors with 3-year change in SCREEN-8 score, stratified by baseline nutrition risk status
Full size table
Table 5 Multivariable linear regression assessing covariates associated with change in SCREEN-8 score stratified by baseline nutrition risk status
Full size table

Participants not at high nutrition risk at baseline

Females experienced a decrease in SCREEN-8 score by 0.57 points as compared to males (CI [−1.04,  −0.10]). Participants who currently smoked experienced a decrease of 1.83 points compared to non-smokers (CI [−3.42,  −0.24]). Those who reported having a mental health condition at both baseline and follow-up experienced a decrease in SCREEN-8 by 0.83 points compared to participants who did not report having a mental health condition at any time (CI [−1.44,  −0.22]). Similarly, reporting a diagnosis of a kidney condition at baseline or follow-up was associated with a decrease in SCREEN-8 score by 1.51 points (CI [−2.81,  −0.21]). Participants who reported using psychologist or social services in the past 12 months at either baseline or follow-up experienced a decrease in SCREEN-8 score by 1.25 points compared to participants who did not report using these services (CI [−2.13,  −0.37]). Participants who did not report going to the dentist at baseline or follow-up or only went to the dentist at baseline experienced a decrease in SCREEN-8 score by 0.91 and 1.32 points, respectively, when compared to participants who visited the dentist at both timepoints (CI [−1.62,  −0.20], [−2.45,  −0.19]). Similarly, participants who consistently experienced oral health problems at baseline and follow-up experienced a decrease in SCREEN-8 by 0.92 points compared to participants who did not report any oral health problems (CI [−1.44,  −0.39]). Experiencing any change in marital status (e.g., single to widowed, or separated to married) was associated with a decrease in SCREEN-8 score by 1.52 points compared to participants who did not experience a change in marital status (CI [−2.74,  −0.30]). Newly living alone at follow-up and not at baseline was also significantly associated with a decrease in SCREEN-8 score by 1.98 points compared to participants who did not live alone at either timepoint (CI [−3.40,  −0.56]).

Participants at high nutrition risk at baseline

Gastrointestinal conditions were associated with SCREEN-8 score, with participants who reported a condition at baseline or a new condition at follow-up experiencing a decrease by 1.10 and 2.20 points, respectively, when compared to participants who did not report gastrointestinal conditions at baseline or follow-up (CI [−2.19. −0.01], [−3.93, −0.46]). Change in experiencing pain was associated with a change in SCREEN-8 score (F = 3.89, p = 0.009); however, significant differences between groups were not apparent when participants who were usually pain free at baseline and follow-up were used as the reference group. Participants who reported a decrease in hearing at follow-up experienced a 1.02 point decrease in SCREEN-8 score (CI [−1.94,  −0.09]). Hospital service use in the last 12 months was associated with a change in SCREEN-8 score, where participants who reported going to the hospital at baseline but not follow-up experienced a significant increase in SCREEN-8 score by 1.35 points compared to participants who did not report going to the hospital at either timepoint (CI [0.12, 2.58]). Participants who reported a decrease in their income experienced an increase in SCREEN-8 score by 1.94 points compared to participants who did not report a change in their income (CI [0.71, 3.18]). Participants who experienced impairment with their activities of daily living at both baseline and follow-up experienced a decrease in SCREEN-8 score by 2.56 points compared to participants who did not experience any impairment at baseline and follow-up (CI [−4.36,  −0.77]). Those who declined in their chair-rise performance between baseline and follow-up also experienced a decrease in SCREEN-8 score by 1.98 points compared to participants who performed adequately at baseline and follow-up (CI [−3.33,  −0.63]). Last, older adults who experienced oral health problems at baseline but not follow-up experienced an increase in SCREEN-8 score by 1.60 points compared to participants who did not experience oral health problems at either timepoint (CI [0.15, 3.05]).

Discussion

High nutrition risk was common in this sample (29.0% at baseline and 35.6% at follow-up) and although most participants did not change their nutrition risk level between baseline and follow-up, more participants were likely to decline than improve in their nutrition. Decline was specifically seen in those not at high risk at baseline, whereas an increase in mean SCREEN-8 score was seen for those with high risk at baseline, suggesting potential impact of interventions or other changes that improved nutrition risk for some in this group.

This longitudinal analysis for the first time distinguished the determinants of a decline in nutrition for those not at risk at baseline (i.e., early in risk trajectory) and those who were initially at risk. Determinants were different among these groups but generally in anticipated directions. Poor oral health, lack of dental care, poor mental health, use of psychological or social services, and change in marital status or living arrangements were associated with a decline in SCREEN-8 scores (i.e., more nutrition risk) in those who were not at risk at baseline. These factors have previously been shown to be important determinants of nutrition risk or malnutrition (Bardon et al., 2021; Keller, 2007; Streicher et al., 2018; Thompson Martin et al., 2006). Smokers, females, and those with a kidney condition at either timepoint were also more likely to have declining SCREEN-8 scores. Alternatively, those participants at high nutrition risk at baseline had further declines in their SCREEN-8 scores and thus declining nutrition if they had a past or new gastrointestinal condition (of note the negative effect was greater for a new diagnosis at follow-up), decline in self-reported hearing from baseline, any impairment with activities of daily living at baseline or follow-up, and a decline in chair rise performance at follow-up. Decline in function has been associated with risk or malnutrition in previous work (Bardon et al., 2021; Keller, 2007; Thompson Martin et al., 2006; O’Keeffe et al. 2019), and points towards a frailer older adult continuing to decline in their overall health and nutrition. Three factors were associated with an improvement in SCREEN-8 scores in those at high risk at baseline: hospitalization in the 12 months prior to baseline but no hospitalization in the 12 months prior to follow-up, a decrease in income from baseline, and oral health problems at baseline that were resolved at follow-up. Hospitalization is known to be associated with high nutrition risk (Streicher et al., 2018) and could have offered an opportunity to identify malnutrition (Allard et al., 2016) and provide supports that could improve nutritional status. A declining income was associated with better nutrition at follow-up and could be attributed to increased social assistance programming or eligibility for specific benefits that supported nutrition, such as food-based social services. As we adjusted for living alone (26.0% at baseline) vs. with others (64.3% living with at least one other person with only 9.7% living with two or more people) in the analysis, any change in income due to a change in living with others would be accounted for in the regression. Further, as all participants were over the age of 65 years, new retirement was less likely to be a reason for this change in income. As sufficiency of income was not ascertained, we cannot further speculate on why a reported lower income was associated with higher SCREEN-8 scores at follow-up for those at nutrition risk at baseline. Meanwhile, improved oral health could have resolved challenges with eating that led to high nutrition risk at baseline.

This analysis demonstrates that declines in SCREEN-8 scores in those not at risk at baseline are associated with indirect factors like mental health and living situation, while those already at nutrition risk at baseline who had declines in SCREEN-8 scores had determinants consistent with frailty and worsening health. This analysis has confirmed the nutrition risk conceptualization (Keller, 2007) of indirect determinants that may not necessarily result in overt malnutrition and/or its effects and the view that several determinants are based on aging processes (Bardon et al., 2021). The DoMAP model also segments determinants by how direct their effect may be on three mechanisms for malnutrition. Several of the items on SCREEN-8 are included in the DoMAP as factors most directly associated with malnutrition through the mechanism of poor food intake (e.g., swallowing problems, poor appetite), but other direct factors on DoMAP include impaired activities of daily living including shopping and cooking (Volkert et al., 2019). Our findings on factors associated with a decline in SCREEN-8 in those already at nutrition risk at baseline are consistent with the DoMAP model. Indirect factors on the DoMAP model include living situation, mental health, health conditions, and poor oral health, which in our analysis were associated with declines in SCREEN-8 scores for those not at risk at baseline. Thus, this analysis of SCREEN-8 stratified by baseline nutrition risk also provides evidence for the consensus-developed DoMAP model of direct and indirect factors leading to malnutrition.

Use of preventive services like meal programs was very low in this sample (11.4% reported receiving help at some point) and we could not model change in professional or non-professional meal preparation or delivery help. However, other services, such as allied healthcare visits, were insignificant in both models. Based on the prevalence of nutrition risk and declines in SCREEN-8 scores seen over time, it can be surmised that there may be an underutilization as well as lack of availability of programs and services that could address low food intake and nutrition risk in community-living older Canadians (Keller et al., 2021). Hospital use and improved oral health resulted in higher SCREEN-8 scores for the high-risk group, suggesting that improvements are possible. However, lack of service use points to the need to operationalize primary care nutrition risk screening to find and treat the determinants of nutrition risk. If nutrition risk screening cannot be a global activity for all of those over the age of 65 in primary care to promote healthy aging, then targeted screening should be considered (Laur & Keller, 2017; Lee et al., 2018). Older adults experiencing life transitions such as a change in living situation or marital status, and those with mental health conditions and oral health problems should be screened for nutrition risk in primary care settings. Older adults with impaired activities of daily living and chair rise performance, as well as gastrointestinal diseases are likely already at high nutrition risk and prone to further declines in nutritional health, necessitating the monitoring and intervention of a dietitian (Keller et al., 2021; Laur & Keller 2017; Lee et al., 2018).

Limitations of the study

This study is not without its limitations. CLSA is not representative of all Canadians, due to the exclusion of some groups (e.g., those who could not speak English or French), poor representation of certain demographics (e.g., race/ethnicity; 96% of the 5031 participants self-identified as white), and previous analyses have identified that participants in the CLSA comprehensive cohort are more educated and have higher household incomes compared to the Canadian population (Raina et al., 2019). Further, the comprehensive CLSA cohort was used for this analysis and was limited to those who could readily visit the urban academic test centres, further under-representing Canadians who live in small urban and rural communities. All data (except for body mass index) were collected by self-report questionnaires and due to missing data, our analyses were based on less than half of the eligible cohort sample and differences in demographic variables like sex, age, and education were detected when the cohort with missing data and the analytic cohort were compared. Finally, as with all longitudinal studies, loss to follow-up (n = 2332 of the original sample) occurred. Thus, the associations identified in this analysis should be considered as potentially biased by the data collection measures and sample constraints. Prior research has demonstrated lack of association in meta-analyses (Streicher et al., 2018) and systematic reviews (O’Keeffe et al., 2019), potentially due to similar measurement and sample issues.

Conclusion

Nutrition risk is common in community-living older adults. Determinants of decline in nutrition over a 3-year period were dependent on whether the older adult was at nutrition risk at baseline. Indirect factors that may be effectively detected and managed through secondary prevention measures (e.g., oral health problems), some sociodemographic factors (e.g., living alone), and healthcare service use (e.g., not visiting a dentist) are associated with declines in nutrition for those not already at high nutrition risk, while those at high risk see further declines in their nutrition when functional changes and gastrointestinal disease are present. Implementation of nutrition screening in primary care, like the consensus-based nutrition care pathway for community-dwelling persons ages  ≥65 (Keller et al., 2021), could provide opportunities for individualized monitoring and timely intervention while accounting for older adults’ current nutritional status. Targeted nutrition risk screening, assessment, and treatment are needed to promote healthy aging for older Canadians.

Contributions to knowledge

What does this study add to existing knowledge?

  • This study demonstrates that determinants contributing to nutritional decline differ for those not at nutrition risk versus those at high risk.

  • Indirect determinants of nutrition risk like living alone and smoking, in addition to determinants that may more directly indicate frailty, like impaired activities of daily living, signal high nutrition risk and continued trajectory towards malnutrition.

  • Hospitalization and improved oral health can improve nutrition in high-risk older adults.

What are the key implications for public health interventions, practice, or policy?

  • This analysis differentiates segments of the older adult population for nutrition risk screening and intervention in primary care.

  • Older adults at high nutrition risk should have dietitian intervention and community supports to improve their food intake due to likely challenges with function.

  • Older adults not yet at high nutrition risk can see declines in nutrition if they are also experiencing a life transition, mental health condition, or oral health problems.

  • Nutrition risk screening and identification of supports to mitigate challenges are needed for older adults. SCREEN-8 is ideal for nutrition risk screening in primary care for early detection and timely intervention.