Abstract
Social and structural determinants of health (SSDOH) must be measured longitudinally to understand how lived experiences shape trajectories of Alzheimer’s disease and related dementias (ADRD). This study evaluated the feasibility of administering an SSDOH survey to cognitively unimpaired older adults, examining response consistency, changes over time, and missing data patterns. A follow-up survey was conducted with participants in the UPenn Alzheimer’s Disease Research Center clinical cohort an average of 1.7 years after the initial survey. The 225-item questionnaire covered domains including education, social networks, and stressors. At follow-up, markers of feasibility included a 60% completion rate (81 of 135 participants), high item completion (>93%), and minimal missing data (<3% missed more than 10% of data). Logistic regression identified gender, social network size, and social readjustment experiences as predictors of nonrandom missingness. Response changes between administrations were likely due to ambiguity in item phrasing, instructions, or changes in participants’ experiences. Overall, repeated administration of the SSDOH survey was feasible. The response rate was reasonable but lower than expected for a volunteer research sample, suggesting multiple modes of completion may increase engagement. Repeated administration also helped identify ambiguous items and methods for improving the validity and reliability of SSDOH measures.
Introduction
Structural and social determinants of health (SSDOH) are environmental conditions that shape health and wellbeing (U.S. Department of Health and Human Services, 2023; World Health Organization [WHO], 2010). SSDOH may explain heterogeneity in patient outcomes and aging trajectories (Adkins-Jackson et al., 2023; U.S. Department of Health and Human Services, 2023; Wang et al., 2019), including risk of developing Alzheimer’s disease and related dementias (ADRD). Longitudinal studies are needed to further understand how social and structural conditions affect ADRD outcomes (Stites et al., 2022). Accordingly, this study evaluated the feasibility and performance of multiple administrations of a questionnaire designed to capture SSDOH data. The findings will inform how this information can be most effectively collected for use in ADRD research.
Prior research suggests differences in rates of cognitive and functional decline in dementia, such as impairments in memory, attention, and executive functioning (American Psychiatric Association, 2013; Cipriani et al., 2020), are influenced by social and structural factors (Centers for Disease Control and Prevention, 2021; Hill et al., 2015; Lyu et al., 2014). Given the heterogeneity in dementia progression and its long preclinical period (Melis et al., 2019; Payton et al., 2023), repeated SSDOH assessments could identify critical periods when these factors are most influential.
ADRD also have significant social and economic burdens (Skaria, 2022). Delaying the onset of advanced disease can lower healthcare costs and caregiving demands (Chandler et al., 2024; Nandi et al., 2024). Identifying specific determinants associated with disease outcomes helps develop treatment plans and informs policy efforts addressing the broader impacts of ADRD.
Most SSDOH data are collected cross-sectionally (Whitman et al., 2022), which fail to capture within-person change or support causal inference. Longitudinal data fills this gap and strengthens findings from cross-sectional work (Caruana et al., 2015). However, important considerations remain about longitudinal collection, such as the feasibility of repeatedly asking participants for information that may not change, assessing domains with unreliable measurability, and determining optimal assessment intervals that capture meaningful changes while minimizing burden. These considerations are pertinent to older adults in ADRD Research Centers, where cognitive changes may complicate data collection.
Key SSDOH domains include social identity and support, education, and socioeconomic status and strain (Stites et al., 2023; Streitz et al., 2022). These factors are critical predictors of overall well-being across the lifespan (WHO, n.d.). For example, social support, defined as perceived access to emotional, informational, or instrumental help from others, has been associated with greater cognitive resilience and function (Ma et al., 2024; Röhr et al., 2020; Salinas et al., 2021). Because older adults report the smallest network size (de Bruin et al., 2020), studying changes in social support over time may offer insight into their long-term effects on cognitive resilience and open opportunities for intervention.
Not all SSDOH domains will vary over time among older adults (e.g., demographic data), so participants may be unwilling to repeatedly answer such items. Still, repeated measurement is often necessary to establish the reliability and validity of measures (Ranganathan et al., 2024). Identifying which domains are most affected, and how repeated assessment influences response behavior, is essential for designing effective longitudinal data collection methods that maintain high completion.
It is also unclear how some domains change over time. For example, subjective social status may appear stable in older adults because of fewer opportunities to improve their social standing (Weiss et al., 2022), yet perceptions of age-related loss may influence their status ratings (Blawert & Wurm, 2021). Including such domains in longitudinal assessments could clarify their stability and how they relate to ADRD outcomes.
This study evaluated the feasibility and performance of readministering the UPenn Life Experiences Survey, a multidomain SSDOH measure, in cognitively unimpaired older adults. Feasibility was assessed using descriptive analyses that examined item completion, participant retention, and changes in responses over time. Patterns and predictors of missing data were also examined to provide a more comprehensive assessment. The findings inform best practices for collecting SSDOH data in this sample.
Methods
Design
This study reports results of the initial and follow-up administration of the UPenn Life Experiences Survey, administered online an average of 1.7 years apart. Initial data were collected between December 9, 2021 and March 14, 2022. Follow-up data from December 19, 2022 to May 7, 2024.
Participants
Participants were recruited from the University of Pennsylvania ADRC, including those co-enrolled in the Aging Brain Cohort Dedicated to Diversity study (ABCD2), which enrolls an additional 160 Black adults from the Philadelphia region to increase diversity in research cohorts.
Study Eligibility
Eligible participants were adults aged 65 and older, native English speakers, cognitively unimpaired, and enrolled with a study partner in the above longitudinal studies. Only those who completed the initial survey were invited for follow-up.
Cognitive status was determined by expert consensus conference and review of data from their annual National Alzheimer’s Coordinating Center (NACC) visit. Data included the Uniform Data Set, version 3 (Besser et al., 2018), Clinical Dementia Rating® scale (Morris, 1993), a neurological exam, and questionnaires assessing emotional, behavioral, and everyday functioning. Participants classified as having unimpaired cognition had intact performance on age and education-normed neuropsychological testing and did not meet criteria for Mild Cognitive Impairment (MCI) or AD dementia (McKhann et al., 1984; Petersen et al., 1997; Shiloh, 2006).
Measures
Social Identity
Basic demographics, sex assigned at birth, sexual orientation, gender identity, and language information are described in Supplemental Table A1.
Education
Supplemental Table A1 describes schooling assessment based on completed years of education and whether participants attended a racially segregated school (Barnes et al., 2012). Health literacy was measured with the Calgary Charter on Health Literacy Scale, a five-item scale that assesses comprehension and literacy skills (Pleasant et al., 2018). Responses range from 1 (never) to 4 (always), with higher scores indicating better health literacy.
Occupation and Social and Socioeconomic Positioning
Cognitive demands of work were determined from a free-text question asking about main occupation (National Alzheimers Coordinating Center [NACC], 2023; Pool et al., 2016; additional information in Supplemental Table A1). Socioeconomic Status and Strain were measured with three items (response range: 2–14), where higher scores indicate greater financial strain (Sternthal et al., 2011). Modified MacArthur Scale of Subjective Social Status asked participants to rank themselves on a 10-interval scale from worst off to best off (Adler et al., 1994). World View rating was assessed by asking participants to compare how they felt about the world during childhood versus now (better, about the same, or worse).
Social Support
The Lubben Social Network Scale includes six questions about family connections and friendships (Lubben et al., 2006). Total scores range from 0 to 30, with higher scores indicating better social support. Participants marked if they regularly engage in hobbies and activities, including physical, community, mental, and/or other. Brief Multidimensional Measure of Religiousness/Spirituality asked about religious and spiritual practices (The Fetzer Institute and National Institute on Aging Working Group, 1999). Responses ranged from 1 (very religious/spiritual) to 4 (not at all). Two additional 5-point Likert items (1 = strongly disagree; 5 = strongly agree) measured the degree to which religious and spiritual beliefs gave their life a sense of purpose/significance.
Social Stressors and Perceived Stress
The Perceived Stress Scale contains 10-items that asks how often participants felt stressed in the past 30 days (Teresi et al., 2020). Responses range from 0 to 4, with higher scores indicating more frequent stress. The Everyday Discrimination Scale is a nine-item measure that assesses how often participants experience instances of discrimination, from 1 (often) to 4 (never); higher scores indicate more discrimination (Williams, 2016). The Social Readjustment Rating Scale measured exposure to stressful life events in the past year using a 43-item checklist; higher scores indicate more stress (Holmes & Rahe, 1967).
Childhood Experiences
The Adverse Childhood Experiences Scale contains 20 multiple-choice and yes/no items. Scores range from 0 to 7, with higher scores indicating more exposure to abuse (Felitti et al., 1998). Childhood caregiver experiences were assessed with questions on caregiver relationships, education level, occupational status, and general well-being (detailed descriptions found in Supplemental Table A1).
Built Environment/Neighborhood
The Modified Life-Course Sociodemographic Neighborhood Exposures Questionnaire includes 12 items assessing neighborhood quality, sense of belonging, and early-life socioeconomic status, using multiple-choice, fill-in-the-blank, and a ten-point rating scale; higher scores indicate worsened perception of neighborhood conditions (Mendes de Leon et al., 2009). Participant’s geographic information (upbringing in the United States, current and childhood zip codes, county, and state) is described in Supplemental Table A1.
Statistical Analysis
Internal consistency of multi-item domains, assessed using Cronbach’s Coefficient Alpha (Cronbach, 1951), are presented in Supplemental Table B1. Completion rates were calculated as the percentage who completed the full survey and retention as the percentage who responded to the follow-up. Descriptive analyses characterize the sample and survey responses, with measures of central tendency and variation reported. Data was reviewed for outliers and erroneous responding. Change estimates involved computing changes within individuals using paired statistical tests. Change over time in key SSDOH domains was examined using mixed-effects models with time as a binary independent variable. Participants’ characteristics of age, race, gender, and education were included as variables that may explain within person differences. Logistic regression was used to test whether Time 1 completion predicted follow-up participation and whether the interval between survey administrations was associated with overall completion. Missing data patterns and predictors were analyzed using Little’s Missing Completely at Random (MCAR) test and logistic regression models (Rezvan et al., 2022). Missingness was coded as a binary outcome (1 = at least one missing response; 0 = no missing responses), and covariates were selected from each SSDOH domain. Statistical significance was determined using a two-sided .05 alpha level. Analyses were conducted in Stata v18.5 and R v4.4.
Results
Participant Sample
Of the 135 older adults who completed the survey at Time 1, 81 completed it at Time 2 (60% response rate; Table 1). Higher completion of the survey items at Time 1 did not predict participation at Time 2 (OR = 1.00, p = .95). The time between survey completions ranged from 1 year to 2.3 years (M = 1.7 years, SD = 0.3 years), but the average age at both times was 74 years, suggesting that younger participants were more likely to complete the survey at Time 2. No other characteristics had statistically significant differences.
Sample Characteristics.
Note. 44/135 (32.6%) responses. SD = standard deviation; DNR = data not reported.
Mean age increased 1.6 years from Time 1 to Time 2 (p = .002 from paired two-tailed t-test). p = .92 from unpaired two-tailed t-test, suggesting differences in average age reflects younger participants being more likely to complete follow-up survey.
Sex assigned at birth as documented on original birth certificate.
Participant’s self-defined identity as a man, woman, transgender man,
transgender woman, gender non-conforming or non-binary individual.
Primary relationship is preferred over marital status to prioritize living arrangements.
‘Partly retired’ means being retired but working reduced hours in a job. ‘Part time employed’ means working reduced hours but not retired.
‘Other’ not specified.
Participant’s identity in relation to the gender(s) they are sexually attracted to.
Clinical tests administered only at Time 1.
134/135 (99.3%) responses.
126/135 (93.3%) responses.
Estimated from Positron Emission Tomography and cerebrospinal fluid testing.
Sample Characteristics Completion Rates
Completion rates on single-question measures were above 85% (Table 2) and were not statistically different between administrations.
Completion Rates and Summary Coefficients for Sample Characteristics, Scales, and Multiple Item Domains.
Note. NA = not applicable.
Denominators represent number of participants who completed survey items at both times.
Total possible score ranges from minimum to maximum.
Two responses excluded due to errors in data collection.
Response changed from male to female and vice versa.
2/4 changed response from living with someone (either spouse or other) to no partner/single.
5/10 changed response from partly retired to retired.
4/6 changed response from bisexual to lesbian/gay.
Item added based on pilot feedback. Late addition resulted in lower completion rate.
6/14 changed response from no to yes; 4/14 yes to no; 2/14 I don’t know to no; 2/14 I don’t know to yes and vice versa.
1/2 changed response from yes to no; 1/2 I don’t know to no.
NACC Uniform Data Set SSDOH item.
Two-item scale (response range: 1–5); higher scores indicate less financial satisfaction and greater difficulty in paying monthly bills.
Select-all-that-apply item, consisting of nine options; higher scores indicate more economic problems.
Yes/no question, yes indicating they engage in hobbies/activities, no indicating they do not. Answering yes prompts further questions to elaborate on hobbies/activities.
2/3 changed response from no to yes.
p < .05.
Multi-Item Scales and Domains
For multi-item scales and domains, completion was defined as answering at least 80% of items. Completion rates improved over time across all domains (Table 2), with statistically significant increases for the MacArthur Scale of Subjective Social Status, Perceived Stress Scale, Hobbies, and World View (all p < .05).
To assess response stability, we evaluated how many participants changed responses on a scale between surveys and the paired mean differences in scale scores. Scores for the Socioeconomic Status and Strain domain remained stable over time for both Financial Situation and Economic Problem items. Less than 50% changed responses, with average score differences of 0.12 and 0.70 points, respectively (Figure 1).
Histogram of scores for socioeconomic status and strain.
The Perceived Stress Scale showed the greatest percentage of changes (100%), with an average score difference of −0.95 points (p < .05), indicating lower perceived stress.
The average score on the Life Course Sociodemographic and Neighborhood Questionnaire increased over time from 4.2 to 7.7 points, with a paired mean difference of 3.53 (p < .05), indicating a worsened perception of socioeconomic and neighborhood conditions.
Scores on the Lubben Social Network Scale decreased overtime from an average of 22.0 to 17.4 points (mean change = −4.69, p < .05), suggesting lower perceived social support. Each item-level response also decreased (all p < .05, Figure 2).
Histogram of scores for Lubben Social Network Scale.
In separate mixed-effects models that adjusted for age, race, gender, and education, study time showed a significant effect on the Lubben Social Network Scale, Perceived Stress Scale, and Life Course Sociodemographic and Neighborhood Questionnaire scores (Estimates = −6.03, −0.90, and 3.60, respectively; all p < .001). Moreover, item-level analyses showed that in some cases, specific items appeared to be driving these differences (Supplemental Appendix C, Figures C1–C3). In a follow-up analysis of the PSS, for example, we found that some items contributed to this change while others remained more stable. Those items that were more inclined to change asked about controlling issues in contrast to other items that assessed feelings or perceptions of events (Supplemental Figure C1).
Participant Caregiver and Geographic Data
Childhood caregiver and geographic history questions had consistently high completion rates at both times, with rates above 79% and 88%, respectively. Detailed completion rates and response changes are presented in Supplemental Table D1.
Missing Data
After excluding questions not applicable to all participants (e.g., secondary caregiver questions), approximately 8.9% of participants at Time 1 were missing more than 10% of data, compared to 2.5% at Time 2. Little’s MCAR test revealed a significant result at both times, suggesting the data are not missing completely at random (Time 1: χ2 (380) = 441, p = .017; Time 2: χ2 (212) = 249, p = .042, data not tabled).
To explore sources of systematic missingness, multivariable logistic regression analyses showed that at both times, males had lower odds of missing data, with stronger effects observed at Time 2 (OR = 0.06, 95% CI [0.004, 0.44], p < .05; Supplemental Table E1). Larger social networks (OR = 0.86, 95% CI [0.75, 0.98]) and more experiences of social readjustment (OR = 0.66, 95% CI [0.46, 0.92]) also indicated reduced odds of missing data. Odds ratios for each variable are presented in Supplemental Table E1.
To assess the robustness of the logistic regression analysis, Little’s MCAR test was repeated after excluding the predictor variables. The resulting test was non-significant (Time 1: χ2 (218) = 246, p = .094; Time 2: χ2 (173) = 197, p = .099; data not tabled), suggesting the data to be MCAR. This strengthens the interpretation that the excluded variables contributed to the initial patterns of missingness.
Discussion
Within a cohort of cognitively unimpaired older adults, we found it feasible to repeatedly administer a multidomain SSDOH survey. Feasibility was evaluated based on participant retention, completion, response changes, and missing data patterns. Retention rate was reasonable, though lower than expected, completion across domains remained high, and missing data was minimal. These findings help advance our ability to capture SSDOH information in ADRD research, allowing for improved understanding of their influence on clinical outcomes.
Contrary to the frequent survey request hypothesis (Eggleston, 2024), which states that the repeated survey exposure may overburden participants and reduce completion, we found that completion rates improved. One possible reason for this is that participants were ADRD research volunteers, so providing SSDOH information was personally meaningful; thus, increasing the survey’s salience and mitigating survey fatigue (Porter et al., 2004; Wu et al., 2022). Familiarity from the first survey may have also increased comfort in providing sensitive information, helping to improve completion (Kartsounidou et al., 2024).
Analyzing missingness patterns helped identify potential sources of bias and inform strategies for improving completion in future administrations. Although participants are offered alternative options for completion (e.g., paper surveys), are given reminder emails, and are compensated for completion of the survey, our results suggest that these efforts for retaining completion across time points could be improved. By anticipating which participants need additional support, targeted retention strategies like reducing barriers by offering assistance with transportation can be utilized (Teague et al., 2018).
We also found that asking participants to repeatedly respond to stable questions (e.g., life history questions) did not reduce engagement. Given the high test-retest reliability in similar populations (Streitz et al., 2022), continuing to collect such data will be helpful in flagging response inconsistences or meaningful changes. For instance, few changes are expected in childhood caregiver information, so discrepancies over time could signal data quality issues that warrant follow-up. Conversely, if stable factors like gender identity shift, then greater attention can be directed toward any associations with outcomes that might be overlooked (Ocasio & Isabel Fernández, 2024).
While most domains remained stable over time, the 1.7-year interval between administrations may be too brief to capture some changes. For instance, while no notable changes were seen in religiousness/spirituality, a 10-year longitudinal study found these factors changed across adulthood, with engagement peaking in the 70s (Nelson et al., 2024). Such changes are important, as shifts in religiousness/spirituality have been associated with well-being (Joiner et al., 2022). These studies highlight the need for continued administrations beyond our single follow-up administration.
Some domains, such as perceived stress, showed statistically significant change over time. The mixed-effect analysis suggests that some of these differences may reflect measurement artifacts, indicating that additional clarification or guidance may be needed during completion. In a follow-up analysis, we found that some items contributed to this change while others remained more stable. Those items that were more inclined to change seemed to relate to factors that could be more likely to change from day to day. The finding highlights the importance of identifying an optimal interval between administrations that capture meaningful changes without increasing participant burden. While existing literature offers some guidance on this concern (Feely et al., 2020; Pullenayegum et al., 2021), they focus on one specific domain, whereas our SSDOH questionnaire adds complexity due to its multidomain component. Addressing this requires optimizing schedules for each SSDOH domain. For example, religiousness/spirituality may warrant longer intervals, whereas perceived stress may require more frequent assessment. This approach would streamline the survey and minimize burden for participants and researchers.
We administered the full version of the survey at both study time points in order to facilitate the goal of this study, which was to evaluate repeated administration of the measures. Our findings showed that some measures did not change between administrations. Ongoing administration of measures that generate limited new information could be an inappropriate burden to research participants. In our research cohort, we have restricted subsequent follow-up administrations of the survey to measures that vary annually. We have also developed a plan to modify the survey – adding, removing, or adapting – measures based on what we learn from analyzing them in order to ensure items posed to participants generate new information.
Our study has important limitations to note. The generalizability of our findings is limited by the characteristics of the sample. Participants were predominantly white, more educated, and limited to cognitively unimpaired and healthy older adults. Our findings are based on participants who have voluntarily enrolled in a research cohort, which is unlikely to reflect the wide breadth of diversity in the US older adult population. In addition, the average time between administrations was 1.7 years, which is a relatively short observation interval for some aspects of SSDOH that may vary over longer intervals, such as religiousness/spirituality or health literacy. Resultantly, it remains unclear whether repeated administration of this questionnaire is feasible for detecting meaningful change in these domains. Moreover, our findings are limited by the specific SSDOH measures assessed and thus cannot generalize to other domains. Larger, more diverse samples with repeated administrations are necessary to support the feasibility of repeatedly capturing SSDOH information.
By addressing practical concerns of longitudinal SSDOH data collection, this study contributes to calls for universally implementing SSDOH assessment within a life course framework (Adkins-Jackson et al., 2023; Stites et al., 2022). Integrating these multidomain questionnaires into ADRD contexts will advance understanding of the social and structural influences on disease risk and progression.
Conclusion
It is feasible to repeatedly administer a multidomain SSDOH measure in cognitively unimpaired older adults. The readministration was well-received by participants, evidenced by improved completion rates, minimal missing data, and reliable responses across domains. Identifying predictors of missingness provides insight into future efforts to boost participant engagement. This work supports the integration of repeated SSDOH assessments, advancing understanding of how these determinants influence health outcomes in ADRD populations.
Supplemental Material
sj-docx-1-ggm-10.1177_30495334261430788 – Supplemental material for Repeated Administration of Social and Structural Determinants of Health (SSDOH) Questions in an Alzheimer’s Disease Research Center: The Aging Brain Study (ABC) Life Experience Survey
Supplemental material, sj-docx-1-ggm-10.1177_30495334261430788 for Repeated Administration of Social and Structural Determinants of Health (SSDOH) Questions in an Alzheimer’s Disease Research Center: The Aging Brain Study (ABC) Life Experience Survey by Shana D. Stites, Daniel S. Lee, Carolyn Kuz, Valerie Humphreys, Marissa L. Streitz, Sharon X. Xie, Jason Flatt, Crystal M. Glover and Dawn Mechanic-Hamilton in Sage Open Aging
Footnotes
Acknowledgements
We thank the research participations for their contributions to this project.
Ethical Considerations
This study was approved by the University of Pennsylvania IRB.
Consent to Participate
Participants provided written informed consent to participate.
Consent for Publication
Participants provided written informed consent to having their data analyzed and published in aggregate.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from the National Institute of Aging (NIA P30AG072979-05, P30AG012836-25-S2, NIA P30AG010124, K23AG065442, P01AG03991, P01AG026276, P30AG066444, R01AG083177, R24AG066599, K01AG056669, K23AG065499, 1RF1NS143766-01, R01AG095017, P30AG066519-06, RF1AG083177); and by the Alzheimer’s Association (AARF-17-528934, AARGD-22-929144). The Aging Brain Cohort Dedicated to Diversity (ABCD2) Study is supported by a grant from the Pennsylvania Department of Health (2019NF4100087335).
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
Data are available upon request for purposes of replication by contacting the corresponding author.
Supplemental Material
Supplemental material for this article is available online.