📄 social_science

Beyond Digital Access: Development and Validation of a Digital Engagement Scale for Older Adults in the United States

This study develops and validates a 14-item, three-factor Digital Engagement Scale using 2022 Health and Retirement Study data to provide a behaviorally grounded, multidimensional instrument for assessing distinct domains of online activity among U.S. adults aged 51 and older.

Original authors: Mary Akinde, Patryk Babiarz, Lu Fan, Diann Moorman

Published 2026-07-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mary Akinde, Patryk Babiarz, Lu Fan, Diann Moorman

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Idea: It's Not Just About Having a Phone

Imagine two neighbors, both in their 70s. Both have a smartphone and both know how to turn it on.

Neighbor A uses their phone only to check the weather and read the news.
Neighbor B uses their phone to video chat with grandkids, pay bills, order groceries, book doctor appointments, and manage their bank account.

If you only asked, "Do you have internet?" both would say "Yes." But their digital lives are completely different. This paper argues that we need to stop just checking if older adults are "online" and start measuring what they actually do while they are online.

The Problem with Old Measurements

For a long time, researchers treated "internet use" like a light switch: it's either ON or OFF.

The Old Way: "Do you use the internet?" (Yes/No).
The New Way: "What kind of digital activities do you do, and how often?"

The authors say that just knowing someone has a device is like knowing someone owns a car. It doesn't tell you if they are driving to the grocery store, racing on a track, or just sitting in the driveway. To understand their real life, we need to know how they drive.

The Solution: The "Digital Engagement Scale" (DES)

The researchers created a new tool called the Digital Engagement Scale (DES). Think of this scale as a menu rather than a checklist. Instead of just checking a box for "uses internet," it sorts activities into three distinct "flavors" or categories:

The "Newsstand" (Informational Engagement):
- What it is: Checking the weather, reading news, looking up recipes, or finding health info.
- The Analogy: This is like browsing a library or reading a newspaper. It's about gathering knowledge.
- Finding: This was the most common activity. Older adults love getting information.
The "Town Square" (Social-Communication Engagement):
- What it is: Sending texts, emailing, posting on Facebook, or sharing photos.
- The Analogy: This is like chatting at a community center or calling a friend. It's about staying connected.
- Finding: This was also very common, especially among women.
The "Bank & Travel Agent" (Transactional Engagement):
- What it is: Paying bills, shopping online, booking a ride (like Uber), or managing travel.
- The Analogy: This is like handing over your wallet or passport. It involves risk, money, and logistics.
- Finding: This was the least common activity. Many older adults avoid this part of the digital world.

How They Built the Tool

The researchers didn't invent new questions from scratch. They took 31 existing questions from a massive, trusted national survey (the Health and Retirement Study) and tested them like a chef tasting a soup to see if the flavors balanced.

They split the data in half: one half to discover the patterns, and the other half to prove the patterns were real.
They found that the 31 questions naturally grouped together into the three "flavors" mentioned above (News, Chat, and Transactions).
They proved that these three groups are different from each other (you can be good at News but bad at Transactions) and that the tool works fairly well for men, women, different ages, and different income levels.

What They Discovered

When they looked at the results, a clear picture emerged:

The "Partial" Digital Life: Many older adults are fully integrated into the "Newsstand" and "Town Square" but are left out of the "Bank & Travel Agent" world.
The "Trust" Barrier: The paper suggests that avoiding transactions (like online banking) isn't always about not knowing how to do it. It's often about trust. Just like you might know how to drive a car but refuse to drive in a storm because it feels unsafe, older adults might avoid online shopping because they fear fraud or making a mistake with their money.
The "Time" Factor: The paper suggests that as people get older, they might prefer activities that give immediate joy (like talking to family) over activities that require taking risks for a future reward (like managing complex finances).

Why This Matters

This study gives researchers and policymakers a better map.

Before: They thought, "We need to get more seniors online."
Now: They can see, "We need to help seniors feel safe enough to use online banking and travel apps."

The paper concludes that digital inclusion isn't just about handing someone a device; it's about understanding which parts of the digital world they are comfortable navigating and which parts they are avoiding, and why.

In short: The paper built a better ruler to measure how older adults use technology. It found that while many are great at reading and chatting online, they are often hesitant to do the "business" parts of the internet, likely due to trust issues rather than a lack of skill.

Technical Summary: Beyond Digital Access: Development and Validation of a Digital Engagement Scale for Older Adults in the United States

Problem Statement
Current measures of digital inclusion for older adults predominantly rely on binary indicators of access, general internet use frequency, or self-rated digital skills. These metrics fail to capture the behavioral reality of what older adults actually do online, obscuring the qualitative differences between being "connected" and being "meaningfully engaged." Existing literature often collapses diverse online activities into a single composite score, which masks the distinct predictors and consequences of specific digital behaviors. Furthermore, many existing instruments are designed for other populations (e.g., students, clinical samples) or measure latent constructs like literacy and acceptance rather than observed behavioral frequency. A critical gap also exists in the lack of measurement invariance testing; without establishing that instruments function equivalently across sociodemographic groups, observed differences in digital participation may reflect measurement bias rather than true disparities.

Methodology
This study utilized a secondary scale-development approach using data from the 2022 Health and Retirement Study (HRS) Leave-Behind Questionnaire, a nationally representative panel study of U.S. adults aged 51 and older.

Sample: The initial structural pool consisted of 3,306 respondents with complete data on 31 candidate digital-activity items. This sample was randomly split into an exploratory factor analysis (EFA) calibration sample ( $n=1,653$ ) and a confirmatory factor analysis (CFA) hold-out sample ( $n=1,653$ ). A broader criterion-validity sample ( $N=3,551$ ) was used for descriptive and regression analyses.
Instrument: The study evaluated 31 ordered-categorical items (reverse-coded 1–5, where 5 indicates daily use) covering a broad range of modern-device activities.
Analytic Strategy:
- Factor Analysis: Given the ordered-categorical nature of the data and floor effects in certain items, the study employed polychoric correlations and the Weighted Least Squares Mean- and Variance-Adjusted (WLSMV) estimator. An EFA with oblique rotation was conducted to identify the latent structure, followed by a CFA to test the resulting model against a unidimensional alternative.
- Validity and Reliability: The study assessed internal consistency (ordinal alpha, McDonald's omega, composite reliability), convergent validity (Average Variance Extracted), and discriminant validity (Heterotrait-Monotrait ratio of correlations, HTMT).
- Invariance Testing: Multigroup CFA was performed to test measurement invariance (configural, threshold, loading, and residual) across gender, age, education, race/ethnicity, and income.
- Criterion Validity: Known-groups validity was examined by comparing DES scores across sociodemographic groups and using multiple linear regression to predict overall engagement.

Key Results

Factor Structure: The EFA and subsequent CFA supported a 14-item, three-factor structure that significantly outperformed a unidimensional model. The three dimensions are:
1. Informational Engagement: (5 items; e.g., checking weather, getting news, searching for ideas).
2. Social-Communication Engagement: (4 items; e.g., messaging, social networking, sharing photos).
3. Transactional Engagement: (5 items; e.g., online banking, shopping, ride-hailing, travel management).
Model Fit: The three-factor model demonstrated excellent fit in the hold-out sample (CFI = .963, TLI = .955, RMSEA = .056, SRMR = .053), compared to the poor fit of the unidimensional model (CFI = .896).
Reliability and Validity: All three domains showed strong reliability ( $\omega$ = .84–.85; CR = .83–.84) and convergent validity (AVE > .50). Discriminant validity was confirmed via HTMT values (all < .85), with the strongest correlation between Informational and Social-Communication engagement (.79) and the weakest between Informational and Transactional (.66).
Engagement Profiles: Descriptive results indicated that Informational Engagement was the most frequent domain ( $M=3.44$ ), followed by Social-Communication ( $M=3.17$ ), while Transactional Engagement was the least frequent ( $M=1.98$ ), with significant floor effects for items like ride-hailing and travel booking.
Sociodemographic Patterns: Regression analyses confirmed that higher engagement was associated with younger age, higher education, higher income, and female gender. The education gradient was the strongest predictor ( $\beta = 0.275$ ).
Measurement Invariance: The three-factor structure demonstrated configural, threshold, loading, and residual invariance across gender, age, education, and race/ethnicity. Invariance for income showed a slightly larger change in CFI ( $\Delta$ CFI = .013) than the conventional .010 cutoff, suggesting income-based comparisons should be interpreted with caution, though the overall structure remained stable.

Key Contributions
The paper makes three primary contributions, framed modestly as a psychometric evaluation rather than a de novo scale development:

Psychometric Evaluation of Existing Items: The study demonstrates that existing HRS behavioral items support a coherent, reliable, and invariant multidimensional structure, validating their use for measuring digital engagement without the need for new item generation.
Partial Investigation of Dimensionality: It recovers the latent dimensions of older adults' technology use that are captured by the HRS module, distinguishing between informational, social, and transactional behaviors. The authors note that domains not represented in the HRS items (e.g., telehealth portals, AI use) remain invisible to this specific measure.
Theory-Guided Interpretation: The study provides a post-hoc theoretical interpretation of the empirically derived structure using Uses and Gratifications Theory (UGT) and Socioemotional Selectivity Theory (SST). UGT helps categorize the content of the three domains (informational, social, instrumental), while SST offers a developmental explanation for why older adults may prioritize low-risk informational/social activities over higher-risk transactional ones.

Significance and Claims
The authors claim that the Digital Engagement Scale (DES) offers a precise, behaviorally grounded tool for distinguishing how older adults participate in digital life, moving beyond the question of if they are online.

Reframing Digital Inequality: The results highlight a "partial inclusion" phenomenon where older adults may be highly engaged in informational and social domains but remain excluded from transactional domains (banking, travel, commerce) that are increasingly essential for daily autonomy. This suggests that digital inequality has shifted from access gaps to disparities in the types of activities performed.
Policy and Research Utility: By establishing measurement invariance, the DES allows for valid comparisons across key sociodemographic groups. The authors argue that interventions should be domain-specific; for instance, older adults with low transactional engagement may need trust-building and risk-reduction strategies rather than generic digital literacy training.
Limitations: The authors explicitly acknowledge that the scale is limited to the activities measured in the HRS module and does not assess skill, confidence, or the quality of use. The cross-sectional design precludes causal inferences, and the theoretical explanations regarding risk and motivation are interpretive rather than empirically tested within this study.

In conclusion, the paper posits that digital inclusion in later life cannot be understood as a single construct. The DES provides a validated mechanism to identify specific areas of digital marginalization, particularly in transactional activities, offering a more nuanced foundation for policy and research on digital inequality.