Introduction
Dementia, characterized by progressive cognitive decline and functional impairment, represents a formidable global health challenge [1,2]. As the world’s population ages rapidly, the prevalence of dementia is projected to surge, with an estimated 153 million individuals affected by 2050 [3]. In the absence of curative treatments, the imperative for effective prevention strategies has never been more pressing. The etiology of dementia is multifactorial, encompassing both nonmodifiable factors such as age and genetic predisposition, and modifiable risk factors including cardiovascular health, lifestyle choices, and social engagement [4-6].
In the context of an increasingly digitalized society, interaction with technology has become integral to modern life. However, the digital revolution has not benefited all segments of the population equally. A significant proportion of older adults find themselves in a state of “digital isolation,” either due to limited access or inadequate digital literacy [7-9]. This concept extends beyond traditional social isolation by emphasizing the absence of digital engagement, including the use of the internet, smartphones, or social media, which can offer additional cognitive and social stimulation [10,11]. Recent studies indicate that older adults’ use of technology-based social platforms or electronic health resources may enhance cognitive outcomes, postpone cognitive decline, and alleviate loneliness. Consequently, individuals who are digitally isolated may miss these protective effects, which could accelerate cognitive decline and elevate their risk of dementia [11-13]. Limited literacy and education levels may further restrict older adults’ ability to engage with digital technologies, hindering their ability to benefit and potentially increasing the risk of cognitive decline [14,15].
While the association between traditional social isolation and dementia risk has been extensively studied, the impact of digital isolation, a phenomenon unique to our technologically driven era, has received comparatively little attention [16-18]. Preliminary investigations suggest a potential link between digital isolation and accelerated cognitive decline, as well as increased dementia risk. However, these studies are often constrained by limited sample sizes and cross-sectional designs, precluding the establishment of causal relationships. Furthermore, many existing studies fail to adequately control for potential confounding factors such as depression, anxiety, chronic comorbidities, and lifestyle variables, potentially biasing their results [19,20].
This study aims to clarify the relationship between digital isolation and dementia risk using a large-scale, longitudinal cohort design. Using a multistage Cox proportional hazards model on discovery and validation cohorts, we control for numerous potential confounders, including sociodemographic characteristics (eg, education level), baseline health status, and lifestyle factors. This study aims to provide strong evidence supporting the hypothesis that higher levels of digital isolation increase dementia risk. It also seeks to identify mechanisms underlying this emerging risk factor and to inform novel dementia prevention strategies, particularly those enhancing digital literacy and technology access. We aim to support a holistic public health approach that integrates traditional and digital aspects of social engagement in aging populations.
Methods
Study Population
This investigation used data from the National Health and Aging Trends Study (NHATS), a nationally representative longitudinal survey of Medicare beneficiaries aged 65 years and older in the United States. To ensure representativeness of the older adult population, appropriate survey weights provided by NHATS were applied to account for the complex sampling design. Our analysis used data spanning from the 3rd wave (2013), when digital product usage assessment was initiated, to the 12th wave (2022). The study cohort was stratified into discovery and validation samples.
In the third wave (2013), 5799 participants remained in the study. We excluded individuals lacking baseline digital isolation data or with preexisting dementia diagnoses. We rigorously controlled potential confounders influencing the digital isolation-dementia risk relationship, including age, education level, gender, race or ethnicity, baseline diseases, depression, anxiety, smoking status, and sleep difficulties. Participants were followed from the fourth wave (2014) through the twelfth wave (2022). During follow-up, some individuals were excluded due to attrition or death before dementia diagnosis. The final analytical sample comprised 4455 individuals.
To validate our findings, we used an independent cohort of 4182 individuals newly recruited in the fifth wave (2015). After applying the same exclusion criteria as the discovery cohort and accounting for attrition and mortality during follow-up, 3734 individuals were included in the validation sample. This cohort was followed from the 5th wave (2015) through the 12th wave (2022). As shown in Figure 1, we provide a detailed overview of the inclusion and exclusion processes for the discovery and validation cohorts, including the study waves, exclusion criteria, and the final number of participants included in the analysis dataset.
Enlarge this image.Figure 1. Flowchart depicting participant inclusion and exclusion criteria for the discovery (wave 3) and validation (wave 5) cohorts, with subsequent follow-up through wave 12 and formation of the pooled sample. NHATS: National Health and Aging Trends Study.
Digital Isolation
Digital isolation was assessed using a composite digital isolation index, which was constructed based on individuals’ usage of various digital devices and the internet. The design of this index was informed by relevant literature in the fields of social isolation and digital health. For instance, Cornwell and Waite [21] quantified social isolation by evaluating the extent of individuals’ social contact and participation and proposed a method for constructing a social isolation index using self-reported data. Similarly, digital isolation, as a modern form of isolation, is primarily reflected in the insufficient engagement with and use of digital technologies.
In this study, we operationalized digital isolation through a composite digital isolation index comprising 7 parameters (Textbox 1).
Textbox 1.Components of the Digital Isolation Index.
Parameters:
Dementia
In the NHATS database, dementia ascertainment is primarily predicated on cognitive function assessments and self-reports or proxy reports. NHATS uses a multifaceted approach to dementia evaluation, encompassing cognitive testing, proxy reports, and clinical records. Specifically, NHATS uses a battery of cognitive tests to assess participants’ memory, attention, and executive function, with these metrics serving as indicators of cognitive status. Concurrently, NHATS collects proxy reports, typically from family members or caregivers, regarding the participants’ cognitive condition. These reports may encompass physician-diagnosed dementia and observe cognitive deficits in activities of daily living. Investigators typically synthesize these data with additional clinical information to determine dementia status and monitor its progression longitudinally. Upon confirmation or report of dementia in any follow-up wave, subsequent inquiries regarding dementia status are discontinued for that participant.
Covariates
This study incorporated a comprehensive set of covariates to ensure a precise estimation of the association between digital isolation and dementia risk. These covariates encompass sociodemographic characteristics, clinical parameters, and health-related behaviors.
Sociodemographic variables include education level, age, gender, and race or ethnicity. Education level was categorized as < high school, high school or general educational development (GED), some college, and college or above, reflecting participants’ highest educational attainment. Age was stratified into 6 cohorts: 65-69 years, 70-74 years, 75-79 years, 80-84 years, 85-89 years, and ≥90 years. Sex was dichotomized as male or female. Race or ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Hispanic, and other. These variables are routinely used as baseline covariates to adjust for sociodemographic heterogeneity in health outcomes [25-27].
Clinical parameters comprise the number of baseline diseases, depressive symptomatology, and anxiety manifestations. The number of baseline diseases was trichotomized based on self-reported chronic conditions (including arthritis, cardiovascular disease, hypertension, diabetes, pulmonary disease, cerebrovascular accident, osteoporosis, and malignancy): no diseases, 1-2 diseases, and ≥3 diseases. Depressive and anxiety symptoms were assessed by validated self-report instruments and operationalized as binary variables in the analyses [28,29].
Health-related behaviors include smoking status and sleep difficulties. Smoking status was dichotomized as current smokers and noncurrent smokers to account for potential cognitive effects of tobacco use [30]. Sleep difficulties were stratified based on the frequency of difficulty falling asleep within 30 minutes: high sleep difficulty (every night or most nights), moderate sleep difficulty (some nights), and low or no sleep difficulty (rarely or never) [31]. Previous research has demonstrated associations between poor sleep quality, cognitive decline, and elevated dementia risk, warranting its inclusion as a key covariate.
Statistical Analysis
To address the complex sampling design of NHATS, we applied person-level sampling weights for longitudinal analyses spanning 2013 to 2022. Clustering and stratification were incorporated by specifying the primary sampling units (PSUs) and strata in the design-based analyses. Baseline characteristics of the study population were summarized using descriptive statistics, with categorical variables reported as frequencies and percentages. Appropriate NHATS survey weights were applied across all analyses to ensure the findings represent the broader US older adult population. To assess the probability of dementia-free survival across different digital isolation groups (low isolation vs moderate to high isolation), we used Kaplan-Meier survival analysis. The resultant Kaplan-Meier survival curves are depicted in Figure 2. To quantify the association between digital isolation and dementia risk, we used Cox proportional hazards regression models, calculating hazard ratios (HRs) with corresponding 95% CIs. A total of 2 models were constructed: model 1 (unadjusted) and model 2 (adjusted for potential confounders including education level, age, gender, race, number of baseline diseases, depression, anxiety, smoking status, and sleep difficulties).
We further investigated the individual components of digital isolation (eg, mobile phone use, computer use, tablet use, frequency of email usage, internet usage, online activities, and health-related online activities) and their respective impacts on dementia risk. Each component was analyzed independently using Cox proportional hazards models to estimate HRs and 95% CIs.
To examine the potential heterogeneity in the association between digital isolation and dementia, we conducted stratified analyses across subgroups defined by gender, age, race or ethnicity, and comorbidity status. Effect modification by these factors was assessed through the inclusion of interaction terms in the Cox models. In addition, a sensitivity analysis was performed to ensure the robustness of our findings. We recalculated the primary models under several alternative conditions, including adjusting the threshold for defining digital isolation, excluding participants with missing key variables, and omitting early dementia cases to reduce potential reverse causality. To enhance the robustness of our findings, we combined the results from the discovery and validation samples using meta-analytic techniques. Pooled HRs and CIs were computed using a fixed-effects model, under the assumption of homogeneity across samples. Statistical analyses were performed using R statistical software (version 4.4.1; R Foundation for Statistical Computing). A 2-sided P value < .05 was considered statistically significant for all analyses.
Enlarge this image.Figure 2. Kaplan‒Meier curves of the relationship between digital isolation and dementia.
Ethical Considerations
The study protocol was approved by the institutional review board of Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China. All methods were conducted in accordance with relevant guidelines and regulations. The NHATS dataset is deidentified and publicly available, with informed consent obtained from participants at enrollment. No additional individual-level consent was necessary for this secondary analysis.
Results
Baseline Characteristics of the Study Population
The study cohort comprised 8189 participants, with 4455 individuals in the discovery sample and 3734 in the validation sample. Survey weights provided by NHATS were applied in all analyses to ensure representativeness of the older adult population in the United States. As shown in Table 1, Females constituted a slight majority, accounting for 4724 out of 8189 (57.68%) of the cohort. The racial or ethnic composition was predominantly non-Hispanic White (5615/8189, 68.58%), followed by non-Hispanic Black (1677/8189, 20.48%) and Hispanic (469/8189, 5.73%). Regarding education level, approximately 1250 out of 8189 (15.27%) of participants had less than a high school education, 2560/8189 (31.28%) completed high school or GED, 2375 out of 8189 (29.00%) had some college education, and 2004 out of 8189 (24.45%) had a college degree or higher. Age distribution analysis revealed that the “70-74 years” age group was most prevalent (1906/8189, 23.27%), succeeded by the “75-79 years” (1689/8189, 20.63%) and “80-84 years” (1479/8189, 18.06%) age groups. With respect to baseline diseases, 3918 out of 8189 (47.85%) of participants reported no chronic conditions, while 2593 out of 8189 (31.67%) had 1-2 diseases, and 1678 out of 8189 (20.49%) had 3 or more chronic conditions. The prevalence of depression and anxiety was 2189 out of 8189 (26.74%) and 2819 out of 8189 (34.43%), respectively. The majority of participants (7582/8189, 92.58%) were nonsmokers. Sleep difficulties were reported as significant by 1707 out of 8189 (20.84%) of participants, and moderate by 1989 out of 8189 (24.29%). In the context of digital isolation, 4473 out of 8189 (54.62%) of participants were categorized as experiencing moderate to high isolation, while 3716 out of 8189 (45.38%) were classified as having low isolation. Digital device use rates were as follows: mobile phones (6615/8189, 80.8%), computers (4857/8189, 59.31%), and tablets (2045/8189, 24.98%). Frequent engagement in digital activities was reported for email or text messaging 6537/8189, (79.83%), internet access (4709/8189, 57.51%), general online activities (7282/8189, 88.93%), and health-related online activities (6315/8189, 77.13%).
Table 1.Baseline characteristics of the study population.
| Variables | Discovery sample (N=4455) | Validation sample (N=3734) | Pooled sample (N=8189) | |
| Gender, n (%) | ||||
| Male | 1843 (41.37) | 1622 (43.44) | 3465 (42.32) | |
| Female | 2612 (58.63) | 2112 (56.56) | 4724 (57.68) | |
| Race or ethnicitya, n (%) | ||||
| White, non-Hispanic | 3162 (70.98) | 2453 (65.69) | 5615 (68.58) | |
| Black, non-hispanic | 915 (20.54) | 762 (20.41) | 1677 (20.48) | |
| Hispanic | 230 (5.16) | 239 (6.4) | 469 (5.73) | |
| Other | 148 (3.32) | 280 (7.5) | 428 (5.22) | |
| Education level, n (%): | ||||
| < High school | 680 (15.26) | 570 (15.27) | 1250 (15.27) | |
| High school or GED | 1380 (31) | 1180 (31.6) | 2560 (31.28) | |
| Some college | 1370 (30.77) | 1005 (26.91) | 2375 (29) | |
| College or above | 1025 (23) | 979 (26.22) | 2004 (24.45) | |
| Age (years), n (%) | ||||
| 65 to 69 | 459 (10.3) | 1000 (26.78) | 1459 (17.82) | |
| 70 to 74 | 1076 (24.15) | 830 (22.23) | 1906 (23.27) | |
| 75 to 79 | 961 (21.57) | 728 (19.5) | 1689 (20.63) | |
| 80 to 84 | 899 (20.18) | 580 (15.53) | 1479 (18.06) | |
| 85 to 89 | 647 (14.52) | 355 (9.51) | 1002 (12.23) | |
| 90 or above | 413 (9.27) | 241 (6.45) | 654 (7.99) | |
| Baseline disease, n (%) | ||||
| No disease | 3568 (80.09) | 350 (9.37) | 3918 (47.85) | |
| 1-2 diseases | 861 (19.33) | 1732 (46.38) | 2593 (31.67) | |
| 3 or more diseases | 26 (0.58) | 1652 (44.24) | 1678 (20.49) | |
| Depressionb, n (%) | ||||
| No depression | 3241 (72.75) | 2759 (73.89) | 6000 (73.26) | |
| Depression | 1214 (27.25) | 975 (26.11) | 2189 (26.74) | |
| Anxiety, n (%) | ||||
| No anxiety | 2908 (65.27) | 2462 (65.93) | 5370 (65.57) | |
| Anxiety | 1547 (34.73) | 1272 (34.07) | 2819 (34.43) | |
| Smoking status, n (%) | ||||
| Nonsmoker | 4152 (93.2) | 3430 (91.86) | 7582 (92.58) | |
| Smoker | 303 (6.8) | 304 (8.14) | 607 (7.42) | |
| Sleep difficultyc, n (%) | ||||
| High difficulty | 848 (19.13) | 859 (23.07) | 1707 (20.84) | |
| Medium difficulty | 1115 (25.15) | 874 (23.47) | 1989 (24.29) | |
| Low or no difficulty | 2470 (55.72) | 1991 (53.46) | 4461 (54.47) | |
| Digital isolationd group, n (%) | ||||
| Low isolation | 1803 (40.47) | 1913 (51.23) | 3716 (45.38) | |
| Medium and high isolation | 2652 (59.53) | 1821 (48.77) | 4473 (54.62) | |
| Items of digital isolation, n (%) | ||||
| Phone use, n (%) | 3446 (77.35) | 3169 (84.87) | 6615 (80.8) | |
| Computer use, n (%) | 2531 (56.81) | 2326 (62.29) | 4857 (59.31) | |
| Tablet use, n (%) | 764 (17.15) | 1281 (34.31) | 2045 (24.98) | |
| Email frequency, n (%) | 3603 (80.88) | 2934 (78.58) | 6537 (79.83) | |
| Internet use, n (%) | 2498 (56.07) | 2211 (59.21) | 4709 (57.51) | |
| Online activity, n (%) | 3945 (88.55) | 3337 (89.37) | 7282 (88.93) | |
| Health-related online use, n (%) | 3484 (78.2) | 2831 (75.82) | 6315 (77.13) | |
aRace or ethnicity was reclassified into 4 categories: White, non-Hispanic; Black, non-Hispanic; Hispanic; and Other.
bThe depression, anxiety, and smoking status variables were reclassified as binary variables. Responses of “Don\'t know,” “Refused to answer,” and similar were treated as missing data.
cSleep difficulty was categorized by the frequency of difficulty falling asleep within 30 minutes, with unknown or refused responses treated as missing data.
dThe Digital Isolation Index was used to create a new variable (Digital Isolation Group) dividing participants into low and moderate and high isolation groups.
Association Between Digital Isolation and the Risk of Dementia
The association between digital isolation and dementia risk was comprehensively examined, adjusting for factors including education level, age, gender, race or ethnicity, number of baseline diseases, depression, anxiety, smoking status, and sleep difficulties, with results presented in Table 2. In the discovery sample, the moderate to high isolation group demonstrated a significantly elevated risk of dementia compared with the low isolation group. The unadjusted Cox proportional hazards model (model 1) yielded a HR of 1.58 (95% CI 1.31-1.89, P < .001) for dementia in the moderate to high isolation group, indicating a 58% higher relative risk of dementia compared with the low isolation group. After adjusting for potential confounders (age, gender, race or ethnicity, number of baseline diseases, depression, anxiety, smoking status, and sleep disorders), the HR in model 2 attenuated to 1.22 (95% CI 1.01-1.47, P=.041), yet remained statistically significant. The validation sample corroborated these findings. The unadjusted model 1 revealed an HR of 2.36 (95% CI 1.86-3.01, P < .001) for the moderate to high isolation group, while the adjusted model 2 showed an HR of 1.62 (95% CI 1.27-2.08, P < .001), consistently indicating a significantly higher dementia risk in the moderate to high isolation group.
Table 2.Association between digital isolation and the risk of dementia.
| Sample and variables | Event, n/N (%) | Model 1a HRb (95% CI) | P value | Model 2c HR (95% CI) | P value | ||||||
| Discovery sample | |||||||||||
| Low isolation (reference range) | 1803/4455 (40.47) | 1 (Ref) | —d | 1 (Ref) | — | ||||||
| Medium and high isolation | 2652/4455 (59.53) | 1.58 (1.31-1.89) | < .001 | 1.22 (1.01-1.47) | .04 | ||||||
| Validation sample | |||||||||||
| Low isolation (reference range) | 1913/3734 (51.23) | 1 (Ref) | — | 1 (Ref) | — | ||||||
| Medium and high isolation | 1821/3734 (48.77) | 2.36 (1.86-3.01) | < .001 | 1.62 (1.27-2.08) | < .001 | ||||||
| Pooled sample | |||||||||||
| Low isolation (reference range) | 3716/8189 (45.38) | 1 (Ref) | — | 1 (Ref) | — | ||||||
| Medium and high isolation | 4473/8189 (54.62) | 1.89 (1.63-2.19) | < .001 | 1.36 (1.16-1.59) | < .001 | ||||||
aModel 1: unadjusted Cox proportional hazards model.
bHR: hazard ratio.
cModel 2: Cox proportional hazards model adjusted for age, gender, race, education level, baseline disease, depression, anxiety, smoking status, and sleep difficulty.
dNot applicable.
To enhance the robustness of our findings, we applied meta-analytic techniques to pool results from both samples. The pooled analysis yielded HRs of 1.89 (95% CI 1.63-2.19, P < .001) in the unadjusted model 1 and 1.36 (95% CI 1.16-1.59, P < .001) in the adjusted model 2, further substantiating the significant risk increase.
Kaplan-Meier survival curves were used to visualize dementia incidence across different digital isolation groups (Figure 2). These curves demonstrated a significantly higher probability of dementia development in the moderate to high isolation group, corroborating the Cox regression analysis results. Figure 3 illustrates the association between digital isolation and dementia risk across various demographic and clinical subgroups. Consistently elevated dementia risk was observed in the moderate to high isolation group across all subgroups, with notable differences in gender, age, and comorbidity status. These findings underscore the pervasive impact of digital isolation on dementia onset, demonstrating a consistent trend of increased risk both in the overall population and within specific subgroups.
Enlarge this image.Figure 3. Association between digital isolation and risk of dementia among subgroups. Reference group: Low digital isolation group. All models were adjusted for age, gender, race, baseline disease, depression, anxiety, smoking, sleep difficulty. P-int represents the heterogeneity among subgroups based on the metaregression analysis.
Association Between Digital Isolation Components and the Risk of Dementia
We investigated the association between individual components of digital isolation and incident dementia risk, with results summarized in Table 3. In the discovery sample, we analyzed the use of mobile phones, computers, tablets, and the internet in relation to dementia risk. The unadjusted Cox proportional hazards model (model 1) revealed that nonusers of mobile phones had a significantly higher dementia risk compared with users (HR 1.78, 95% CI 1.41-2.24, P < .001). After adjusting for covariates (model 2), the HR attenuated to 1.45 (95% CI 1.14-1.84, P=.002). For computer use, the adjusted HR was 1.15 (95% CI 0.96-1.38, P=.12), approaching but not reaching statistical significance. Similarly, adjusted HRs for tablet use were not statistically significant across samples, with a pooled HR of 1.05 (95% CI 0.78-1.42, P=.73).
Table 3.Association between digital isolation components and the risk of dementia.
| Component and variables | Discovery sample | Validation sample | Pooled sample | |||||||||
| Model 1a, HRb (95% CI) | P values | Model 2c, HR (95% CI) | P values | Model 1, HR (95% CI) | P values | Model 2, HR (95% CI) | P values | Model 1, HR (95% CI) | P values | Model 2, HR (95% CI) | P values | |
| Phone use (nonuse vs use) | 1.78 (1.41-2.24) | < .001 | 1.45 (1.14-1.84) | .002 | 3.05 (2.28-4.08) | < .001 | 2.12 (1.55-2.9) | < .001 | 2.05 (1.6-2.62) | < .001 | 1.65 (1.35-2.02) | < .001 |
| Computer use (nonuse vs use) | 1.22 (1-1.5) | .052 | 1.10 (0.88-1.36) | .386 | 1.50 (1.15-1.97) | .003 | 1.28 (0.98-1.66) | .071 | 1.25 (0.98-1.59) | .074 | 1.15 (0.96-1.38) | .118 |
| Tablet use (nonuse vs use) | 1.39 (1.02-1.88) | .035 | 1.28 (0.94-1.73) | .108 | 1.19 (0.88-1.6) | .264 | 1.10 (0.82-1.48) | .503 | 1.10 (0.8-1.53) | .548 | 1.05 (0.78-1.42) | .734 |
| Email frequency (nonuse vs use) | 0.74 (0.5-0.99) | .043 | 0.78 (0.58-1.04) | .090 | 1.03 (0.74-1.44) | .848 | 1 (0.72-1.38) | .998 | 0.95 (0.7-1.28) | .707 | 0.9 (0.66-1.23) | .51 |
| Internet use (nonuse vs use) | 1.42 (1.12-1.79) | .004 | 1.28 (1-1.64) | .047 | 1.75 (1.3-2.34) | < .001 | 1.56 (1.15-2.11) | .005 | 1.6 (1.24-2.07) | < .001 | 1.42 (1.15-1.76) | .001 |
| Online activity (nonuse vs use) | 1.12 (0.78-1.59) | .541 | 1.1 (0.77-1.57) | .612 | 1.93 (1.24-3.02) | .004 | 1.78 (1.14-2.79) | .012 | 1.6 (1.1-2.31) | .013 | 1.32 (1-1.74) | .048 |
| Health-related online use (nonuse vs use) | 1.08 (0.78-1.48) | .656 | 1.15 (0.83-1.59) | .402 | 0.92 (0.61-1.4) | .707 | 0.9 (0.58-1.36) | .594 | 0.95 (0.7-1.28) | .745 | 0.9 (0.64-1.28) | .576 |
aModel 1: unadjusted Cox proportional hazards model.
bHR: hazard ratio.
cModel 2: Cox proportional hazards model adjusted for age, gender, race, education level, baseline disease, depression, anxiety, smoking status, and sleep difficulty.
In the validation sample, nonusers of mobile phones exhibited a significantly higher dementia risk, with an unadjusted HR of 3.05 (95% CI 2.28-4.08, P < .001), which decreased to 2.12 (95% CI 1.55-2.9, P < .001) after adjustment. Nonusers of the internet demonstrated a robust association with dementia risk, with a pooled adjusted HR of 1.42 (95% CI 1.15-1.76, P=.001). For individuals not engaging in online activities, the pooled adjusted HR was 1.32 (95% CI 1-1.74, P=.05), further underscoring the importance of digital engagement. However, nonusers of health-related online platforms showed no significant association with dementia risk (pooled HR 0.9, 95% CI: 0.64-1.28, P=.58).
These findings underscore that specific digital behaviors, particularly mobile phone and internet use, are significantly associated with a lower risk of dementia. In contrast, the influence of other digital components, such as computer or tablet use, appears less consistent or pronounced.
Sensitivity Analyses
To evaluate the robustness of our findings and address potential confounders, we conducted several sensitivity analyses. First, we repeated the Cox regression models after excluding participants with incomplete education data, confirming that the significant association between digital isolation and dementia risk remained robust even after adjusting for educational level (adjusted HR range 1.20-1.45, all P < .05). Second, we applied alternative thresholds for defining digital isolation, such as categorizing digital isolation index ≥4 as high isolation, and observed results consistent with the primary analysis. Third, we excluded individuals who developed dementia or died within the first 2 years of follow-up to reduce potential reverse causation, finding HRs comparable with those in the main analysis (HR changes < 10%). Finally, we replicated the main models without applying survey weights. Although the unweighted models yielded slightly smaller standard errors, the overall trends remained consistent. Collectively, these sensitivity analyses indicate that our key conclusions are robust across varying analytical assumptions and sample selections.
Discussion
Principal Results
This study examined the association between digital isolation and dementia risk, demonstrating that higher levels of digital isolation significantly increase the risk of dementia. These findings emphasize the critical role of digital engagement in promoting cognitive health among older adults and provide valuable insights for public health policy development. By using the nationally representative NHATS dataset and its survey weights, we ensured that our results reflect the broader US older adult population, thereby enhancing the robustness of our conclusions. Our analysis identified a significant relationship between digital isolation and dementia risk. In both discovery and validation samples, individuals who did not use basic digital devices (eg, mobile phones and the internet) or who lacked online activity participation exhibited a substantially higher dementia risk compared with their digitally engaged peers. Specifically, nonusers of mobile phones had a 1.65-fold increased risk (95% CI 1.35-2.02, P < .001), while nonusers of the internet demonstrated a 1.42-fold increased risk (95% CI 1.15-1.76, P=.001) in the pooled analysis. In addition, the pooled analysis showed that individuals with a higher digital isolation index faced a 1.36-fold increased risk of dementia (95% CI 1.16-1.59, P < .001). These results are consistent with previous research identifying social isolation and lack of social interaction as independent risk factors for dementia [13,32].
However, some digital activities, such as emailing or texting, showed no significant association with dementia risk. This might be because these activities are less interactive or infrequently used by older adults, offering limited cognitive or social stimulation compared with more engaging digital behaviors, such as frequent phone or internet use.
Furthermore, this study validates digital isolation as a novel form of isolation that may influence cognitive health through multiple mechanisms. Digital technologies can promote social interaction and participation, crucial for maintaining cognitive function [33-35]. Conversely, a lack of digital engagement may exacerbate social isolation, thereby increasing cognitive decline risk. In addition, digital technology use can provide cognitive stimulation, potentially helping older adults maintain brain vitality [36,37]. Thus, digital isolation not only reflects diminished social interaction but may also represent insufficient cognitive stimulation. These findings have significant public health implications. As supported by recent studies, educational attainment can significantly influence individuals’ ability to engage with digital tools, highlighting the importance of prioritizing older adults with lower educational backgrounds in targeted interventions [38-40]. As society undergoes digital transformation, issues related to digital engagement among older adults are becoming increasingly salient. Our study suggests that promoting digital technology use among older adults, particularly within vulnerable populations, could effectively mitigate dementia risk. This implies that enhancing digital literacy and expanding access to digital resources should be integral components of dementia prevention strategies. While other studies may not have used the specific “digital isolation index,” research examining older adults’ online engagement or technology use consistently demonstrates that digitally engaged older adults report better cognitive and psychosocial outcomes [10,41,42]. These findings reinforce the protective role of digital engagement in cognitive health and underscore the importance of encouraging digital technology adoption among older adults.
Limitations
However, this study has limitations. Despite controlling for multiple covariates, the observational nature of our study design precludes complete elimination of confounding factors. For instance, individuals’ health behaviors and cognitive abilities may simultaneously influence their digital technology use and dementia risk, potentially leading to reverse causality [19,43]. Furthermore, educational attainment could discourage older adults from adopting digital devices, complicating the interpretation of whether digital isolation is a cause or result of declining cognitive function. In addition, our reliance on self-reported data may introduce information bias, particularly in the preclinical stages of dementia when individuals’ recollection of their digital usage may not be entirely accurate.
Notwithstanding these limitations, this study provides important empirical support for exploring the relationship between digital isolation and dementia risk. Future research should further investigate factors that promote or hinder digital engagement among older adults to develop more effective interventions. In addition, prospective longitudinal studies, randomized controlled trials, and qualitative investigations are crucial for further validating our findings and examining the causal relationship between digital technology use and cognitive health.
Conclusions
This study, examining the association between digital isolation and dementia risk in older adults, underscores the critical role of digital engagement in maintaining cognitive health. Our findings demonstrate that digital isolation significantly increases dementia risk, particularly among individuals who do not use basic digital devices, lack online communication, and abstain from online activities. These findings indicate that strategies focused on enhancing digital literacy, promoting equitable access to digital tools, and addressing educational disparities may effectively mitigate dementia risk in an increasingly digitalized society.
Our research not only provides new evidence of the detrimental effects of digital isolation but also offers valuable insights for future public health interventions. For instance, community-based digital literacy programs and targeted technology access initiatives could help reduce digital isolation among older adults, especially those from underserved or low-education backgrounds. Enhancing digital literacy and expanding accessibility to digital resources among older adults may effectively reduce digital isolation, potentially lowering dementia incidence. Future research should explore the specific mechanisms through which digital engagement affects cognitive health and further validate the causal relationship between digital isolation and dementia through longitudinal studies and intervention trials.
In conclusion, this study highlights the importance of digital technology in older adults’ health and advocates for the integration of digital literacy education, user-friendly technological interfaces, and widespread digital resource availability in public health policies. These measures will facilitate older adults’ integration into the digital society, thereby potentially improving their cognitive health and overall quality of life. Ensuring that these interventions are adapted to individual educational levels may be pivotal for maximizing their preventive impact.
Acknowledgments
We would like to express our sincere gratitude to everyone who contributed to this study. Finally, we acknowledge the assistance of ChatGPT in refining the language and grammar of this manuscript.
Authors' Contributions
CD and NS, as co–first authors, were instrumental in the conceptualization, data analysis, and drafting of the manuscript. GL provided essential software support and was pivotal in data visualization. KZ offered valuable supervision and secured the necessary resources and funding for the project. SY, as the corresponding author, provided overall guidance, ensuring the study’s success through his expertise in project administration and manuscript review.
Conflicts of Interest
None declared.
------------
Abbreviations
| GED: general educational development |
| HR: hazard ratio |
| NHATS: National Health and Aging Trends Study |
| PSU: primary sampling unit |
------------
Edited by A Mavragani; submitted 14.08.24; peer-reviewed by H Shah, AJ Santos; comments to author 29.12.24; revised version received 11.01.25; accepted 12.01.25; published 19.02.25.
Dementia U. What is Dementia. London. Dementia UK[Google Scholar]; 2019.
Kumar A, Vandana. Dementia: an Overview. J. Drug Delivery Ther. 2013;3(3):163-167.
GBD 2019 Dementia Forecasting Collaborators. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the global burden of disease study 2019. Lancet Public Health. 2022;7(2):e105-e125.
Licher S, Ahmad S, Karamujić-Čomić H, Voortman T, Leening MJG, Ikram MA, et al. Genetic predisposition, modifiable-risk-factor profile and long-term dementia risk in the general population. Nat Med. 2019;25(9):1364-1369.
Fillit H, Nash DT, Rundek T, Zuckerman A. Cardiovascular risk factors and dementia. Am J Geriatr Pharmacother. 2008;6(2):100-118.
Ngandu T, Lehtisalo J, Solomon A, Levälahti E, Ahtiluoto S, Antikainen R, et al. A 2 year multidomain intervention of diet, exercise, cognitive training, and vascular risk monitoring versus control to prevent cognitive decline in at-risk elderly people (FINGER): a randomised controlled trial. Lancet. 2015;385(9984):2255-2263.
Mubarak F, Suomi R. Elderly forgotten? Digital exclusion in the information age and the rising grey digital divide. Inquiry. 2022;59:469580221096272.
Reine I, Ivanovs A, Mieriņa I, Gehtmane-Hofmane I, Koroļeva I. Overcoming social isolation with digital technologies among ageing populations during covid-19. SIE. 2021;4:171-178.
Sin F, Berger S, Kim I, Yoon D. Digital social interaction in older adults during the COVID-19 pandemic. Proc. ACM Hum.-Comput. Interact. 2021;5(CSCW2):1-20.
Zhan R, Mpofu E, Olanrewaju S, Sutherland L. Community engagement in older adults with cognitive decline: the usage of information communication technology. Innovation in Aging. 2023;7((Supplement_1)):1007.
Byrne K, Ghaiumy Anaraky R. Identifying racial and rural disparities of cognitive functioning among older adults: the role of social isolation and social technology use. J Gerontol B Psychol Sci Soc Sci. 2022;77(10):1779-1790.
Neter E, Chachashvili-Bolotin S, Erlich B, Ifrah K. Benefiting from digital use: prospective association of internet use with knowledge and preventive behaviors related to alzheimer disease in the israeli survey of aging. JMIR Aging. 2021;4(2):e25706.
Yu B, Steptoe A, Chen Y, Jia X. Social isolation, rather than loneliness, is associated with cognitive decline in older adults: the China health and retirement longitudinal study. Psychol Med. 2021;51(14):2414-2421.
Wiese L, Park J. Digital learning and online chair yoga for rural underserved older adults at risk of cognitive decline. Innovation in Aging. 2022;6((Supplement_1)):94.
Lu SY, Yoon S, Yee WQ, Heng Wen Ngiam N, Ng KYY, Low LL. Experiences of a community-based digital intervention among older people living in a low-income neighborhood: qualitative study. JMIR Aging. 2024;7:e52292.
Horoszkiewicz B. Digital dementia and its impact on human cognitive and emotional functioning. J Educ Health Sport. Jan 2022;12(11):290-296.
Talbot CV, Briggs P. The use of digital technologies by people with mild-to-moderate dementia during the COVID-19 pandemic: a positive technology perspective. Dementia (London). 2022;21(4):1363-1380.
Rai HK, Kernaghan D, Schoonmade L, Egan KJ, Pot AM. Digital technologies to prevent social isolation and loneliness in dementia: a systematic review. J Alzheimers Dis. 2022;90(2):513-528.
Jin Y, Jing M, Ma X. Effects of digital device ownership on cognitive decline in a middle-aged and elderly population: longitudinal observational study. J Med Internet Res. 2019;21(7):e14210.
Xavier AJ, d'Orsi E, de Oliveira CM, Orrell M, Demakakos P, Biddulph JP, et al. English longitudinal study of aging: can internet/e-mail use reduce cognitive decline? J Gerontol A Biol Sci Med Sci. 2014;69(9):1117-1121.
Cornwell EY, Waite LJ. Social disconnectedness, perceived isolation, and health among older adults. J Health Soc Behav. 2009;50(1):31-48.
Kraut R, Patterson M, Lundmark V, Kiesler S, Mukopadhyay T, Scherlis W. Internet paradox. A social technology that reduces social involvement and psychological well-being? Am Psychol. 1998;53(9):1017-1031.
Makizako H, Shimada H, Tsutsumimoto K, Lee S, Doi T, Nakakubo S, et al. Social frailty in community-dwelling older adults as a risk factor for disability. J Am Med Dir Assoc. 2015;16(11):1003.e7-1003.11.
Wei K, Nyunt MSZ, Gao Q, Wee SL, Yap KB, Ng TP. Association of Frailty and Malnutrition With Long-term Functional and Mortality Outcomes Among Community-Dwelling Older Adults: Results From the Singapore Longitudinal Aging Study 1. JAMA Netw Open. 2018;1(3):e180650.
Brookmeyer R, Johnson E, Ziegler-Graham K, Arrighi HM. Forecasting the global burden of Alzheimer's disease. Alzheimers Dement. 2007;3(3):186-191.
Henderson VW. Estrogen, cognition, and a woman's risk of alzheimer's disease. Am J Med. 1997;103(3A):11S-18S.
Lim U, Wang S, Park S, Bogumil D, Wu AH, Cheng I, et al. Risk of alzheimer's disease and related dementia by sex and race/ethnicity: the multiethnic cohort study. Alzheimers Dement. 2022;18(9):1625-1634.
Burton C, Campbell P, Jordan K, Strauss V, Mallen C. The association of anxiety and depression with future dementia diagnosis: a case-control study in primary care. Fam Pract. 2013;30(1):25-30.
Becker E, Orellana Rios CL, Lahmann C, Rücker G, Bauer J, Boeker M. Anxiety as a risk factor of alzheimer's disease and vascular dementia. Br J Psychiatry. 2018;213(5):654-660.
Ott A, Slooter AJ, Hofman A, van Harskamp F, Witteman JC, Van Broeckhoven C, et al. Smoking and risk of dementia and alzheimer's disease in a population-based cohort study: the rotterdam study. Lancet. 1998;351(9119):1840-1843.
Shi L, Chen SJ, Ma MY, Bao YP, Han Y, Wang YM, et al. Sleep disturbances increase the risk of dementia: a systematic review and meta-analysis. Sleep Med Rev. 2018;40:4-16.
Guo L, Luo F, Gao N, Yu B. Social isolation and cognitive decline among older adults with depressive symptoms: prospective findings from the China health and retirement longitudinal study. Arch Gerontol Geriatr. 2021;95:104390.
Dodge HH, Zhu J, Mattek NC, Bowman M, Ybarra O, Wild KV, et al. Web-enabled conversational interactions as a means to improve cognitive functions: results of a 6-week randomized controlled trial. Alzheimers Dement (N Y). 2015;1(1):1-12.
Berner J, Comijs H, Elmståhl S, Welmer AK, Sanmartin Berglund J, Anderberg P, et al. Maintaining cognitive function with internet use: a two-country, six-year longitudinal study. International Psychogeriatrics. 2019;31(7):929-936.
Wang J, Liu J, Wang X, Zhu J, Bai Y, Che Y, et al. Association between change in social participation and improved cognitive function among older adults in China: a national prospective cohort study. Health Soc Care Community. 2022;30(6):e4199-e4210.
Gottlieb S. Mental activity may help prevent dementia. BMJ. 2003;326(7404):1418.
Wu Z, Pandigama DH, Wrigglesworth J, Owen A, Woods RL, Chong TTJ, et al. Lifestyle enrichment in later life and its association with dementia risk. JAMA Netw Open. 2023;6(7):e2323690.
Shin J, Kang H, Choi S, Choi J. Profiles of digital literacy among community-dwelling Korean older adults: a latent profile analysis. Innovation in Aging. 2023;7:1101.
Pelegrini LNDC, Casemiro FG, Bregola A, Ottaviani AC, Pavarini SCI. Performance of older adults in a digital change detection task: the role of heterogeneous education. Appl Neuropsychol Adult. 2023:1-9.
Ali Hafez S, Mohamed El-Sayed El-sheikh S, Abd El Fattah Ibrahim N, youssef Gaid J. Exploring dgital literacy among community- dwelling older adults. Alexandria Scientific Nursing Journal. 2024;26(2):186-197.
Kiselica AM, Stojanovic M, Waters A, Benge J. Association of technology use with subjective cognitive decline among older adults. Alzheimer's & Dementia. 2023;19(S23).
Ngiam NHW, Yee WQ, Teo N, Yow KS, Soundararajan A, Lim JX, et al. Building digital literacy in older adults of low socioeconomic status in Singapore (Project Wire Up): nonrandomized controlled trial. J Med Internet Res. 2022;24(12):e40341.
Smith S, Splonskowski M, Jacova C. Self-experienced cognitive function in the digital era: are older adults at risk of subjective cognitive decline? Innovation in Aging. Innovation in Aging. 2021;5((Supplement_1)):752.
© 2025. This work is licensed under https://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
Details
; Shen, Na
; Li, Guangzhou
; Zhang, Ke
; Yang, Shijun