This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
As for all individuals, the Internet is important in the everyday life of older adults. Research on older adults’ use of the Internet has merely focused on users versus nonusers and consequences of Internet use and nonuse. Older adults are a heterogeneous group, which may implicate that their use of the Internet is diverse as well. Older adults can use the Internet for different activities, and this usage can be of influence on benefits the Internet can have for them.
The aim of this paper was to describe the diversity or heterogeneity in the activities for which older adults use the Internet and determine whether diversity is related to social or health-related variables.
We used data of a national representative Internet panel in the Netherlands. Panel members aged 65 years and older and who have access to and use the Internet were selected (N=1418). We conducted a latent class analysis based on the Internet activities that panel members reported to spend time on. Second, we described the identified clusters with descriptive statistics and compared the clusters using analysis of variance (ANOVA) and chi-square tests.
Four clusters were distinguished. Cluster 1 was labeled as the “practical users” (36.88%, n=523). These respondents mainly used the Internet for practical and financial purposes such as searching for information, comparing products, and banking. Respondents in Cluster 2, the “minimizers” (32.23%, n=457), reported lowest frequency on most Internet activities, are older (mean age 73 years), and spent the smallest time on the Internet. Cluster 3 was labeled as the “maximizers” (17.77%, n=252); these respondents used the Internet for various activities, spent most time on the Internet, and were relatively younger (mean age below 70 years). Respondents in Cluster 4, the “social users,” mainly used the Internet for social and leisure-related activities such as gaming and social network sites. The identified clusters significantly differed in age (
Older adults are a diverse group in terms of their activities on the Internet. This underlines the importance to look beyond use versus nonuse when studying older adults’ Internet use. The clusters we have identified in this study can help tailor the development and deployment of eHealth intervention to specific segments of the older population.
In Western societies, Internet use is widespread and is increasingly important in diverse aspects of everyday life. For instance, Internet is indispensable in communication; access to news and information; and administrative applications such as applying for allowance, tax declaration, or Internet banking. Age is known to be strongly related to the likelihood that individuals use the Internet. In the Netherlands, among adults aged 65 years and older, 77.8% have access to the Internet, compared with 94% of Internet users among the whole population [
The digital divide framework [
The objective of this study was to identify and describe the diversity in older adults’ activities on the Internet and whether this diversity is related to social and health-related variables. The following research questions were formulated: (1) Which subgroups or clusters can be identified among older adults based on their Internet activities? (2) What are the features of these subgroups and how do they differ in their Internet activities, time spent on the Internet, and demographic variables? (3) Is there a difference between the subgroups concerning social and emotional loneliness, psychological health, and activities of daily living (ADL) of older adults?
We used data collected by an existing Internet panel that is representative of the Dutch population, namely, longitudinal Internet studies for social sciences (LISS) panel. This panel is administered by CentERdata, a Dutch research institute specialized in data collection. Panel members receive questionnaires every month and completed questionnaires are rewarded. The panel is based on a true probability sample of households drawn from the population register by Statistics Netherlands. Households are invited to participate in the panel, and people without an appropriate computer or Internet connection are provided equipment, insuring a representative sample. The LISS panel consists of 4500 households with approximately 7000 individuals.
For this study, data from 2 different questionnaires that are annually completed by LISS panel members were combined, namely, the LISS core studies “social integration and leisure” (data collection in October and November 2015) and “health” (data collection in July and August 2015). Demographic information such as age, gender, and marital status were measured in November 2015. Panel members aged 65 years and older were selected if they completed the “social integration and leisure” questionnaire that included questions regarding Internet use (N=1608). In addition, respondents were included in the analyses when they reported to have access and use the Internet, which was the majority 88.18% (1418/1608).
All measures were taken from annual core studies among the LISS panel members and therefore, questions were developed and tested by CentERdata. The core study, “social integration and leisure,” provided information regarding Internet use of the respondents. Web-based activities were assessed by 17 dichotomous items (never or ever spend time on this particular functionality or Web-based activity). Web-based activities included financially related activities (eg, “comparing products and searching product information,” “Internet banking”), functional and more traditional activities (“emailing,” “searching for information”), and social and leisure-related Internet activities (eg, “reading and viewing social media,” “playing Internet or Web-based games”). The amount of time spent on the Internet was asked by the following items: “Can you indicate how many hours you use the Internet on a computer or laptop/tablet/smartphone per week, on average (including emailing), besides when completing questionnaires of this panel? These items were added up to an overall amount of hours using the Internet. Answers ranged between 0 and 175 h per week. Since 175 h per week is an obvious outlier, we categorized the answers into 7 categories ranging from 1: ≤5 h, 2: 5-10 h, 3: 10-15 h...7: ≥30 h per week. Frequency of downloading apps was assessed by the items: “How often do you download apps on your tablet / smartphone” (1=never to 7=almost every day). We calculated mean scores based on both items with higher scores representing more frequent downloading apps.
Social and emotional loneliness was measured with the 6-item version of the loneliness scale of de Jong Gierveld [
The first step in the analyses was performing a latent class analysis (LCA) to identify underlying structure of the categorical data about Internet use among older people. LCA is a statistical and probabilistic method that can be used to classify individuals from a heterogeneous group into smaller more homogenous unobserved subgroups [
The second step in the analyses was describing and comparing the identified clusters. SPSS version 22 was used to conduct these analyses. A probability level of
In total, N=1418 respondents were included for the analyses who were individuals aged 65 years and older and using the Internet. Of these respondents, 52.82% were men (749/1418). Mean age of the respondents was 71.8 (standard deviation, SD 5.7). Of the selected sample, 8.04% (114/1418) was provided with equipment from LISS panel to be able to fill in the questionnaires monthly.
Background data of the study sample (N=1418).
Variable | n (%) | |
Age (mean 71.79, SD 5.68, range 65-93) | 1418 (100.00) | |
1418 (100.00) | ||
Men | 749 (52.82) | |
Women | 669 (47.18) | |
1418 (100.00) | ||
Married | 931 (65.66) | |
Separated | 4 (0.28) | |
Divorced | 173 (12.20) | |
Widow or widower | 235 (16.57) | |
Never been married | 75 (5.29) | |
1416 (99.86) | ||
Low education | 618 (43.64) | |
Middle education | 346 (24.44) | |
High education | 452 (31.92) | |
1409 (99.37) | ||
Dutch background | 1240 (88.01) | |
First generation foreign, Western background | 61 (4.33) | |
First generation foreign, non-Western background | 20 (1.42) | |
Second generation foreign, Western background | 82 (5.82) | |
Second generation foreign, non-Western background | 6 (0.43) | |
1413 (99.65) | ||
Extremely urban | 160 (11.32) | |
Very urban | 378 (26.75) | |
Moderately urban | 310 (21.94) | |
Slightly urban | 357 (25.27) | |
Not urban | 208 (14.72) |
aLow education refers to primary education or prevocational secondary education. Middle education refers to preuniversity education or secondary vocational education. High education refers to higher professional education or university education.
We compared the model fit indices, number of parameters, and classification error for models ranging from 1-8 clusters (see
Results of the latent class analysis (N=1418).
Model | LLa | BICb (LL) | AIC3c (LL) | # parameters | Classification error |
1-cluster | −9040.46 | 18168.00 | 18116.91 | 12 | 0 |
2-cluster | −8380.13 | 16941.69 | 16835.27 | 25 | 0.08 |
3-cluster | −8182.31 | 16640.38 | 16478.61 | 38 | 0.13 |
4-cluster | −8116.41 | 16602.92 | 16385.81 | 51 | 0.15 |
5-cluster | −8076.15 | 16616.74 | 16344.29 | 64 | 0.16 |
6-cluster | −8040.57 | 16639.93 | 16312.14 | 77 | 0.19 |
7-cluster | −8014.25 | 16681.63 | 16298.50 | 90 | 0.20 |
4-cluster with direct effectsd | −8039.69 | 16478.51 | 16244.37 | 55 | 0.15 |
aLL: Log likelihood.
bBIC: Bayesian information criterion.
cAIC3: Akaike’s information criterion 3.
dFour direct effects were included in the model based on bivariate residuals, namely (1) newsgroups—reading Web-based news and magazines, (2) searching for information—email, (3) product information—searching for information, and (4) reading Web-based news and magazines—watching Web-based films or TV programs.
Cluster 1 included 36.88% (523/1418) of the respondents, and these respondents can be described as the “practical users.” The majority of respondents in this cluster used the Internet for functional and financially related activities such as “comparing products and searching product information,” “purchasing items,” and “Internet banking.” In addition, “email” and “searching for information” was a frequently mentioned activity for which these practical users used the Internet. Among the practical users, the amount of men was high compared with the other clusters (65.0%, 340/523,
ANOVA tests were carried out to compare the 4 clusters on social and health-related variables, namely, social and emotional loneliness, psychological well-being, ADL, and iADL. Overall, no big differences were found in social and health-related variables between the identified clusters since effects sizes were all rather small (see
Frequency (%) of respondents ever spending time on an Internet activity per cluster.
Internet activity | Practical users | Minimizers | Maximizers | Social users | Chi-square |
Cramer |
n=523 | n=457 | n=252 | n=186 | |||
99.6 | 83.2 | 100 | 98.9 | <.001 | .33 | |
Searching for information | 98.5 | 79.0 | 98.0 | 91.4 | <.001 | .31 |
Comparing products or product information | 94.8 | 33.9 | 100 | 53.2 | <.001 | .64 |
Purchasing items | 81.8 | 8.5 | 100 | 14.0 | <.001 | .79 |
Watching Web-based films or TV programs | 15.1 | 4.6 | 38.5 | 17.2 | <.001 | .31 |
Downloading software or music or filmsa | 15.1 | 2.8 | 28.2 | 10.2 | <.001 | .26 |
Internet banking | 98.1 | 46.4 | 95.6 | 68.8 | <.001 | .55 |
Playing Internet or Web-based games | 20.8 | 19.3 | 40.5 | 52.7 | <.001 | .27 |
Reading Web-based news or magazines | 55.3 | 19.7 | 73.0 | 48.9 | <.001 | .39 |
Newsgroups | 18.4 | 10.1 | 29.8 | 24.7 | <.001 | .18 |
Reading and viewing social media | 23.5 | 8.5 | 99.6 | 93.5 | <.001 | .77 |
Reading or writing blogsa | 7.3 | 1.8 | 21.8 | 14.0 | <.001 | .25 |
Posting messages or photos or short films on social media | 1.0 | 2.4 | 59.9 | 57.5 | <.001 | .67 |
Chatting or video calling or sending messages | 33.3 | 5.9 | 80.6 | 52.7 | <.001 | .55 |
Dating websitesa | 1.5 | 0.9 | 2.8 | 3.8 | .05 | .07 |
Visiting forums and communitiesa | 3.3 | 0.9 | 11.9 | 3.8 | <.001 | .19 |
Other activities | 15.1 | 5.9 | 32.1 | 14.0 | <.001 | .25 |
aNot included in the latent class analysis because frequency of activity mentioned by <15% of the respondents.
Comparison (chi-square tests) of the identified clusters on demographic variables.
Demographic variables | Practical users |
Minimizers |
Maximizers |
Social users |
Χ2 | Cramer |
||
63.4 | <.001 | .21 | ||||||
Men | 340 (65.0) | 206 (45.1) | 136 (54.0) | 67 (36.0) | ||||
Women | 183 (35.0) | 251 (54.9) | 116 (46.0) | 119 (64.0) | ||||
27.4 | .007 | .08 | ||||||
Married | 348 (66.5) | 301 (65.9) | 170 (67.5) | 112 (60.2) | ||||
Separated | 1 (0.2) | - | 2 (0.8) | 1 (0.5) | ||||
Divorced | 64 (12.2) | 37 (8.1) | 39 (15.5) | 33 (17.7) | ||||
Widow or widower | 81 (15.5) | 93 (20.4) | 29 (11.5) | 32 (17.2) | ||||
Never married | 29 (5.6) | 26 (5.7) | 12 (4.8) | 8 (4.3) | ||||
90.0 | <.001 | .18 | ||||||
Low education | 187 (35.8) | 257 (56.2) | 72 (28.6) | 102 (54.8) | ||||
Middle education | 121 (23.1) | 96 (21.0) | 83 (32.9) | 46 (24.7) | ||||
High education | 215 (41.1) | 104 (22.8) | 96 (38.1) | 37 (19.9) |
Comparison (analysis of variances) of the identified clusters on age, Internet variables, and social and health-related variables.
Variables, mean (SDa) | Practical users1f |
Minimizers2f |
Maximizers3f |
Social users4f |
Welch |
ω2 | |
Age | 71.3 (5.3)2,3 | 73.8 (6.3)1,3,4 | 69.6 (4.4)1,2,4 | 71.1 (4.9)2,3 | 36.7e (3610) | <.001 | 0.07 |
Amount of hours spend on Internet per week | 2.4 (1.6)2,3 | 1.6 (1.3)1,3,4 | 3.4 (2.1)1,2,4 | 2.5 (1.7)2,3 | 63.3e (3550) | <.001 | 0.12 |
Frequency downloading apps | 1.7 (1.8)2,3 | 0.7 (1.4)1,3,4 | 2.6 (1.7)1,2,4 | 1.5 (1.8)2,3 | 87.2e (3567) | <.001 | 0.14 |
Psychological well-being | 79.9 (13.6)2 | 76.7 (15.4)1 | 78.7 (14.6) | 76.3 (15.5) | 5.0e (3558) | .002 | 0.01 |
Emotional loneliness | 0.5 (0.9) | 0.5 (0.9) | 0.5 (0.9) | 0.6 (1.0) | 1.7e (3576) | .17 | 0.00 |
Social loneliness | 1.1 (1.2) | 1.0 (1.2) | 1.0 (1.12) | 1.0 (1.1) | 0.1 (31,411) | .96 | 0.00 |
ADLc | 6.8 (1.9) | 7.1 (2.2) | 6.7 (1.8) | 6.8 (1. 6) | 3.9e (3591) | .009 | 0.01 |
iADLd | 8.3 (2.3)2 | 9.2 (3.4)1,3 | 8.2 (2.1)2 | 8.8 (2.4) | 9.5e (3588) | <.001 | 0.02 |
Experienced health | 2.9 (0.7) | 2.8 (0.7)3 | 3.0 (0.7)2 | 2.9 (0.7) | 4.0 (31,383) | .007 | 0.01 |
aSD: standard deviation.
bdf: degrees of freedom.
cADL: activities of daily living.
diADL: instrumental activities of daily living.
eWelch
fThe superscript numbers 1-4 indicate significant differences (<.01) between the clusters on Bonferroni and Games-Howell posthoc test.
The results of this study show that older adults are a diverse group concerning their activities on the Internet. We identified 4 clusters of older adults based on the activities for which they use the Internet. First, the minimizers are the oldest respondents (mean age 74 years old) and spend the least time on the Internet. The minimizers report the lowest frequency on most of the Internet activities and mainly use the Internet for traditional purposes such as email. Second, on the other end of the spectrum are the maximizers, who are relatively young (mean age below 70 years old), spend the most time on the Internet, and are reported to spend time on almost all of the Internet activities. Among the maximizers, the amount of men and women, as well as the different education levels, are equally distributed. The third and fourth clusters are in between the minimizers and maximizers: the practical users and social users. These clusters score in between the maximizers and minimizers regarding both time spent on the Internet and their age (mean age 71 years old). The practical users, in contrast with the social users, use the Internet mainly for financial and practical matters such as searching for information, comparing products, and Internet banking. As opposed to practical users, the social users mainly use the Internet for social and leisure related activities (social media, games, etc). The amount of men is higher among the practical users and the amount of women higher among the social users.
The clusters did not differ to a large extent in social and health-related variables. However, the minimizers reported lower psychological well-being compared with the practical users, more problems with iADL in comparison with the maximizers and practical users, and lower experienced health compared with the maximizers. In sum, it appeared that the minimizers show a somewhat lower health, but this cluster also comprised the oldest respondents (mean age 74 years old). As causality between the variables is unclear, it is unknown whether age causes older adults to be only minimally active on the Internet or that a lower health status causes lower Internet activity.
It has been established that physical and mental limitations may form a barrier for older adults to use computers and the Internet [
One study [
To our knowledge this is the first study that identifies clusters of older adults based on the activities for which older adults use the Internet. The large sample size strengthens the findings of this study. We strongly recommend other studies to consider using LISS panel data since the quality of the data is excellent. The Dutch population is considered to be comparable to other Western populations in terms of Internet use; therefore, we expect that the findings of this study apply to a large extent to other Western populations. Nevertheless, attention should be paid to the following limitations. Data of two surveys were combined and data collection took place on two different moments within a time span of 3 months. We are of the opinion that the variables included in this study are quite stable and are not expected to fluctuate to a large extent in a period of 3 months. The information on which the LCA was based was limited to dichotomous information whether respondents ever spend time on a particular Internet activity or not. We did not have information about the time spent on each of the Internet activities, nor did we have information about attitudes of the respondents with regard to Internet use. In addition, no information was available about support that older adults receive in using technologies which is known to be related to older adults’ use of technologies [
In the Netherlands, considerable emphasis is placed on increasing the use of eHealth, in particular among older adults and patient with chronic illnesses [
The findings of this study establish that older adults are a diverse group in terms of their activities on the Internet. This underlines the importance to look beyond use versus nonuse when investigating older adults’ Internet use. The heterogeneity in activities for which older adults use the Internet is widespread and is vital to consider when attempting to stimulate or facilitate Internet use among older adults. The clusters we have identified in this study can be useful in creating awareness of eHealth interventions among specific segments of the older population.
Akaike’s Information Criterion 3
activities of daily living
analysis of variance
Bayesian Information Criterion
instrumental activities of daily living
information and communication technology
latent class analysis
longitudinal Internet studies for social sciences
log likelihood
In this paper we used data from the longitudinal Internet studies for the social sciences (LISS) panel administered by CentERdata (Tilburg University, The Netherlands). The authors would like to thank CentERdata for providing the data.
None declared.