The goal of study 1 was to initially assess the explicit and implicit attitudes of health care professionals toward older adults’ abilities to use DT and the association between these attitudes and ageism. As specific measures are currently unavailable, an additional goal was to develop and test the use of new direct and indirect measurements of DT-related attitudes and ageism. We hypothesized that health care professionals would express negative attitudes toward the abilities of older adults (compared to younger adults) to use DT. We further hypothesized that higher levels of ageism would be correlated with more negative attitudes.

We recruited physiotherapists working in the Netherlands and fourth year physiotherapy students who had already gained professional experience during their internships using available mailing lists between April 2019 and May 2019. Out of the 155 who were contacted, 97 participants voluntarily completed the questionnaire. A statistical power analysis was performed for sample size estimation, using G*Power 3.1.9.4 (Heinrich Heine University Düsseldorf). We used the assumption of a minimal effect size of 0.2 (as no prior knowledge of this scale is available), with an α of .05 and a power of 0.9. The projected sample size needed for a multiple regression with 4 predictors was 82, 75% (73/97) of participants were female, M_age = 32.39 (SD 11.24), 23% (22/97) were fourth year students, 34% (33/97) had 1-5 years of work experience, 28% (27/97) had 6-20 years of experience, and the remaining 15% (15/97) exceeded 20 years of experience. Finally, 47% (46/97) indicated that most of their patients were 65 years or older.

Participants received an invitation through email to participate in a study about how older adults use technology in health care and everyday life. Participants were directed to an online questionnaire on Qualtrics, where they gave consent to participate. Afterward, they answered questions about demographic information and to which age group most of their patients belonged. Following this, they answered the DT-related ageism measures (vignettes and ATOAUT scale) and the FSA.

Additionally, we inserted an unrelated validity item (“For this question only, mark the number 2”) in the middle of the FSA and ATOAUT scales to check actual reading of the question.

SPSS Statistics 26 (IBM Corp) was used to perform the analysis. In order to examine the ATOAUT scale, a principal component analysis was performed, and modifications were made to the scale. In order to examine our hypothesis on the attitudes of physiotherapists on the indirect measure of the vignettes, a repeated measures analysis of variance (ANOVA) was performed, with the sum of vignettes as the dependent variable, the age group assessed as the within-subject independent variable, primary age group of patients the physiotherapists work with as the between-subject independent variable, and the age of the physiotherapists as a covariate. Finally, to examine the correlation between ATOAUT and ageism (FSA scale) and other variables, we used a correlation matrix and a multiple regression.

The goal of study 2 was to test how age salience and stereotype activation might moderate the correlation between ageism and ATOAUT. Specifically, we wanted to address the limitations of study 1: that by merely asking all participants to rank the abilities of every age group to use DT, we actually primed social categorization [47] and age-based stereotypes. Therefore, we sought to control the age group salient in the vignettes, so that participants would need to rate the ability of a young or contrastively an old person to use the DTs described in the vignettes. Additionally, we sought to broaden our findings to a more diverse group of health care professionals. We hypothesized that older adults would be assessed as less able than younger adults in using health care DTs. Furthermore, we hypothesized that the “old” salience condition would prime age stereotypes and the need to categorize and differentiate oneself from the older group, leading to attitudes that are more negative. In contrast, rating the “young” condition would allow participants to affirm their self-concept without categorizing and comparing themselves to the older group. Therefore, we hypothesized that the age salient manipulation would act as a moderator in the correlation between ageism and ATOAUT.

We recruited 93 health care professionals and fourth year health care students in the Netherlands between December 2019 and February 2020. A statistical power analysis was performed for sample size estimation, using G*Power 3.1.9.4 (Heinrich Heine University Düsseldorf). We used the assumption of a minimal effect size of 0.2, with an α of .05 and a power of 0.9. The projected sample size needed for a multiple regression with 6 predictors was 94. Participants were recruited among the students and staff at a university of applied science, and the questionnaire was further distributed within their networks of health care professionals. Of the respondents, 67% (62/93) of the participants were female, M_{age =} 37.01 (SD 11.89), 38% (35/93) were physiotherapists, 25% (23/93) were speech therapists, 17% (16/93) were medical doctors, and 20% (19/93) belonged to other health professions. Moreover, 18% (17/93) were students, 33% (31/93) had 1-5 years of experience, 29% (27/93) had 6-20 years, and the remaining 20% (18/93) had more than 20 years of experience. For patient age, 41% (38/92) of the professionals indicated that most of their patients are 65 or older.

Participants received an invitation through email to participate in a study about health care professionals’ perspectives about using DT in everyday life and in health care. Participants were directed to an online questionnaire on Qualtrics, where they gave consent to participate and answered demographic questions. Participants were then randomly assigned to rate one of the contrastive age groups (young or old) in the 3 vignettes, and then completed the ATOAUT-10 and ERA-12 scales, and the question used in study 1 about the belief of age being a limiting factor. Finally, participants indicated the percentage of their patients who are above the age of 65 years. One validity check item was inserted in the middle of the ATOAUT-10 scale.

SPSS 26 (IBM Corp.) was used to perform the analysis. In order to examine our hypothesis that older adults would be assessed as less able than younger adults to use health care DTs on the indirect measure, we used a correlation matrix and multiple regression. In order to examine our hypothesis that age salience would moderate the correlation between ATOAUT and ageism (ERA-12), we employed a regression procedure using the PROCESS bootstrapping macro [54] in SPSS (model 1: 5000 iterations). Figure 1 presents the assumed moderation model.

We found that 27 participants did not answer the validity items correctly. Consequently, concern regarding their attention in answering was raised. Analysis including and excluding these participants, revealed that the results would not be different. We therefore included all 97 participants in this study.

The initial Cronbach α coefficient of the 15-item scale was .77. We then used a principal component analysis in order to examine the latent components of the scale. Bartlett test confirmed no violation of sphericity (approximate χ² = 379.1; P<.001). Five factors with eigenvalues larger than 1 were identified and accounted for 62.07% of the variance (see Table 1). Eight items loaded mainly on the first factor (all loadings above 0.43), representing stereotypes and prejudice toward older adults’ abilities to use DT (eg, “Using digital technology is harder for most older adults”), with the exception of item 2, were also strongly loaded on the fourth factor. Two additional items (5 and 12), loaded on the second factor, represented attitudes toward older adults’ access to DT and online digital services. Examining the remaining 5 items that did not load on the first and second factor revealed that the phrases used might have been ambiguous and interpreted variably. Consequently, answering them might not necessarily reflect stereotypes or prejudice toward older adults, but rather general attitudes toward the role of DT in improving well-being (item 8), matters of privacy (item 9), accessibility of the design (items 4 and 11), and how playful people of different ages are (item 3).

Item 10, which loaded on the first factor, seemed to have a weaker loading compared to the other items. This might have been related to confusion in the use of negation in this item. We therefore concluded to change the phrasing for future use to “Most older adults can give useful feedback about new digital technologies.” Finally we formed the new ATOAUT-10 scale, comprising items 1, 2, 5, 6, 7, 10, 12, 13, 14, and 15. The Cronbach α coefficient of the new 10-item scale was higher (.82), compared to the 15-item scale coefficient (.77), and explained 91.2% of the variance of the 15-item scale.

Table 2 presents the means, SDs, and correlations of the sum of the 3 vignettes for each age category, ATOAUT-10, FSA, and other variables. Physiotherapists assessed older age groups as less able to use the health care DTs described in the vignettes (Figure 2). We further examined these differences using a repeated measures ANOVA, with the sum of vignettes as the dependent variable, age group assessed as the within-subject independent variable, primary age group of patients the physiotherapists work with as the between-subject independent variable, and age of the physiotherapists as a covariate. By interpreting the Greenhouse-Geisser test (due to violation of sphericity), we found a significant main effect of the age group assessed (F_{2.33, 211.74}=34.66; P<.001; ηp²=0.28). Post hoc comparisons using the Bonferroni adjustment revealed that each age group above 31-50 years was assessed as significantly less able to use health care DT than the younger age group before it (all P values <.001). The interaction of age group assessed and age of the physiotherapists reached significance (F_2.33,211.74=5.79; P=.002; ηp²=0.06), such that younger physiotherapists assessed the ability of older adults to use the DTs as lower than that of older physiotherapists. The interaction between age group assessed and the primary age group of patients the physiotherapists work with was not significant.

As seen in Table 2, a more negative ATOAUT score was significantly correlated with higher levels of ageism (FSA), a belief that age can be a barrier to using DT, and lower perceptions of older adults’ abilities to use health care DTs (sum of vignettes for the 65-79 years and 80+ years age groups). Additionally, higher scores of ageism as measured by the FSA were correlated with lower perceptions of older adults’ abilities to use health care DTs (80+ years age group). To further examine the correlation between ATOAUT and ageism, we performed a multiple regression with ATOAUT-10 as the dependent variable, and FSA, age, gender, and the primary age group physiotherapists work with as independent variables. A significant model was found (adjusted R²=0.14; F_4,92=4.75; P=.002), indicating that beyond all predictors, FSA was the only significant predictor (β=.36; t=3.73; P<.001).

One participant did not answer the validity item correctly. Analysis that included this participant revealed that the results would not be altered, and the participant was hence included in the analysis. Table 3 presents the means and SDs of the main variables by condition and correlations. No significant differences were found between participants in the different age salient conditions (young or old) regarding background variables of age, gender, health care profession, and percentage of patients over the age of 65 years that the professional works with. We then analyzed the indirect measure of assessing the ability in using actual health care DTs, using a multiple regression with the sum of the vignettes as the dependent variable and age salience condition and background variables (age, gender, and percentage of patients over the age of 65 years) as independent variables. As observed in Table 3, younger adults were perceived as more likely to use the described DTs compared to older adults. This was qualified by a significant regression model (adjusted R²=0.59; F_4,87= 34.15; P<.001), with the age salience condition being the only significant predictor (β=–.78; t =–11.61; P<.001).

Examining the direct measure of the ATOAUT-10 scale revealed that negative ATOAUT scores correlated with the younger age of the health care professionals and belief that age is a barrier to using DT. Yet, the simple correlation with ageism (ERA-12) was insignificant. We then tested our hypothesis that age salience would moderate the correlation between ATOAUT and ageism. Table 4 and Figure 3 present the regression coefficients and interaction of age salience X ERA-12. A significant moderation model was found (R²=0.19, F_6,85=3.35; P=.005) and the age salience X ERA-12 interaction qualified as a significant moderator, adding significant explanatory variance to the model (R² change=0.05; F_1,85=4.90; P=.03). For the conditional effect of the old age salience condition, negative ERA-12 (higher ageism) was (marginally) associated with a more negative ATOAUT score (β=.34; t=0.82 P=.07; CI –0.03 to 0.72), whereas for the young age salience condition, it was not (β =–.22; t=–1.23; P=.22; CI –0.57 to 0.13). Moreover, the age of the professional was identified as a significant predictor of ATOAUT score, with the younger age of the professional being associated with a negative ATOAUT score.

The aim of these studies was to explore attitudes of health care professionals toward older adults’ abilities to use DT, as a specific domain of ageism. In study 1 we developed the ATOAUT-10 scale, a direct measurement of stereotypes and prejudice toward use of DT by older adults, and an indirect measurement (vignettes), which were thereafter used in study 2. In the process of developing the ATOAUT-10, we identified ten items that represent stereotypes and prejudice toward older adults in the context of using DT. As hypothesized, significant correlations were found between negative ATOAUT and higher levels of ageism, measured by 2 different ageism scales (FSA in study 1 and ERA-12 in study 2). More so, negative ATOAUT correlated with lower perception of older adults’ abilities to use actual health care DTs, measured by the vignettes; belief that age is a barrier to using DT; and younger age of the health care professionals (in study 2). Thus, enhancing its construct validity. The correlation with ageism measured by the FSA in study 1 was significant and accounted for additional predictors in a multiple regression. Yet, this correlation was small to medium, suggesting that ageism alone does not fully explain physiotherapists’ attitudes toward older adults’ abilities to use DT.

By using an indirect measure of the vignettes, we were able to assess health care professionals’ perceptions on how they believe older adults use health care DT. Our main finding in study 1, that individuals from older age categories are perceived as less able to use DTs is not surprising [10]. Yet still, the difference found between age groups was quite dramatic, with a significant difference between each age group above 31-50 years with the previous age group. Hence, a DT-specific age threshold [23] for negative assumptions about the ability to use health care DTs, might be as early as 50 years, which includes people in their working age who are considered to use DT on a daily basis, and not only people above 80 years. This finding alone might confirm an assessment that is based on stereotypes and not facts, as the majority of older adults in some countries use smartphones and the internet [30]. In study 1, rating all age groups by the participants might have activated age stereotypes by means of categorization and social comparison [47], but in study 2, the age group salient in the vignettes was manipulated (young or old). Once more, a significant difference was found in how the technological abilities of older adults were perceived by health care professionals. More importantly, as hypothesized, this subtle manipulation of age salience moderated the correlation between ageism and ATOAUT, such that higher levels of ageism were associated with a more negative ATOAUT score but only when old age was made salient beforehand. Although the moderator of age salience was significant, it should be noted that the effect size of the moderator’s addition to the explained variance was small.

These results reveal perspectives that are quite ageist, considering the accumulating evidence on the increasing prevalence of using DT by older adults. Furthermore, they demonstrate how DT-specific ageism can operate implicitly by merely inducing social comparison or making the concept of age salient. Consequently, this raises the concern that older adults might be discriminated in how they receive (or do not receive) technology-based treatments, as found in other studies with nontechnological treatments [37,38]. This is worrying, considering the discussed benefits of DT in facilitating health care and reducing costs [2]. Notably, we did not explicitly ask health care professionals about their intentions to offer DT-based treatments; therefore, the behavioral and discriminatory aspect of ageism was not addressed in this study. However, it can be assumed that attitudes and beliefs might influence intentions to use and actual use, as is often emphasized by technology acceptance models [7,9,10]. Interestingly, the correlation between ageism and ATOAUT was not significant for participants in the young age salient condition. This could be explained by the matching of the age category (young vs old) and the specific attributes of the context (eg, competency of older adults in using DT) that occurs only in the old age salient condition and may activate specific age-based stereotypes [55]. Although people might be unaware of the underlying processes that influence their attitudes, there is still a matter of controllability of induced behavior and actual use of stereotypes [15]. Explicit and implicit stereotypical evaluations are both, to some extent, prone to belief-based learning processes [18]. Therefore, control over behavioral expressions (namely discrimination) is also a matter of social norms and social acceptability. As ageism is still relatively socially acceptable, expressions of discrimination, especially in domains where older adults are highly stereotyped, such as DT, might be prevalent.

Although previous studies reported gender and more experience of treating older patients [41] as possible predictors of attitudes of health care professionals, these effects on ATOAUT or ageism were not found in this study. Younger age of the professional was found to predict negative attitudes only to a limited extent. Unfortunately, these findings do not shed new light into the inconclusive findings in the literature [35] and do not strongly support the idea that increased social contact with older adults [43] can reduce negative attitudes in relation to health care DTs. This might be due to the higher exposure of health care professionals to older adults who are ill or suffer from chronic conditions [45]. Subsequently, these findings suggest that stereotype activation might be a stronger predictor of negative ATOAUT score. However, the effect sizes found were relatively low. Perhaps a stronger manipulation of stereotype activation, such as priming negative age stereotypes, could lead to stronger effects. Alternatively, other characteristics should be taken into account, such as professionals’ desire to work with older adults, the valence of their professional and social contact with older adults, or previous experience of using DT in health care.

Knowledge and training are important aspects when trying to control or combat automatic stereotype activation. Gawronski et al [56] found that training people to acknowledge information that contradicts stereotypes may reduce automatic stereotype activation. Training might offer positive outcomes in reducing ageism in health care and enhance positive contact [57]. Nevertheless, formal training of health care professionals regarding ageing (more so stereotype activation) is still lacking in curricula, and the esteem of working with older patients is still low [58]. We therefore suggest that training should include modules to raise awareness on biases and how easily our behavior is affected by them.

This study focused mainly on the perspective of health care professionals. It is however important to consider that indirect expressions of ageism, including patronizing speech [17], “Elderspeak” [59] or other nonverbal communications [36], might in turn lead to stereotype activation within older patients. These stereotypes are often embodied [21] and directed toward oneself, further affecting participation [19] and actual use of DTs. Hence, attitudes of older adults, as well the reciprocal nature of the interaction between them and others, should be a focus of future research on DT-related ageism.

The findings of this study provide a preliminary basis for future research and validation of the ATOAUT-10 scale. However, there were 3 main limitations regarding our measurements. First, our sample size in study 1, while sufficient for the regression analysis, can be considered to be on the lower boundary of minimal sample sizes for factor analysis [60]. Furthermore, the sample in study 1 consisted only of physiotherapists. The sample in study 2, therefore, was broadened to a wider range of health care professionals. However, in order to expand the scale’s validity, future studies focusing solely on the validation of the scale with an appropriate sample size and including a diversity of health care professionals and other stakeholders, is needed. Second, it can be claimed that our sample was biased in gender and experience. Although early career and experienced professionals were represented in both studies, a better balanced and planned sampling might have enhanced the validation of the scale. Third, in study 1, a large number of participants missed the validity check item inserted in the FSA, which raised concerns they did not answer the FSA seriously enough. This could be due to the length of the FSA, and some outdated items which might be less suited to Dutch society. A recent review by Ayalon et al [11] found that the psychometric properties of the FSA are considered low compared to other ageism scales. This was the reason we used a different ageism scale in study 2 (ERA-12). Therefore, continued use of the ERA-12 in future studies aimed at replicating these findings is warranted. Finally, as mentioned, our study did not focus on the actual intentions of professionals to use health care DT, or, in other words, the behavioral aspect of discrimination and adoption of DT. This would also be a recommendation for future research with this scale, which could also be used to broaden theoretical models of technology adoption [7,9,10].

The highly negative attitudes of health care professionals toward older adults’ abilities to use DTs raise the question of how DTs are used in treatment. Activation of age-based stereotypes seems to play a pivotal role in these attitudes, possibly suggesting that nonadoption of DT is not entirely attributable to chronological age or individual characteristics. Using health care DTs indeed presents an opportunity to improve treatment of chronic diseases and well-being. However, when categorizing the lower access of older adults to use DTs and discussing the digital divide as a purportedly well-established fact [28], we are often preserving a negative view that might hamper the successful implementation of DTs in health care. Instead, it is essential to acknowledge how the field of DT is constantly evolving and includes individuals from all ages with different wants and needs. It is suggested that in order to increase adoption of DT, the focus ought to shift toward how we can change stereotypes and their activation on the individual and societal level.