This study showed an eHealth disability digital divide in that participants with impairment reported less use and more difficulty in the use of eHealth compared to participants without impairment. When subgrouping impairments, which to our knowledge has not been previously done in this detailed way, the least use and most difficulty using eHealth were shown among participants with communication, language and calculation impairments, and intellectual impairments.
In line with our results, other studies have shown that people with impairment use eHealth services less  and report more difficulties  than people without impairment. The one study that did not show less use of the internet for health-related activities, did not include people with such impairments, who in the present study had the least use of, and the most difficulty in using eHealth, i.e. communication, language and calculation impairments, and intellectual impairments .
Booking healthcare appointments online and the SSIA website were the most difficult eHealth service in this study. Using these services is complex, as they require digital identification to log in, require understanding of digital functions and features and require considerable executive functions to operate. Twice as many participants with impairment avoided booking healthcare appointments online and reported difficulty in using the SSIA website, as compared to participants without impairment.
Among participants with communication, language and calculation impairments, intellectual impairments and all visual impairments (visual impairment, blindness and deaf-blindness), most of the eHealth services were reported as difficult to use: i.e., digital identification, the 1177.se web-portal and booking healthcare appointments online.
Participants with impairments relating to communication, language and calculation (Fig. 2) were grouped together based on similar co-morbidity and functioning regarding working memory, symbol interpretation and comprehension [27,28,29]. Indeed, the results confirmed that they presented a similar pattern of less use and more difficulties in the use of eHealth. Previous literature has discussed the limitation of eHealth lacking non-verbal communication (such as eye-contact) and communicative emotion (such as vocal intonation) , and concerns have been raised about the patient—care provider relationship and therapeutic alliance of sporadic contacts , which could plausibly be important among people with communication, language and calculation impairments. Design features to increase their accessibility are standards for understandable texts, text-to-speech technology, audio, speech-to-text technology and the possibility to import numbers from a reliable source rather than to enter them manually , as well as visualisations  and allowing for longer duration of communication .
The least use and most frequent difficulties in use of eHealth overall in this study were reported among participants with intellectual impairments. This is consistent with findings in other studies showing that people with intellectual impairments struggle with using the web [35, 36]. Improvements in the digital design that increase accessibility in relation to intellectual impairments are plain language with short sentences, visualisations, clear icons, calm background, audio including narrative information, intuitive navigation and language options [35, 37], whereas disablement by design can be caused by updates requiring new learning .
Participants with visual impairment reported less use of digital identification and more frequently avoiding booking healthcare appointments online, as compared to all other participants. Previous studies showed that not complying with design guidelines made digital services visually inaccessible, and that accessibility improved after re-design [38, 39]. People with progressive visual impairment can usually appreciate assistive devices such as magnifiers to reinforce sight . Assistive device compatibility could aid their access to eHealth , but also technical options for magnifying, such as contrast and brightness . Childhood onset severe visual impairment, on the other hand, makes communicative development rely on senses other than vision. Among participants with blindness, few reported use of digital identification or booking healthcare appointments online. Text-to-speech technology and audio could increase accessibility of eHealth [32, 42]. However, people with blindness often use their own assistive device with e.g. functions for screen reading, since built-in audio functions on webpages may need visual functioning to initiate. Thus, assistive device compatibility, in the 1177.se web-portal and for digital identification, is important. In the present study, participants with deaf-blindness had the most difficulty using digital identification and the 1177.se web-portal, as compared to all other participants. The diagnosis deaf-blindness defines when the level of visual and/or hearing impairment is too severe for one to compensate for the other. When people with deaf-blindness have residual hearing or vision, the same design functions as for other sensory impairments can make eHealth accessible for them. Otherwise, the tactile sense is important for their communication and eHealth use [9, 43].
In the present study, the only exception to the disability digital divide was booking healthcare appointments online, which was used by a higher proportion of participants with neurological and musculoskeletal impairments, deafness, and hearing impairment, as compared to participants without impairment. Also, participants with hearing impairment alongside participants without impairment were least likely to avoid booking healthcare appointments online. This finding is reasonable, since the telephone is difficult with hearing impairment and written information or administrative procedures can be a secure option avoiding potential loss of spoken communication . Therefore, eHealth may be an important tool for improved overall healthcare accessibility for people with hearing loss.
Our results show that people with impairment have more difficulties using eHealth than people without impairment. At the same time, people with impairment are under-represented in eHealth research . The results in the present study demonstrated that people with similar functional impairments report similar use of eHealth and difficulties using eHealth. Henni et al. recently reported the same findings in a scoping review, by combining data from multiple studies . Our analyses showed that the overlaps due to participants reporting several impairments did not exceed the predetermined cut-off, and thus people with multi-morbidity could be represented in the results. Hence, the purposeful subgrouping of impairments can be a useful tool to understand who will probably perceive difficulties in the use of eHealth. This can be used to inform designers and policy makers as to who should be involved in the design process of eHealth services. Our subgrouping of impairments shows that it is especially important that people with communication, language and calculation impairments, and intellectual impairments, are involved in the design of eHealth, as they reported the least use and the most difficulties. Evaluations show that many eHealth services do not comply with accessibility standards [8, 10]. Further, there is a critique that accessibility standards are too narrow, lacking cognitive accessibility [9, 34, 35], which was shown to be a prominent aspect for the impairments of the participants who in this study reported the most difficulties in the use of eHealth services. We suggest that including people with these types of impairment in co-design processes when developing eHealth services, would increase the focus on cognitive accessibility and complement existing accessibility standards. The importance of user participation in eHealth development is acknowledged by both research and policy makers, as it can improve accessibility [46,47,48,49]. Co-design of eHealth by user participation involving people with the most significant difficulties, will thereby produce eHealth services that are usable and accessible to the widest range of people, i.e. universal design [7, 9]. In short, designing for people with the most difficulties will produce eHealth for the whole population. Our subgrouping of impairments can also be used when evaluating effects on eHealth investments. If large proportions of the population avoid eHealth services, the favourable effects of those services will be lower than anticipated . Thus, the use of eHealth among people with impairment is important for accurately measuring the effects of eHealth and for obtaining maximal gain on eHealth investments.
Strengths and limitations
It is a major strength of this study that, by use of the snowball sampling method, we managed to achieve substantial participation of people with impairment, i.e. among populations considered hard-to-reach by conventional survey sampling methods . However, the use of snowball sampling mainly through online survey, plausibly reached more digitally literate people, which limits the generalisation of the findings to all people with impairment in Sweden. In addition, previous research has shown that self-assessment of digital literacy, might result in underestimation of difficulties [50, 51]. In summary, our results succeeded in measuring the disability digital divide of eHealth, but plausibly underestimated its severity.
A strength of this study is that we showed differences not only between people with and without impairment, but also differences between subgroups of impairments. We believe it to be a rigorous strategy to survey all significant impairments, if managing methodological challenges in the statistical analysis. Since having multiple impairments was more common than having a single diagnosis, the risk of misclassification bias in selecting a primary impairment would be substantial. Multicollinearity did not exceed the predetermined cut-off, sensitivity analyses did not have significant impact on odds ratios and outcomes were not associated with number of reported impairments. This contributed to a differentiated knowledge of the heterogeneity in use and difficulty in use of eHealth between purposefully grouped impairments. The proposed subgrouping of impairments is a first attempt that will need to be validated in future studies.
It is notable that 14% of those in the general population who responded to the survey reported having impairment. This roughly corresponds to estimates of prevalence of impairment in the population . It indicates that how we constructed the question on impairment was successful in attaining appropriate information of impairment in the general population. There are multiple approaches to constructing questions on impairment, which should be guided by the purpose, e.g. in medicine or for legal definitions . We want to stress the importance of asking about impairments in relation to the outcome. Functioning is indeed continuous, not dichotomous, and differs depending on the activity and the context. We asked participants to report impairment that ‘clearly affect how you live your life’ since eHealth usage requires high level of functioning. We developed the question on impairment in collaboration with people having different types of impairment, which strengthens the validity of the question. Altogether, by this approach we believe we present results that are more credible as compared to other national surveys which have used generic questions, without specifying type of impairment [16, 17]. We suggest that the question used in this study is favourable for reporting impairment and should also be used in other studies of accessibility of eHealth.
The body of literature is still scarce on accessibility of eHealth. However, there are more studies on digital accessibility in general [11, 19, 35,36,37,38, 43]. We find it reasonable to discuss our findings in relation to studies of web accessibility and digital technology in general. However, eHealth involves complex services and the interplay with health literacy compels caution in comparability. Further, there might be other factors not investigated in this study that are associated with accessibility of eHealth, for example eHealth literacy [6, 53] and socioeconomic factors [14, 15]. This study was undertaken prior to the COVID-19 pandemic. The pandemic has forced community services online  and it is plausible that digital participation has changed in the population, which raises questions on whether the findings are valid in a post-pandemic context. We are currently undertaking a survey to investigate changes in the disability digital divide related to the COVID-19 pandemic.