Chronic care for frail older people is fragmented, with involvement from a large and constantly changing group of professionals who are frequently unaware that they provide care to the same patient [1]. Such professionals include home care professionals, general practitioners (GPs), clinicians, physiotherapists, and case managers dedicated to long-term care of the patients in the community. Frail elderly often suffer from comorbidities, which results in care by multiple health care professionals [2]. Therefore lack of communication between professionals leads to a fragmented and ineffective health care delivery for frail elderly [3]. To reduce fragmentation and promote continuity of care, better coordination and communication between professionals and with patients is necessary. Online health communities (OHCs) have been recognized as an effective mechanism for supporting continuous care for frail older people [4], allowing better coordination and more efficient communication with patients and among professionals. OHCs consist of Internet-based platforms that unite groups of individuals with a shared goal or similar interest, including both professionals and patients [5]. The main strength of OHCs is that they allow communication between people who would not have met each other otherwise [5]. Thus, OHCs are particularly suited for improving the coordination of care for frail elderly who have multiple professional caregivers. For this purpose, we developed and evaluated the Health and Welfare Information Portal (Zorg en Welzijns Informatie Portaal, ZWIP, in Dutch) [1,6] on its effectiveness.

Overall, 290/622 (46.6%) of all frail persons identified within 11 practices participated in the intervention group. From 6 practices 392 frail older people participated in the control group. At 12-month follow-up, in the intervention group 179/290 (61.7% of original) patients provided data at follow-up, versus 270/392 (68.8% of original) patients in the control group. At baseline, participants in the intervention group were more likely to have completed primary education only, have more informal caregivers, and have higher complexity of care compared to the participants in the control group. Further, participants in the intervention groups also had a higher average frailty index score, and GPs had lower experience with coordination of care (Table 1).

One quarter 45/117 (25%) of all patients in the intervention group used ZWIP at least once a month during a period of 12 months. Controlling for frailty and other unbalanced baseline characteristics, we found no significant differences in primary patient outcomes (Table 2). Change in coordination of care as reported by GPs improved in the control group.

There were 290 patients who participated in the intervention group and 392 in the control group. In the intervention group 76/290 (26.2%) of the patients actively used ZWIP. After a follow-up of 12 months, we observed no significant improvement on primary patient outcomes, ADL, IADL, and mental health.

The online ZWIP platform was specifically developed for reducing fragmentation of care delivery in older people. Almost half of a frail elderly population without exclusion criteria could be included in the intervention group for using the online ZWIP tool [6]. This is modestly higher than what can be expected in the Dutch context, where 39% persons older than 75 years report having Internet access [22]. This study has two important limitations that can impact results. First, due to the observational nature of the study, comparability between the intervention and the control groups was limited. Despite adjusting for a range of covariates, there may be residual confounding.

Observational, controlled before-after designs are common for complex interventions, where randomized controlled trials (RCTs) are often not appropriate or feasible for evaluation [23]. In the case of ZWIP, contamination between patients would have made individual-level randomization inappropriate. Cluster randomization was not feasible because the level of commitment required from a number of local stakeholders could not be sustained in the control group.

A second limitation was the fact that actual usage of ZWIP was low, even though the implementation of ZWIP was prepared systematically during the development of ZWIP, as this is a structural part of intervention mapping [1,6,7,24,25]. Additionally, implementation strategies were added or adapted when needed during the actual implementation phase. A wide range of implementation strategies were used to encourage uptake; for example, a training program was developed for professionals and an active recruitment phase led to a high participation of older persons. Therefore, low levels of use were attained not because of the lack of, but despite using state of the art implementation techniques. Failure to integrate eHealth interventions in health care is widespread [26], and therefore the low levels of use of these frail older subjects is not surprising. This is especially true for sustained usage of an eHealth intervention [27]. As in other studies [26], further efforts should be focused on improving usability of the intervention, in terms of compatibility for frail older people in chronic disease trajectories [28].

In addition to further refinement, it is essential to identify those who benefit most from ZWIP and eHealth applications in general. The use of eHealth applications in frail populations could be increased by first identifying frail people with a high likelihood of early adoption of the eHealth intervention, such as people with high computer literacy. Which frail elderly are likely adopters requires further research [26]. Therefore, we plan to perform a quantitative and qualitative evaluation of ZWIP usage as well, going beyond the scope of this paper. We must recognize that in the early stages of evaluation, we take more of an efficacy approach to the evaluation, rather than a pragmatic trial approach. Although the efficacy approach limits generalizability, it allows a thorough investigation of the intervention’s working mechanisms under more controlled, laboratory-like conditions. Such work may also reveal ideal levels of use of ZWIP, as it is possible that communication was already adequate in the case of some patients, making ZWIP usage superfluous. Using both quantitative and qualitative methods in this development phase may elicit remaining barriers and reveal more effective implementation strategies. Only after adapting to this group and proven efficacy is large-scale implementation warranted. Successful wide-scale implementation is a precondition for investigating the effectiveness of eHealth interventions. Otherwise finding no differences between treatment arms cannot be interpreted as a lack of effectiveness. These arguments show that, sufficient time and resources are required to develop, test, and retest new eHealth interventions before finally evaluating their effectiveness in pragmatic trials [29,30].

Overall, the study confirmed that introducing eHealth interventions in the elderly is a difficult task. Despite using a theory-driven intervention design and state of the art implementation techniques, usage remained low and effectiveness was not observed. Performing a thorough proof of principle study in early adopters may be crucial to improving the use of eHealth interventions in the elderly before evaluating effects on a larger scale.