A one-group pre/posttest online prospective study was conducted over a university academic semester (July to November 2011). Inclusion criteria were (1) aged 18 or above, and (2) with access to the Internet and email at least on a monthly basis.

Analysis was conducted on an intention-to-treat basis. Sequential logistic regression analyses were undertaken to prospectively examine the crude and adjusted odds ratios (ORs) for participants’ health service utilization and help-seeking behaviors for physical and emotional well-being matters [44]. Independent variables assessed included whether participants accessed (or did not access) each specific PCHMS feature (journey, personal health record, forum, poll, diary, and online appointment service), controlling for participant’s gender, age, and potential confounders (eg, whether the participant was a university service patient/client prior to the study) to provide a stratified estimate of intervention effect. The Pearson product-moment correlation coefficient was used to examine correlations among usage of features that were associated with consumers’ behaviors.

Participants’ health service utilization rates (ie, visits to a health professional, University Health Service, or the University Counselling and Psychological Services), and their help-seeking behaviors for physical or emotional well-being matters were compared at different PCHMS login frequency thresholds (zero logins, once only, two to five times, six to 10 times, more than 10 times). The rationale for selecting these login frequency cutoffs is based on using heuristics to ensure important login frequency thresholds are covered (ie, zero, once only, and ≥ a high login frequency threshold) and that there are sufficient data points in each frequency threshold to conduct analyses.

Between group analyses were conducted using chi-square analysis. Participants’ pre-study characteristics (namely use of the Internet to find health information, use of social networking websites, visits to a health care professional in the past 6 months, and their self-rated well-being ratings classified as over or below 50 at pre-study) were compared between different PCHMS login frequencies using chi-square to assess whether these characteristics were associated with PCHMS usage levels. Descriptive analyses were conducted on participants’ reasons for not seeking help during study.

Data analysis was performed using IBM SPSS Statistics 20 [45]. Tests performed were two-tailed and assumed a cutoff of P<.05 for statistical significance.

A total of 1985 participants met inclusion criteria and were recruited into the study. All completed the pre-study questionnaire. Of those, 709 completed the post-study questionnaire (Figure 3). Analyses were conducted on 709 eligible participants who completed both the pre-study and post-study questionnaires. Of these, 81% (572/709) participants logged into the PCHMS at least once. No significant differences were found between questionnaires completers and noncompleters in their pre-study characteristics or the number of PCHMS login sessions (P>.05). Among questionnaire completers, no significant differences were found between different PCHMS login frequencies (P>.05).

Baseline characteristics of eligible participants are presented in Table 1. Participants’ well-being concerns during the study, their help-seeking behaviors, and their purpose for visiting a health care professional are outlined in Table 2.

Overall, 50% (358/709) of participants visited a health care professional (for themselves or others) for a physical well-being concern and 13% (95/709) for emotional well-being during the study (Table 2). Health service utilization rates (ie, visits to a health professional, University Health Service, and University Counselling and Psychological Services) during the study are outlined in Table 3 and Figure 4 according to PCHMS login frequency.

In absolute terms, participants who logged into the PCHMS more than 10 times were 22 percentage points more than those who never logged in to visit a health care professional during the study: χ² ₄=11.80, P=.01, n=708; zero logins: 44% (60/136, 95% CI 36-53) vs ≥10 logins: 66% (39/59, 95% CI 53-77). Relative to participants who never logged in to the PCHMS, the proportion of participants visiting a health care professional during study was 50% higher for those who logged into the PCHMS more than 10 times: RR is 1.50 (95% CI 1.15-1.95).

Visits to the University Health Service did not differ significantly between different PCHMS login frequency thresholds: χ² ₄=2.79, P=.59, n=708. However, participants who logged into the PCHMS more than 10 times were 10 percentage points more than less-engaged PCHMS users to visit the University Counselling and Psychological Services during the study: χ² ₄=10.26, P=.03, n=708; once only was 4% (10/287, 95% CI 2-6) vs ≥ 10 logins at 14% (8/59, 95% CI 7-25). Relative to participants who logged into the PCHMS only once, the proportion of participants visiting the University Counselling and Psychological Services during the study was 289% higher for those who logged into the PCHMS more than 10 times: RR is 3.85 (95% CI 1.60-9.44).

Overall, 54% (386/709) of participants sought formal or informal help (for themselves or others) for physical well-being and 32% (225/709) for emotional well-being during the study (Table 2). Participants’ help-seeking behaviors during the study between different PCHMS login frequency thresholds are described in Table 3 and Figure 4.

Help-seeking behaviors for physical well-being matters did not differ significantly between different PCHMS login frequencies: χ² ₄=8.94, P=.06, n=708. However, participants who logged into the PCHMS 10 times or more were 20 percentage points more likely to seek help for an emotional well-being matter than less-engaged users, ie, χ² ₄=10.70, P=.03, n=708; once only: 29% (82/287, 95% CI 24-34) vs > 10 logins: 49% (29/59, 95% CI 37-62). Relative to participants who logged into the PCHMS only once, the proportion of participants seeking assistance for an emotional well-being matter during the study was 72% higher for those who logged into the PCHMS more than 10 times, ie, RR is 1.72 (95% CI 1.25-2.36).

This is the first study that shows a dose-response effect of using a PCHMS on consumers’ health service utilization (primary care and counselling) and help-seeking behaviors for emotional well-being. To our knowledge, this is also the first study that contributes an understanding of which bundles of PCHMS features are associated with consumer help-seeking and health service utilization behaviors.

The nature of this study only allows associational inferences to be drawn and specifically, we cannot say that it was the usage of these features that drove user behaviors. An alternate reading of our results is that those individuals who are most likely to use health services are also the population most likely to be drawn to use a PCHMS. Both causal readings are of interest, and it is also likely that both probably are to some extent in operation in the results reported here. As the pre-study characteristics and well-being ratings of users were uniformly distributed across different PCHMS login frequency thresholds, we could not in this study detect differences in users to explain differences in their behaviors, suggesting the dose-response reading of our results is more likely the explanation. Untangling these two alternate readings will undoubtedly be resolved with further research.

The dose-response effect observed in this study may be explained by the availability heuristic, which describes the “situation in which people assess the frequency of a class or the probability of an event by the ease with which instances or occurrences can be brought to mind.” [46]. At times when participants experienced emotional and/or physical well-being matters, they may have utilized the availability heuristic to recall mechanisms they have been exposed to (in this case possibly the PCHMS), and accessed it to seek help and/or engage with health services. In addition, by providing informational cues in the PCHMS that are directly linked to an action (such as a “Book now” button embedded within consumer specific content) [15,43], we may have helped participants overcome their perceived barriers to visiting a health service (such as no time, inconvenience, or having to explain one’s emotional concern over the phone), which may have hindered the step in making an appointment. Moreover, the PCHMS may have provided a supplementary tool for participants who are already patients at participating health services to manage their health care concerns in conjunction with the services.

A PCHMS in practice offers a “bundle” of eHealth services and features including but not limited to a PHR. However, no studies to date have examined which bundles of features might motivate consumers’ health behaviors. Past studies that examined user engagement investigated the whole website [47-49], but not individual or bundles of features and not on associating PCHMS engagement with health behaviors.

While help-seeking for physical well-being was only correlated with use of the personal health record, a bundle of PCHMS features were correlated with emotional well-being help-seeking. Providing an environment that allows self-reflection (diary), social feedback (poll), and reducing the barriers to engage with health services (online appointment booking) appeared to work in combination for emotional well-being help-seeking. As suggested by Coiera, one reason such bundles might work is that they are programmatic, bringing together components that reinforce each other’s value and use [19]. Unpacking the use and impact of complex eHealth interventions from a “bundle” perspective should help us understand the right type, number, and complexity of features needed for consumers’ health behaviors. Identifying effective feature bundles thus appears to be a logical principle to follow in designing the next generation of PCHMS and PHR-related systems.

There are strong theoretical reasons why the features tested in this study could drive behavioral change:

Our findings are in line with the emerging body of literature that associates eHealth interventions with consumers’ health behaviors (such as personalization, tailoring, and behavioral feedback [13,14,51,52]). Previously, we have demonstrated that these PCHMS features have broad utility in a number of health areas including in vitro fertilization and influenza vaccination [42,43]. The current study extends those applications to help-seeking behaviors for physical and emotional well-being.

Our findings also extend previous studies that describe models, guidelines, and definitions on Internet interventions for consumers’ health behaviors [9-12], especially PHR-related systems. For PCHMS to be valued by consumers, clinicians, and health service providers, they need to (1) address pressing needs faced by patients and consumers, (2) enable patients and consumers to accomplish tasks without further complicating their lives, and (3) avoid unnecessarily disrupting clinicians’ workload or increasing pressure on health service providers [13].

Attrition is a significant concern in consumer eHealth research [5,53]. The Internet has the capacity to reach many individuals who may never seek formal treatment at physical or mental health services. Stumbling across an eHealth application, even for a short period of time, may trigger the much-needed opportunity for those in need to reflect and seek appropriate help. As suggested by Christensen and Mackinnon, the primary role of the Internet in disease prevention and early intervention could possibly be in the delivery of short positive health messages [53]. This study has identified bundles of features in a PCHMS that are associated with consumers’ help-seeking and health service utilization rates. Encouraging participants to engage in “relevant” features, rather than the whole intervention, at times of need may lead to the desired health behaviors and outcomes. In fact, website adherence or “stickiness” may cease to be an issue when “sufficient” engagement with the “right” bundle of features are demonstrated to lead to similar health outcomes and help-seeking behaviors [53].

Key strengths include the large number of participants, a multifaceted PCHMS with connectivity to health service providers that model many of the generic PHR systems and the use of PCHMS usage metrics to associate with consumers’ health behaviors. Some limitations include:

Our online prospective study provides evidence that PCHMS usage is associated with consumers’ utilization of health services and help-seeking behaviors for emotional well-being concerns. The features in this PCHMS are sufficiently general to be applicable to a variety of help-seeking and preventative health tasks.

While there is evidence that Web interventions can trigger significant consumer health behaviors, the empirical and theoretical basis for developing PCHMS features in general is still weak. Abandoning an eHealth application is a common and significant phenomenon. Asking participants to engage in all features of an eHealth application may not be an effective strategy to reduce attrition or influence health behaviors. Strategies that alert participants to bundles of features that are of immediate relevance or benefit to them could possibly result in higher engagement with the intervention and achieve the desired health actions. Future studies should investigate whether participants’ engagement with “tailored” bundles of PCHMS features are effective in achieving the desired health behaviors and outcomes, and that more correlation studies are needed to show which bundles of features are best for which kinds of consumers’ health tasks, conditions, and help-seeking stages.