We developed the CRCS-PHR as a stand-alone, web-based tool for patient convenience, portability, and dissemination potential. An iterative, user-centered design approach was followed during development, including the creation of clinical content, program design, and web design usability testing. The design process began with the creation of content and technical parameters, culminating in a web-based interactive prototype. Product development included the application of established usability methods [32]. Stakeholders (consisting of patients, caregivers, and health care providers) were asked to participate in scenario-based evaluations with direct observation and debriefing interviews to gather data on user performance and preferences as described elsewhere [33]. Changes were not made in the web-based design until data from at least 4-6 stakeholders had been collected; consistent with previous work by Nielsen et al [34], a total of 17 stakeholders participated in data collection.

The CRCS-PHR includes the following clinical information: CRC surveillance guidelines, treatment received (surgery; adjuvant therapy; and lab, radiology, and procedure results), and potential future toxicities of the treatment received. The CRCS-PHR also had the ability to collect personal observations from the CRC survivor in an electronic journal or blog, relationships with providers and family members or friends, and communities with other CRC survivors.

Surveillance guidelines included information about guideline-concordant surveillance care, including bowel surveillance (colonoscopy), CEA tests, and CT scans [2], with reminders for individual surveillance tests based on individuals’ needs. Recommendations for surveillance care were adapted from the guidelines of professional organizations [2]. Information about CRC surveillance guidelines were automatically tailored to the CRC survivor’s disease stage (eg, CEA testing was not to be recommended for patients with stage I CRC). In the CRC-PHR, 2 tables related to surveillance care were delivered to the patient (Figure 1). First, a table was generated, which indicated what surveillance tests were appropriate for the patient and by what dates the tests should be performed. This informational table was designed to increase patient knowledge about surveillance and its benefits, as well as to prompt CRC survivors to seek the receipt of surveillance tests [35]. Second, a table was created wherein patients could self-enter information about the surveillance test received (date completed, type of test, and a brief description of results). This interactive table was designed both to enable the tracking of completed tests and to promote patient self-efficacy; interactivity is widely believed to enhance user involvement, commitment, and learning [36].

A 30-minute training session was developed to introduce patients to the content and functionality of the CRCS-PHR and was conducted in person by the research assistant at the time of recruitment. In addition, virtual training tools were embedded in the CRCS-PHR, including a 5-minute narrated training video and a detailed help section describing the purpose of all links, data-entry forms, and features of the tool.

A pre- and posttest intervention trial was conducted to test the feasibility and to determine the ability of the targeted PHR intervention to increase patient knowledge, self-efficacy, beliefs, and receipt of surveillance tests among CRC survivors. Participants were recruited between March and October 2012, with the postintervention follow-up survey occurring 6 months after the baseline survey.

Patients with CRC were eligible to participate if they had received curative-intent therapy and had been diagnosed with American Joint Committee on Cancer Stage I-III adenocarcinoma at least 9 months (but no more than 24 months) earlier. Participants were excluded from the study if they had metastatic disease. We approached CRC survivors for recruitment at an academic medical center and the Veterans Affairs (VA) hospital in Indianapolis. At the academic medical center, patients were seen in surgery clinics led by CRC surgeons.

Data were collected via patient self-report. Self-reported data are a valid, widely accepted source about clinical service use due to cost and time efficiency, particularly for early-phase studies such as this feasibility study, as well as large-scale epidemiologic studies [37-40]. Presurvey measurements were collected by a research assistant at baseline immediately after the patient was provided with access to the CRCS-PHR. Postsurvey measurement was then collected 6 months after initial enrollment over the telephone, with a written survey being mailed beforehand to patients in order to facilitate answering the questions. In addition to the measures discussed below, participants were asked how they used the PHR and what features they found to be most and least useful after the intervention.

Approval for this study was obtained by the Indiana University Institutional Review Board (1201007805), as well as the Indianapolis VA Research & Development committee. The procedures used in this study adhere to the guidelines of the World Medical Association Declaration of Helsinki. Prior to enrollment in the study, the purpose of the study and each participant’s role were explained. Written consent was obtained from everyone who participated.

A total of 28 patients with CRC completed the baseline survey, with 22 patients completing the follow-up survey at 6 months after the intervention. The majority of the sample was recruited at Indiana University Health (25/28, 89%), with 3 patients being recruited through the Indianapolis VA Medical Center. Table 1 describes the patient sociodemographics. The average age of the sample was 58 (SD 9.9) years. Two-thirds (18/28, 64%) of the patients had rectal cancer vs colon cancer (9/28, 33%). The majority of the patients were male (16/28, 57%), married (20/28, 71%), and were employed full-time (15/28, 54%); they also had an annual household income of >US $60,000 (16/28, 57%). Slightly less than half of patients (13/28, 46%) had a college education or greater.

Descriptive statistics and paired t test results for the 4 categories of behavior outcomes, including self-efficacy, perceived benefit, perceived barriers, and patient knowledge, both before and after the intervention are reported in Table 2, while Tables 3-5 provide a more in-depth view of how patients answered the baseline questions for the 3 categories of knowledge, barriers, and benefits. For knowledge, patients were asked about the correct intervals of recommended follow-up times for various surveillance tests (physical exam, CEA test, colonoscopy, and CT scan). Out of the 4 knowledge questions, patients answered on average just under 2 of the 4 correctly, with no change in knowledge between the two surveys (P=.69). Self-efficacy (range 10-40) saw little change between pre- (32.2) and postintervention surveys (31.8; P=.66). Patients rated barriers (range 3-15) at both intervals fairly low, with 4.7 before the intervention and 4.9 at after the intervention P=.81). Benefits (range 5-25) is the only beliefs category in which we saw a significant change, and this was only benefits for CEA test, as benefits for all tests were rated fairly high. For colonoscopy, patients rated benefits before the intervention at 22.6, with no change after the intervention at 22.8 (P=.75). CEA test was rated as 20.9 before the intervention, with 22.0 after the intervention (P=.04). CT scan saw little change with preintervention rating at 21.6 and postintervention rating at 22.1 (P=.45).

Table 6 reports the prevalence and comparison (paired t test) for each of the 3 primary surveillance tests at each time point. Since having surgery, only 52% had a colonoscopy since their CRC surgery, while that number increased to 86% (18/21) after the intervention (P=.005). Similarly, 67% (14/21) had a CEA test from the time of their surgery prior to the preintervention survey, while the proportion increased to 95% (20/21) after the intervention (P=.01). CT scan was the only surveillance test in which we did not see a significant uptick, with 67% (14/21) reporting having had a CT scan at the beginning of the study and 86% (18/21) having had one after the intervention (P=.10).

The purpose of our study was to test the feasibility of a stand-alone PHR for CRC survivors’ post resection and examine its impact upon receipt of recommended surveillance testing and behavior outcomes. CRC survivors are at an increased risk of recurrence, especially within the first 2-3 years after treatment. Interventions targeted toward increasing surveillance rates for CRC survivors would help to detect signs of recurrence early in its progression, and thus potentially decrease morbidity and mortality.

For our primary outcome, we found an overall significant impact on receipt of CRC surveillance tests. From baseline to the 6-month postintervention follow-up, we saw a significant impact for both colonoscopy (P=.005) and CEA testing (P=.01). There was no significant increase in CT scans between before and after the intervention, although we did observe an increase from 67% (14/21) at baseline to 86% (18/21) of the sample receiving a CT scan at 6 months. The effect of CRC surveillance tests is commonly clinically approached as a bundle of care, that is, recommending the combination of colonoscopy, CEA testing, and imaging [44]. At baseline, only 42% (12/28) of participants reported having received all 3 tests, with 29% (8/28) not having received any. Whereas, at the 6-month follow-up, 77% (16/21) had received all 3 surveillance tests, and all patients had received at least one. Both CEA testing and CT scans have been associated with increased rates of surgical treatment of recurrence, suggesting that increases in either type of surveillance testing may be associated with more salvage surgery with curative intent [45].

Existing frameworks provide some guidance about what behavioral mechanisms may explain these main effects. The Health Belief Model [46] posits that self-efficacy, perceived barriers, and perceived benefits (ie, belief about the effectiveness of surveillance in reducing risk) mediate changes in the health behavior of individuals. Further, other investigators have postulated that patient-centered portals, with many features in common with personal health records, will have a positive effect upon patient self-efficacy [47-49]. In addition, Lo et al [50] found screening knowledge, perceived barriers to care, and social norms to be significant mediators of sociodemographic differences in the uptake of CRC screening. The CRCS-PHR may act through similar mechanisms. Our patient-centered technology had design features intended to increase screening knowledge and the perceived benefit of surveillance, including both clinical reminders to the patients about the next surveillance test due and web-based educational materials to explain the nature and purpose of each test. Participants found the summary and schedules of their cancer treatment and follow-up appointments to be the most useful in the PHR, along with side effects of treatment and community resources [31,47].

Due to the conceptual and empiric importance of patient knowledge, perceived benefits and barriers, as well as self-efficacy to the uptake of CRC screening [51], we explored the effect of the CRCS-PHR upon these secondary outcomes. Many patients did not know the answer to individual knowledge questions (Table 3); the proportion varied by test, from 25% (7/28) to 75% (21/28) for physical exam to colonoscopy, respectively. These findings suggest that patient-centered technologies have the potential to increase patient knowledge but can be further tailored to tests about which patient have the least awareness (eg, colonoscopy or CT scans). Nonetheless, patient knowledge is commonly not associated with changes in patient screening behavior [52]; our observations that surveillance test use increased, whereas knowledge about the tests often did not, reinforced this weak association.

Overall, patients largely agreed about the benefits of CRC surveillance, with the proportion who reported individual tests (colonoscopy, CEA test, or CT scan) as beneficial ranging from 75% to 100% (find recurrence early); from 68% to 88% (decrease chance of dying from recurrence); and from 71% to 89% (help you not to worry). With patient beliefs, the only domain that significantly increased was the perceived benefits of CEA testing, as most were quite high at baseline. These relatively high proportions are similar to the perceived benefit of CRC screening tests among a general population not already diagnosed with cancer [43]. High perceived benefit limited the potential for improvement in these perceptions among the CRC survivors enrolled in our trial. Conversely, low perceived barriers to care at baseline likely limited the potential for improvement in these domains. Moreover, the potential barriers of cost and transportation are challenging to address [53-58], and our CRCS-PHR implementation needs to be accompanied by changes in the health care systems, and policy needs to be adequately addressed.

We found no significant differences in patient self-efficacy. Systematic reviews conducted by Han et al [47] and Lancaster et al [48] found that eHealth tools such as provider-patient communication functionalities, case management, and other forms of clinical support may increase self-efficacy and self-management. Empiric findings from previous studies have been mixed. Secure messaging had a positive impact upon medication self-efficacy among patients with diabetes [59,60], but other studies have shown no association [60]. These mixed findings suggest that the influence of patient portals upon self-efficacy may vary depending upon both the functions used and the populations targeted.

Future studies in the field of cancer control should assess new populations of patients with cancer, including prostate, ovarian, and skin cancers, as they are underrepresented in this area of the literature and patient-centered technology [31]. However, with differences across sites of cancer, and more specifically, across the cancer continuum, patient-centered technology interventions will need to be targeted toward specific sites and continuum levels as efficacy and effectiveness may vary [31]. This may be due in part to the complexity of cancer care across the care trajectory versus other chronic disease management, as well as the need for tailored functionalities by cancer type. Treatment plans and surveillance testing will likely differ for each cancer type in terms of the tests and intervals recommended. Future studies should also consider what types of recruitment strategies may be optimal in this type of research, such as recruitment through clinicians versus registry-based outreach. We employed a recruitment strategy involving clinician engagement to recruit individuals, as we felt this approach would better identify eligible patients and decrease attrition over time due to ongoing engagement with their clinicians. Testing this type of engagement was important to assess initial intervention feasibility. However, there are advantages to registry-based recruitment, which is more likely to lead to increased access to a larger number of potential participants. Researchers should weigh the advantages and disadvantages of different recruitment strategies in relation to the specific needs of their studies.

While this work provides strong insights and evidence to inform the development of the CRCS-PHR, our work is not without its limitations. First, without a control group not exposed to the CRCS-PHR, we are limited in our ability to make inferences of the intervention effect. Additionally, the pre- and posttest design has the potential for a temporal effect on the patient beliefs and surveillance receipt over time, which is important to note. With feasibility established, future studies should use a randomized controlled trial design, which will account for the potential of a temporal effect and increase the strength of our causal inferences with the introduction of a control group. The relatively small sample size of our feasibility study also limits our ability to test mediation pathways. Our sample was primarily White, younger on average compared with the national CRC population (58 years versus 66 years old) [1], and more highly educated than the population of Indiana and that of the United States (13/28, 46% of the sample having college degree). The young age and higher education of our sample is not unusual among early adopters of new patient technologies and cancer survivors recruited at academic medical centers [60-62]; however, the results are not directly generalizable to other populations, including patients treated in community-based oncology clinics. Future work needs to continue to focus upon how best to engage CRC survivors who tend to be older adults in use of these new technologies. While our sample had a majority of rectal patients, nearly an inverse of the US CRC survivor population (70% of national CRC survivors being colon), we believe it is important to understand the use of this technology among both rectal and colon cancer survivors, and do not have a reason to believe its use would differ between these closely related cancer types. The higher proportion of rectal to colon patients in the sample was due in part to the local expertise in rectal cancer surgery at one of the academic sites.

Patient-centered technologies such as the CRCS-PHR represent an important potential approach to improving the receipt of guideline-concordant care such as surveillance tests among cancer survivors. In assessing these rapidly emerging technologies, we encourage investigators and evaluators to continue measuring behavioral constructs that might serve as plausible mechanisms to explain observed effects. With this approach, we can grow to understand not only if new technologies improve the quality of care but how this improvement takes place. Future research in this area should also assess the effect of personal health records with quasi-experimental and randomized controlled study designs when possible. Finally, survivors of different types of cancers should be enrolled in future research, given that the clinical and supportive care needs of patients may vary widely among different populations.