This study included patients from the intervention groups of two separate randomized controlled trials: patients with breast cancer during neoadjuvant chemotherapy (ClinicalTrials.gov; NCT02479607) and patients with locally advanced prostate cancer during treatment with radiotherapy (ClinicalTrials.gov; NCT02477137). The study comprised logged data from patient reports and interviews with patients. Ethical approval was obtained from the Regional Ethical Review Board of Stockholm (registration no. 2013/1652-31/2 and 201712519-32).

The patients were consecutively recruited at two university hospitals in Stockholm, Sweden. Patients with breast cancer or prostate cancer undergoing neoadjuvant chemotherapy and radiotherapy, respectively, who were able to speak and understand Swedish, who were presumed cognitively able to use a mobile app for symptom reporting, who agreed to participate, and who signed a written informed consent form were eligible. Of the 75 patients in the breast cancer group, 1 patient had a change of treatment to surgery instead of neoadjuvant chemotherapy, and 1 patient withdrew their consent to participate in the interview. In the prostate cancer group, 58 patients of the 75 patients participated in the interviews, due to organizational circumstances and difficulties reaching the patients after their treatment had ended (Figure 1).

Patients downloaded Interaktor onto their smartphone or tablet. Patients who did not own one were lent a smartphone by the research group (2 in the breast cancer group, 2 in the prostate cancer group). The patients also received an individual log-in for access to the content of the app. The patients received verbal and written instructions on how to use the app and were asked to report their symptoms daily on weekdays during treatment. Patients with breast cancer were asked to start using the app on their first day of neoadjuvant chemotherapy and continue until 2 weeks after treatment had ended—a total of approximately 18 weeks. Patients with prostate cancer started using the app on their first day of radiotherapy and continued until 3 weeks after treatment had ended—a total of approximately 9 weeks. All patients were informed that nurses at the clinic would survey and respond to alerts triggered on weekdays (8 AM to 4 PM). If patients needed support at other times of the day, they were instructed to contact the clinic according to the standard procedure. If a report had not been submitted before 2 PM, a notification was sent out to remind the patient to report. The breast cancer version of the app included a notification that was sent to the patient, suggesting the patient read self-care advice related to alerted symptoms. This function was not included in the prostate cancer version of the app.

Data management was performed using Excel, and statistical analyses were carried out using SPSS statistical software (version 24.0; IBM Corp). Differences in demographic and clinical characteristics between the two groups were analyzed using two-tailed independent t test, Fischer exact test, and the chi-square test. Usage was analyzed with descriptive and inferential statistics on the variables reports sent, alerts triggered, self-care advice section views, and free text messages sent; group median values were calculated and are reported. The intended use was measured by daily symptom reporting during weekdays. Adherence to daily reporting was calculated as the number of weekday reports (excluding multiple daily reports) divided by the total number of reportable weekdays for each patient.

Multiple regression analysis was conducted (using the enter method) to see if the independent variables (predictors) age, comorbidity, marital status, and education level predicted usage of the app. The usage variables (dependent variables) were adherence to daily reporting as intended, total number of alerts triggered, total views on self-care advice, and total number of free text messages sent. Since all of the usage variables were positively skewed, these were normalized using a natural logarithm transformation [47]. Level of significance was determined as P<.005.

The interview notes were analyzed by conventional content analysis [48]. First, two authors read the complete interview notes from each group (breast cancer and prostate cancer) individually, to grasp the entire data. Both authors compiled a data sheet for each group with the patients' answers. The data sheets were reviewed repeatedly, and initial codes were derived. Thereafter, all authors reviewed both sets of data sheets and codes. Due to substantial similarities of codes in the two groups (breast cancer and prostate cancer), one data set containing all coded responses (with credentials) was assembled. Subsequently, all text material were analyzed. In an iterative process, the codes were sorted based on similarity into subcategories and combined based on content into overarching categories. Some exemplifying quotes from the patients’ statements are presented in the Results section. During the analytical process, all authors met continually to discuss and come to a consensus.

The patients’ sociodemographic (Table 2) and clinical data (Table 3) are presented below. Patients in the breast cancer group were significantly younger than the patients in the prostate cancer group (P<.001). Moreover, the patients in the breast cancer group had a statistically significant lower comorbidity score (P<.001) and a higher self-reported education level (P=.005).

All patients reported with the app at least once during the study period. Due to differences in treatment schedules, the reporting period in the breast cancer group ranged from 22 to 183 days (median 106, IQR 7). The reporting period in the prostate cancer group ranged from 54 to 89 days (median 63, IQR 11). Median adherence in the breast cancer group was 83% (IQR 36%). In the prostate cancer group, the median adherence percentage was also 83% (IQR 34%). Graphs of adherence patterns over time show that the level of adherence remained stable over time, although it dropped somewhat after day 49 in the breast cancer group and day 39 in the prostate cancer group (Figure 2).

In the breast cancer group, 96% (71/74) of patients triggered at least one alert during the study period; in the prostate cancer group, 72% (54/75). Patients in the breast cancer group triggered a median of 7 alerts (IQR 21, range 1-210). In the prostate cancer group, the median number of alerts triggered was 2 (IQR 9, range 1-60). The distribution of yellow and red alerts was 86% yellow to 14% red in the breast cancer group and 90% yellow to 10% red in the prostate cancer group.

Among patients in the breast cancer group, 100% (74/74) viewed self-care advice at least once. Patients in the breast cancer group viewed a median of 11 (IQR 15) self-care advice topics at least once during the study period, out of 17 self-care advice available. The total number of views was 1075, and 34% (362/1075) were views made after the patient was notified with the suggestion to read related self-care advice after an alert had been triggered. In the prostate cancer group, 87% (65/75) of patients viewed self-care advice at least once and the median number of various self-care advice topics viewed at least once was 5 (IQR 11) out of 16 self-care advice available. The total number of self-care advice views by patients in the prostate cancer group was 697.

Most of the patients used the free text function at least once: 93% (69/74) in the breast cancer group and 75% (56/75) in the prostate cancer group. The free text messages were mainly about symptoms, requesting or declining contact, care-related information, and issues linked to the app or to reporting. There was a variation in how the function was used, some patients wrote short, condensed messages while some wrote longer, richer descriptions.

In the breast cancer group none of the multiple regression models were statistically significant (Table 4). Higher age predicted a lower total number of free text messages sent (P=.04). In the prostate cancer group the multiple regression model showed that the total number of views on self-care advice was statistically significant (F_4,70=3.811, P=.007, adjusted R²=.132) (Table 5). Higher age (P=.01) and higher education level (P=.04) predicted a higher total number of views of self-care advice. Higher comorbidity score predicted fewer self-care advice views (P=.004). Furthermore, being married or cohabitating predicted a higher adherence to daily symptom reporting (P=.02) (Table 5).

The findings of this study show that the app was largely used as intended and appreciated by patients undergoing treatment for breast cancer and prostate cancer. The app gave the patients a feeling of assurance by offering a convenient method to contact their health care professionals and the security of being monitored via the symptom reports. Furthermore, the app promoted self-care by facilitating self-monitoring and learning about symptoms.

In this study, adherence to daily symptom reporting in the app was 83%. This is high in comparison to a review [49] in which it was concluded that around half of participants in web-based health interventions for chronic conditions and lifestyle and mental health management adhered to the interventions. Plausible explanations for the high adherence to the Interaktor app are that the app was interactive and easily available on a smartphone or tablet when compared to a computer-based system. Previous research [49,50] shows that intervention characteristics such as intended usage frequency, updates, and persuasive design increase patients’ adherence to web-based interventions. Moreover, the app may reflect patients’ need for frequent, continuous contact with health care professionals during treatment for cancer [51].

Most patients perceived the content of the Interaktor app as relevant, and the fact that Interaktor was developed in collaboration with patients and health care professionals is likely to have contributed to this result. Two previous studies [21,26] of the Interaktor app among patients treated for cancer yielded results in line with this study, both in terms of adherence levels above 80% and interviews revealing that patients appreciate and perceive the app as supportive in their symptom management.

There was a temporal aspect to how patients perceived using the app. At the beginning of treatment, when the situation was new or they experienced a new symptom, daily symptom reporting was considered especially meaningful. Later, as patients became more experienced and familiar with their symptoms, some noted a lower inclination to report each day. Also, patients commented that after some time, as they acquired general knowledge about the disease and treatment, they felt a desire for more individualized and in-depth information than that contained in the app.

The observation that memory impairment and feeling ill influenced engagement adversely contrasts with a study [52] that showed that increased use of a web-based symptom management system was predicted by higher levels of symptom distress among men with prostate cancer, and it may be of value to investigate the effect of symptom burden on app usage in future studies.

The interviews indicated that some patients with comorbidities felt the need to add clarifications to responses available in the self-assessment form by free text. This finding warrants further study, considering the expanding community with multiple conditions, and it contrasts somewhat from the findings of a study [52] that showed that increased use of the system was associated with the absence of comorbidities.

The findings described above are in line with a conceptual model where perceived credibility and personal relevance influence engagement [36]. This study accumulates existing evidence, which imply the need to develop cancer-supportive digital interventions that are interactive and tailored [53]. Tailoring can be performed by the individual patient before the intervention (pretailoring), by preference to promote autonomy (self-tailoring), and within-person as health status or needs evolve [54]. In future technological development of the app, it might be useful to integrate an option for the individual patient to add or exclude symptoms or concerns in the self-assessment component to make it more person-centered. Expanding the interactive features in Interaktor by adding or updating information as patients progress through treatment may also be a way to maintain and enhance patients' experience of personal relevance [36].

The adjusted R² values in the regression models in this study were low, indicating that none or only one predictor was correlated to each dependent variable. The only variable tested that predicted usage for both groups was age; higher age predicted a decrease of the total number of free text messages sent in the breast cancer group, while a higher age predicted more self-care advice views in the prostate cancer group. It has previously been suggested that demographic variables may be too broad to indicate usage motives and preferences for mHealth [55]. Higher education level predicted usage in the prostate cancer group, specifically, predicting a higher number of views of self-care advice. Furthermore, in the prostate cancer group, being married or cohabiting predicted higher adherence to daily reporting as intended than that of patients who were single. These findings are in line with theory as well as research suggesting that social support and education level influence the adoption and usage of web-based interventions [17,56]. Both social support and higher education level have been associated with a significant increase in engagement [15].

In the breast cancer group, it was suggested by patients that they should be responsible for contacting health care professionals. This finding is noteworthy, as it pertains to patients’ sense of control and self-reliance. From the perspective of a review on person-centered participation [57], this may signify a patient with confidence in their own experience-based knowledge and expertise, voicing a willingness for increased responsibility and management of their care. This is an essential precondition if we are to achieve equality and partnership among health care professionals and patients, but it is always preceded by phases of dialogue, knowledge building, and information sharing [57]. Previous studies have demonstrated that being female, having a higher education level, and being younger are predictive for preferring a more active role in care [58,59].

Analysis of usage metrics enables a systematic assessment of, and insight into, patients’ exposure to and behavior throughout an intervention relating to frequency, depth, and breadth of use [17] but cannot disclose how use was perceived [60]. On the other hand, interviews or self-reports are associated with a risk of social desirability and challenges in communication, such as discrepancies in researcher and respondent terminology [60].

To counteract the limitations of each method, a combination was used in this study. Below are two examples of findings that reinforce how relying on only one source of data to study engagement is not sufficient. In the interviews, patients described aspects that were perceived to influence their engagement, which were not visible in the logged data. Patients stated that remembering to report daily was sometimes difficult, and moreover, patients noted that daily symptom reporting was perceived as more meaningful in the beginning. But the logged data show that adherence to daily symptom reporting is high and stable throughout the time of the intervention as adherence patterns do not decrease markedly over time. Relying on either one of the methods would have resulted in a less comprehensive account of patient engagement.

It may be a limitation that all the interviews took place after the intervention was completed and collecting data during the intervention as well as afterward may have added additional insights relating to usability and learning [61]. Symptom graphs and self-care advice features went undiscovered by a small number of patients in the significantly older prostate cancer group, indicating that enhanced patient training and further reviews of usability may be profitable in future studies. Research contributing to an understanding of how adherence can be promoted is starting to emerge [50,61,62]. However, a comparison of adherence levels between studies is debatable due to limited consistency in how adherence is reported [37,38]. The results of this study provide valuable knowledge, as it relates both to the patients’ actual usage versus intended use of the intervention and to how the patients thought about using it. The strength of this study consolidating investigations of two distinct patient groups is somewhat limited as the two versions of Interaktor slightly differed. The prostate cancer version of the app was developed and clinically trialed before the planning of this study and did not have the function that automatically prompted reading of specific self-care advice based on symptom reports. It is also a limitation that the use of links and graphs were not logged. Both are, in a sense, reflective of a learning curve in innovation research and will be taken into consideration in forthcoming studies.

High patient engagement in the Interaktor app was achieved. Using the app promotes self-care by facilitating self-monitoring and timely advice. Furthermore, it provides assurance through continuous and convenient contact with health care professionals.

The study supports the notion that interactivity enhances patients’ feelings of personal relevance and thus increases engagement. The predictive ability of demographic variables differed between patient groups, but higher age and a higher educational level predicted the usage of specific app functions for both patient groups.

Taken together, the findings suggest a role for the use of an interactive app, such as Interaktor, to promote patients' participation in their care. The findings may have relevance for outpatients undergoing other cancer treatments associated with a risk of toxicities.