This was a single-arm pilot feasibility study of the MyPainPal mHealth app (NCT03717402, registration date: October 10, 2018). Our interdisciplinary research team developed MyPainPal using an iterative process combining the Agile [20] and mHealth Development and Evaluation Frameworks [21] that we previously described [22]. The intervention consists of a patient-facing smartphone app featuring a library of comprehensive cancer pain education and a “virtual medicine cabinet” that hosts patients’ specific analgesic and laxative medications. Daily surveys deliver tailored symptom support (Figures 1 and 2). MyPainPal also offers a HIPAA (Health Insurance Portability and Accountability Act)-compliant web portal tailored for providers.

In Figure 1, daily symptom and medication use surveys were pushed at 10 AM each calendar day and were available for completion until midnight. The Ecological Momentary Assessments (EMAs) drew from the Brief Pain Inventory [23] to assess pain severity and interference in the past 24 hours, with adaptive survey logic for “red flag” symptoms (eg, new or worsening chest pain, fever, severe vomiting, and symptoms of bowel obstruction) when patients reported severe pain (>7 on Numeric Rating Scale [NRS]). Including the name and dose of patients’ specific short-acting opioid, they indicated doses taken, degree of pain relief, and whether their pain control was acceptable in the past 24 hours. An additional 4 items from the Positive and Negative Affect Schedule (PANAS) scale [24] assessed mood in the past 24 hours. To enable self-management support for opioid-induced constipation, patients indicated the number of days since their last bowel movement, stool characteristics, and laxative doses used in the past 24 hours. Once per week, the survey included additional items assessing use of other pain medications, opioid side effects, and psychological factors, including pain catastrophizing [25] and sleep quality [26]. Just-in-Time Adaptive Intervention (JITAI) technology provided patient-facing feedback in response to symptoms on a summary screen presented immediately following survey completion.

In Figure 2, from left on the first row, there is a screenshot of the home screen, the Medication cabinet, and examples of medications. In the second row, there are screenshots of pain ratings, opioid medication reporting, and end-of-survey summary feedback that follow survey completion.

EMAs were pushed at 10 AM each day, assessing symptoms and medication use over the past 24 hours. The EMAs drew from the Brief Pain Inventory [23] to assess pain severity and interference (0-10, NRS); adaptive survey logic assessed for “red flag” symptoms (eg, new or worsening chest pain, fever) if participants rated their pain severity 7 or higher. Surveys included participants’ prescribed opioid medications, and they reported the number of doses taken in the past 24 hours, the degree of pain relief, and whether their pain control was acceptable. An additional 4 items from the PANAS scale [24] assessed mood. To enable self-management support for opioid-induced constipation, participants reported constipation severity, days since their last bowel movement, and doses of laxatives taken. Once per week, a slightly longer version of the survey also assessed use of other pain medications, side effects, pain catastrophizing [25], and sleep quality [26].

MyPainPal uses JITAI technology to tailor feedback in response to pain and constipation reporting (Figure 1). Algorithms based on pain severity and pain acceptability categorize patients’ responses into good, fair, and poor pain control [22]. Drawing from message banks tailored to each of these conditions, a summary screen (Figure 2) presents a brief, empathic statement, an educational “pearl,” and links to relevant content. For constipation, the summary screen includes laxative dosing suggestions based on days since last bowel movement and doses of senna and polyethylene glycol taken in the preceding 24 hours, adapted from clinical decision support algorithms developed by Yackel et al [27]. The summary screen additionally instructs participants to contact their care team for poorly controlled pain or constipation, or for any severe symptoms (eg, pain severity >9/10, any “red flag” symptoms). A HIPAA-compliant web portal displays participants’ symptom reports and medications, and includes a dashboard to facilitate triage and participant management.

The study was conducted in the outpatient palliative care clinic of the Dana-Farber Cancer Institute (DFCI), a comprehensive cancer center in Boston, MA. English-proficient adults were eligible if they had an advanced solid malignancy or multiple myeloma, were being treated with palliative intent, had chronic pain related to their cancer or cancer treatment, reported an average pain severity of >4/10 (NRS), took opioid pain medications, and owned a compatible smartphone (iOS or Android). Exclusion criteria included cognitive impairment, history of opioid use disorder, recent bowel obstruction, hospice enrollment, or use of buprenorphine (not app supported).

Participants were recruited between June and October 2021 and from February through April 2022, after pausing recruitment to correct an app bug. Potentially eligible patients were identified by screening clinic schedules or direct referrals. After receiving permission from the treating clinician, a research assistant approached patients for participation in person or by phone and obtained written informed consent [28]. After participants completed a baseline survey, study staff conducted an onboarding visit over video (Zoom; Zoom Video Communications) or telephone during which they helped participants download the app and conducted a brief tutorial. Patients were encouraged to complete the app’s daily symptom survey and use other app features. Our goal was to allow patients flexibility to use the program in the way that seemed most useful to them, rather than mandate a particular level of engagement. Participants were reminded that the dashboard was not monitored 24/7 and to contact their care team for any severe symptoms.

Participants were on-study for 28 days and received 2 protocol-directed outreach calls 1 week after onboarding. First, the research assistant assessed and provided support for any technical difficulties. Second, a palliative care clinic nurse reviewed participants’ pain self-management challenges and provided advice. Thereafter, nurses monitored the portal on weekdays. The research assistant also sent the care team a symptom “alert” summary via the electronic health record (EHR) for poorly controlled pain, constipation, or any “red flag” symptoms. Nurses were advised to contact patients using their clinical discretion and to document triage phone calls in the EHR using a study template.

Patients completed surveys at baseline and end-of-study using REDCap (Research Electronic Data Capture; Vanderbilt University) [28]. Clinical factors were abstracted from the EHR. At baseline, patients self-reported demographics and completed the 5-item Media and Technology Usage and Attitudes Scale [29]. In an optional semistructured debriefing interview conducted over video (Zoom) or phone conducted within 4 weeks of completing their MyPainPal testing, a trained interviewer not personally acquainted with the participant (DRA and AE) assessed patients’ experiences using the app, features they appreciated and disliked, perceived impacts on symptom self-management, and recommended changes or features they disliked (refer to Multimedia Appendix 1 for interview guide).

Study procedures were approved by the Dana-Farber and Harvard Cancer Center Institutional Review Board (IRB# 18-504) and conducted in accordance with the 1964 Declaration of Helsinki and later amendments. All participants provided written informed consent before trial participation. All data have been deidentified and remain anonymous. Upon consent, participants were provided a numerical identifier which was paired with their data thereafter. Participants received a US $50 gift card for participation.

Of 193 patients identified on initial screening, we received permission for and contacted 61, of whom 47 were eligible, 26 consented, and 20 enrolled in the study. Common reasons patients declined participation included technology challenges (eg, lack of comfort with smartphones), having stable pain, being overwhelmed or not being interested in participation. All 20 participants completed baseline and end-of-study surveys (Figure 3). Participants had a mean age of 57 (SD 12.3) years; 55% (11/20) were female; and most were non-Hispanic White (16/20, 80%; Table 1).

The most common cancers were breast and colorectal; 90% (18/20) of patients were on active cancer treatment; all patients were taking short-acting opioids and 75% (15/20) were taking long-acting opioids. Five (25%) patients were hospitalized during the study, ranging from 1 to 15 days.

Patients logged into the app a median of 14 (IQR 9-17) times and completed a median of 8 EMAs (IQR 5-14) over the 28-day study. The mean rate of symptom survey completion was 34% (SD 21%; mean 9.40, SD 5.5) as a function of days on study, and 36% (SD 20%) as a function of eligible days (ie, out-of-hospital). A total of 188 symptom surveys were completed.

Of the 19 participants who completed usability and acceptability measures (one did not use the app and omitted these items), 79% (15/19) rated their overall satisfaction as greater than or equal to 4/5. The mean Acceptability E-scale score was 24 (SD 4.4), the mean SUS score was 78.3 (SD 16.2), meeting our usability benchmark of 68/100, and 65% (13/19) were likely or very likely to recommend the app to a friend (Table 2).

Nineteen percent of symptom surveys (36/188 total) triggered an alert (Table 2) across 16 individual patients, with a median of 2 (IQR 1-2) alerts per patient. The most common symptoms triggering the alerts were pain (78%), red flag symptoms (25%), and constipation (14%). Nurses called patients all alerts, with some resulting in ≥1 call. Analysis of EHR documentation found that nearly all (97%) triage calls involved medication counseling, most frequently centered on optimizing the timing of medication use (eg, using analgesics before going to sleep or sequencing medications differently), and encouraging patients to take short-acting opioids more frequently or at the higher prescribed range. Triage calls directly resulted in prescription medication changes for 5 unique patients (25%), with 2 patients having >1 medication change. Additionally, the nurse calls identified and corrected errors in medication self-administration for 2 patients.

Although we did not test for pre-post differences in this exploratory, single-arm study, pain severity, pain interference, pain catastrophizing, physical well-being, social and familial well-being, functional well-being, and constipation ratings remained relatively stable from baseline to end-of-study. We observed slight decrements in emotional well-being and quality of life, which is unsurprising for this seriously ill population (Table S1 in Multimedia Appendix 2).

Fourteen of the 20 (70%) participants participated in optional debriefing interviews lasting between 15 and 47 minutes. Dominant themes were (1) aspects of the app that patients appreciated and used most, (2) dislikes and suggested improvements, (3) perceived effects of the app, and (4) perspectives on the optimal conditions for app use (Table 3).

This single-arm pilot study demonstrated that MyPainPal was usable, acceptable, and showed preliminary signals of potential clinical impact on the management of cancer pain. Participants reported symptoms on the app approximately one-third of study days, which resulted in timely nurse outreach for poorly controlled symptoms. It is uncertain whether patients might have experienced similar symptom care without MyPainPal; however, in debriefing interviews, many directly credited the app with reducing barriers to symptom communication and recalled specific instances of timely clinician outreach, and reported feeling supported and validated, while gaining insights into symptom self-management. The study also identified several areas for app refinements.

While remote, PRO-based symptom monitoring interventions have proliferated in oncology [13,14] and are beginning to be applied to cancer pain [34-37], their benefits for pain have been modest [15,38]. This may be because they offer patients limited direct symptom support, tailored medication management advice, and lack the nuanced content needed to educate users and sustain engagement [39,40]. Successful pain management requires that patients and providers engage together to evaluate, understand, and treat the dynamic symptom of pain. MyPainPal addressed these foundational aspects of pain management by including robust psychoeducation on cancer pain and opioid-induced constipation, evaluating pain-related symptoms and medications, delivering tailored feedback and algorithmic self-management support, and providing care teams with nuanced symptom and medication use data in a structured format that supported outreach and population management. We found this approach to be feasible, with patients reporting good intervention usability and acceptability ratings.

In qualitative findings, patients credited MyPainPal with allowing them to gain new insights into pain self-management and promoting timely interactions with care teams, as has been seen with other PRO-based remote symptom monitoring interventions [41,42]. Approximately 1 in 5 symptom surveys triggered a clinical alert, which aligns with rates observed in other PRO interventions [43,44]. While a meaningful proportion of these alerts resulted in adjustments to patients’ pain medications before routine clinic visits, this must be balanced with the added burden to nursing staff. Future studies could evaluate varying alert thresholds or could investigate the potential of artificial intelligence to provide clinicians with management recommendations. While the exact frequency of optimal app use and symptom reporting is unclear, our findings suggest that this may depend on the severity and lability of patients’ symptoms. Interestingly, several patients described that the app helped normalize their need for opioids, reduced stigma, and may have enhanced their willingness to take opioids when needed. Opioid stigma can be a major deterrent to advanced cancer patients’ pain self-management [11,45] and may reduce clinicians’ willingness to prescribe [46-48]. Our findings suggest that standardized assessment of opioid use and education may destigmatize opioid use and potentially improve pain for patients with advanced cancer.

The study also identified several opportunities to further optimize the intervention [49], which we are implementing. Specifically, some patient-facing education was underrecognized and underused. Thus, we are creating the capability to assign educational content to bolster patient engagement and deliver behavioral evidence-based pain interventions such as cognitive behavioral therapy [50] and mindfulness. Second, we are reducing symptom survey frequencies and adding free-text entries to allow patients to communicate more nuanced information. We are also creating a feature that enables patients to log symptoms and medication use in real time to further facilitate self-monitoring and engagement. Finally, to promote scalability, we are initiating the integration of MyPainPal into the electronic medical record.

This study has several limitations. First, this single-arm pilot was unable to evaluate efficacy and its small sample size lacked statistical power to detect meaningful pre-post differences in pain; however, our encouraging findings of feasibility, acceptability, and clinical impact justify a randomized clinical trial to evaluate its impact on pain severity and pain-related health care use. Second, we tested the app in a single academic center, among a predominantly non-Hispanic White and affluent palliative care population. The generalizability of our feasibility and acceptability findings to other settings and populations is uncertain. Moreover, our population had a high symptom burden and many were near the end-of-life. While several patients suggested that MyPainPal would have been more useful earlier in their pain trajectory, this requires confirmation in a more general oncology population. Additional pilot studies are needed to evaluate the feasibility of the intervention and of a clinical trial in different clinical settings and among patients at earlier points within their care. Interestingly, a few patients relied on their caregiver to help them use the app. Caregiver-focused symptom reporting interventions have shown promise in the hospice context [51] and could be explored in future work.

We found that this novel cancer pain mHealth intervention that combines symptom and medication reporting, adaptive feedback, symptom education, and care team monitoring was feasible and acceptable among patients with advanced cancer using opioids for chronic pain. Chart abstraction data and qualitative feedback suggested that patients may have benefited from the intervention through insight-building, improved care team communication, and timely medication changes. Following app refinements, randomized testing will determine its impact on relevant outcomes, including pain severity, pain interference, and pain-related acute care use.