Introduction
Breast cancer is the second most common cancer and the second leading cause of cancer death in women (Centers for Disease Control and Prevention 2025). Pain in women with breast cancer is prevalent with pain estimates varying by surgeries and treatments received, time since treatment, and pain severity (Poleshuck et al. Reference Poleshuck, Katz and Andrus2006; Gartner et al. Reference Gartner, Jensen and Nielsen2009; Glare et al. Reference Glare, Davies and Finlay2014). Sociodemographic characteristics and psychosocial stressors among women with breast cancer living in medically underserved areas in the southern United States contribute to higher pain burden (Warner et al. Reference Warner, Tamimi and Hughes2015; Janevic et al. Reference Janevic, McLaughlin and Heapy2017).
Behavioral pain interventions can improve pain outcomes in women with cancer and pain (Syrjala et al. Reference Syrjala, Mendoza and Yi2014; Ruano et al. Reference Ruano, García-torres and Gálvez-lara2022). Pain Coping Skills Training (PCST) is one type of behavioral intervention designed to provide patients with pain education and coping skills to improve their confidence in their ability to manage their pain. PCST has been shown in randomized trials to improve pain and other outcomes associated with pain (Somers et al. Reference Somers, Winger and Fisher2023). However, provision of PCST and other behavioral interventions is largely restricted to major medical centers with limited reach to patients living in medically underserved areas. Adapting these interventions to serve women with cancer pain living in medically underserved areas could help to fill this gap in care and improve their health-related quality of life (HRQOL).
An existing PCST protocol was adapted to be delivered via videoconferencing and to address limited health literacy and other concerns identified by women with breast cancer living in medically underserved areas (Dorfman et al. Reference Dorfman, Kelleher and Winger2019). The ICAN NC trial was designed to test whether the adapted mobile health Pain Coping Skills Training protocol (mPCST) would significantly improve pain and other pain-related outcomes compared to an education attention-control intervention among women with breast cancer residing in medically underserved areas in North Carolina (Kelleher et al. Reference Kelleher, Winger and Fisher2021; Somers et al. Reference Somers, Winger and Kelleher2026). In this paper, we report on a prospectively planned cost-effectiveness analysis of the behavioral pain intervention evaluated in ICAN NC using 2 perspectives: (1) the U.S. health care system perspective, which considers all direct medical costs, inclusive of intervention costs, and (2) the societal perspective, which expands the analysis to include the value of patients’ time spent receiving and engaging in intervention-related activities.
Methods
Trial design
This trial was preregistered on ClinicalTrials.gov (NCT04175639). The design and primary results of the ICAN NC trial have been described in previous publications (Kelleher et al. Reference Kelleher, Winger and Fisher2021; Somers et al. Reference Somers, Winger and Kelleher2026). Briefly, the study was designed to randomize women with breast cancer who were receiving cancer care in medically underserved areas to mPCST or an education attention control (mHealth-Ed). Participants in both study groups received four 50-minute one-on-one sessions with a trained psychologist (mPCST) or nurse (mHealth-Ed) via videoconferencing over about 8 weeks after randomization. To ensure that all patients had access to and support with videoconferencing, remote sessions were conducted at local community clinics. The trial included assessments at baseline, post-intervention (up to 10 weeks), and at 3 and 6 months after completing the intervention. Participants in both arms also used a mobile application.
The trial revealed significant improvements in pain severity and pain interference along with lower levels of pain catastrophizing in the mPCST group versus mHealth-Ed group following the intervention, and at 3 and 6 months later, among 149 patients who completed all 4 pain severity assessments through the follow-up period (Somers et al. Reference Somers, Winger and Kelleher2026).
Data collection to support the cost-effectiveness analysis
To estimate costs associated with delivery of the mPCST intervention, therapists’ time spent delivering the video-based sessions, as well as the time devoted to pre- and post-intervention activities related to intervention delivery, such as preparation and documentation, was recorded for all the intervention sessions by the study therapists and nurses. To estimate patient-time cost, time spent receiving the intervention was automatically captured using the videoconferencing platform. In addition, patients’ estimates of their travel time to and from sessions were collected via a one-time survey question. The amount of time patients spent interacting with the mobile app was automatically tracked.
Participant-level all-cause and pain-related medical resource use data were collected at the post-intervention assessment and at 3- and 6-month post-intervention follow-up assessment using patient report and available medical records. Medical resource data included hospitalizations, outpatient clinic visits, emergency department visits, and urgent care visits. The trial case report form captured the counts of each type of medical resource utilized along with the whether use was attributable to cancer pain-related, non-cancer pain-related, and other non-pain causes. In addition, the 5-level EuroQol 5-Dimension (EQ-5D-5L) instrument, a standardized, preference-based instrument of HRQOL that measures mobility, self-care, usual activities, pain/discomfort, and anxiety/depression, was administered to participants at baseline and at all follow-up visits.
Cost assignment
All costs are reported in 2025 US dollars. Unit costs sourced from other years were adjusted to 2025 costs for inflation using the U.S. Bureau of Labor Statistics Consumer Price Index (U.S. Bureau of Labor Statistics 2025).
Intervention (therapist) costs
Intervention costs were computed using the time spent by study therapists (i.e., PhD-level clinical psychologists) responsible for the delivery of the PCST interventions. Personnel time was valued using the 2025 national non-facility payment amount from the Centers for Medicare & Medicaid Services (CMS) for HCPCS code 90836, representing evaluation and management services of psychotherapy and behavior modification (\$89.39 per 45 minutes) (U.S. Centers for Medicare & Medicaid Services 2025). In a sensitivity analysis, therapists’ time was valued by converting the 2024 national average annual pay of \$95,830 for clinical and counseling psychologists to an hourly wage plus an additional 45.6% for fringe benefits (U.S. Bureau of Labor Statistics 2024a).
Participant time cost
Participant time was valued using the 2024 average US hourly wage of \$32.66, which was inflation-adjusted to \$33.54 for 2025 (U.S. Bureau of Labor Statistics 2024b).
Medical resources
Costs assigned to value inpatient care and emergency department visits were based on estimates generated from the Healthcare Cost and Utilization Project (HCUP) for 2021 (most recent data available at the time of analysis), updated to 2025 values (Agency for Healthcare Research and Quality 2021a). To assign inpatient costs, each hospitalization in the trial was assigned a Major Diagnostic Category code based on the documented reason for admission, and the corresponding mean hospitalization cost from HCUP was assigned. The cost assigned to each emergency department visit reported in the trial was based on the average cost of \$1,110 reported from 2021 HCUP Nationwide Emergency Department Sample for the people 65 years and older (Agency for Healthcare Research and Quality 2021b).
Costs for outpatient visits were based on 2025 national average payment rates for Medicare available from the Centers for Medicare and Medicaid Services (CMS 2025 Medicare Fee Schedule). The Current Procedural Terminology (CPT) code 99214, representing evaluation and management of an established patient in an office or outpatient location for 30-39 minutes, was assigned to reported counts of outpatient visits. The national average payment for non-facility fees from the CMS for this CPT code was \$125.18. Unit cost assigned to urgent care visits was based on a published estimate of \$168 (Ho et al. Reference Ho, Metcalfe and Dark2017) and updated to 2025 dollars (Federal Reserve Bank of St Louis 2024).
Health utilities and quality-adjusted life years
Responses on the EQ-5D-5L were converted to US preference weights using a scoring system from Pickard et al. (Reference Pickard, Law and Jiang2019), where the value of 1 represents perfect health, 0 represents death, and negative values correspond to health states considered worse than death. Within-trial quality-adjusted life years (QALYs) represent the duration of time between EQ-5D-5L measures, from baseline to the final study visit, multiplied by corresponding preference weights. The trapezoidal method was used in which a linear rate of change in preference weights is applied between study visits.
Within-trial incremental cost-effectiveness ratios
To quantify the additional cost required to achieve 1 additional QALY gained, incremental cost-effectiveness ratios (ICERs) were calculated by dividing the differences in mean costs between mPCST and mHealth-Ed groups by the difference in mean QALYs accrued during the trial period. Although the mHealth-Ed group served as an attention-control group in the trial, videoconference-based patient education (e.g., overview of medical treatments for cancer, healthy eating and physical activity tips)is not standard care. For the cost-effectiveness analysis, we are counting the full cost of the mPCST intervention as an incremental cost relative to standard care (i.e., \$0 cost was assigned to the mHealth-Ed group). We also conservatively assume that incremental benefits, in terms of additional QALYs with mPCST vs. mHealth-Ed, do not extend beyond the trial’s observation period of about 9 months. The resulting ICERs are compared to a willingness-to-pay (WTP) per QALY thresholds of \$50,000 per QALY, a historical standard, and \$120,000 per QALY, a newly endorsed threshold (Neumann et al. Reference Neumann, Cohen and Weinstein2014; Kazi et al. Reference Kazi, Abdullah and Arnold2025).
Statistical analysis
Continuous variables are described with means, SDs, medians, and 25th and 75th percentiles. Categorical variables are presented as frequencies with percentages. To transparently report on statistical uncertainty associated with between-group comparisons, mean differences and corresponding 95% confidence intervals (CIs) in medical resource use and costs between study groups were generated using bias-adjusted nonparametric bootstrapping. Generalized linear models were applied to compare medical resource use and costs incurred during the follow-up period between study groups. The models were specified with negative binomial error distributions and log links for comparisons of medical resource use. Comparisons of health utilities by the intervention group were based on a multilevel mixed-effect linear regression model. In the model, individual patients were modeled as random effects. Time, intervention group assignment, and an interaction between time and intervention group were modeled as fixed effects. The economic evaluation included all participants who were not withdrawn from the study (n = 162) (Somers et al. Reference Somers, Winger and Kelleher2026).
The study protocol was approved by the Duke University Health System Institutional Review Board (protocol Pro00103527), and all patients provided written informed consent.
Results
Demographic characteristics for the 162 patients included in the economic evaluation are available in Supplemental Table S1. Across both study arms, the mean age is 60 (SD = 12) years with 42% identified as a racial minority. Most did not have a college degree (67%) and reported annual household incomes below \$60,000 (68%).
Intervention time and costs
In the mPCST group, on average, the therapist delivered 3.7 intervention sessions, a mean total of 3.8 hours per participant (Table 1). After accounting for the therapist’s time before and after each mPCST session (i.e., preparing and documenting session activities), a mean of 0.4 hours, therapists spent a mean of 4.2 total hours per participant delivering the intervention and related activities. The average therapist cost for delivering the mPCST intervention is estimated at \$500 per participant in our base-case analysis using Medicare payment rates. With the hourly wage method, the therapist cost is estimated to be \$368 per participant.
Interventionist and participant time
Table 1 Long description
MPCST time metrics are summarized for 79 participants, including number of sessions, intervention hours, therapist time, and participant time with and without travel. Participants completed a mean of 3.7 sessions (SD 1.0) and a median of 4 sessions (interquartile range 4 to 4). Total intervention hours averaged 3.8 (SD 1.1) with a median of 4 hours (interquartile range 3.7 to 4.3). Therapist pre and post session time averaged 0.4 hours (SD 0.3) with a median of 0.3 hours (interquartile range 0.1 to 0.7), bringing total therapist time to a mean of 4.2 hours (SD 1.2) and a median of 4.4 hours (interquartile range 4.0 to 4.8). Participant time on the intervention and app, excluding travel, averaged 4.8 hours (SD 2.1) with a median of 4.7 hours (interquartile range 3.7 to 6.0). Including travel, participant time increased to a mean of 7.4 hours (SD 3.2) and a median of 7.3 hours (interquartile range 5.7 to 9.1). Travel time appears to add roughly two and a half hours to the median participant time, and variability is larger for participant time than for session counts.
Participant time and costs
Participants in the mPCST group spent an average of 4.8 hours receiving the intervention and on the mobile app, which corresponds to a mean patient-time cost of \$291 over the entire follow-up period. With the travel time included, participants’ total time increases to 7.4 hours, raising the estimated value for patient-time costs to \$441.
Medical resource use and costs
Over the intervention period and 6-month follow-up period, the proportions of participants incurring at least 1 outpatient visit for pain related to cancer, for pain unrelated to cancer, for any type of pain are lower in the mPCST group than in the mHealth-Ed group, but differences are not statistically significant. Similar patterns are observed urgent care and emergency department visits (Supplemental Table S2). However, the mean numbers of all outpatient visits are similar at about 6.4 in both groups (Table 2). The mean numbers of urgent care and ED visits are slightly higher in the mPCST group than in the mHealth-Ed group (ED visits, 0.54 vs 0.43; urgent-care visits, 0.41 vs 0.35).
Within-trial medical resource use and costs
Table 2 Long description
Within-trial healthcare use and costs are compared between mPCST (79 participants) and mHealth-Ed (83 participants), reporting means with standard deviations and medians with interquartile ranges. Outpatient visits averaged about six and a half in both groups, with nearly no difference in mean visits and similar median counts. Urgent care use was low in both groups, averaging under one visit, with medians of zero. Emergency department visits were also low on average, under one visit in each group, and medians were zero with similar spread. Hospitalizations averaged about half a hospitalization in mPCST versus about three tenths in mHealth-Ed, with a small positive difference and a confidence interval that includes no difference. Mean hospitalization costs were higher in mPCST than mHealth-Ed, and mean total costs were also higher in mPCST, though the confidence interval for the total-cost difference includes no difference. Overall, routine visit use appears comparable across groups, while higher average costs in mPCST are driven mainly by hospital-related costs and show substantial variability.
a Costs valued in 2025 US dollars.
More patients in the mPCST group (27.8%) than the mHealth-Ed group (20.5%) were hospitalized at least once, though the difference was nonsignificant (p = 0.28). When accounting for readmissions through the follow-up period, the mean number of all-cause hospitalizations is higher in the mPCST group (0.53) than in the mHealth-Ed group (0.29), with a nonsignificant difference of 0.24 (95% CI: −0.02 to 0.54; Table 2).
Total mean medical costs are \$10,559 for the mPCST group and \$6,248 for the mHealth-Ed group, a difference of \$4,311 (95% CI: −\$674 to \$10,301) (Table 2). Hospitalization costs account for the majority of the total medical costs (\$8,917 vs \$4,762; 84% and 76% of the total medical costs for mPCST and mHealth-Ed, respectively). When excluding 1 individual who was hospitalized 6 times in the mPCST group, the mean total cost in the mPCST group is \$9,167.
HRQOL preference weights and QALYs
At baseline, after mapping responses to the EQ-5D-5L to HRQOL preference weights, both groups have the same mean at 0.61 (Table 3). In the mPCST group, mean preference-weighted HRQOL increase to 0.67 following the intervention, improvements that are maintained at 3 months post-treatment (0.70) and 6 months post-treatment (0.68). In contrast, the mHealth-Ed group had stable HRQOL of 0.62 (SD, 0.25) across all post-treatment visits.
Preference-weighted HRQOL across study visits
Table 3 Long description
Preference-weighted health-related quality of life scores are summarized for two groups, mPCST and mHealth-Ed, at baseline, post-intervention, 3 months, and 6 months. At baseline, both groups have the same mean score of 0.61, with medians 0.67 for mPCST and 0.66 for mHealth-Ed. Post-intervention, mPCST increases to a mean of 0.67 and median 0.71, while mHealth-Ed remains similar with mean 0.62 and median 0.68. At 3 months, mPCST reaches its highest mean of 0.70 and median 0.74, whereas mHealth-Ed stays at mean 0.62 and median 0.68. At 6 months, mPCST remains higher than baseline with mean 0.68 and median 0.73, and mHealth-Ed remains near baseline with mean 0.62 and median 0.68. Sample sizes are similar but vary slightly by visit, ranging from 77 to 78 for mPCST and 80 to 83 for mHealth-Ed. Standard deviations and interquartile ranges indicate overlapping variability, so differences should be interpreted cautiously without additional statistical testing details.
HRQOL, health-related quality of life.
Cost components and ICERs
a Therapist costs valued using 2025 national Medicare reimbursement rates are applied in all scenarios except in the scenario in which therapist costs are based on 2024 national mean salary from the US Bureau of Labor Statistics.
b Incremental cost for mPCST represents the full cost of the intervention (i.e., \$0 cost for mHealthEd).
CI, confidence interval; ICER, incremental cost-effectiveness ratio; mPCST, mobile pain coping skills training; SD, standard deviation.
Model-based intervention effects on EQ-5D-5L preference weights reveal an incremental improvement in HRQOL at the post-intervention visit of 0.059; p = 0.11. The mPCST vs. Mhealth-Ed intervention effect increased to 0.077 (p = 0.04) at 3 months post-intervention and slightly waned at the 6-month post-intervention visit (0.063, p = 0.09). However, the average intervention effect across all follow-up visits reveals a statistically significant improvement in HRQOL of 0.066 (p = 0.04).
When accounting for time between EQ-5D-5L assessments, mean within-trial QALYs are 0.438 (SD, 0.118) in the mPCST group and 0.398 (SD, 0.142) in the mHealth-Ed group, a within-trial difference of 0.039 QALYs (95% CI: −0.002 to 0.076).
Within-trial incremental cost-effectiveness analysis
When accounting for the full cost of therapist time to deliver the mPCST intervention (i.e., assigned \$0 cost to mHealth-Ed), the ICER is estimated to be \$12,725 per QALY (95% CI: 5,566–69,343) when therapist time was valued using the Medicare payment rates and \$9,373 per QALY (95% CI: 4,100–51,080) in a scenario where therapist time is valued using the national average hourly wage (Table 4). Joint distributions of incremental differences in mean costs and QALYs across 1,000 bootstrap replications are plotted in Fig. 1 for both scenarios. The percentages of bootstrap replications that are \$50,000 or lower are 95.0% and 97.1% in the scenarios representing use of Medicare payment rates and national salaries, respectively. Corresponding percentages below a WTP threshold of \$120,000 per QALY are 98.5% and 99.2%, respectively.
Scatterplot of paired incremental costs and QALYs between mPCST versus mHealth-Ed from 1,000 bootstrap replications.
Including the value of patient time with and without accounting for travel time to attend videoconference sessions at the local community clinic, ICERs increase to \$20,438 per QALY and \$17,812 per QALY, respectively. When accounting for higher medical costs in the mPCST group that are primarily attributable to higher inpatient costs, the ICERs increase to approximately \$122,498 per QALY to \$130,211per QALY when excluding and including patient time costs.
Discussion
The ICAN NC trial found that the mobile behavioral health intervention, consisting of 4 remote videoconferencing sessions with a trained psychologist and use of a mobile app, significantly improved pain outcomes for women with breast cancer-associated pain who were residing in medically underserved areas across North Carolina, South Carolina, and Virginia. Our economic evaluation revealed that when compared to an attention-control educational intervention (mHealth-Ed), the mPCST intervention is cost-effective at about \$9,000–\$13,000 per QALY from the healthcare sector perspective and at about \$20,000 per QALY from the societal perspective that also accounts for the time patients spend on intervention-associated activities. All of these ICERs are consistent with good value for money even when using conservative WTP thresholds recommended 50 years ago (i.e., \$50,000 per QALY) (Neumann et al. Reference Neumann, Cohen and Weinstein2014).
Our cost-effectiveness analysis adheres to good practice recommendations for economic evaluations in health (Neumann et al. Reference Neumann, Sanders and Russell2017). It uses an incremental approach to evaluating costs and gains in QALYs, thereby facilitating comparisons with other studies. Also as recommended, we applied a healthcare sector perspective and a societal perspective, the latter of which accounts for the value of patients’ time spent on intervention-related activities (Russell Reference Russell2009; Neumann et al. Reference Neumann, Sanders and Russell2017). We also evaluate uncertainty associated with estimated differences in costs, QALYs, and ICERs along with graphical representations of uncertainty. Although the recommended time horizon for cost-effectiveness analyses should include the full duration of time when costs or outcomes are affected by the intervention being evaluated, extrapolating results beyond the follow-up period in the trial requires assumptions about how long the effects will last. Our analysis based on observed patient-level data collected within the ICAN NC trial allows us to avoid such assumptions (Ramsey et al. Reference Ramsey, Willke and Glick2015). We consider this a strength of the analysis. If the observed HRQOL benefits continue to accrue beyond the 10-month follow-up period, its cost-effectiveness would continue to improve.
In a cost-effectiveness analysis of a patient-centered behavioral health intervention, it is unusual to limit intervention costs to the value of time required by the provider because it ignores the cost of overhead. However, because the intervention is delivered virtually, overhead costs are minimal and non-facility Medicare payment rates used in the base-case analysis are intended to cover such costs. In addition, we included estimated time therapists spend, on average, before and after each videoconference session to account for preparation, scheduling, and documentation to avoid underestimating the full amount of time required by psychologists to deliver the mPCST intervention. There may be some additional provider time associated with referring patients to the intervention, and staff assistance with scheduling, helping with videoconferencing issues, and other activities. However, with automated order sets available within electronic medical record systems, we anticipate that the ability to refer patients to an effective pain-management program could ultimately save providers time. Regarding technology required to deliver the intervention, associated costs will vary depending on whether patients participate in videoconferencing sessions at local oncology clinics or at their homes and whether they have required hardware and adequate-speed internet access. Practices considering implementation of a mPCST program could estimate their expected costs. As long as the cost per patient served does not exceed \$1,950, the program would be considered cost-effective at a WTP threshold of \$50,000 per QALY.
There are few other CEAs of nonpharmacologic interventions for chronic pain management. We previously reported on the cost-effectiveness of 8 possible dosing strategies for a PCST intervention, which delivered good value, particularly for dosing that allowed for telephone-based maintenance calls for people who responded to an initial single PCST session (Li et al. Reference Li, Reed and Winger2023). Other studies have evaluated the cost effectiveness of multidisciplinary approaches to managing chronic musculoskeletal pain. A 2022 review of 7 CEAs found mixed results, and only 1 was based in the U.S. The U.S. study found that care for chronic low-back pain provided at an integrative care clinic vs. other clinics at the same hospital was not cost-effective based on their estimated ICER of \$436,676 per QALY, but they found lower costs of care between 6 and 12 months with integrative care (Wayne et al. Reference Wayne, Buring and Eisenberg2019).
Findings from our cost-effectiveness analysis complement implementation measures collected in the trial on accrual, participant retention, and intervention protocol adherence. Among the 998 women who could be contacted and were not ineligible, 45% completed verbal screening. Among the 289 women who were eligible after verbal screening, 61.2% (177) were randomized, and 86.4% (153) of those randomized completed all 4 sessions.
The results of this work demonstrate that a nonpharmacologic behavioral pain intervention can deliver meaningful improvements in quality of life at a cost well within commonly accepted thresholds for good value. The consistently favorable ICERs across healthcare and societal perspectives indicate that mPCST provides efficient use of resources while addressing pain in patients with cancer receiving care in medically underserved communities where adjunctive pain interventions, particularly nonpharmacologic, are limited. The use of remote delivery of this intervention minimizes infrastructure demands and enhances accessibility without compromising economic value. These features of the intervention also increase the scalability of mPCST. The study results support broader dissemination and implementation of mPCST as a high-value component of comprehensive cancer care.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1478951526103393.
Acknowledgments
The authors wish to thank the women who participated in this study and the research and medical staff who contributed to this study.
Funding
This study was funded through an NIH/NCI grant R01CA237892 awarded to T.J.S.
Competing interests
The authors declare no competing interests.
Open access
