Cost evaluation of an exercise oncology intervention: The exercise in all chemotherapy trial

Abstract Background There is strong evidence supporting the efficacy of exercise oncology programs to improve physical and psychosocial outcomes during active treatment. However, there is a paucity of evidence on the effect of exercise on healthcare utilization and cost analyzes of exercise oncology programs. Aims Our objective was to assess the effects of a pragmatic exercise oncology program (ENACT) during active chemotherapy treatment on healthcare utilization and associated costs. Methods We conducted post‐hoc analyzes on 160 ENACT participants and 75 comparison participants matched on cancer site, stage, age range, and gender. We obtained complete healthcare utilization histories for each patient (specific to emergency department [ED] visits and hospital admissions) coinciding with their participation in ENACT. A sub‐analysis was conducted for advanced stage breast, gastrointestinal, and pancreatic cancer patients. Results Healthcare costs for patients who participated in the ENACT exercise oncology intervention were numerically lower than healthcare costs for the comparison group, even after accounting for the cost of the intervention. However, the differences were not statistically significant. Conclusion Our findings suggest that an exercise oncology program during active chemotherapy treatment are at least cost neutral for all cancer patients, including advanced stage cancers. Additional research is warranted to evaluate the potential for exercise oncology programs to reduce healthcare utilization, particularly in advanced cancer patients.


| INTRODUCTION
Patients who receive chemotherapy average two emergency department (ED) visits and one hospitalization per year compared to the national average of 0.17 ED visits and 0.055 hospitalizations among the general population. 1 Fifty percent of the ED and hospitalization visits in patients who receive chemotherapy are potentially avoidable [2][3][4] and pose a high morbidity and mortality risk. 5 Reducing avoidable ED visits and hospitalizations has implications in: improving the financial health of health systems that provide cancer care, calculating hospital performance, and lessening the economic burden for patients, payers, and society. 6 Previous supportive care interventions that have attempted to address this pressing healthcare utilization issue have utilized more costly approaches such as additional nursing time, and/or advanced monitoring systems. [6][7][8] Over 680 randomized controlled trials have clearly demonstrated that exercise is beneficial and effective for numerous health-related outcomes during chemotherapy for multiple cancer diagnoses. 9,10 These include fatigue, physical function, quality of life, anxiety, depression, sleep, and breast-cancer related lymphedema. 10 Despite the extensive body of scientific evidence, only 30%-47% of cancer survivors are adequately meeting exercise guidelines. 11,12 Further, there is no similar body of scientific evidence that provides data on the proportion of patients who are adequately active during chemotherapy. Contributing to the low number of exercising patients is the low proportion of exercise referrals made by health care providers during infusion therapy: it is estimated that 9% of nurses and up to 23% of physicians refer their patients on active cancer therapy to exercise programs. [12][13][14][15] A major barrier to referral and implementation to exercise oncology programs is cost. 9 Exercise oncology programs are not presently covered by third party payers in the United States and are seen as an unrecoverable cost on the part of patients, payers, and society.
The perception of exercise oncology programs as an expensive burden could potentially be eliminated if it could be demonstrated that exercise programming can reduce healthcare utilization (i.e., ED visits, hospitalizations) during chemotherapy. Even if the costs of an intervention were neutral compared to the elevated costs of healthcare utilization, this could be of interest to providers, health systems, and payers. To address this pressing issue, our team used data from a pragmatic pilot trial, Exercise iN All ChemoTherapy (ENACT). The ENACT trial demonstrated that embedding a cancer exercise trainer into the chemotherapy infusion suite to counsel and prescribe home-based resistance and aerobic exercise is highly acceptable to healthcare providers and patients, and is feasible for patients to complete. 16 For this post-hoc analysis, we retrospectively examined healthcare utilization within trial participants and matched comparisons, to discern if exercise might alter these potentially avoidable costs.

| METHODS
The ENACT trial was a mixed-methods pre-post single group pragmatic trial to assess the feasibility and acceptability of embedding an exercise trainer into the chemotherapy infusion suite from the perspective of clinicians and patients at the Penn State Cancer Institute (PSCI) (NCT03461471). For this post-hoc analysis of the trial, we further identified gender-, age-, and cancer site-and stage-matched patients through the electronic medical record. The Penn State Human Subjects Protection Office approved this protocol and all patients provided written consent prior to any study-related activities.

| Comparison group
Patients who served as the comparison group were selected from the ENACT patient directory of approached patients who were either not interested in participating or were excluded for nonmedical reasons such as timing (e.g., identified as they were nearing completion of cancer infusion therapy). Therefore, patients in the comparison control group shared identical inclusion and exclusion criteria as patients in ENACT. Further, for the sub analysis of advanced breast, gastrointestinal (GI), and pancreatic cancer ENACT patients, comparison group patients were matched based on cancer site, stage, age range, and gender.
We extracted ED visits and hospitalizations from the electronic medical record for all participants. We recorded information on ED visits and inpatient hospitalization stays based on the start date of when the patient was approached for participation in ENACT until the end of their primary treatment for nonmetastatic patients or 6 months later for metastatic patients, coinciding with study duration of ENACT patients. The length of time for comparison patients was matched.

| Exercise intervention
The exercise intervention has been described previously. 16 Briefly, the home-based exercise intervention consisted of a mix of five elements: aerobic, resistance, balance, flexibility, and rest.
The resistance training prescription contained specific frequency, intensity, and time for each exercise. Although resistance training was the main exercise prescription for the majority of

| Measurements
To identify ED visits and hospitalizations, staff reviewed all encounter notes from the start date (ENACT intervention start date or a matched comparison start date) and forward by 6 months or to the end of infusion therapy (if less than 6 months). Reports of ED visits and hospitalizations at all locations (Penn State and other health systems) were treated the same.
ED costs were calculated using the average cost of $1389.00 per ED visit. 19 Hospitalization costs were calculated using the average cost of $10 700.00 per night of hospitalization admission. 20 Intervention costs were estimated by evaluating the cost of staff time for screening patients in the medical record, the interventionist time per patient, multiplied by hourly wage including fringe, and equipment cost. We have previously reported this intervention cost. 16 Overall total cost from the payers' perspective were calculated by summing up the costs of the ED visits, hospitalizations, and the intervention.

| Statistical analysis
Descriptive statistics were used to report participant characteristics.
The t tests and chi-square tests were used to compare the intervention and comparison group patients, as appropriate. Because the cost data display right skewness and contain numerous zero values, we assumed a zero-inflated lognormal distribution for the cost data.
We derived maximum likelihood estimates of the parameters (mean and variance on the natural logarithmic scale, and the probability of a zero cost) for each of the ENACT and comparison groups. We then estimated the overall geometric mean for each group as the probability of a nonzero cost times the exponentiation of the mean on the logarithmic scale. Finally, we compared the geometric means via an approximate t test. All analyzes were adjusted for number of comorbidities. We used PROC NLMIXED of SAS, Version 9.4, for these calculations. Both arithmetic and geometric means are presented.

| RESULTS
The characteristics of the study participants are described in Table 1.
They ranged from 30 to 82 years, and on average were 60 years of age. There were slightly more women than men and 94% selfreported white race. The most common tumor types included in the study were breast, GI, and pancreatic. Forty four percent of the consented patients had stage 4 cancer. Patients in ENACT had an average of 1.18 (SD = 1.29) comorbidities at baseline. Table 1, we provide the same demographic and clinical description of the 75 participants who served as concurrent comparisons. They shared a similar average age of 60 years, a slightly higher proportion of women to men, 92% self-reported white race, common tumor types of breast, GI, and pancreatic cancer.

| All patients
An examination of healthcare utilization and associated costs in all ENACT participants and matched comparisons show that even after accounting for intervention costs and adjusting for the number of preexisting comorbidities, the overall costs were lower among the ENACT participants as compared to the match comparisons (Table 2).
Although these differences were not statistically significant, it can be noted, descriptively, that ENACT patients had fewer instances of ED visits (0.42 ± 0.07) and days spent in the hospital (1.40 ± 0.37) com-

| Advanced stage patients
In Table 3, we re-examined healthcare utilization and associated costs in patients with Stages 3 and 4 breast, GI, and pancreatic cancer. After accounting for intervention costs and adjusting for the number of preexisting comorbidities, the overall costs were lower among the ENACT participants as compared to the match comparisons. However, there were no statistically significant differences in healthcare costs between groups. Although these differences were not statistically significant,

| DISCUSSION
This study demonstrates that for patients who underwent chemotherapy, participation in the ENACT exercise oncology program was associated with numerically fewer ED visits and hospitalizations. These comparisons were made after including the cost of the intervention.
Therefore, at the very least, the exercise oncology intervention may be cost neutral.
Our sub-analysis of advanced stage breast, GI, and pancreatic can-   Table 2). Our prior publication noted that the exercise intervention was highly acceptable and feasible from the perspective of clinicians and patients alike, 16 suggesting potential for future larger cost-effectiveness trials.
Because the ENACT trial was designed to be pragmatic, the analysis conducted was appropriate. However, we acknowledge our analysis had some limitations. ENACT was not a randomized controlled trial, thus for this report, we conducted a post hoc secondary analysis of data from a single group pragmatic trial. Despite this, we were successful in identifying matched comparisons for ENACT trial participants (Table 1). There was one variable which the matched comparisons differed (comorbidities) and the analysis was adjusted for this difference. We also recognize this cost analysis does not include all direct healthcare costs, direct non-healthcare costs, indirect costs, and travel costs, therefore, it cannot be viewed from the societal perspective.
Despite international guidelines promoting the implementation of exercise in cancer treatment from the American College of Sports Medicine, 9,10 little is known about the effects of exercise oncology interventions on healthcare utilization and healthcare costs. Understanding effects on healthcare costs would facilitate incorporating exercise as a standard and reimbursable form of treatment. Future research is needed to conduct a more thorough cost-effectiveness analysis that would include indirect costs and provide a more detailed and holistic picture of the cost effectiveness of exercise during treatment.

| CONCLUSION
Exercise is an effective strategy for symptom control with several international guidelines suggesting exercise should be implemented into standard of care during chemotherapy. However, only 23% of cancer patients and survivors are being referred to exercise by oncology physicians and only 30%-47% of cancer survivors are participating in exercise. A primary major barrier to referral and implementation to exercise oncology programs is the perceived burden of cost. This preliminary analysis suggests that an exercise oncology program is at least cost neutral, thus alleviating the financial burden of patients, payers, and society. Larger, well-powered trials are required to determine whether there may be cost savings.

ACKNOWLEDGMENT
This trial was funded by the Penn State Cancer Institute.