Impact of Ostomy Self-Management Telehealth training for rural cancer survivors on health care utilization and economic outcomes in the United States

Abstract

Purpose:

To assess the impact of a structured educational curriculum Ostomy Self-Management Telehealth (OSMT) treatment among cancer survivors residing in rural areas of the United States on selected measures of health care utilization, cost, and employment status.

Methods:

This was a multi-site randomized controlled trial comparing OSMT treatment group against a control group receiving usual care (UC) in rural populations. OSMT treatment consisted of virtual group sessions led by trained peer ostomates delivered once a week over a 5-week period via video conferencing platforms. Surveys related to health care utilization were administered up to four times: baseline, post-session, 3-month and 6-month follow-up.

Results:

Compared to the UC group, the OSMT group was associated with lower frequencies of in-person nurse (−57.2%; p = 0.015) and physician (−76.1%; p = 0.024) visits in the post-session follow-up survey; no significant differences were observed in the subsequent follow-up surveys. Moreover, the OSMT treatment group was also associated with lower ostomy-related emergency department visits (−88.3%; p = 0.119), lower direct out-of-pocket health care (−25.8%; p = 0.405) and travel costs (−47.7%; p = 0.105), as well as higher probability of full-time employment (18.9% vs. 12.3%; p = 0.179) and lower probability of claiming disability (14.3% vs. 18.9%; p = 0.459) in the 6-month follow-up; these differences, however, were not statistically significant.

Conclusion:

While not all statistically significant, the OSMT treatment was associated with some notable changes in the patterns of health care utilization and selected economic outcomes among ostomates residing in rural communities. This suggests that the OSMT treatment likely contributes to more efficient and cost-effective care in the target population.

Synopsis:

Ostomy Self-Management Telehealth (OSMT) program seeks to reduce barriers to care and improve self-management skills especially among ostomates residing in rural communities. This study reports that OSMT was associated with lower in-person health care provider visits, suggesting OSMT may lead to more efficient and cost-effective care.

INTRODUCTION

An ostomy can adversely affect health-related quality of life in a diverse population of cancer survivors.^1-3 For cancer, ostomies are most commonly created for rectal cancers, followed by urinary bladder cancer.⁴ Ostomies may be needed for other cancers related to bulky or metastatic disease, or in emergencies. To cope with the specific challenges of ostomy management, cancer survivors need strong self-management skills, which are often long-term, complex and multidimensional. Trial-and-error ostomy self-care, nurse counseling, and community referral have been the primary modes of self-management education and support in the peri-operative period.

In reality, ostomates face many obstacles in coping with their condition, not least of which is medical care that often does not meet their needs for effective information, clinical management, psychological support, and patient activation. Rural cancer survivors in particular face significant barriers to ongoing cancer care including lack of access to specialists and cancer-specific support, longer travel distances to access follow-up care, and transportation challenges due to financial barriers or lack of a vehicle.^5,6 Individuals in rural areas travel a median of 51–59 min to reach specialized oncology care. ⁶ Additionally, some studies suggest that rural cancer patients have worse quality of life and mental health than urban counterparts, and up to a quarter of rural cancer patients have unmet cancer information needs.⁷ Previous studies have demonstrated the high post-surgical cost of care burden from the perspectives of payers and health systems, largely due to high utilization of acute care associated with post-surgical complications.^8-10

Telehealth is broadly defined as providing health care via communications technologies. Telehealth delivery of an ostomy self-management training program offers an opportunity to reduce access disparities among rural ostomates by mitigating access barriers and/or burdens associated with patients’ geographic, economic, social, and/or demographic characteristics, or with systemic factors, such as provider shortages.¹¹ The current usual care (UC) in peri-operative and long-term settings is often not standardized for ostomy patients. It typically consists of an ostomy care nurse who works with patients and caregivers on technical issues (fitting, emptying, supplies, surrounding skin care, etc.) while the patient is still an inpatient. Moreover, ostomates often do not receive specific appointments to ostomy support providers postoperatively, but receive ostomy care or advice in the context of other scheduled clinic appointments only on an “as-needed” basis.¹² Often, there is no recurring support group or formal peer support systems for ostomates, and little or no matching of specific products to individual patient needs. These challenges are likely more difficult in rural populations, where formal peer support systems are likely less available and in-person consultations may present access barriers. Potential consequences of such challenges include, among others, avoidable acute care utilization due to complications (i.e., emergency department visits and inpatient admissions), higher out-of-pocket costs for patients, and reduced quality of life.

This study examines the impact of the Ostomy Self-Management Telehealth (OSMT) program among ostomates residing in rural communities on patterns of post-surgical health care utilization, direct health care cost, and indicators of economic activity (i.e., employment). OSMT is a model of care that has been developed specifically for ostomates who are unable to receive necessary care through the usual care model.^12,13 Specifically, this study tests the following hypotheses: Compared to those who receive usual care, ostomates residing in rural communities who receive OSMT are less likely to rely on in-person nurse and physician visits while incurring lower rates of ostomy-related acute care utilization and lower direct health care costs. Furthermore, it is also hypothesized that the OSMT is associated with greater likelihood of reporting full-time employment and lower likelihood of reporting disability among ostomates, because their time and travel burdens associated with in-person ostomy care would be reduced due to OSMT.

METHODS

Previous studies have described the design and implementation of OSMT in detail.^12,13 For the purposes of this study, the OSMT model has been modified specifically to accommodate the rurality of the study participants, who were drawn from rural populations of eight participating sites: Geisinger Medical Center, Danville, PA, University of North Carolina at Chapel Hill, NC, City of Hope, Duarte, CA, University of New Mexico, Albuquerque, NM, University of North Dakota—Sanford Health, Fargo ND, Loma Linda University Health, Loma Linda, CA, University of South Carolina Greenville (Prisma Health), Greenville, SC Lancaster General Hospital, Lancaster PA.

Eligible patients were identified by a study coordinator in each site who verified eligibility, obtained patient consent, and randomized patients into either the OSMT treatment group or the usual care (UC) control group. The inclusion criteria included (1) cancer or pre-cancer (e.g. carcinoma in situ or severe dysplasia) survivors over 18 years of age having undergone a procedure that needed an intestinal stoma (fecal or urinary), (2) residence within a zip code that is designated as non-metropolitan or non-urban, (3) Participant must have been able to complete the study questionnaires and sessions in English, (4) all participants attended their first training session at least six weeks after their operation, (there was no maximum time since surgery), and (5) eligible patients with temporary ostomies could not undergo ostomy reversal: (i) While they are participating in the training sessions (intervention arm). (ii) During the corresponding time of the training sessions (UC arm). To enhance accrual to the cancer survivorship study, the study team included patients with carcinoma in situ, because there is evidence that those with “pre-cancer” identify as cancer survivors.¹⁴ In addition, the study team recognized that the operation, while rarely performed for those with carcinoma in situ, is the same for those with invasive cancer and thus believed that it was reasonable to include them in this study. Figure 1 summarizes the flow of patients from eligible patient identification and randomization to survey completion and thus the final sample size in each arm.

FIGURE 1.

Consort diagram.

Patients were randomized by ostomy type into either the OSMT treatment group or usual care arm. Those in the OSMT treatment group received a local and national resource list as well as virtual group sessions led by ostomy nurses and trained peer ostomates delivered once a week over a 5-week period via secure video conferencing platforms, while those who were randomized into the UC group received the same resource list and in-person ostomy nurse visits on an “as-needed” basis as arranged by home institutions.

Patient surveys were administered up to four times over the study period: baseline, post-session, 3-month follow-up, and 6-month follow-up. The baseline and 6-month follow-up surveys shared one set of identical questions, while the post-session and 3-month follow-up surveys shared another set of identical questions. Several questions related to nurse and physician in-person visits, however, were included in all four survey rounds to capture more detailed changes over time. Other questions considered in this study were included only in the baseline and the 6-month follow-up surveys (refer to Table A1 for the list of survey questions considered for the purposes of this study’s aims).

From the survey questions, the following dependent variables of interest were constructed: frequencies of in-person nurse and doctor visits, ostomy care-related emergency department (ED) visits, acute care visits (defined as sum of ED, urgent care visits, and inpatient hospital admissions), as well as out-of-pocket (OOP) direct health care and travel costs, that were incurred during the “past four week” period as of the time of the survey completion. Because the survey questions provided multiple choice answers indicating pre-specified value ranges (e.g., “How much do you typically pay out of pocket for each nurse visit?: Nothing; $1–$10; $11–$25; $26–$50; $51–$75; $76–$100; $101+”) rather than asking for actual values, the study team took the mid-points of the respective value ranges for the purposes of calculating total and mean values.

To estimate the OSMT treatment effects on the dependent variables, the difference-in-difference approach was used to specifically account for potential differences at the baseline between the OSMT treatment group and the usual care control group.^15,16 Even though randomization ensured balance between the OSMT treatment group and the usual care control group, the actual survey respondents (who were a subset of those who had participated in the trial) might be different in terms of their baseline characteristics between the two groups. The difference-in-difference method was implemented via a set of multivariate regression models in which the key explanatory variables were a binary indicator for the OSMT treatment group versus control, a set of indicator variables for the three post-baseline survey rounds, and a set of interaction terms between the OSMT treatment group and the survey round indicator variables. The coefficients of the interaction terms represented the OSMT post-baseline treatment effects on the respective dependent variables (see Appendix for the description of the multivariate regression model specification).

For the dependent variables that were count variables in nature (i.e., frequency of in-person nurse and doctor visits, emergency department visits, etc.), Poisson model with patient random effects (to account for repeated observation for each patient in the data) was used; for those dependent variables that were continuous in nature (i.e., out-of-pocket direct health care and travel costs), linear regression model with patient random effects was used; and for those that are binary in nature (i.e., full-time employed or not; on disability or not), probit model with patient random effects were used. Although there is no a priori reason to use probit instead of logit model in this context, logit was not used because the standard odds ratio transformation and interpretation of logistic regression model coefficients do not apply when estimating interaction effects via non-linear models such as logit, as articulated by Ai and Norton.¹⁷ Other covariates in the models included patient age, gender, race (White vs. other), insurance type (commercial, Medicare, or other), and ostomy type (colostomy, ileostomy, urostomy, or other).

The magnitudes of the OSMT treatment effects were represented as “observed” versus “expected” values. “Observed” values represented the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group. “Expected” values were obtained by setting the interaction effects between the OSMT treatment group and the survey round indicator variables to zero and re-calculating the regression-adjusted values of the respective dependent variables for the OSMT treatment group; these values represented the expected values of the dependent variables that would be expected if the OSMT treatment group had followed the same post-baseline trends as the control group. That is, the “expected” values capture what would have happened to the OSMT treatment group if they had not been exposed to the OSMT treatment. The differences between the “observed” and “expected” values therefore captured the regression-adjusted magnitudes of the OSMT treatment effects (refer to Tables A2-A5 for the full regression model outputs). The percentage changes were then calculated by subtracting the observed value from the expected value and the dividing by the expected value, which was then multiplied by 100. The p-values were obtained from the corresponding p-values of the regression coefficients as shown in the Appendix, as the percentage changes are equivalent to the regression model coefficients re-transformed to their original scale. This study was reviewed approved by University of Pennsylvania Internal Review Board.

RESULTS

A total of 204 patients participated in the study, of which 101 were randomized to the OSMT group, and 103 in UC. Of those randomized, 76 (75.2%) in OSMT completed all surveys, and 88(85.4%) in UC completed surveys. Table 1 summarizes the baseline characteristics among the survey respondents of OSMT and the UC groups. There were no significant differences in demographics after randomization; however, they differed from each other in terms of the baseline frequency of acute care visits and direct out-of-pocket costs. Such baseline differences, therefore, suggest that use of the difference-in-difference approach is justified.

TABLE 1.

Descriptive Statistics.

Patient Characteristics	Usual Care (n=88*) Mean (SD) / %	OSMT (n=76*) Mean (SD) / %	p-value
Demographic
Mean Age (SD)	62.8 (13.4)	62.6 (13.5)	0.884
Female	43.6%	44.7%	0.875
Race: White	78.2%	76.7%	0.795
Race: Other	21.8%	23.3%	0.795
Insurance Type**
Commercial	30.7%	25.2%	0.386
Medicare	45.5%	41.7%	0.585
Other	27.7%	33.0%	0.412
Ostomy Type
Colostomy	37.6%	40.8%	0.645
Ileostomy	23.8%	15.5%	0.139
Urostomy	37.6%	34.0%	0.587
Other	3.0%	1.0%	0.303
Baseline Utilization & Cost
Mean# Nurse Visits	1.01 (2.03)	1 (1.48)	0.968
Mean # Doctor Visits	0.21 (0.56)	0.4 (1.04)	0.153
Mean # ED Visits	0.03 (0.18)	0.09 (0.37)	0.215
Mean # Acute Care Visits***	0.07 (0.26)	0.23 (0.65)	0.040
Mean Healthcare OOP Cost ($)	7.87 (24.25)	21.19 (53.72)	0.044
Mean Travel OOP Cost ($)	5.2 (16.46)	13.82 (37.44)	0.060
Employment Status
Full-time	10.9%	11.7%	0.464
On Disability	13.9%	12.6%	0.338

Note: OSMT = ostomy self-management telehealth; SD = standard deviation; ED = emergency department; OOP = out-of-pocket.

^*The sample sizes represent the numbers of completed baseline surveys; the total numbers of patients in the usual care group and OSMT control groups are 101 and 103, respectively.

^**The sum of percents may exceed 100% because patients may have switched insurance type over the follow-up period.

^***“acute care” is defined as the sum of ED visits, urgent care visits, and inpatient admissions.

Table 2 summarizes the unadjusted mean and the corresponding standard deviation of each of the relevant outcome variables considered for this study. Comparing the post-intervention outcome variables to their respective baseline values over the 3 follow-up periods, both the OSMT and the usual care groups appear to follow similar trajectories over time. However, Table 2 also indicates significant baseline differences between the OSMT and the usual care groups, which provides justification for the use of difference-in-difference approach in this case. That is, because the baseline values differ significantly between the OMST and the usual care groups despite randomization, simple comparisons of the post-intervention outcome variables between the two groups are confounded and thus cannot be interpreted as causal effects without explicitly accounting for the different baseline values between the two groups.

TABLE 2.

Summary of Unadjusted Outcome Variables.

Outcome Variable	USUAL CARE				OSMT
	Baseline	Pre-session	3-month Follow-up	6-Month Follow-up	Baseline	Pre-session	3-month Follow-up	6-Month Follow-up
Mean # Nurse Visits	1.011 (2.033)	0.500 (1.168)**	0.316 (0.679)**	0.058 (0.291)***	1.000 (1.476)	0.224 (0.487)***	0.558 (1.634)	0.140 (0.441)***
Mean # Doctor Visits	0.212 (0.558)	0.171 (0.577)	0.089 (0.345)**	0.070 (0.258)**	0.399 (1.040)	0.075 (0.317)**	0.221 (1.014)	0.071 (0.260)**
Mean # ED Visits	0.035 (0.185)	n/a	n/a	0.065 (0.437)	0.091 (0.369)	n/a	n/a	0.018 (0.132)
Mean # Acute Care Visits	0.071 (0.258)	n/a	n/a	0.065 (0.437)	0.23ˆ (0.653)	n/a	n/a	0.073 (0.325)*
Mean Healthcare OOP Cost ($)	7.87 (24.25)	n/a	n/a	2.25 (13.13)*	21.19ˆ (53.72)	n/a	n/a	8.92 (31.23)
Mean Travel OOP Cost ($)	5.20 (16.46)	n/a	n/a	5.40 (22.41)	13.82 (37.44)	n/a	n/a	6.66 (22.45)
% Full-time Employed	10.9%	n/a	n/a	12.0%	11.7%	n/a	n/a	21.3%**
% on Disability	13.9%	n/a	n/a	22.7%**	12.6%	n/a	n/a	14.8%

Note: Standard deviation in parentheses.

^*Statistically significant at 10% level (p<0.1) relative to respective baseline estimates.

^**p < 0.05.

^***p < 0.001; ˆ Statistically significant difference (p<0.05) in baseline estimates between the usual care and OSMT groups.

Figure 2 indicates that, compared to the usual care group, the OSMT treatment group was associated with lower frequencies of in-person nurse (57.2%; p = 0.015) and physician (76.1%; p = 0.024) visits at the first follow-up survey; no significant differences were observed in the subsequent follow-up surveys. Figure 3 indicates that the OSMT treatment group was also associated with lower ostomy-related emergency department visits (−88.3%; p = 0.119) and lower direct out-of-pocket health care cost (−25.8%; p = 0.405) and travel costs (−47.7%; p = 0.105), although the differences were not statistically significant. Moreover, Figure 4 indicates OSMT treatment group was associated higher probability of full-time employment (18.9% vs. 12.3%; p = 0.179) and lower probability of claiming disability (14.3% vs. 18.9%; p = 0.459) in the 6-month follow-up, albeit these differences were also not statistically significant.

FIGURE 2.

OSMT impact on in-person provider visits. Note: ED = emergency department; OOP = out-of-pocket; “acute care” is defined as the sum of ED visits, urgent care visits, and inpatient admissions; “Observed” represents the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group; “Expected” represents the values of the dependent variables that would be expected if the OSMT treatment group had followed the same trends as the control group.

Note: “Observed” represents the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group; “Expected” represents the values of the dependent variables that would be expected if the OSMT treatment group had followed the same trends as the control group

FIGURE 3.

OSMT impact on health care utilization and cost at 6-month follow-up. Note: “Observed” represents the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group; “Expected” represents the values of the dependent variables that would be expected if the OSMT treatment group had followed the same trends as the control group.

Note: ED=emergency department; OOP-out-of-pocket; “acute care” is defined as the sum of ED visits, urgent care visits, and inpatient admissions; “Observed” represents the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group; “Expected” represents the values of the dependent variables that would be expected if the OSMT treatment group had followed the same trends as the control group

FIGURE 4.

OSMT impact on employment status at 6-month follow-up.

Note: “Observed” represents the regression-adjusted values of the dependent variables as observed in the data among the OSMT treatment group; “Expected” represents the values of the dependent variables that would be expected if the OSMT treatment group had followed the same trends as the control group

DISCUSSION

The OSMT treatment aims to close gaps in post-surgical care among cancer survivors with an ostomy who often face significant challenges due to limited training, support, and access to care, which are more acutely experienced among those who reside in rural communities. The OSMT treatment leverages telehealth via readily available secure video conferencing to deliver real-time ostomy self- management training and support to a patient population that is in clear need of an alternative approach to receive support. Despite the lack of statistical significance in most cases, the findings from this study are suggestive of potentially altered patient behaviors and outcomes that are indicative of greater efficiency in care delivery.

As shown in Figure 2, the OSMT treatment group was associated with statistically significant reductions in in-person health care provider visits at least during the time when the patients were undergoing the telehealth sessions. This implies that the ostomates felt their immediate needs were met by OSMT without resorting to in-person health care encounters. The two groups, however, converged over the subsequent follow-up surveys. This may be because all ostomates are likely to become skilled at self-management over time with or without further clinical support.

Although other survey results did not display statistical significance, the overall directions of the estimated OSMT effects all consistently suggest benefit for the OSMT intervention. The OSMT treatment group made fewer in-person visits, and their out-of-pocket direct costs associated with in-person visits (i.e., copays and travel costs) were lower than the UC control group. In theory, this could help cancer survivors with ostomies retain full-time employment and be less likely to claim disability. At the same time, the OSMT treatment may have reduced ED visits and other acute care utilization due to improved self-management skills attributable to OSMT. Possible future research should include more detailed examination of potential changes in post-surgical health care utilization using objective secondary data such as health insurance claims data and/or electronic medical records. Moreover, implementation of non-English versions of OSMT to target ostomates who are more comfortable in other non-English languages is also an area of future expansion and evaluation research.

This study is subject to several limitations: Due to the small sample sizes and the statistical noise in the data attributable to the process of converting interval responses into single numerical values, the statistical precision of the data is lacking. As a result, most of the estimates are not statistically significant. Nevertheless, as noted above, the overall directions of the estimates are consistent with the hypotheses. Also, because the respondents were allowed to skip any questions or to opt out of any given survey, this non-random sample attribution may have led to unobserved biases in the sample. Although the differencein-difference method implemented via multivariate regression models was used to mitigate such biases, the results may nonetheless be subject to unobserved confounding. The generalizability of our findings may also be limited depending on accessibility and availability of internet in rural areas. Previous studies, however, have suggested that magnitudes of such problems may be lower than what is commonly perceived about using these technologies in rural communities.¹⁸

To our knowledge, this is a unique study in the context of telehealth self-management coaching intervention for ostomates residing in rural areas for which there is a lack of comparable studies in the current literature. In general, telehealth for ostomy patients has shown to decrease the burden of care with high degrees of patient satisfaction. ^19,20 Another study has reported that while ostomy nurse follow-up did not show a difference in resource use, there was a preference to be followed via telehealth.²¹ Moreover, a previously published study by this study team has examined the same intervention as described in this current study but without a specific focus on rural communities. ²² This focus on rural ostomates is unique and important because ostomates residing in rural communities often experience a lack of clinical resources.²³ The rapid technological advancements allowing telehealth in these communities may prove especially beneficial over time, and the data presented in this study are consistent with such expectations.

CONCLUSION

The OSMT treatment was associated with some notable changes in the patterns of post-surgical health care utilization and selected economic outcomes among ostomates residing in rural communities, suggesting that the OSMT treatment may have contributed to potentially more efficient and effective care delivery in the target population. Future research will seek to validate these results via detailed examination of large administrative secondary data such as health insurance claims data and/or electronic medical records.

APPENDIX

Difference-in-difference regression model specification:

$$Y_{it}={\beta }_0+{\beta }_1{\text{OSMT}}_i+\beta 2{\text{SR}}_t+{\beta }_3({\text{OSMT}}_i\times {\text{SR}}_t)+{\beta }_4{X}_i+{\epsilon }_{it}.$$

$Y_{it}$ represents the outcome variable of interest; ${\text{OSMT}}_i$ is a binary indicator that equals 1 if patient $i$ belongs to the OSMT intervention group and zero otherwise; SR represents a set of indicator variables corresponding to the survey round at time $t$ (i.e., baseline (referent), post-session, 3-month follow-up, or 6-month follow-up survey); ${\text{OSMT}}_i\times {\text{SR}}_t$ represents the set of interaction terms between the OMST and SR variables; $X_i$ represents the set of patient characteristics (e.g., age, sex, primary payer type, and ostomy type). The main coefficients of interest are represented by ${\beta }_3$, which captures the OSMT intervention effect above and beyond any baseline difference between the OSMT and the usual care groups as well as any secular changes in the outcome variables over the survey period.

TABLE A1.

Survey questions included for analysis.

Selected survey question for analysis	In baseline survey?	In post-session survey?	In 3-month survey?	In 6-month survey?
How many times did you visit nurse in person in the past 4 weeks?	Yes (n=88; 76)	Yes (n = 78; 67)	Yes (n = 57; 43)	Yes (n = 69; 57)
How many times did you visit physician in person in the past 4 weeks?	Yes (n = 85; 74)	Yes (n = 79; 67)	Yes (n = 56; 43)	Yes (n = 71; 56)
If needed, how many times did you visit the ER for stoma and/or bowel function problems in the past 4 weeks?	Yes (n = 86; 77)	No	No	Yes (n = 69; 57)
IF YES: How many times were you admitted for stoma and/or bowel function problems in the past 4 weeks?	Yes (n = 85; 75)	No	No	Yes (n = 70; 57)
IF YES: How many times did you visit an Urgent Care center for stoma and/or bowel function problems in the past 4 weeks?	Yes (n = 86; 76)	No	No	Yes (n = 70; 55)
How much do you typically pay out of pocket for each nurse visit (copayments, deductibles, or other costs for the visits that weren’t covered by your insurance)?	Yes (n = 88; 76)	No	No	Yes (n = 69; 57)
How much do you typically pay out of pocket for each physician visit (copayments, deductibles, or other costs for the visit that weren’t covered by your insurance)?	Yes (n = 85; 74)	No	No	Yes (n = 71; 56)
How much did you pay out-of-pocket for your most recent ER visit (copayments, deductibles, or other costs that weren’t covered by your insurance)?	Yes (n = 86; 77)	No	No	Yes (n = 69; 57)
How much did you pay out-of-pocket for your most recent Urgent Care visit (copayments, deductibles, or other costs for the visit that weren’t covered by your insurance)?	Yes (n = 86; 76)	No	No	Yes (n = 70; 55)
How much did you pay out-of-pocket for your most recent hospital admission (copayments, deductibles, or other costs for the visit that weren’t covered by your insurance)?	Yes (n = 86; 75)	No	No	Yes (n = 70; 57)
How much do you typically pay out of pocket to get to and from each nurse visit (such as the cost for taxi, bus or parking)?	Yes (n = 88; 76)	No	No	Yes (n = 69; 57)
How much do you typically pay out of pocket to get to and from each physician visit (such as the cost for taxi, bus, or parking)?	Yes (n = 85; 74)	No	No	Yes (n = 71; 56)
How much did you pay out-of-pocket to get to and from this ER visit (such as the cost for taxi, bus, or parking)?	Yes (n = 86; 77)	No	No	Yes (n = 69; 57)
How much did you pay out-of-pocket to get to and from the urgent care visit (such as the cost for taxi, bus or parking)?	Yes (n = 86; 76)	No	No	Yes (n = 70; 55)
How much did you pay out-of-pocket to get to and from the hospital you were admitted to (such as the cost for taxi, bus or parking)?	Yes (n = 86; 75)	No	No	Yes (n = 70; 57)
How would you describe your current employment? (full-time, part-time, disability, other)	Yes (n = 88; 74)	No	No	Yes (n = 70; 58)

Note: Number of valid responses for respective question in parentheses (n = usual care; intervention).

TABLE A2.

Random effects Poisson regression model output for in-person nurse and doctor visits.

Covariate	In-person nurse visits				In-person doctor visits
	Coefficient	p-value	95% CI UB	95% CI LB	Coefficient	p-value	95% CI UB	95% CI LB
OSMT	0.092	0.730	−0.429	0.612	0.677	0.145	−0.234	1.588
Round: post-session	−0.657	0.001	−1.043	−0.270	−0.298	0.420	−1.023	0.426
Round: 3-month FU	−1.183	<0.001	−1.702	−0.663	−0.815	0.116	−1.831	0.202
Round: 6-month FU	−2.761	<0.001	−3.769	−1.754	−0.885	0.088	−1.900	0.130
OSMT X post-session	−0.850	0.015	−1.532	−0.167	−1.430	0.024	−2.667	−0.193
OSMT X 3-month FU	0.558	0.124	−0.152	1.269	0.059	0.930	−1.273	1.392
OSMT X 6-month FU	0.846	0.185	−0.405	2.097	−0.887	0.250	−2.399	0.624
Female	−0.116	0.652	−0.618	0.387	0.557	0.205	−0.304	1.418
Age	0.016	0.137	−0.005	0.038	0.009	0.622	−0.028	0.047
Insurance: commercial	−0.173	0.602	−0.825	0.478	0.001	0.998	−1.064	1.066
Insurance: Medicare	−0.467	0.150	−1.104	0.169	−0.435	0.421	−1.495	0.625
Race: White	−0.693	0.028	−1.313	−0.073	−0.152	0.779	−1.216	0.911
Colostomy	0.498	0.165	−0.206	1.201	0.210	0.738	−1.020	1.439
Ileostomy	0.703	0.050	−0.002	1.407	0.408	0.510	−0.804	1.619
Urostomy	0.338	0.313	−0.318	0.994	−0.191	0.752	−1.371	0.990
Other ostomy	−0.850	0.347	−2.623	0.922	−16.895	0.996	−6140.606	6106.816
Constant	−1.349	0.103	−2.972	0.274	−3.378	0.022	−6.273	−0.483

Note: The coefficient estimates shown in bold correspond to ${\beta }_3$ in the regression model as described in the difference-in-difference regression model above.

TABLE A3.

Random effects Poisson regression model output for ED visits and acute care utilization.

Covariate	ED visits				Acute care
	Coefficient	p-value	95% CI UB	95% CI LB	Coefficient	p-value	95% CI UB	95% CI LB
OSMT	0.943	0.276	−0.754	2.641	1.251	0.045	0.028	2.474
Round: 6-month FU	0.599	0.430	−0.888	2.085	0.026	0.968	−1.244	1.297
OSMT X 6-month FU	−2.149	0.119	−4.850	0.552	−1.088	0.227	−2.854	0.678
Female	0.205	0.786	−1.275	1.686	0.292	0.603	−0.809	1.394
Age	−0.034	0.245	−0.090	0.023	−0.018	0.440	−0.062	0.027
Insurance: commercial	−0.920	0.384	−2.992	1.151	−0.313	0.676	−1.782	1.156
Insurance: Medicare	0.343	0.712	−1.477	2.162	−0.100	0.892	−1.542	1.342
Race: White	−0.705	0.411	−2.388	0.977	−0.067	0.921	−1.395	1.260
Colostomy	−0.089	0.940	−2.406	2.228	−0.052	0.951	−1.709	1.604
Ileostomy	−0.546	0.658	−2.960	1.868	−0.435	0.634	−2.224	1.354
Urostomy	−0.046	0.967	−2.239	2.147	−0.037	0.964	−1.622	1.548
Other ostomy	−16.464	0.998	−14865.570	14832.640	−18.092	0.999	−23440.060	23403.880
Constant	−2.039	0.388	−6.670	2.593	−2.564	0.177	−6.288	1.159

Note: The coefficient estimates shown in bold correspond to ${\beta }_3$ in the regression model as described in the difference-in-difference regression model above.

TABLE A4.

Random effects linear regression model output for out-of-pocket direct costs.

Covariate	OOP health care cost				OOP travel cost
	Coefficient	p-value	95% CI UB	95% CI LB	Coefficient	p-value	95% CI UB	95% CI LB
OSMT	13.541	0.028	1.456	25.625	8.948	0.042	0.333	17.564
Round: 6-month FU	−3.721	0.298	−10.729	3.287	0.485	0.862	−4.986	5.955
OSMT X 6-month FU	−4.528	0.405	−15.197	6.141	−6.881	0.105	−15.191	1.429
Female	3.088	0.607	−8.664	14.839	6.834	0.104	−1.411	15.078
Age	−0.187	0.474	−0.699	0.325	−0.111	0.547	−0.470	0.249
Insurance: commercial	6.259	0.438	−9.549	22.067	5.352	0.345	−5.747	16.450
Insurance: Medicare	−8.634	0.288	−24.549	7.281	2.602	0.648	−8.582	13.787
Race: White	−8.092	0.299	−23.364	7.180	−16.641	0.002	−27.366	−5.916
Colostomy	6.766	0.455	−11.000	24.532	0.304	0.962	−12.194	12.803
Ileostomy	4.766	0.591	−12.620	22.151	2.544	0.683	−9.658	14.746
Urostomy	12.528	0.136	−.961	29.017	−0.623	0.916	−12.215	10.970
Other ostomy	−12.185	0.529	−50.083	25.713	−11.800	0.386	−38.460	14.860
Constant	17.896	0.367	−21.023	56.815	19.181	0.169	−8.181	46.544

Note: The coefficient estimates shown in bold correspond to ${\beta }_3$ in the regression model as described in the difference-in-difference regression model above.

TABLE A5.

Random effects probit regression model output for employment status.

Covariate	Full-time employed				On disability
	Coefficient	p-value	95% CI UB	95% CI LB	Coefficient	p-value	95% CI UB	95% CI LB
OSMT	0.070	0.925	−1.403	1.543	−0.043	0.928	−0.973	0.888
Round: 6-month FU	0.157	0.760	−0.851	1.166	0.605	0.102	−0.120	1.331
OSMT X 6-month FU	1.027	0.179	−0.470	2.525	−0.403	0.459	−1.470	0.664
Female	−1.095	0.178	−2.687	0.498	0.593	0.185	−0.284	1.470
Age	−0.060	0.080	−0.127	0.007	−0.121	<0.001	−0.182	−0.060
Insurance: commercial	2.847	0.001	1.123	4.572	0.265	0.616	−0.772	1.301
Insurance: Medicare	−0.598	0.509	−2.371	1.175	1.637	0.011	0.376	2.899
Race: White	0.353	0.698	−1.428	2.134	1.111	0.072	−0.099	2.321
Colostomy	−0.411	0.668	−2.290	1.468	0.414	0.485	−0.748	1.577
Ileostomy	0.640	0.497	−1.207	2.487	−0.151	0.797	−1.301	0.999
Urostomy	0.449	0.626	−1.359	2.257	1.026	0.095	−0.178	2.230
Other ostomy	−0.194	0.940	−5.266	4.878	1.017	0.440	−1.561	3.595
Constant	−1.088	0.612	−5.300	3.123	2.403	0.082	−0.307	5.112

Note: The coefficient estimates shown in bold correspond to ${\beta }_3$ in the regression model as described in the difference-in-difference regression model above.