Peer mentoring reduces unplanned readmissions and improves self-efficacy following inpatient rehabilitation for individuals with spinal cord injury

Abstract

Objective

Evaluate effectiveness of peer interventions on self-efficacy, unplanned hospital readmissions, and quality of life for patients with spinal cord injury (SCI) undergoing inpatient rehabilitation.

Design

Interrupted time-series analysis (ITSA) examined effects of peer interventions on unplanned readmissions. Intervention variables added to ITSA regression examined relationships with exposure to peer interventions. Heterogeneity of treatment effects (HTE) analysis examined differences in intervention effectiveness for patients with quadriplegia and paraplegia.

Setting

Rehabilitation hospital specializing in SCI and brain injury.

Participants

SCI inpatients (n = 1117) admitted for rehabilitation whose discharge location was home (77% male, 71% Caucasian, mean age 38.2 (SD 16.8)). A subsample of 799 patients participated in secondary analyses examining relationship between peer interventions, readmissions, changes in patient-reported outcomes, and HTE.

Interventions

One-to-one mentoring and participation in peer-led self-management classes.

Main outcome measures

Unplanned readmissions, general self-efficacy (GSE), and depressive symptoms 30, 90, and 180 days post discharge; satisfaction with life at 180 days.

Results

After implementing the peer interventions, we observed a significant decrease in both level and slope of number of patients readmitted, and level only of unplanned hospital days 30-days post-discharge. Reduction in the number of patients and unplanned hospital days was associated with number of peer visits but not peer-led education classes attended. Higher self-efficacy (GSE) was associated with greater exposure to peer mentoring, and a significant relationship between improvement in GSE and reduced hospital readmissions was observed.

Conclusions

One-to-one peer mentoring improves self-efficacy and reduces unplanned hospital readmissions following inpatient rehabilitation for persons with SCI.

Introduction

Spinal cord injury (SCI) results in functional limitations, contributing to acute and long-term care needs and associated costs.¹ Inpatient rehabilitation is the standard of care for SCI,² one goal of which is to impart the knowledge and skills needed for effective (self-) management of care needs to avoid complications resulting in hospital readmissions. Despite the importance of effective self-management, rehabilitation providers struggle to find time (with shorter lengths of stay³) and effective strategies to ensure patients are prepared for discharge. Patients often leave the hospital feeling emotionally overwhelmed, socially isolated, and lacking in the competence and confidence needed to assume responsibility for their care needs.⁴ The result has been a high prevalence of secondary complications and unplanned readmissions.⁵ Recent estimates indicate 36% of SCI patients will have an unplanned readmission, and over 12% will have multiple readmissions, within the first year of injury.⁶

Transformative learning theory helps explain why patients may not understand the importance of effective self-management early after injury. Transformative learning is the process by which individuals transform how they think about their lives following a major life event.⁷ It involves critically examining the “disconnect” between existing beliefs and the changes that result from a life-altering event.⁸ This transformation in perspective helps regain meaning in life and is important to achieving a “new normal” after SCI.⁴

Self-efficacy – belief in one's ability to execute actions required to achieve desired outcomes – is another important concept, founded in social learning theory.⁹ Self-efficacy beliefs provide the foundation for human motivation, well-being, and personal accomplishment.¹⁰ It is especially applicable to the acquisition of new skills and has been shown to be a powerful mediator in self-management of chronic health conditions.¹¹ Self-efficacy can be enhanced through opportunities that support skill mastery, and include modeling of desired behavior by others and social support for behavior change.¹²

Peer mentoring and peer-led instruction have been shown to foster transformative learning and improve self-efficacy. People learn more and try harder when they learn from people they perceive to be like themselves.^11–14 The best-researched self-efficacy enhancing intervention is the Stanford Chronic Disease Self-Management Program (CDSMP) developed by Lorig and colleagues.¹³ Research shows that the CDSMP enhances participants’ self-efficacy in self-management tasks, thereby improving health outcomes and reducing hospitalizations.^11–14 Variations of the program, including group, one-to-one, and internet and telephone-delivered training, have shown effectiveness across numerous chronic conditions, including diabetes, arthritis, heart disease, lung disease, stroke, and chronic SCI.^15–20

Because of its large referral area, the hospital in this study devoted considerable effort to develop effective transition support services throughout the 1990s. These efforts culminated in a resource-intensive, donor-funded transition support program that operated from 2001 to 2012. While effective in reducing unplanned readmissions, the program was not economically sustainable. A systems-change initiative began in 2012 to identify effective, sustainable approaches to transition support that could be adopted by other healthcare systems. Hospital leaders gathered input from patient, family, and clinical stakeholders to examine existing transition support services and identify change initiatives to enhance patients’ capacity for self-management.

Based on stakeholder input, which validated the importance of transformative learning, we developed two systems change interventions intended to improve patients’ self-efficacy in care management and reduce unplanned hospital readmissions: (1) one-to-one peer mentoring, and (2) peer-led patient self-management education. We previously reported on a randomized trial evaluating one-to-one peer mentoring for persons with SCI. Exposure to peer mentoring was associated with greater growth in self-efficacy and fewer days rehospitalized post-discharge.²¹ We also previously reported that SCI patient education classes led by peers resulted in greater patient engagement when compared with conventional education classes.^22,23

This article reports on our evaluation of the combined effects of both interventions once implemented into standard practice. The primary aims of the study were to examine (1) the impact of these interventions on hospital readmissions and (2) the relationship between exposure to the peer interventions and self-efficacy, hospital readmissions, and other patient-reported outcomes. Our hypothesis was that peer interventions would support transformative learning in newly-injured patients, leading to greater self-efficacy in care management and fewer unplanned readmissions.

Methods

Design

Peer-mentoring and peer-led education classes were introduced as standard practice in June 2015. We examined the effects of these two systems changes using an interrupted time series analysis (ITSA) to evaluate changes across two time segments. Segment A consisted of the 24 months (June 2013–May 2015) prior to implementation and Segment B consisted of the 24 months (June 2015–May 2017) after implementation of both interventions.

We conducted a power analysis to calculate the necessary number of observations (months) needed of readmissions 30 days post-discharge. We used data for patients discharged from June 2013 to January 2016 and based our power on a conservative estimate (0.6%) in level change from pre- to post-intervention. Our power analysis (conducted in Stata²⁴ correcting for multiple covariates²⁵) returned estimates of 45 months of data yielding statistical power of 0.90. We included 48 months to achieve an equivalent number of months (24) pre- and post-intervention.

Setting and participants

Eligible participants included all patients admitted to the SCI inpatient rehabilitation program whose discharge location was home and who had no co-occurring brain injury. Consent was obtained for participation. The Inpatient Rehabilitation Facility-Patient Assessment Instrument (IRF-PAI) Case-Mix Group (CMG) classification,²⁶ which groups patients into 4 tiers, was used to identify patients with similar impairments based on injury, motor and cognitive functioning at admission. Tier 1 is most severe and tier 4 is least.

Interventions

Both interventions have been described previously^21–23 and are summarized below. Once implemented as standard practice, all patients in the inpatient rehabilitation program were expected to participate in peer-led self-management training and one-to-one peer mentoring. Although an expected part of inpatient rehabilitation, patients could refuse to participate in either intervention.

Peer mentoring was provided by the hospital-employed peer support team (3 FTEs split among 4 staff) and volunteer mentors. Potential volunteer mentors were identified by hospital staff or the peer support team, from among individuals who had been patients at the host hospital, served as volunteers, or participated in hospital-sponsored sports, recreation or other community activities. All mentors completed the Christopher and Dana Reeve Foundation training program, the hospital's volunteer training program, and a brief “preceptorship.” Over 70 volunteers completed the peer mentor training; however, a core group of about 20 mentors plus the 4 SCI peer support staff provided most of the mentoring to patients. Staff mentors were salaried employees; volunteer mentors received a $20 stipend for each documented peer visit.

Patients received an initial introduction to peer support by a member of the peer support team and were assigned a peer mentor (staff or volunteer) based on injury level, age, sex, and personal interests (e.g. hobbies). The mentor met with the patient (in person or by phone) weekly during the inpatient stay and up to 90 days post-discharge.

The peer-led education program consisted of four 1-hour classes focused on self-management of conditions associated with SCI – bowel, bladder, and skin management, and special concerns (e.g. respiratory, cardiovascular). Classes (replacing conventional classes presented by nurse educators) were led by one of the hospital's staff peer mentors (time since injury ranged from 3 to 18 years); a nurse educator was present as a medical content expert. The peer-led classes incorporated a blended learning approach. Attendance in conventional classes and the peer-led classes that replaced them, was mandatory for all patients undergoing inpatient rehabilitation.

In addition to one-to-one mentoring and the peer-led education classes, therapy teams often requested peer mentors to demonstrate self-management activities (e.g. wheelchair skills) or to join patients on community outings. Patients were also encouraged to join the SCI Peers Facebook page, a community of over 2500 active members, with monitored forums on various topics related to self-management and life after SCI. We documented participation in self-management classes and peer mentoring contacts but did not document the extent of participation in peer support social media.

Patient-reported outcomes

We documented four patient-reported outcomes: (1) hospital readmissions, (2) perceived self-efficacy, (3) depression symptoms, and (4) satisfaction with life. We used two measures of hospital readmission: (1) number of patients readmitted to the hospital, and (2) number of days rehospitalized during the designated time interval (e.g. 30 days post-discharge). We added an offset variable (the log of the number of patients discharged each month) to account for differences in number of patients discharged each month.

The General Self-Efficacy scale (GSE), developed for use with chronic disease populations,^11–14 was trialed and adapted for persons with SCI. The GSE has a maximum score of 10, based on average scoring of all items on a 10-point scale.

The Patient Health Questionnaire (PHQ-8)^27,28 is a validated diagnostic and severity measure for depressive disorders, which rates depression symptoms over the past two-week period. Scores of 5, 10, 15, and 20 represent mild, moderate, moderately severe, and severe depression, respectively.

The Satisfaction with Life Scale (SWLS)²⁹ is a global measure of life satisfaction. Respondents indicate how much they agree or disagree with each of 5 items using a 7-point scale. A total score of 20–24 is considered “average” life satisfaction.

Participants were called for follow-up interviews 30, 90, and 180 days post-discharge; the SWLS was collected only in the 180-day call. Interviewers were trained in the interview process and had experience conducting interviews. Interviewers placed calls to all numbers on record for the participant and made calls in the evening and on weekends. Up to five attempts were made to contact patients at each time interval.

Data analysis

We used ITSA to calculate quasi-Poisson regressions³⁰ (with Newey–West corrections for standard errors³¹) examining changes in level and slope for number of patients and number of days hospitalized 30 days post-discharge between pre-intervention and post-intervention months. We added intervention variables to examine relationships between readmission and exposure to peer mentoring and peer-led education classes attended. We also used Pearson chi² test to compare differences in the expected and actual number of months with no hospital readmissions pre- and post-intervention.

Ordinary least squares (OLS) regressions examined relationships between hospital readmissions and self-efficacy, depression, satisfaction with life scale, and peer exposure. We used a generalized, additive model for location, scale and shape to account for zero-inflated hospital readmission data. We included a subset of 799 patients discharged during a 34-month time frame in which we collected 30-, 90-, and 180-day post discharge outcome measures.

We examined differences in hospital readmissions and patient-reported outcomes post-discharge as a function of exposure to peer mentoring. We looked only at peer mentoring based on results of the ITSA, which indicated limited effects of peer-led education on hospital readmissions. We plotted a local polynomial regression curve (using GGPlot in R³²), also known as a LOESS curve, to determine the optimal cut point in number of peer mentor visits. The regression curve indicated a marked change between 6 and 7 visits. This is illustrated in Fig. 1 noting the cut point between 6 and 7 visits. Only 10 of 204 patients (4.9%) rehospitalized within 30-days post discharge received 7 or more peer mentor visits, with an average of 0.24 days rehospitalized. This compares with 51 of 595 patients (8.6%) with fewer than 7 peer visits who were rehospitalized for an average of 0.57 days. Seven also was the median number of contacts received. Thus, we defined “high” exposure as ≥7 contacts. Thus, we defined “high” exposure as ≥7 contacts.

Figure 1.

Scatter plot of days rehospitalized × number of peer mentor visits, and noting the LOESS regression curve cut-off between 6 and 7 peer visits, N = 799.

The high exposure group (≥7 interactions) consisted of 204 patients and 595 patients comprised the low exposure (<7) group. To equalize groups, we applied propensity score matching, a statistical technique where a smaller group is matched 1-to-1 with a subject from a larger group based on identified criteria, such as demographics.^33–35 We used injury level, age, sex, race, marital status, and length of stay as matching variables so that the group of patients who received less peer interaction was similar in number and characteristics to the high group. After propensity score matching, 204 subjects were included in our low exposure group and 391 subjects were excluded from analyses. T-tests confirmed there were no significant differences between our final high and low exposure groups included in the analyses. As may be expected, the 391 subjects excluded after propensity matching differed significantly in all characteristics (except sex) from subjects included in the analysis. We used t-tests of mean differences, to compare differences in hospital readmissions and patient-reported outcomes for patients who received high versus low peer mentoring exposure. We also conducted an exploratory heterogeneity of treatment effects (HTE) analysis to examine differences in intervention effectiveness for patients with quadriplegia vs. paraplegia.

Results

A total of 1144 patients were eligible to participate; 1117 (98%) were successfully contacted for the 30-day post discharge interview. Participants were 77% male, 71% Caucasian/24% Black, mean age 38.2 (SD 16.8). We had 799 participants with complete data at 90 and 180 days post discharge who were included in secondary analyses. We examined missing data for our primary (hospital readmissions) and secondary (GSE, PHQ-8, SWLS) outcome measures at each data collection interval (30, 90, 180 days post-discharge). For readmissions, we had data for 98%, 87% and 79% of participants at each respective time interval. For secondary outcomes, we had full data for 79%, 73% and 59% of participants at each interval. We compared characteristics of participants with missing and complete data, and only marital status was significantly different between groups – 46% of participants with full data were married vs. 34% of participants with missing data. Missing data for secondary outcomes were imputed using the Multivariate Imputation by Chained Equations package in R,³⁶ which imputes the dataset five times before returning a final and best dataset. After imputation, we completed a sensitivity analysis using three different delta-adjustments to create various scenarios of our analyses.³⁶ Independent means tests showed no significant differences, suggesting our data are sufficiently robust to withstand multiple imputations.

Table 1 presents results of the ITSA regression comparing segment A (pre-intervention) to segment B (post-intervention) for the two measures of hospital readmissions. There were no significant differences in age, sex, race, or marital status between participants in segments A and B, but significantly more persons with quadriplegia and IRF-PAI CMG Tier 2 classification in Segment B. A small effect size (Cohen’s d = 0.238) was observed for number of patients and a medium effect size (Cohen’s d = 0.446) for number of days rehospitalized.³⁷ For number of patients readmitted, we observed a statistically significant change in level (P = 0.002) and slope (P = 0.48). For number of hospital days, we observed a statistically significant change in level (P < 0.001) but not slope (P = 0.087). A statistically significant relationship was noted between one-to-one peer mentoring and days rehospitalized (P = 0.01); the relationship between peer mentoring and number of patients with readmissions almost achieved significance (P = 0.050). There was no significant relationship between peer-led classes attended and either measure of readmissions (P = 0.500 for number of patients and 0.822 for number of days).

Table 1. Quasi-Poisson regressions* for unplanned hospital readmissions. N = 1117.

Variable	Number of patients**		Number of days***
	Estimate	P-value	Estimate	P-value
Intercept	−3.111	<.001	−1.249	<.001
Pre-slope	0.060	0.017	0.088	0.002
Level change	−1.016	0.002	−1.465	<.001
Slope change	−0.066	0.048	−0.069	0.087
One-to-one peer mentoring	−0.069	0.050	−0.110	0.010
Peer-led classes	0.141	0.500	−0.055	0.822

*With Newey–West standard errors.

**Effect Size (Cohen’s d) = 0.182; (small = 0.2, medium = 0.5, large = 0.8).

***Effect Size (Cohen’s d) = 0.569; (small = 0.2, medium = 0.5, large = 0.8).

Figs. 2 and 3 present the number of patients discharged each month (gray bars) and the number of patients readmitted and days rehospitalized 30 days post-discharge. It is important to note the ascending trends in Segment A. Donor-funded transition support services were phased out beginning in December 2012 and, as anticipated, unplanned readmissions began trending up with the loss of these services. The average number of patients readmitted each month decreased from 1.75 pre-intervention to 1.46 post-intervention. Average hospital days readmitted decreased from 13.08 to 8.83. There were also significantly more months with no unplanned readmissions post-intervention (P = 0.010), increasing from one of 24 months pre-intervention (4%) to 8 of 24 months post-intervention (33%).

Figure 2.

Number of patients rehospitalized 30 days post discharge, N = 1117.

Figure 3.

Number of days rehospitalized 30 days post discharge, N = 1117.

Using a subset of participants with data collected at the 90- and 180-day evaluation periods (n = 799), OLS regressions indicated a significant relationship between self-efficacy and hospital readmissions at all three evaluation time points (P < 0.001). See Table 2. No significant associations were found with PHQ-8, SWLS, or peer exposure. Table 3 presents 30, 90, and 180-day post discharge outcomes for 408 patients with high vs. low exposure to the peer mentoring intervention. Significant differences in days hospitalized (P = 0.047) are noted between groups at 30 days post discharge. At 180 days post discharge, there are significant differences in cumulative days rehospitalized (P = .023) and in GSE (P < 0.001). No differences are noted in PHQ-8 or SWLS.

Table 2. Unplanned hospital readmission and self-efficacy, Ordinary least squares regression. N = 799.*.

GSE	30-day Readmission		90-day Readmission		180-day Readmission
	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value
30 days post discharge	−0.211	0.000	−0.374	0.001	−0.550	<.001
90 days post discharge			−0.470	<.001	−0.687	<.001
180 days post discharge					−0.794	<.001

*Subset of full sample that had complete 180-day post discharge data.

Table 3. Peer mentor exposure, independent means t-tests. N = 408.*.

	≥7 Peer mentor contacts n = 204	<7 Peer mentor contacts n = 204*	P-value
30 Days post discharge
Hospital days	0.24 ± 1.45	0.61 ± 2.23	0.047
General self-efficacy score	8.12 ± 1.31	8.15 ± 1.32	0.830
PHQ-8	6.89 ± 4.88	6.97 ± 4.89	0.871
90 Days post discharge
Hospital days	0.63 ± 3.05	1.09 ± 3.76	0.180
Cumulative hospital days	0.87 ± 3.36	1.70 ± 5.09	0.053
General self-efficacy score	8.37 ± 1.25	8.15 ± 1.47	0.111
PHQ-8	6.91 ± 4.90	7.40 ± 5.24	0.329
180 Days post discharge
Hospital days	0.61 ± 2.75	1.05 ± 3.36	0.148
Cumulative Hospital days	1.49 ± 4.71	2.75 ± 6.41	0.023
General self-efficacy score	8.56 ± 1.11	8.10 ± 1.52	<.001
PHQ-8	6.54 ± 4.84	7.05 ± 5.37	0.309
Satisfaction with life scale	19.40 ± 8.01	18.44 ± 7.61	0.212

*Subset of 799 with complete 180-day post discharge data after propensity score matching applied to identify subject group with <7 peer mentor contacts.

HTE analyses compared intervention effectiveness for patients with paraplegia and quadriplegia, separately. Patients with paraplegia exposed to ≥7 peer contacts demonstrated greater improvement in GSE 180 days post-discharge (P = 0.034). Patients with quadriplegia who received ≥7 peer contacts also experienced greater improvement in GSE (P = 0.005), and fewer days readmitted within 30 days of discharge (P = 0.050).

Discussion

The role of peers as change agents during rehabilitation and recovery after a significant health event is increasingly being recognized.³⁸ However, there is a paucity of evidence to support the effects of peer mentoring. The current study contributes evidence suggesting a positive relationship between peer mentoring, improved GSE, and fewer days of hospitalization post-discharge.

Our findings concerning relationships between peer-led interventions and self-efficacy build on the work of Lorig and colleagues in chronic disease self-management, and lend further evidence for the link between self-efficacy and improved health outcomes.^11–20,39 Our studies offer the first evidence of the effectiveness of peer interventions implemented during early stages of recovery when patients typically have not undergone the transformative learning needed to appreciate the need to manage personal care needs.⁸ This poses a challenge for healthcare systems that are increasingly being held accountable for post-discharge outcomes, yet struggle to engage patients in efforts to acquire knowledge and skills needed for effective self-management.

Evidence suggesting a positive association between peer mentoring interventions and fewer days of unplanned readmissions provides support for inclusion of peer support in rehabilitation. The percent of patients rehospitalized dropped from 8% to 6% after the inclusion of peer services at the host facility; however, nationally, about 30% of persons with SCI are rehospitalized within any given year post injury. More importantly, the number of months in which there were no unplanned readmissions increased dramatically from 1 to 8 pre- to post-intervention. These changes have meaningful clinical and financial implications.

For ITSA, we used a metric of cumulative days rehospitalized which may provide finer delineation because patients may be readmitted more than once and hospitalizations vary in length. A significant decrease in the level of hospital days post-intervention indicates shorter lengths of stay and may signify less severe conditions requiring rehospitalization. Self-efficacy may be linked with utilization; persons who are better able to manage conditions associated with injury may seek appropriate healthcare services but need less time for resolution when issues are identified early. This decrease in days rehospitalized is especially noteworthy given that there were significantly more people with quadriplegia and in CMG Tier 2 post-intervention than pre-intervention.

We allowed all study participants to be included in peer intensity analyses, which probably resulted in a lower than expected median number of peer visits (7), as peer interventions were not a part of standard practice at the beginning of the study. Propensity score matching was used to equalize group size and characteristics as it reduces selection bias and strengthens generalizability of findings.^33–35

Study limitations

This research was conducted in a single, possibly atypical care setting. Generalizability to other settings may be limited. The importance of organizational culture that endorses peer mentors as part of the rehabilitation team cannot be overstated. We benefited from a patient-centered culture and strong leadership support that may not be available elsewhere.

There are limitations to ITSA suggesting limited benefits from exposure to peer-led classes and any conclusions drawn from it. Four classes may be insufficient exposure to make a difference and quantifying the role of peers in education based on class participation alone may not capture the full impact of peers. Patients also participated in the peer social media community and peers became more active in patients’ rehabilitation as organizational culture shifted to embrace peers as integral members of the rehabilitation team. We were not able to quantify this exposure nor track the level of exposure at the patient level. Thus, the impact of this broader peer milieu is unknown. One challenge for future research will be quantifying elements of this broader peer influence and evaluating its impact.

Another potential limitation relates to the timing of the study. As noted, we implemented the peer interventions to replace donor-funded transition support services phased out prior to the study. As anticipated, there was an upward trend in 30-day readmissions noted in Segment A. While this may have inflated baseline slope and level of unplanned readmissions, the immediate reversal of this trend observed with introduction of the peer interventions supports their effectiveness as an acceptable alternative.

Despite these weaknesses, a strength of the project is in demonstrating the potential for peer interventions as part of transition supports for patients with complex care needs. Showing benefit with acutely injured/disabled individuals suggests that other patient populations might benefit from this approach. Healthcare systems may consider these findings in making decisions aimed at reducing hospital readmissions for medically complex patients transitioning home.

Future research

Replication of the peer interventions is needed across multiple settings to establish external validity and to identify setting characteristics important to successful replication. Additional research is needed to capture the full nature of peer interventions and their impact in transition support, especially the role of social media and involvement of peers as members of the rehabilitation team. Finally, research is warranted to examine use of peer interventions with other acutely injured/disabled populations.

Conclusion

One-to-one peer mentoring improves self-efficacy and reduces unplanned readmissions and number of days rehospitalized after discharge from inpatient rehabilitation. Findings will be of interest to healthcare systems seeking to implement transition services that support successful discharge and reduce readmissions for patients with complex needs.

Disclaimer statements

Contributors None.

Conflicts of interest The authors have no conflicts of interest to report.

Funding Statement

This work was supported through a Patient-Centered Outcomes Research Institute (PCORI) Award # IH-12-11-5106 and the Robert W. Woodruff Foundation.

Clinical trial registration number

NCT02746978