Association of a Health Coaching and Transportation Assistance Intervention at a Free Ophthalmology Homeless Shelter Clinic With Follow-up Rates

Key Points

Question

What is the association of a health coaching and transportation assistance intervention at a free ophthalmology homeless shelter clinic with follow-up rates?

Findings

In this cohort study of 71 patients, 39% of patients were referred for free eyeglasses, 20% to the county hospital for advanced care, and 10% to both. Of those referred, the difference in follow-up rates from the postintervention to preintervention groups was 54%.

Meaning

A health coaching and bus token intervention improved follow-up by at least 40%; this improvement supports considering implementation of this intervention when developing public assistance programs.

Abstract

Importance

Eye health in the homeless population is important, yet follow-up to referral appointments in this population remains low.

Objective

To investigate the association of health coaching and transportation vouchers with follow-up rates at a free ophthalmology homeless shelter clinic.

Design, Setting, and Participants

A prospective cohort study was conducted from January 9, 2019, to March 4, 2020, among all 71 patients evaluated at a free ophthalmology clinic at a single homeless shelter in San Francisco, California.

Exposures

If indicated, patients were referred for advanced ophthalmologic care at a county hospital and free eyeglasses from a nonprofit organization.

Main Outcomes and Measures

The primary outcome was follow-up rates to referral appointments. The secondary outcomes were prespecified baseline variables hypothesized to be associated with follow-up. The intervention began September 4, 2019. Follow-up rates were compared between the preintervention (n = 37) and postintervention (n = 34) groups. The hypothesis was formulated before data collection.

Results

Among the 71 patients, 50 (70.4%) were men, and the mean (SD) age was 51.9 (12.4) years. A total of 28 patients (39.4%) were referred for free eyeglasses, 14 (19.7%) to the county hospital for advanced care, and 7 (9.9%) to both. Of those referred, the difference in follow-up from the postintervention to preintervention groups was 53.8% (95% CI, 39.8%-67.9%; P < .001). Compared with patients who did not follow up, those who did had a mean difference of 59 more days at the shelter (95% CI, 39-80 days; P = .003). Among patients with a visual acuity of 20/40 or worse in the better-seeing eye, the mean difference between those who did not follow up and those who did was 61% (95% CI, 44%-78%; P = .003). The mean difference in follow-up between patients who were born in the US and patients not born in the US was 89% (95% CI, 79%-98%; P = .02). Of those in the postintervention group, the difference in presentation to follow-up for patients with a high school diploma compared with those without was 59% (95% CI, 37%-81%; P = .001).

Conclusions and Relevance

This study suggests that a health coaching and bus token intervention improved follow-up rates at a free ophthalmology homeless shelter clinic by at least 39.8%; this improvement supports considering implementation of this intervention when developing public assistance programs if independent corroboration is provided. Barriers to follow-up included a shorter duration of stay at the homeless shelter, visual acuity better than 20/40, not being born in the US, and lower educational level, although the size of this study does not permit determining if some or all of these are associated with one another.

This cohort study investigates the association of health coaching and transportation vouchers with follow-up rates at a free ophthalmology homeless shelter clinic.

Introduction

Homelessness is a significant public health problem in the United States.^1,2 Poor vision in those living without permanent housing makes this population particularly vulnerable and adds significantly to their social and health burden. The homeless population has more eye problems compared with the general population,³ and refractive error, in particular, has been shown to be the leading cause of visual impairment in this population.⁴ Visual impairment is associated with unemployment,^5,6 and both screening for visual problems and providing free eyeglasses can improve the quality of life and earning potential of homeless individuals.⁷

In 2017, a free ophthalmology clinic was created at a homeless shelter in San Francisco, California, to bridge some of these gaps.⁸ The clinic refers patients to the local county hospital for advanced ophthalmologic care and to a partnering nonprofit organization for free eyeglasses; however, patient follow-up rates to referral appointments were low. Accordingly, interviews were conducted with both staff and patients at the shelter⁸ to design a health coaching and transportation voucher intervention aimed to improve patient follow-up to referral appointments.

Methods

Study Setting

This prospective cohort study took place at the monthly University of California, San Francisco (UCSF) Ophthalmology Homeless Shelter Clinic at the St Vincent de Paul Society Division Circle Navigation Center in San Francisco. This clinic was created in January 2019 to meet the eye health needs of its residents. The initial protocol with a hypothesized intervention and power calculation was created in November 2018. The health coaching materials and identification of funding for bus tokens were facilitated in concurrence with working on the protocol. The protocol was presented to the UCSF Ophthalmology Resident Research Committee and finalized in May 2019. Thus, the initial protocol was designed prior to January 2019; however, the intervention was implemented as soon as the health coaching materials were validated and funding was identified because the implementors did not want to delay a potential impactful intervention in this vulnerable population. At each clinic, the UCSF medical student volunteers collected demographic, social, and medical information from the participants. With the oversight of UCSF ophthalmology residents and an attending physician, comprehensive eye examinations were performed by the UCSF medical student volunteers. All data were collected and uploaded into the database by 2 premedical student volunteers. This prospective cohort complied with the Health Insurance Portability and Accountability Act and followed the tenets of the Declaration of Helsinki.⁹ This study received institutional review board (IRB) approval from UCSF and was registered through the Open Science Framework Registry through the Center for Open Science prior to intervention implementation. Informed consent was waived by the IRB. It was stated in the IRB outcome letter that “this is a project that includes program evaluations, quality improvement activities, or other activities that do not require further IRB oversight according to the federal regulations summarized in 45 CFR 46.102(d).” The data were deidentified and protected by privacy safeguards. Participants did not receive a stipend nor were they offered any incentives to participate in the study.

If indicated, patients were referred to the partnering nonprofit organization, Project Homeless Connect, for free eyeglasses and the local county hospital, Zuckerberg San Francisco General Hospital, for advanced ophthalmologic care and follow-up. Zuckerberg San Francisco General Hospital is a level 1 trauma center that serves as a safety-net hospital for an ethnically diverse population. The eyeglass prescription could include single-vision lenses or bifocals. At the time of referral, patients were provided with a prearranged, follow-up appointment date that was within 4 days. Follow-up data were provided by the lead service coordinator at Project Homeless Connect and the clinic coordinator at Zuckerberg San Francisco General Hospital.

Study Participants

This prospective cohort study included all patients seen at the monthly UCSF Ophthalmology Homeless Shelter Clinic from January 9, 2019, to March 4, 2020. No patients were excluded. A sample size calculation using the power of 2 proportions was performed using Stata/SE, version 15.0 (StataCorp LLC). Based on our existing referral and patient follow-up rates from a previous ophthalmology homeless shelter clinic at a different location, this calculation determined that at least 25 preintervention patients and 23 postintervention patients needed to be referred to assess a difference in follow-up rates between the 2 groups.

Intervention

A health coaching and transportation voucher intervention aimed to improve patient follow-up to referral appointments was implemented in September 4, 2019. This intervention was designed based on interviews with staff and individuals residing at the homeless shelter between November and December 2018 that identified barriers to accessing eye care.⁸

For the health coaching portion of the intervention, 5 health coaching documents were designed and reviewed by experts in the field to serve as a guide for medical students when performing health coaching (eFigures 1-5 in the Supplement). These documents were designed based on the UCSF Urban Underserved Health Coaching Curriculum¹⁰ to promote behavior change associated with eye health. The health coaching documents were validated by 3 UCSF medical students. Using the methods for training clinicians in health coaching by Miller et al,¹¹ a UCSF ophthalmology resident held a 2-hour training session for the medical student volunteers on how to use the health coaching documents. All medical students had received formal health coaching training through their UCSF School of Medicine curriculum prior to this training session. The health coaching documents were available for use at each ophthalmology clinic. During the intervention period, UCSF medical student volunteers performed relevant health coaching with their patients with the oversight of a UCSF ophthalmologist attending physician at each ophthalmology clinic.

For the transportation assistance portion of the intervention, the homeless shelter provided patients with bus tokens for use to and from their referral appointments. At the time of referral, the patients were given a physical paper with their referral appointment date and time as a reminder. The bus tokens were distributed to the patients the morning of their appointment.

Statistical Analysis

The primary outcome was the preintervention follow-up rate compared with the postintervention follow-up rate. Because patients were provided with a prespecified referral appointment date and time within 4 days prior to leaving our clinic, a patient was determined to have followed up if he or she presented to that prespecified appointment. Secondary outcomes included prespecified variables that were hypothesized to be associated with follow-up rates, including having a primary care clinician; current or prior alcohol, tobacco, and recreational drug use; earning a high school diploma; employment status; duration of stay at the homeless shelter; and presenting visual acuity (logMAR), defined as visual acuity with or without correction and with pinhole.

Follow-up rates were calculated using Microsoft Excel (Microsoft Corp). A t test was used to compare the mean preintervention follow-up with the mean postintervention follow-up. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05. P values were not adjusted for multiple analyses. Prespecified variables were compared using the χ² test for categorical variables or the t test for continuous variables to assess their association with follow-up rates. The small sample size did not allow a regression analysis to assess whether 1 or more of the findings were associated with one another. All statistical analysis was conducted with Stata/SE, version 15.0.

Results

A total of 71 patients were included in the study. No patients were excluded. The baseline characteristics of the 71 patients included in this study are summarized in Table 1. The mean (SD) age was 51.9 (12.4) years. There were no differences between the 2 groups, and there were no patients included in both the preintervention and postintervention groups. There were also no differences in the patient diagnoses between the preintervention and postintervention groups (Table 2). The most common vision-threatening diagnosis was cataract judged to be clinically relevant (13 [18.3%]), followed by glaucoma or suspected glaucoma (12 [16.9%]), whereas the most common non–vision-threatening diagnosis was refractive error (61 [85.9%]).

Table 1. Baseline Characteristics of Patients.

Characteristic	Patients, No. (%)
	Preintervention group (n = 37)	Postintervention group (n = 34)
Age, mean (SD), y	53.5 (12.5)	50.2 (12.3)
Female	13 (35.1)	8 (23.5)
Hispanic	1 (2.7)	2 (5.9)
Race
White	12 (32.4)	13 (38.2)
Black or African American	8 (21.6)	8 (23.5)
American Indian or Alaska Native	1 (2.7)	1 (2.9)
Asian	2 (5.4)	0
Native Hawaiian or Pacific Islander	1 (2.7)	0
Multiple	4 (10.8)	1 (2.9)
Othera	2 (5.4)	3 (8.8)
History of type 1 or 2 diabetes	6 (16.2)	12 (35.3)
History of hypertension	18 (48.6)	17 (50.0)
History of eye trauma	4 (10.8)	4 (11.8)
Current or prior tobacco use	26 (70.3)	25 (73.5)
Illegal drug use	11 (29.7)	13 (38.2)
Marijuana use	12 (32.4)	13 (38.2)
Preexisting PCP	17 (45.9)	18 (52.9)
Born in United States	29 (78.4)	26 (76.5)
Days at shelter, mean (SD)	51.8 (66.6)	62.6 (50.7)
Years homeless, mean (SD)	9.7 (8.5)	12.5 (13.0)
High school diploma	24 (64.9)	17 (50.0)
Unemployed	26 (70.3)	25 (73.5)
Uncorrected visual acuity
OD, logMAR	0.21	0.33
Snellen equivalent (SD)	20/32 (0.21)	20/40 (0.29)
OS, logMAR	0.30	0.30
Snellen equivalent (SD)	20/40 (0.28)	20/40 (0.25)
Best-corrected visual acuity
OD, logMAR	0.23	0.23
Snellen equivalent (SD)	20/32 (0.20)	20/32 (0.14)
OS, logMAR	0.20	0.28
Snellen equivalent (SD)	20/32 (0.17)	20/40 (0.22)

Abbreviation: PCP, primary care physician.

^aOther races/ethnicities not specified.

Table 2. Diagnoses of Patients.

Diagnosis	Patients, No. (%)
	Preintervention group (n = 37)	Postintervention group (n = 34)
Vision-threatening diagnoses
Cataract judged as clinically relevant	6 (16.2)	7 (20.6)
Glaucoma or suspected glaucoma	6 (16.2)	6 (17.6)
Afferent pupillary defect	5 (13.5)	2 (5.9)
Visual field defect	2 (5.4)	3 (8.8)
Acute posterior vitreous detachment	3 (8.1)	1 (2.9)
Proliferative diabetic retinopathy	0	2 (5.9)
Nonproliferative diabetic retinopathy	4 (10.8)	2 (5.9)
Age-related macular degeneration	1 (2.7)	0
Macular hole	0	1 (2.9)
Suspicious choroidal lesion	1 (2.7)	0
Proptosis	1 (2.7)	0
Aphakia	1 (2.7)	0
Retinal detachment	1 (2.7)	0
Macular scar	1 (2.7)	0
Non–vision-threatening diagnoses
Refractive error	32 (86.5)	29 (85.3)
Presbyopia	29 (78.4)	25 (73.5)
Myopia	9 (24.3)	10 (29.4)
Astigmatism	5 (13.5)	5 (14.7)
Hyperopia	4 (10.8)	1 (2.9)
Dry eye syndrome	3 (8.1)	0
Epithelial basement membrane dystrophy	1 (2.7)	0
Epiphora	1 (2.7)	0
Strabismus	1 (2.7)	1 (2.9)
Choroidal nevus	1 (2.7)	0
Lattice degeneration	1 (2.7)	0
Ptosis	0	1 (2.9)
Blepharitis	1 (2.7)	0
Epitheliopathy	1 (2.7)	0

The Figure illustrates the referrals and follow-up rates in the preintervention and postintervention groups. A total of 28 patients (39.4%) were referred for free eyeglasses, 14 (19.7%) to the county hospital for advanced care, and 7 (9.9%) to both. There was no difference in referral rates between the 2 groups: 64.9% (24 of 37) in the preintervention group and 73.5% (25 of 34) in the postintervention group (P = .43). Of those referred, the difference in follow-up from the postintervention to preintervention groups was 53.8% (95% CI, 39.8%-67.9%; P < .001). Compared with patients who did not follow up, those who did had a mean difference of 59 more days at the shelter (95% CI, 39-80 days; P = .003). Among patients with a visual acuity of 20/40 or worse in the better-seeing eye, the mean difference between those who did not follow up and those who did was 61% (95% CI, 44%-78%; P = .003). The mean difference in follow-up between patients who were born in the US and patients not born in the US was 89% (95% CI, 79%-98%; P = .02). Of those in the postintervention group, the difference in presentation to follow-up for patients with a high school diploma compared with those without was 59% (95% CI, 37%-81%; P = .001). There were no differences in follow-up rates based on history of preexisting primary care clinician, current or prior tobacco or illegal drug use, and employment status.

Figure. Referrals and Follow-up Rates in the Preintervention and Postintervention Groups.

Discussion

Principal Findings

A health coaching and bus token assistance intervention was associated with improved follow-up rates by at least 39.8% (P < .001); this improvement supports considering implementation of this intervention when developing public assistance programs to meet the eye health needs of the homeless population. This study included patients who were referred to both a partnering nonprofit organization for free eyeglasses and the local county hospital for advanced ophthalmologic care and follow-up. Our health coaching intervention was performed by UCSF medical student volunteers with attending physician oversight, and the bus tokens were provided by the homeless shelter.

Health Coaching

Relatively brief interventions have been shown to trigger significant change, and a single empathetic counseling session may promote long-lasting behavioral change.¹² When discussing health behavioral change with physicians, patients of lower socioeconomic status are more likely than middle-income and higher-income patients to report attempted behavioral change based on physician recommendations.¹³ Despite this fact, patients of lower socioeconomic status receive less information, directions, and partnership-building information from their physicians.¹⁴

Health coaching has been shown to be particularly promising in improving medication adherence, self-monitoring, alcohol and tobacco use cessation, confidence in change, and treatment engagement.¹⁵ Within the field of ophthalmology, health coaching of patients with glaucoma has been associated with improved adherence to glaucoma eye drops,^16,17 motivation,¹⁷ and perceived improvement in eye and general health self-care.¹⁶ Our intervention was likely useful given the empathetic health counseling sessions focused on engaging with patients, evoking talk of change, partnership building, and collaborative planning (eFigures 1-5 in the Supplement). To our knowledge, our health coaching intervention is the first to apply discussions of behavioral change to a broad range of ophthalmologic diseases in the homeless population. We hope our findings increase the use of health coaching in free ophthalmology homeless shelter clinics and that our health coaching documents may be valuable to ophthalmologists who perform health coaching in this and other vulnerable populations.

Transportation Assistance

Financial incentives, including vouchers, can improve patient compliance with medications, medical advice, and medical appointments.¹⁸ Even when patients receive free medical care, the use of some form of financial incentive has been shown to increase compliance. In fact, small financial incentives may be more efficacious at improving compliance compared with other methods, such as peer support; a $5 incentive for homeless individuals with tuberculosis was more effective than peer support in improving adherence to a first follow-up appointment.¹⁹ In our study, the homeless shelter provided patients with bus tokens to be used to and from their referral appointment, which were valued at $6 total ($3 each way). This low-cost intervention should be considered when applying for grants and when developing public assistance programs. Patients received the bus tokens the morning of their appointment if they presented to the front desk of the shelter to ask for the bus tokens. This additional act of distributing the tokens may also have been associated with improved follow-up rates.

Barriers to Accessing Eye Care

We found that patients were more likely to follow up if they had a longer duration of stay at the shelter, visual acuity of 20/40 or worse in the better-seeing eye, and a high school diploma. Prior studies have demonstrated that orientation to clinic services improves access to a primary care clinician for homeless adults residing in an urban setting.²⁰ Although access to an ophthalmologist has not been studied specifically, we hypothesize that patients with a longer duration of stay followed up because they were more involved in the social and clinic services provided at the shelter. Worse far visual acuity²¹ has been associated with an increased likelihood of obtaining vision care in the homeless population. It would be useful to design future interventions with particular attention to patients with these risk factors to maximize follow-up.

We also found that patients born in the US were more likely to follow up compared with patients not born in the US. Most patients seen at our clinic spoke English, and, when necessary, a telephone interpreter was used to interact with patients. Even if patients did speak English, perhaps English was not their primary language of choice, and thus a language barrier was present that affected their ability to completely understand the health coaching and subsequently follow up. We did not inquire about citizenship status but postulate that patients not born in the US may have been hesitant to seek health care if they were recent immigrants²² or not as comfortable using the health care system, thereby affecting their ability to follow up. Future studies that assess the barriers to presentation faced by patients not born in the US are warranted.

Limitations

This study has some limitations. The prospective cohort design allows for possible confounders that may threaten our ability to make an association between the intervention and improving follow-up rates. Furthermore, our study design may introduce unknown confounders that would not be the case if we were performing a randomized clinical trial. Despite this fact, we believe that randomizing health coaching and transportation assistance may have questionable ethics in such a vulnerable population, especially if bus tokens were provided to only one group when enough bus tokens available for all patients. Furthermore, the preintervention and postintervention groups were not different in demographic characteristics or diagnoses as outlined in Table 1 and Table 2, which increases the likelihood that the 2 groups were comparable.

Given that our clinic may improve in both operations and efficiency over time, the improvement in follow-up rates may hypothesized to be associated with the improvement in operations and efficiency as opposed to the intervention itself. That being said, the desire to volunteer in our clinic is high, and thus medical students typically volunteer only once in an academic year to maximize medical student exposure to our clinic. This situation limits the likelihood that this potential confounder (ie, improving health coaching skills over time) was associated with follow-up rates because most medical students at our clinic were performing ophthalmology-related health coaching for the first time.

Because both health coaching and transportation assistance interventions were initiated at the same time in our study, it is difficult to know whether one intervention was associated with follow-up rates more than the other. We took the approach that it was best to study the ideal scenario initially (ie, both a health coaching and transportation assistance intervention) before isolating each individual intervention. Future studies would be useful to identify the association of health coaching vs transportation assistance with follow-up rates. Our findings also demonstrate wide 95% CIs, and with several outcomes and no adjustment of P values for multiple analyses, it is likely that only outcomes with a significance level of P < .001 should be considered as potentially ruling out chance findings.

Finally, the study sample is relatively small and our study includes a single homeless shelter, which may make our findings less generalizable to the homeless population as a whole or at least to those outside of San Francisco. Future studies with larger sample sizes that include patients from multiple homeless shelters and cities would be helpful.

Conclusions

A health coaching and bus token assistance intervention at a free ophthalmology homeless shelter clinic was associated with improved follow-up rates. This improvement supports considering implementation of this intervention when developing public assistance programs to meet the eye health needs of the homeless population.

Supplement.

eFigure 1. Homeless Shelter Clinic – Motivational Interviewing Template

eFigure 2. Homeless Shelter Clinic – Diabetic Retinopathy Coaching

eFigure 3. Homeless Shelter Clinic – Glaucoma/Glaucoma Suspect Coaching

eFigure 4. Homeless Shelter Clinic – Coaching on Cataracts

eFigure 5. Homeless Shelter Clinic – Coaching on Refractive Errors