Description of the Evidence
We identified 3463 potentially relevant citations. Fifty-three publications underwent full-text review, of which 41 publications were excluded (see Appendix B). A total of 12 publications were identified as meeting inclusion criteria (see Figure 1). Of these 12 studies, 3 were Randomized Controlled Trials (RCTs), 1 was a controlled clinical trial, and 8 were observational studies. One study was conducted in England, the rest were conducted in the United States. Details of included studies are presented in the Evidence Table (see Table 1).
All controlled trials were judged as having at least 1 domain as being at high risk of bias, but this was because it is not possible to blind participants and personnel to the intervention (see Appendix Table 1). Two of the four trials were judged as being at low risk of bias for all other domains.(17, 18) Two of the three controlled observational studies were judged as being at low risk of bias in all domains.(19, 20) See Appendix Table 2. All pre-post studies were limited by one or more of small sample size or loss to follow-up (see Appendix Table 3)
Associations with missed clinic visits and Emergency Department visits
Nine studies reported on made or missed clinic visits. (17, 19, 21-27) Three of these studies were controlled trials,(17, 18, 23) 1 was a controlled before-and-after study,(19) and the rest were pre/post studies.(21, 22, 25-27) The enrolled populations were a heterogenous mix of patients with specific conditions (Human Immunodeficiency Virus [HIV], or in need of cervical cytology follow-up, or prenatal care) or patients who were poor, or both. Six studies used as their transportation intervention taxicab or rideshare services (Lyft or Uber),(18, 19, 23, 25-27) 2 used van rides or bus tickets,(17, 21) and in one study the intervention was advice and assistance with transportation.(22) One study reported its outcomes as means.(21) Seven studies measured utilization in terms of the proportion of clinic appointments, either missed or kept. The random effects pooled estimate of these 7 studies on missed appointments was an odds ratio of 0.63 (95% CI [0.48, 0.83]) in favor of the intervention (see Figure 2). The I2 statistic was 76%. Pooled results from only the 3 controlled trials (OR=0.71, 95% CI [0.44, 1.14] were not statistically different from pooled results of the remaining 4 observational studies (OR=0.58, 95% CI [0.41, 0.82]. There was no statistical evidence of publication bias (Eggar’s test p=0.24, Begg’s test p=0.38). The two studies that could not be included in the pooled analysis showed no statistically significant difference in 1) the self-report of the number of HIV visits before and during an intervention that consisted of giving free medical van transportation to and from the clinic for HIV positive women who had been poorly compliant to keeping medical appointments, although the self-reported missed clinic appointments did decline (21); and 2) the number of patients making same-day cancellations for Magnetic Resonance Imaging (MRI) appointments; in this study only 2% of patients used an offer of a ride-sharing service, making any difference in cancellations due to the intervention very hard to detect.
Two additional studies reported utilization in terms of Emergency Department (ED) visits. The first was a controlled clinical trial that offered 394 Medicaid beneficiaries free rideshare services to come for scheduled clinic visits and compared this to 392 other clinic patients as control.(23) Almost all participants were Black females, the mean age was 46 years. The study found no differences in ED visits at 7 or 30 days; however only 20% of eligible patients in the intervention group actually used the rideshare service. The second study assessed the effect of the use of transportation brokerage services for non-emergency medical transportation in adult Medicaid beneficiaries with diabetes, and found that the use of brokerage services did not significantly reduce the probability of ED visits for diabetes.(20)
Associations with healthcare resource utilization and costs
Six studies reported costs outcomes, one was an RCT,(28) one was a controlled before-and-after study,(20) three were pre/post studies,(20, 26, 27) and one was the controlled clinical trial of rideshare services for Medicaid patients mentioned in the prior section.(23) In the latter study, the mean cost of the rideshare per patient who consented was $14.00; in one pre/post study of a ride share for patients scheduled for MRI visits the mean cost was $17.92,(26) and in another pre-post study the mean cost for sickle cell patients was $67.(27) In a study of 4 general practices in England,(28) intervention practices were given an additional 1500 pounds sterling plus technical assistance to improve over the next 6 months their system for making appointments and helping patients with transportation barriers, which included links to community transport, making appointment times convenient for existing bus schedules, providing charging stations for electric scooters, and at one clinic creating an appointment slot for patients requiring taxi services. Two of the three intervention clinics spent £2262 and £930 of additional money on developing their intervention. Staff time devoted to the intervention was estimated at between £112 and £2651. In a database analysis of the effect of transportation brokerage services on Medicaid expenditures in Georgia and Kentucky, the use of the brokerage service was estimated at decreasing the monthly per person expenditure (inpatient plus outpatient) by about $18 for adults with diabetes and for children with asthma, despite increases in use of health services and prescription drugs (for diabetic patients).(20) All of these interventions were for non-emergency medical transportation to clinic. In one study that included non-medical transportation, discussed below, the cost was $500 per month per patient.(24) One additional evaluation of the use of a ride-sharing program was reported in a blog but could not be included as evidence because it did not report information on the sample sizes, which precluded statistical testing of differences between groups.(29)
Associations with health outcomes
Two studies assessed health outcomes. One of these studies was the assessment of transportation brokerage services, discussed above.(20) In this study, the use of brokerage services for adult patients with diabetes decreased the probability of having an ambulatory care sensitive condition admission by a statistically significant 0.1 percent a month, whereas for children with asthma there were no such benefits seen. In the only study that assessed an intervention that included non-medical transportation, 150 patients of an urban academic medical center who were over the age of 60 (mean age = 72), had a chronic disease, and reported transportation barriers were offered unlimited ridesharing for 3 months.(24) The mean number of rides during this time period was 69, and the mean cost per subject was $500 per month. Patients also received a device to measure step counts, these did not significantly change from before the intervention to during the intervention. A post-intervention-only assessment of health status, limited by a 31% non-response rate, was reported as showing 92% of subjects having improved quality-of-daily life, but no data are provided in the publication. Also, the published paper reports no significant differences in pre- and post-intervention measures of the Satisfaction with Life survey and the Geriatric Depression Scale; however again no data are reported.
Certainty of Evidence
We judged the certainty of evidence that providing free Non-Emergency Medical Transportation is associated with a decrease in missed clinic appointments is High, based on the reasonably consistent results seen in controlled trials and observational studies and the strong mechanism that the intervention of providing free transportation might mitigate transportation barriers to care. We judged all other outcomes as being Low or Very Low certainty evidence, due to limitations in study design, and/or execution (see Table 2).