Our study on ridesharing by a managed care transportation broker found no change in ride quality compared with traditional nonemergency medical transportation, but differences were observed for access-to-care measures.
Objectives: Some managed care companies are testing rideshare services as an approach to providing transportation to health care for Medicaid enrollees. The objective of this study was to assess whether more rideshare transportation to health care was associated with improved self-reported ride experiences and fewer late/failed passenger pickups for Medicaid enrollees.
Study Design: We surveyed a random sample of Medicaid enrollees in a northwestern US state on their experiences with nonemergency medical transportation (NEMT) in the past year. We linked survey responses to administrative data on NEMT utilization from the state’s transportation broker to obtain an objective measure of rideshare utilization.
Methods: We used bivariate tests and multivariable logistic regressions to examine associations between enrollee perspectives on the quality of and access to health care and rideshare use, defined as none, some, or many NEMT trips through rideshare services.
Results: More than 35% of respondents received NEMT from rideshare services at least once. Perceptions of the ride experience, driver, and vehicle did not differ based on the proportion of rideshare trips received. Having more rideshare trips was associated with reporting late and failed pickups. In multivariable regression, the statistical significance held for failed pickups. Sensitivity analyses showed similar results.
Conclusions: This study suggests that rideshare to health care programs can meet similar goals of quality compared with traditional NEMT services but may have implications for health care access for Medicaid enrollees. Future evaluations need to include the perspectives of enrollees and explore potential differences among different Medicaid subpopulations.
Am J Manag Care. 2020;26(9):e276-e281. https://doi.org/10.37765/ajmc.2020.88492
Ridesharing is an understudied service delivery method deployed by managed care organizations for nonemergency medical transportation (NEMT) for Medicaid enrollees. Our study found that:
One substantial barrier to accessing health care is the lack of consistent transportation.1-3 Transportation challenges to health care are disproportionately experienced by individuals who are low-income, older adults, non-White, women, and less educated, many of whom are Medicaid enrollees.4 An estimated 25% to 55% of Medicaid enrollees missed, arrived late to, or did not try to go to a health care appointment because of transportation issues.5-7 Adults and children who missed medical appointments due to transportation issues had extensive comorbidities and a significantly higher prevalence of health conditions compared with those who missed care for other reasons.4 Inconsistent care due to transportation challenges can negatively affect health and increase preventable emergency department visits, particularly for individuals with chronic conditions.8,9
Nonemergency medical transportation (NEMT) is a mandatory benefit provided through Medicaid to travel to and from health care appointments.10 NEMT is provided by nonmedical personnel through a range of vehicles, including wheelchair-accessible vans.8 The most common model of NEMT administration is through a third-party broker and/or managed care organization (MCO), in which the broker or MCO receives capitated payments by the state to broker, coordinate, manage, and/or administer NEMT.11,12
Despite historical support for NEMT, this program is particularly susceptible to service and funding cuts. Some states have used Section 1115 waivers to exclude Medicaid expansion populations from NEMT benefits.13 CMS drafted a proposed rule that would allow states more flexibility in providing NEMT,14 and the current administration’s budget proposals for fiscal years 2019 and 2020 proposed making NEMT an optional benefit.15 The rule could reduce patient access to NEMT and, subsequently, to needed medical services. This tension reflects the demands inherent in the Triple Aim of health care—cost, quality, and access16: It is extremely challenging to reduce NEMT costs without affecting quality and access or to improve quality without raising costs. It is important to consider how quality and access are affected by new cost-reducing models in NEMT.
Rideshare-based medical transportation (RMT) is a program in which NEMT is provided by drivers using their personal vehicles, similar to rideshare companies like Uber and Lyft. RMT can be combined with traditional NEMT to also provide rides via prearranged vans or taxis. RMT is appealing because it may provide more flexibility for passengers; is well suited for last-minute rides, like hospital discharges; and may reduce wait times and cost.8 Additionally, RMT can better track rides and collect data, potentially addressing quality, fraud, waste, and abuse.8
Conversely, critiques of RMT include lower pay for drivers, lack of access in rural areas, inadequate driver screening, and safety issues for drivers and riders.17,18 Because the NEMT population is more likely to be low-income, older adults, and individuals with disabilities,4 additional specialized training for drivers is needed. Ridesharing companies not specific to NEMT have also faced criticism and even legal action for a lack of accessibility for individuals with disabilities.19,20 Rideshare vehicles are typically not equipped to provide rides to those using wheelchairs/scooters.
Some early evaluations of RMT implementation suggest mixed results regarding health care access and service quality. Preliminary results from pilot tests suggested that RMT leads to decreased missed appointments21 and high safety and satisfaction ratings (> 95%).22 However, recent studies contradict these pilots. A randomized controlled trial found no significant effect of RMT on missed appointments.23 A recent analysis of Twitter posts suggested that passengers had overwhelmingly negative experiences with rideshare drivers.24 Based on Andersen’s conceptual model of health care access,25 perceptions of the ride experience, driver, and vehicle appropriateness may affect individuals’ willingness to use NEMT. It is important to understand consumers’ perceptions of RMT because perceptions likely influence NEMT service utilization and overall access to health care.
In this paper, our aim was to determine whether RMT was associated with users’ perceptions of quality and access to care. In a northwestern state in the United States, the Medicaid transportation broker included rideshare services as part of its NEMT. Unlike typical rideshare services, there was no passenger-side smartphone app; the rides were requested on behalf of the passenger directly from the NEMT broker (ie, passengers may not have known whether or not they received a rideshare driver). In this way, the passenger did not change their usual practice for scheduling rides. Enrollees’ trips were simply assigned to RMT or traditional NEMT based on availability of rideshare drivers and origin/destination. Thus, factors that affect a patient’s willingness and ability to use ridesharing and the associated smartphone app did not confound our analysis.
We assessed whether having a greater proportion of rides from rideshare drivers was associated with greater satisfaction and better access to care. To our knowledge, no other studies have examined the association between rideshare use and passenger perspectives through a systematic independent evaluation controlling for multiple potential confounding factors. We examined the following research questions: (1) Was receiving more rides through RMT associated with a higher quality of service (vehicle appropriateness, safety, and cleanliness)? and (2) Was receiving more rides through RMT associated with a lower likelihood of reporting late and/or failed pickups?
We obtained administrative data on all NEMT rides from the NEMT broker’s administrative database for the years 2016 to 2018. To assess the experiences of NEMT users, we developed a survey that was distributed to a stratified random sample of individuals eligible for Medicaid NEMT within the state. The questionnaire included 29 questions that covered transportation utilization, access, experiences, satisfaction, and demographics. Many survey questions were drawn from standard Consumer Assessment of Healthcare Providers and Systems surveys, a national standardized survey tool developed by the Agency for Healthcare Research and Quality.26
We employed proportionate stratified random sampling to ensure that perspectives from a variety of groups were included in the survey. We stratified sampling based on having a legal guardian (for those younger than 18 years and those with a developmental disability), prior NEMT utilization, and county of residence. Based on a power analysis, we estimated that a sample of 1101 was needed. Surveys were distributed to the selected enrollees through the mail at least twice. Up to 3 telephone reminders were made, with the option to complete the survey over the telephone. If requested, the survey and accompanying materials were available in Spanish.
Overall, the response rate was 28.3%, consistent with other Medicaid mail surveys.27,28 Compared with nonresponders, responders were older (mean age, 43 vs 35 years), took more NEMT trips (median number of trips, 38 vs 21), and had a lower proportion with a legal guardian (23.3% vs 46.0%) (eAppendix Table 1 [eAppendix available at ajmc.com]). The differences were not a threat to internal validity because we were primarily interested in responses for enrollees who had taken NEMT.
Dependent variables.The dependent variables were responses to 7 questions related to transportation quality and access (eAppendix Table 2). For most questions, responses were dichotomized into usually/always and never/sometimes. For the failed pickup question, the response options were dichotomized into sometimes/usually/always and never, because a failed pickup is a more extreme event that may have a large impact on access to care and acceptability of the service.
Independent variable of interest.The transportation broker provided a data set of one-way trip details for each respondent. We selected all trips made within 1 year of the month that the survey was received, reflecting the time frame of the survey question wording. For each respondent, we calculated the proportion of total NEMT trips that were provided by a rideshare driver. A categorical variable was coded as “many” for having at least 50% of trips with a rideshare driver, “some” for having 1% to 49%, and “none” for having no rideshare trips. Nonrideshare trips were provided through ambulatory vehicles (sedans), wheelchair-accessible vans, public transportation, and mileage reimbursement. However, 95% of rides came from ambulatory vehicles or wheelchair-accessible vans.
Covariates.We included several covariates to control for confounding, including age, sex, race/ethnicity, frequency of health care visits, total NEMT trips (one-way), and trip distance. Through interviews with advocates, we learned that RMT was not working well for the population with developmental disability (DD), so we included a dummy variable for that group based on administrative data from the state (unpublished data). Finally, we included a dummy variable on mobility disability, which was defined as needing any type of specialized equipment or services to travel outside the home (eg, assistance from another person, interpreter, manual wheelchair).
We computed descriptive statistics (frequencies and counts) for all items. We examined bivariate correlations between the receipt of RMT (none, some, and many) and the dependent variables using Fisher’s exact tests. We conducted subanalyses to compare results for those with and without mobility disability. The sample sizes for the other covariates were too small for meaningful subanalyses.
For variables that were significant in bivariate analyses, we used multivariable logistic regressions to determine the odds of rating the outcome variables positively while controlling for confounders. The variable for on-time pickup was reverse-coded for easier interpretation of results and will be referred to as “late pickup.” To correct bias from the small sample size,29 we bootstrapped the standard errors with 500 repetitions to increase confidence in the statistical significance of our findings. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit test.
To determine if our choice of timing affected the results, we tested both the bivariate association and full regression models utilizing a 6-month time period before the survey was received instead of a 1-year time period.
Table 1 shows the sample characteristics. The majority of respondents were aged between 18 and 64 years (69.2%), female (61.2%), and White (71.1%). More than 18% of the respondents were on the DD waiver, and 44% had a mobility disability. Among all respondents, 18.0% had many RMT trips, 17.3% had some RMT trips, and 64.7% had no RMT trips. A large majority of respondents had positive ratings for the ride quality questions. Just over half of respondents (50.2%) reported having a failed pickup often.
Table 2 shows the bivariate associations between the proportion of rideshare trips and each of the questions on quality and access. Ride quality measures did not differ across the proportion of RMT trips provided. In contrast, having more rideshare trips was associated with reporting late pickup (P = .012) and failed pickup (P < .001). For late pickup, 47.8% of individuals with many RMT trips reported they often had a late pickup compared with 27.3% for those with some RMT trips and 25.0% for those with no RMT trips. For failed pickup, 65.2% of those with many RMT trips reported they often had a failed pickup and 67.6% of those with some RMT trips agreed compared with 39.7% of those with no RMT trips. The sensitivity analyses using NEMT data from the last 6 months instead of from the last year yielded similar results (eAppendix Table 3). Results and tables for the bivariate analysis by mobility disability are shown in eAppendix Table 4 and reveal that, for this subgroup, responses for some of the dependent variables differed by proportion of RMT.
Multivariable Logistic Regression Analyses
Table 3 shows the results for 2 logistic regression models for the late pickup and failed pickup outcomes. All models passed the goodness-of-fit tests. In the first model, having some or many RMT trips compared with no RMT trips was associated with increased odds of reporting a late pickup, but the association was no longer statistically significant. There was a small (odds ratio [OR], 0.970; 95% CI, 0.949-0.992) but significant decrease in the odds of late pickup for every 1-year increase in age. There were no significant associations between the other variables and reporting late pickup.
In the second model, having some RMT trips increased the odds of failed pickup by a factor of 3.44 compared with those with no RMT trips, and having many RMT trips increased the odds of failed pickup by a factor of 3.06. There was also a small decrease in the odds of failed pickup for every year increase in age (OR, 0.979; 95% CI, 0.959-1.000). In our sensitivity analyses (eAppendix Table 3), both models with a shorter time window (6 months instead of 1 year prior to the survey) had results similar to the main models. One exception was that in the shorter time window, having some RMT trips was no longer statistically significant in the failed pickup model.
This study sought to identify whether having more RMT trips was associated with better quality ratings of NEMT and improved access to care for a sample of Medicaid enrollees. We found that having more RMT trips was not associated with reported quality of NEMT in terms of appropriateness, safety of the vehicle, or driver courteousness. In contrast, having more rideshare trips was associated with reporting late and failed pickups of NEMT riders. The statistical significance of the associations held in multivariable analysis for reporting failed pickup.
The appropriateness of the rideshare vehicle, safety, and driver attitudes are major concerns for use of RMT.17,18 Some preliminary results from an RMT pilot in New York City and California indicated a high level of safety and satisfaction, yet there was no control group for comparison and only pilot results have been reported.22 In this study, we were able to compare groups with different levels of rideshare trips (none, some, and many). Across groups, the ratings for ride quality were generally high. We did not find significant differences in responses to the ride quality measures between those with some or many trips with RMT compared with those with traditional NEMT only; this finding can be interpreted both positively and negatively. On one hand, RMT use had similar ratings of driver and ride quality. If maintaining quality was the goal, it would be met. On the other hand, RMT may be less attractive if improving quality was an important outcome for a state’s Medicaid program.
As the proportion of RMT increased, the likelihood of late and failed pickup of NEMT riders also increased: Those who received RMT more frequently were more likely to report late pickup or failed pickup compared with those who received RMT less frequently or used only traditional NEMT. These findings suggest that access to health care may be affected by RMT trips; more research is needed to determine why these differences exist.
RMT may affect health care access for various reasons. One potential explanation is that rideshare drivers may not receive adequate training and may not face consequences for a failed or late pickup. Rideshare dispatch technology problems can lead to access issues. Additional measures may be needed when providing RMT to enrollees with mobility disability, such as building a larger pool of accessible vehicles. Additionally, lower access may be related to cost-reduction strategies used by the NEMT broker. We learned that the NEMT broker in the state under study had a lower bid for its contract and drivers were generally dissatisfied by the pay rate (unpublished data). As costs are reduced, quality or access to care is often affected.16 The evaluation was completed during the second year of the broker’s contract. It is possible that access may improve over a longer period of time when both drivers and enrollees are more familiar with RMT.
Transportation brokers have a plethora of data on shared ride logistics like pickup time and location. However, it is important to understand patients’ perceived access (in this case, late or failed pickup) because these perceptions could be reasons for why consumers may or may not continue to use transportation services. In the course of our evaluation, we also found that the NEMT broker could track late pickups but not failed pickups. Although drivers could report consumers who do not show, consumers may be underreporting when drivers do not show. In a previous evaluation, we found that consumers dissatisfied with an NEMT service sometimes do not bother calling the broker but focus on finding alternative transportation. Understanding the experiences of patients with new services like RMT is critical to tease out patient satisfaction and the likelihood of repeated use. Lower satisfaction with NEMT threatens consistent attendance of medical appointments by the enrollees who are in the most need of care.25 One group in particular to consider is individuals with mobility disabilities. In subanalyses, individuals with mobility disabilities with more RMT trips had significantly lower ratings for some of the quality and access measures than those with no RMT. This may reflect problems with vehicle accessibility, which have also been cited in lawsuits against rideshare companies in Chicago and parts of California.19,20 Future research should evaluate RMT for other transportation-disadvantaged subgroups. This research would be useful for policy makers and other stakeholders in understanding access and experiences with RMT.
Strengths and Limitations
Our study had several strengths. This paper was novel because we linked survey data on consumer experiences to administrative records of health care trips for Medicaid recipients. Our measure for the proportion of RMT was not biased by patient behavioral factors but focused the analysis on the rideshare trip. Finally, the research was based on an independent evaluation of NEMT that was not associated with any rideshare company.
Our study also had some limitations. Like many Medicaid surveys, our response rate was low, at 28.3%.27 We were unable to reach many enrollees because changes of residence and phone number are common among the Medicaid population.28,30 Our comparison of administrative data for responders and nonresponders indicated significant differences, which affects the generalizability of our results. Respondents did not answer all the questions, which reduced the analytic sample for some of the analyses. Clients may have become aware that the driver was not from a traditional transportation company. Because our analysis was cross-sectional and lacked any causal approaches to address omitted variable bias, the results can only reflect associations between RMT and quality and access.
Rideshare companies continue to expand into transportation to health care appointments. As more states incorporate ridesharing into their NEMT delivery models, it is critical to evaluate patient experiences and perceptions. Although RMT may be attractive for its efficiency and lower costs, additional research is needed in diverse settings and varied populations to understand how RMT differs from traditional NEMT and how RMT affects quality and access to care.
The authors thank Helen Rottier and Amy Hofstra for their contributions to this paper by editing and adding pertinent literature when needed.
Author Affiliations: Department of Disability and Human Development, University of Illinois at Chicago (YE, RO, CC, MM), Chicago, IL.
Source of Funding: This evaluation was funded by the state department that houses the Medicaid agency in the state where the research was conducted.
Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.
Authorship Information: Concept and design (YE, RO, CC); acquisition of data (RO, CC); analysis and interpretation of data (YE, RO, CC, MM); drafting of the manuscript (YE, RO, CC, MM); critical revision of the manuscript for important intellectual content (YE, RO, CC); statistical analysis (YE, RO, CC); administrative, technical, or logistic support (YE, MM); and supervision (YE, RO).
Address Correspondence to: Yochai Eisenberg, PhD, Department of Disability and Human Development, University of Illinois at Chicago, 1640 W Roosevelt Rd, MC 626, Chicago, IL 60608. Email: email@example.com.
1. Bellamy GR, Stone K, Richardson SK, Goldsteen RL. Getting from here to there: evaluating West Virginia’s rural nonemergency medical transportation program. J Rural Health. 2003;19(suppl 5):397-406. doi:10.1111/j.1748-0361.2003.tb01060.x
2. Norris TL, Aiken M. Personal access to health care: a concept analysis. Public Health Nurs. 2006;23(1):59-66. doi:10.1111/j.0737-1209.2006.230109.x
3. Syed ST, Gerber BS, Sharp LK. Traveling towards disease: transportation barriers to health care access. J Community Health. 2013;38(5):976-993. doi:10.1007/s10900-013-9681-1
4. Wallace R, Hughes-Cromwick P, Mull H, Khasnabis S. Access to health care and nonemergency medical transportation: two missing links. Transp Res Rec. 2005;1924(1):76-84. doi:10.1177/0361198105192400110
5. Silver D, Blustein J, Weitzman BC. Transportation to clinic: findings from a pilot clinic-based survey of low-income suburbanites. J Immigr Minor Health. 2012;14(2):350-355. doi:10.1007/s10903-010-9410-0
6. Yang S, Zarr RL, Kass-Hout TA, Kourosh A, Kelly NR. Transportation barriers to accessing health care for urban children. J Health Care Poor Underserved. 2006;17(4):928-943. doi:10.1353/hpu.2006.0137
7. Hobson J, Quiroz-Martinez J. Roadblocks to Health: Transportation Barriers to Healthy Communities. Transportation and Land Use Coalition; 2002.
8. Powers BW, Rinefort S, Jain SH. Nonemergency medical transportation: delivering care in the era of Lyft and Uber. JAMA. 2016;316(9):921-922. doi:10.1001/jama.2016.9970
9. Starbird LE, DiMaina C, Sun CA, Han HR. A systematic review of interventions to minimize
transportation barriers among people with chronic diseases. J Community Health. 2019;44(2):400-411. doi:10.1007/s10900-018-0572-3
10. Rosenbaum S, Lopez N, Morris MJ, Simon M. Medicaid’s medical transportation assurance: origins, evolution, current trends, and implications for health reform. Policy Brief George Wash Univ Cent Health Serv Res Policy. 2009;1-24.
11. Musumeci M, Rudowitz R. Medicaid non-emergency medical transportation: overview and key issues in Medicaid expansion waivers. Kaiser Family Foundation. February 24, 2016. Accessed January 7, 2020. https://www.kff.org/medicaid/issue-brief/medicaid-non-emergency-medical-transportation-overview-and-key-issues-in-medicaid-expansion-waivers/
12. Adelberg M, Simon M. Non-emergency medical transportation: will reshaping Medicaid sacrifice an important benefit? Health Affairs. September 20, 2017. Accessed January 7, 2020.
13. Musumeci M, Rudowitz R. The ACA and Medicaid expansion waivers. Kaiser Family Foundation. November 20, 2015. Accessed January 7, 2020. https://www.kff.org/report-section/the-aca-and-medicaid-expansion-waivers-issue-brief/
14. Dickson V. CMS is developing a rule that could curtail Medicaid transportation access. Modern Healthcare. November 7, 2018. Accessed January 7, 2020. https://www.modernhealthcare.com/article/20181107/NEWS/181109932/cms-is-developing-a-rule-that-could-curtail-medicaid-transportation-access
15. Parfaite-Claude D. CMS possibly delaying publication of NEMT rule. American Network of Community Options and Resources (ANCOR). July 1, 2019. Accessed January 8, 2020. https://www.ancor.org/newsroom/news/cms-possibly-delaying-publication-nemt-rule
16. Berwick DM, Nolan TW, Whittington J. The triple aim: care, health, and cost. Health Aff (Millwood). 2008;27(3):759-769. doi:10.1377/hlthaff.27.3.759
17. Ganuza A, Davis R. Disruptive innovation in Medicaid non-emergency transportation. Center for Health Care Strategies, Inc. February 2017. Accessed January 7, 2020. https://www.chcs.org/media/NEMT-Issue-Brief-022717.pdf
18. Chaudhry B, Yasar AUH, El-Amine S, Shakshuki E. Passenger safety in ride-sharing services. Procedia Comput Sci. 2018;130:1044-1050. doi:10.1016/j.procs.2018.04.146
19. Said C. Lyft sued by disability advocates over wheelchair access. San Francisco Chronicle. March 13, 2018. Accessed January 9, 2020. https://www.sfchronicle.com/business/article/Lyft-sued-by-disabled-advocates-over-lack-of-12750101.php
20. Lien T. California lawsuits accuse Uber and Lyft of discriminating against wheelchair users. Los Angeles Times. March 15, 2018. Accessed January 7, 2020. https://www.latimes.com/business/technology/la-fi-tn-uber-lyft-wheelchair-20180315-story.html
21. Chaiyachati KH, Hubbard RA, Yeager A, et al. Rideshare-based medical transportation for Medicaid patients and primary care show rates: a difference-in-difference analysis of a pilot program. J Gen Intern Med. 2018;33(6):863-868. doi:10.1007/s11606-018-4306-0
22. Powers B, Rinefort S, Jain SH. Shifting non-emergency medical transportation to Lyft improves patient experience and lowers costs. Health Affairs. September 13, 2018. Accessed January 7, 2020. https://www.healthaffairs.org/do/10.1377/hblog20180907.685440/full/
23. Chaiyachati KH, Hubbard RA, Yeager A, et al. Association of rideshare-based transportation services and missed primary care appointments. JAMA Intern Med. 2018;178(3):383-389. doi:10.1001/jamainternmed.2017.8336
24. Pratt AN, Morris EA, Zhou Y, Khan S, Chowdhury M. What do riders tweet about the people that they meet? analyzing online commentary about UberPool and Lyft Shared/Lyft Line. Transp Res Part F Traffic Psychol Behav. 2019;62:459-472. doi:10.1016/j.trf.2019.01.015
25. Andersen RM. Revisiting the behavioral model and access to medical care: does it matter? J Health Soc Behav. 1995;36(1):1-10.
26. About CAHPS. Agency for Healthcare Research and Quality. Accessed January 8, 2020.
27. Gallagher PM, Fowler FJ Jr, Stringfellow VL. The nature of nonresponse in a Medicaid survey: causes and consequences. J Off Stat. 2005;21(1):73-87.
28. Owen R, Heller T, Bowers A. Health services appraisal and the transition to Medicaid Managed Care from fee for service. Disabil Health J. 2016;9(2):239-247. doi:10.1016/j.dhjo.2015.10.004
29. Nemes S, Jonasson JM, Genell A, Steineck G. Bias in odds ratios by logistic regression modelling and sample size. BMC Med Res Methodol. 2009;9(1):56. doi:10.1186/1471-2288-9-56
30. Fredrickson DD, Jones TL, Molgaard CA, et al. Optimal design features for surveying low-income populations. J Health Care Poor Underserved. 2005;16(4):677-690. doi:10.1353/hpu.2005.0096