Cost-Benefit Analysis of Providing Non-Emergency Medical Transportation (2005)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Chapter 9: Conclusions While there is substantial uncertainty in the precise computations at the condition level, a strong case can be made that improved access to NEMT for transportation- disadvantaged persons is cost-effective in terms of better healthcare. In some cases, this cost-effectiveness translates directly into decreases in healthcare costs that exceed the added transportation costs. In other cases, longer life expectancy or improved quality of life for those suffering from the studied conditions justify the added costs of improved access to NEMT cost-effectiveness. The latter is not a “soft finding.” To be cost-effective under the well-accepted QALY method, added costs to extend a healthy life must be below a reasonable cost standard, and such is the case for all twelve of the analyses detailed in Chapter 7. Conversely, many healthcare interventions, when carefully analyzed, result in added quality-adjusted life years costing hundreds of thousands of dollars, and thus are clearly cost-ineffective. In this final chapter, we review our principal findings, discuss limitations of the study, and suggest promising avenues for future research on this novel and important topic. 9.1 Principal Findings For the twelve conditions that were analyzed, improved access to NEMT is cost- effective for the transportation-disadvantaged population that misses medical care due to transportation barriers. For four conditions – prenatal care, asthma, congestive heart failure, and diabetes – improved transportation produces net cost savings. For six other conditions – influenza vaccinations, dental care, chronic obstructive pulmonary disease, hypertension, depression/mental health, and end-stage renal disease – improvements in life expectancy or quality of life are easily large enough to justify increased net costs. For the two remaining conditions – breast cancer screening and colorectal cancer screening – net transportation costs fall comfortably below the conventionally accepted $50,000 QALY standard (see Table 9-1). Table 9-1: Summary of Condition-Specific Cost-Effectiveness Results Condition Type Cost per QALY Result Influenza Vaccinations Preventive $31 / QALY Highly Cost-Effective Prenatal Care Preventive $367 Cost Saving Cost Saving Breast Cancer Screening Preventive $34,176 / QALY Moderately Cost-Effective Colorectal Cancer Screening Preventive $22,735 / QALY Moderately Cost-Effective Dental Care Preventive $590 / QALY Highly Cost-Effective Asthma Chronic $333 Cost Saving Cost Saving Heart Disease (Congestive Heart Failure, CHF) Chronic $2,743 Cost Saving Cost Saving Chronic Obstructive Pulmonary Disease (COPD) Chronic $1,272 / QALY Highly Cost-Effective Hypertension (HTN) Chronic $6 / QALY Highly Cost-Effective Diabetes Chronic $927 Cost Saving Cost Saving Depression / Mental Health Chronic $675 / QALY Highly Cost-Effective End-Stage Renal Disease (ESRD) Chronic $410 / QALY Highly Cost-Effective Final Report 89

In the sections that follow, we describe other important results as they relate to the objectives of the study and to the spreadsheet tool that was developed to foster local and regional analyses. 9.1.1 The Transportation-Disadvantaged Population and Access to Healthcare Defining the transportation-disadvantaged population is a complex and contentious problem, made more difficult when trying to identify members of this population who miss non-emergency medical care due to transportation shortcomings – this project’s study population. Similarly, while it may seem clear who requires healthcare, and thus who misses needed care for one reason or another, such a determination is affected by factors such as “persistence” – not needing care in one period but becoming ill in the next, newly emerging disease management strategies that place a premium on careful disease monitoring, and more effective health promotion and disease prevention activities. Making use of available, nationally representative healthcare datasets (NHIS and MEPS), we found 3.6 million people in a given year who are both transportation disadvantaged and miss non-emergency medical care due to a lack of transportation. Thus, these 3.6 million persons became the target population for this study. Although this target population estimate of 3.6 million Americans is an important study finding, our analysis revealed that a larger population is at risk of missing care due to transportation barriers, thus it is conservative and should be seen as a lower bound estimate. There is response bias inherent in these studies that lowers the estimate, and some populations are totally ignored in the data. This bias will tend to lower the estimate than if the studies truly represent the entire U.S. population. Furthermore, because people can fall into and out of transportation-disadvantaged status over time, as well as change healthcare status (e.g., healthy or not, have insurance or not), our results suggest that more Americans are at risk of missing non- emergency care due to a lack of transportation, but that only some of this at-risk population does miss in a given year. This phenomenon is shown in Figure ES-1. Moreover, this number can and should be benchmarked against the number of people suffering from a given condition, and the current resources that are devoted to ameliorating this suffering – or the number of people that stimulates a broad-based public health campaign. For example, the target population is approximately 40% larger than the entire end-stage renal disease group in the U.S. In sum, even with a “small” transportation-disadvantaged population, the healthcare dollars at risk are substantial, and the impact on affected lives is great. Finally, several factors and trends – population growth of groups in the target population, the graying of America, more expensive healthcare, rising prevalence of health conditions – will almost certainly lead to a larger target population in the future as discussed in the following section. 9.1.1.1 Growth of the Target Population The U.S. Census Bureau projects that the total U.S. population will grow from 282,125,000 in 2000, to 308,936,000 in 2010, and to 335,805,000 in 2020. The population will not grow uniformly, however. The total population that is African American is projected to rise from 12.7 percent to 13.1 percent in 2010, and to 13.5 percent in 2020. The share that is Hispanic is projected to rise in the corresponding years from 12.6 percent to 15.5 percent and then to 17.8 percent. While Final Report 90

transportation status is certainly not driven by race, it is associated with it. In the target population, African Americans and Hispanics are disproportionately represented (see Table 9-2). Furthermore, the share of the U.S. population that is aged 65 and over is projected to rise from 12.4 percent, to 13.0 percent, to 16.3 percent over the same time frame (U.S. Census Bureau, 2004). While projecting income changes is fraught with uncertainty, recent trends have clearly gone in the direction of higher income (and wealth) disparities. Roughly 125 million Americans had one or more chronic conditions in 2000, and they accounted for approximately 75 percent of health spending, and by 2020, a projected 157 million Americans will have one or more chronic conditions and account for roughly 80 percent of total U.S. healthcare spending (Anderson and Knickman, 2001). Moreover, the obesity epidemic itself will certainly lead to high growth in the risk and occurrence of diseases such as diabetes, high blood pressure, arthritis, and some cancers (Kolata, 2005) – many of the conditions analyzed in this study. These trends conspire to dramatically increase the future projection of transportation- disadvantaged individuals at risk of missing health care. Table 9-2: Racial Composition of Target Population and All-Others Population Group Percent of Target Population Percent of All-Others White 80.9 82.3 Black 13.5 12.6 Hispanic 16.7 13.2 Total 1.2 98.8 9.1.2 Characteristics of the Target Population The target population differs from the larger U.S. population in several important demographic and socio-economic dimensions. It is poorer, older, less well educated, has higher minority representation, and is more female, on average. The target population also suffers from disease at a higher rate than the rest of the U.S. population and accesses more healthcare in total. A careful review of medical conditions affecting the target population revealed that a focus on 12 conditions would best inform the cost-effectiveness studies. These 12 were prevalent in the target population and amenable to disease management. Of these 12, seven are chronic conditions, while four are preventive care. 9.1.3 Cost of Non-Emergency Medical Transportation We found that the cost of providing NEMT varies considerably across the U.S., with this variation driven primarily by type of transportation service provided (e.g., ambulatory, wheelchair, or stretcher). We also found variation across modes as a whole, with fixed-route public transportation clearly less expensive than any of the paratransit service types. In addition, we found that NEMT is somewhat more expensive to provide in rural areas than it is in urban areas. Final Report 91

9.1.4 Optimal Method for an Economic Evaluation of Healthcare Costs, Outcomes, and Offsetting Transportation Costs In completing this study, including review of relevant literature and investigation and analysis of available data sets, we determined that cost-effectiveness analysis is a more appropriate method for meeting the objectives than is cost benefit analysis. As a result, we conducted twelve cost-effectiveness analyses based on identified medical conditions from which members of the target population suffer. The lack of a global cost benefit result is more than counterbalanced by the value of these specific analyses (see Chapter 6 and Appendix C). 9.1.5 Value of a User-Based Spreadsheet Tool The analyses clearly showed that, for the conditions examined, providing access to NEMT for the target population is cost-effective. Yet, given shortcomings in available datasets and large variations in both transportation and healthcare costs across the U.S., we determined that developing and providing a spreadsheet tool to facilitate local and regional analysis is warranted. Working together, local and regional transportation and healthcare professionals are likely to have access to data that more accurately reflect local and regional conditions. By using the provided tool, these professionals will be able to produce more reliable results for study regions of interest to them. 9.2 Discussion and Suggestions for Further Research This study has multiple strengths in dealing with a difficult, novel, and important human services problem. Throughout this report, we have stressed that this study was challenging along several important dimensions, presenting both conceptual and analytical difficulties. In this section, we discuss these aspects and present suggestions for future research to further our understanding of the intersection between transportation-disadvantaged status and missed non-emergency healthcare. At the conceptual level, precisely defining the target population is complex, though in the end direct estimate of this population using NHIS and MEPS data proved successful. Nonetheless, transportation and healthcare officials and researchers would all benefit from a shared, standardized, operational definition of transportation disadvantage. This shared operational definition would facilitate better analysis of this population and increase understanding of its many characteristics, including but not limited to access to non-emergency medical care. The well versus poorly managed care method points to another issue regarding transportation provision more generally. While transportation is a focus of this study, seen as a significant input to the “production of health,” ultimately it is only one such input. Many elements are required to produce good health, and one should not ascribe undue influence to the role played by improved transportation access. Specifically, one should not assume that solving someone’s transportation access problem will ensure that this person will obtain needed care and adhere to a prescribed course of treatment. Several other barriers, such as lack of insurance, the cost of medications, the difficulty of scheduling visits around work, and whether patients adhere to their medical therapy once received, may remain. Two strengths of Final Report 92

the current study in this regard are (1) the use of MEPS to select the target population based on primary and secondary reasons for missing care only and (2) reduction in prospective benefits by applying compliance factors and other adjustments to avoid over estimating benefits. As discussed in several places throughout the report and in a forthcoming paper (Wallace et al., in press), shortcomings in available data hinder the ability to investigate the intersection of health and transportation. Simply put, healthcare data lack sufficient information on transportation and transportation access to care, while transportation data contain little on healthcare utilization and nothing on utilization by medical condition. To allow more detailed study of the nationally important questions and hypotheses addressed in this study, both transportation and healthcare professionals and researchers need better data. This could occur through a special supplement connected to one of the major existing studies (such as NHIS or MEPS) or through a dedicated data collection effort. Another possibility is the accumulation of many local case studies (such as promoted by the spreadsheet tool), though these will inherently lack the desirable quality of being nationally representative. Currently, the prospects for improved data are declining, as evidenced by the status of the 2002 MEPS. Although the 2002 MEPS data became available in December 2004, we were unable to use them for this study due to the loss of a key transportation question. This affects how recent the data available for analysis are, and it also limited our ability to study the selected diseases longitudinally. In addition, NHIS has made the MSA field inaccessible, hindering the ability of researchers to distinguish urban and rural respondents. Sample sizes of both NHIS and MEPS are falling because of budgetary pressure to reduce survey costs. Selecting the most relevant conditions to analyze in detail presented a challenge. Clearly, important conditions (such as pain management, need for physical therapy, and cancer treatment) were neglected and are good candidates for follow-up study. Also, the high prevalence of multiple conditions (co-morbidities) in the target population introduced noise into the study. There is no quick and easy way to accurately disentangle the effects of these co-morbidities, and we certainly would not want to create a healthcare system that easily disentangled them by, for example, restricting treatment to a single condition at each visit. Although 12 cost-effectiveness studies were conducted for this study, each of them could easily comprise a full-scale study of its own. We have presented illustrative cases of how healthcare provision could be enhanced by additional transportation resources, but each condition that we studied could be extensively expanded by a stand-alone, detailed economic analysis that includes a longitudinal element to account for time-dependent costs and benefits. To compliment this additional detail, greater analysis of transportation costs would be warranted. This study used selective data to compute transportation costs by mode. A well-constructed survey of transit providers would increase accuracy. In addition, future research should be conducted on a key question: to what extent would the provision of new transportation services involve added costs by luring those who currently use alternative modes? This would allow direct assessment of induced demand. There is significant ambiguity in using an expenditure-based dataset to make the well versus poorly managed care distinctions that are at the heart of our chronic condition Final Report 93

analyses. In addition, when using the MEPS, one finds people with a condition by finding condition-related expenditures and then working backwards to the individuals. This can create identification problems for those without expenditures in a particular period or for those with expenditures with ambiguous condition identifiers. Again, the issue of persistence arises, because even someone with a chronic condition will not necessarily have expenses, or need care, for that condition in subsequent time periods. There are selection issues at several junctures of the cost-effectiveness analysis. A compliance factor was used in Chapter 7 to serve as a way to produce more conservative estimates, i.e., reducing benefits by assuming fewer individuals would comply with recommended treatment. A more serious selection problem concerns the comparison between those individuals identified in the target population and those used to make the poor and well-managed designations (and estimate their expenditures). Additional research could analyze these selection issues to determine whether transportation-disadvantaged individuals differ in important ways from those who we use to estimate the benefits of well-managed care. Further complicating matters, the inability to use MEPS to determine the length of time that an individual has suffered from a given health problem poses difficulties for anyone who would try to add a longitudinal element to this type of study. Hence, we were unable to parse out the cases that would be particularly well suited for early treatment and thus foster early disease management – the very same cases that the literature indicates are extremely cost-effective under a well-managed regime. The same argument pertains to severity demarcation. In both cases, we believe that these limitations are consistent with the conservative stance of our approach. The positive findings that we obtained would be larger still if we had been able to examine cumulative net healthcare benefits over time and if we had been able to focus on the most pertinent set of individuals who lack access to care. To the extent that local providers of transportation and healthcare can identify these people in their region, the resulting net benefits created, using the spreadsheet model, should exceed those outlined in this report. Final Report 94