Geographic Variations in Cardiac Rehabilitation Use: Regional Variations in Medical Care or in Patient Behaviors?

The article entitled “Geographic Variation in CR Participation in Medicare and Veterans Affairs Populations: An Opportunity for Improvement?” in this issue of Circulation presents a trove of data relevant to efforts aimed at expanding utilization of cardiac rehabilitation (CR) for appropriate patients¹. Yet, these data are sobering. From 1997 to 2011, despite a seeming uptick in CR referral from 2007 to 2012², there has been no improvement whatsoever in enrolling appropriate Medicare patients to participate in CR (19% in 1997 vs 16% in 2007-11), with the lower rate due in part, to a lower rate of coronary bypass surgery in the more recent cohort³. The situation for patients in the V.A. system is yet more concerning, with a 10% participation rate in 2007-11 despite a slightly younger population and the seeming advantage of the V.A. to systematically deliver care compared with the more heterogeneous U.S. health care system¹. Even when only veterans age 65 or older are examined, the participation rate was lower in the V.A. system than for Medicare patients. These persistently low participation rates are striking in light of the robust literature on the multiple benefits of CR—a 13% reduction in total mortality, a 26% reduction in cardiovascular mortality, a 31% decrease in re-hospitalizations, and an improved quality of life^4,5.

In line with the landmark analysis of Suaya et al of CR participation rates in 1997 for Medicare patients³, regional variations in CR participation have persisted, with the highest participation rates both for Medicare and V.A. systems in North Central states (34% CMS, 17% VA) and lowest rates for both systems in Pacific and Southern states¹. It is not coincidental that the density of CR programs mirrors participation, with the highest density of CR programs per state population is in the North Central U.S. and the lowest in the Pacific and Southern regions⁶. Program density (programs per population) is also an obvious reason for the much lower participation rates in V.A. patients, with only 35 CR programs nationally compared with over 2,500 programs for Medicare patients^1,6. Indeed, with a population of almost 40 million individuals, California has only two V.A. based CR programs and only130 programs available to CMS patients which is the lowest density per population state in the U.S.⁶. The highest density of CR programs are located in Nebraska, South Dakota, Montana, Wyoming, North Dakota, Iowa and Kansas, corresponding quite well to the highest participating states^6,3. Even the best referral practices and high-quality care cannot overcome the barriers created by having few geographically available CR programs. Low CR program density directly impacts participation in at least two ways: 1) creating travel times/distances not palatable or possible for many patients and 2) creating longer wait times until a patient can start the program. Whereas VA patients could be referred to a non-VA CR program (“purchased care”), this almost certainly constitutes a barrier to participation in CR.

With the extremely low VA participation rates, there is no question that this should be viewed as both a crisis and an opportunity for improvement. However, it remains uncertain whether or not one can transport what is being done in North Central states to a low participation state. Certainly, if programs can be made more geographically available (“build it and they will come”), participation will increase, and this is particularly relevant to the VA system. However, whether programs and hospitals can somehow “behave” like programs and hospitals in North Central States and thereby increase CR participation is not clear. One highly likely alternative hypothesis is that CR participation rates reflect the characteristics of the patients themselves given the variability in health-related behaviors across regions. For example, smoking varies by geographical region and the concept of diabetes-, obesity-, and stroke-belts across parts of the US are well known⁷. Additionally, health-related behaviors are significantly correlated with measures of socioeconomic status (SES) such as educational attainment and income⁸. Furthermore, both SES and other health-related behaviors (i.e. smoking) are significant predictors of participation in CR⁹. Accordingly, it is highly likely that patient characteristics (SES, other health-related behaviors) vary by region and account for a significant portion of the variance seen in CR participation rates. Indeed, in their analysis, Beatty et al found that state-level SES was associated with CR participation, with the OR associated with state level variation in Medicare CR participation decreasing from 2.29 to 1.81 when adding socioeconomic variables to the model. These findings confirmed a prior analysis demonstrating that a significant amount of the variation in CR use by state is associated with SES¹⁰. Given that health-related behaviors are associated with both SES and other health-related behaviors we would expect to find a similar pattern here. For example, state-level measures of both high school graduation rates and rates of no leisure time physical activity using the high school graduation rates reported by Beatty and physical inactivity rates for 2010 from the CDC¹¹ are significantly correlated with the adjusted rates for Medicare CR participation. If you run a simple regression including graduation rates and lack of leisure time physical activity the model accounts for 46% of the variance in CR participation (R squared 0.459). These findings suggest that it well may be personal behaviors more than the quality of medical care that explains much of the state-by-state variations, at least in the Medicare system. This would make it less likely that changes in medical practice patterns will result in a substantial increase in CR participation rates. However, we note that the relationship between patient characteristics and CR participation rates is less strong for the VA (R squared 0.245) suggesting that the lack of programs may be limiting participation and increasing the number of available VA CR programs should have a favorable effect.

Opportunities for VA and CMS patients overlap with some additional challenges (and opportunities) for the VA. Both systems need to operationalize EMR-based referrals for patients with early contact from the CR program^12,13. Other proven interventions outlined by the Million Hearts Cardiac Rehabilitation Initiative¹⁴ to increase referral and enrollment of patients into CR include an in-hospital liaison who coordinates referral and uptake into CR, an initial group visit at the CR program prior to leaving the hospital, a home-based or hybrid CR option for patients who live far from the CR program or who have work obligations, and flexible hours of operation of the CR program. Finally, designating CR referral and CR participation rates as a performance measure for appropriate patients has an extremely favorable effect^2,15. It should be noted that a calculation of the benefits of increasing CR participation from 20% to 70% would save 25,000 lives and prevent 180,000 hospitalizations annually in the U.S.¹⁴.

The VA as a coordinated system of care may have a greater ability to systematically deal with this problem. Yet, the extremely low program density within the VA creates additional challenges such that it needs to either build more programs or streamline referral of V.A. patients to non-VA CR programs. If patients need to travel more than 30 miles to a CR program, participation rates at an on-site program plummets³. Additionally, they should consider operating hybrid home programs, possibly with mobile health technology coordinated from a VA-CR center for low to moderate risk patients¹⁶. Another intervention that may increase CR participation in higher risk groups such as low SES individuals is the use of financial incentives¹⁷ although the Medicare Incentives program for CR, originally planned for 2018¹⁸ has been temporarily cancelled by the current C.M.S. administration.

We could not agree more with the author’s statement that “the adoption of new strategies is needed to reduce variation and achieve high levels or participation in CR programs nationwide in all hospitals and healthcare systems”. This should include the development of hybrid and home CR programs, and for the V.A., a more streamlined referral of patients to non-VA CR programs (“purchased care”). Until that time, the broad utilization of automatic EMR referral to CR with an early contact from the CR program will be a “rising tide that lifts all boats”.