• Karen Joynt on hospital readmissions

    Karen Joynt is a practicing cardiologist in the Veterans Health Administration and an Instructor at Harvard Medical School and the Harvard School of Public Health. Her research focuses on understanding differences in quality, outcomes, and costs between hospitals, and the policies that may impact these metrics. She is an expert on Medicare’s Hospital Readmission Reduction Program (HRRP) and has published several papers relevant to the readmission rate model that underlies it, as well as its limitations. We (Aaron and Austin) asked her some questions about the HRRP via email. Our exchange is below, followed by the full references she cites in her responses.

    Note: Harlan Krumholz offers a different perspective in response to a similar set of questions.

    1. Based on our reading of the literature, it seems like the purpose and motivation of Medicare’s Hospital Readmissions Reduction Program (HRRP) is to use financial penalties and rewards to motivate hospitals to improve discharge planning and transitions of care. Do you think the HRRP is well designed for this role? If not, what are some better alternatives, in your view?

    If done right, the HRRP could really help push hospitals to forge new connections with their communities, create partnerships with primary care practices, and innovate around how we define the continuum of care. I see some major problems with the HRRP, however.

    a) Incenting hospitals to improve readmissions is one thing; comparing hospitals to one another on readmission rates and penalizing those that do worse is quite another.

    It just doesn’t have face validity to argue that a hospital with a patient population that struggles with homelessness, limited literacy, lack of access to primary care, and a high burden of substance abuse and mental health issues should be able to achieve the same readmission rate as a hospital with a wealthy patient population with a great deal of resources. Once a patient leaves the hospital, there are myriad factors that will influence their likelihood of returning to the hospital. Some of those may be medical; some social; some due to poor adherence or poor understanding – but regardless, penalizing a hospital for taking on the care of vulnerable populations sets up potentially harmful incentives, and seems to me to be the wrong approach.  We need to find ways to help hospitals and the communities in which they are located create a more comprehensive safety net for their patients, and it’s not clear that these penalties will do that.

    b) There are a number of confounding factors that make readmission rates hard to interpret.

    • Hospitals with high mortality rates may have low readmission rates because the patients who die can’t be readmitted (though this is likely not a big enough problem to explain much).
    • Hospitals with a tendency to admit less-sick patients may have lower readmission rates than hospitals with a higher threshold for admission.  Note that in the current fee-for-service environment, admitting less-sick patients is a win-win for dealing with penalties (increase inpatient volume to offset dollars lost from penalties AND decrease readmission rates to avoid next year’s penalties).
    • Improving access to care for a population may increase readmission rates (Weinberger, Oddone et al. 1996).
    • Hospitals that implement programs to improve longitudinal community care may then only admit the sickest patients, and thus have higher readmission rates.  Again, note that under fee-for-service these hospitals could lose twice (decrease volume AND worsen next year’s penalties).

    c) Incenting hospitals to improve readmissions – given that they have many competing goals and responsibilities – means that resources are not being spent on other things, like reducing medical errors, or improving inpatient quality.

    There are a few fixes to the HRRP that could improve some of these issues, though some are easier than others given that this program is written into law. We could take socioeconomic status into account. We could compare hospitals to a group of peer hospitals, or to themselves (i.e. assess improvement). We could weight nearer-term readmissions more highly, since there is some evidence that the very near-term readmissions are more likely preventable.

    2. Obviously, some readmissions are a good thing, or a necessary thing. How will the HRPP account for these, or differentiate them from bad readmissions?

    Right now, it won’t. The metric used is all-cause readmissions, meaning that a rehospitalization for any reason at any point within the 30 days following a discharge “counts” as a readmission. We know from prior work that only a fraction of readmissions are preventable (van Walraven, Bennett et al. 2011; van Walraven, Jennings et al. 2011), but we know little about what a “good” readmission might look like – that’s a very interesting thought.

    3. Is there a realistic danger that the HRRP could encourage hospitals to resist readmissions, even if that practice is to the detriment of patients? Might hospitals dump patients to alternate facilities instead of readmitting them? Are there mechanisms in place to monitor or prevent such practice?

    I think this is a realistic danger. People respond to incentives, and if the signaling is strong enough, some may respond to them in ways that aren’t in patients’ best interest. Two major ways that hospitals could “game” the readmissions measure are putting patients on observation status rather than full admission status (Feng, Wright et al. 2012), and declining transfers of particularly ill patients. Both are phenomena that we will hopefully be able to track in Medicare data over the coming years, in order to determine if these are real problems or just theoretical ones. There are no formal mechanisms in place to prevent such practice, to my knowledge, though I hope there are folks at CMS who are tracking these types of outcomes as well.

    4. You’ve argued that hospital readmission rates are sensitive to socioeconomic characteristics, yet the HRRP doesn’t adjust for them. Which characteristics have been examined in the literature? How sensitive are they and why aren’t they included in CMS’s calculation?

    The literature has been fairly consistent that socioeconomic characteristics matter in terms of readmissions. Specific characteristics that have been examined include race/ethnicity (Alexander, Grumbach et al. 1999; Rathore, Foody et al. 2003; Jiang, Andrews et al. 2005; Silverstein, Qin et al. 2008; Jencks, Williams et al. 2009; Joynt, Orav et al. 2011; Rodriguez, Joynt et al. 2011), hospital racial makeup (Joynt, Orav et al. 2011; Rodriguez, Joynt et al. 2011), patient poverty (Weissman, Stern et al. 1994; Kangovi, Grande et al. 2012), poverty of the neighborhood in which the patient lives (Foraker, Rose et al. 2011), poverty of the community in which a hospital is located (Joynt and Jha 2011), Medicaid versus private insurance status (Jiang and Wier 2010; Foraker, Rose et al. 2011; Kangovi, Grande et al. 2012), having limited education (Arbaje, Wolff et al. 2008), and things like living alone and requiring help with basic functional needs (Arbaje, Wolff et al. 2008).

    Adjusting for these factors is feasible, at least to the degree to which information about them is available, but this idea has met with resistance in the policy community and thus hasn’t been done.

    My concern is that hospitals that serve a higher proportion of patients who face these challenges are more likely to be penalized under the program, specifically safety-net hospitals (Joynt and Jha 2013).

    5. What other aspects of the HRRP concern you?

    The models used for risk adjustment are not very good at predicting readmissions.  They likely underestimate risk at hospitals that serve a very medically complex group of patients, such as major teaching hospitals. The models generally indicate whether or not a patient has a comorbidity in a binary fashion, which may not capture the complexity at a major referral center. We did find that teaching hospitals and large hospitals were more likely to be penalized under the HRRP than their non-teaching and smaller counterparts, though we can’t be certain that this is due to risk-adjustment alone (Joynt and Jha 2013).

    Also, the models employed use a Bayesian hierarchical shrinkage approach that makes it very unlikely that small hospitals will ever be identified as outliers (Silber, Rosenbaum et al. 2010), though this is probably a little more of a tech-y answer than you were looking for!

    6. No, we like the weeds! Getting back to risk adjustment, proponents of the HRRP argue that socioeconomic risk adjusters would be inappropriate because they could relate to quality. When and why is it appropriate or inappropriate to control for socioeconomics in a hospital or system performance measure?

    My personal opinion is that it is inappropriate to control for socioeconomics when one is measuring processes of care. There is no reason that a hospital should have any lower rate of use of revascularization for an acute MI in poor compared to wealthy patients or in black compared to white patients – and controlling for these factors would give an inappropriate “free pass” to provide low-quality care.

    However, the case of a complicated outcome measure like readmissions is different. There is plenty of evidence that socioeconomics impact readmissions (whereas as far as I know there is no evidence that socioeconomics impact a patient’s benefit from revascularization, or aspirin, or appropriate antibiotics, etc.). Given that we are asking hospitals to do vastly different jobs in preventing readmission, recognizing these differences seems reasonable. A hospital with a high proportion of patients who are homeless, or who cannot afford medications, or who have severe mental illness or substance abuse, will have a harder time preventing readmissions than a hospital with a wealthy, stable population. We shouldn’t penalize hospitals for caring for vulnerable populations – we should “level the playing field” to some degree, while we work to determine better ways to provide care for these populations.

    References

    Alexander, M., K. Grumbach, et al. (1999). “Congestive heart failure hospitalizations and survival in California: patterns according to race/ethnicity.” Am Heart J 137(5): 919-927.

    Arbaje, A. I., J. L. Wolff, et al. (2008). “Postdischarge environmental and socioeconomic factors and the likelihood of early hospital readmission among community-dwelling Medicare beneficiaries.” Gerontologist 48(4): 495-504.

    Feng, Z., B. Wright, et al. (2012). “Sharp rise in Medicare enrollees being held in hospitals for observation raises concerns about causes and consequences.” Health Aff (Millwood) 31(6): 1251-1259.

    Foraker, R. E., K. M. Rose, et al. (2011). “Socioeconomic status, Medicaid coverage, clinical comorbidity, and rehospitalization or death after an incident heart failure hospitalization: Atherosclerosis Risk in Communities cohort (1987 to 2004).” Circ Heart Fail 4(3): 308-316.

    Jencks, S. F., M. V. Williams, et al. (2009). “Rehospitalizations among patients in the Medicare fee-for-service program.” N Engl J Med 360(14): 1418-1428.

    Jiang, H. J., R. Andrews, et al. (2005). “Racial/ethnic disparities in potentially preventable readmissions: the case of diabetes.” Am J Public Health 95(9): 1561-1567.

    Jiang, H. J. and L. M. Wier (2010). All-Cause Hospital Readmissions among Non-Elderly Medicaid Patients, 2007, HCUP Statistical Brief #89. Rockville, MD, United States Agency for Healthcare Research and Quality.

    Joynt, K. E. and A. K. Jha (2011). “Who has higher readmission rates for heart failure, and why? Implications for efforts to improve care using financial incentives.” Circ Cardiovasc Qual Outcomes 4(1): 53-59.

    Joynt, K. E. and A. K. Jha (2013). “Characteristics of hospitals receiving penalties under the Hospital Readmissions Reduction Program.” JAMA 309(4): 342-343.

    Joynt, K. E., E. J. Orav, et al. (2011). “Thirty-day readmission rates for Medicare beneficiaries by race and site of care.” Jama 305(7): 675-681.

    Kangovi, S., D. Grande, et al. (2012). “Perceptions of readmitted patients on the transition from hospital to home.” J Hosp Med 7(9): 709-712.

    Rathore, S. S., J. M. Foody, et al. (2003). “Race, quality of care, and outcomes of elderly patients hospitalized with heart failure.” Jama 289(19): 2517-2524.

    Rodriguez, F., K. E. Joynt, et al. (2011). “Readmission rates for Hispanic Medicare beneficiaries with heart failure and acute myocardial infarction.” Am Heart J 162(2): 254-261 e253.

    Silber, J. H., P. R. Rosenbaum, et al. (2010). “The Hospital Compare mortality model and the volume-outcome relationship.” Health Serv Res 45(5 Pt 1): 1148-1167.

    Silverstein, M. D., H. Qin, et al. (2008). “Risk factors for 30-day hospital readmission in patients >/=65 years of age.” Proc (Bayl Univ Med Cent) 21(4): 363-372.

    van Walraven, C., C. Bennett, et al. (2011). “Proportion of hospital readmissions deemed avoidable: a systematic review.” CMAJ 183(7): E391-402.

    van Walraven, C., A. Jennings, et al. (2011). “Incidence of potentially avoidable urgent readmissions and their relation to all-cause urgent readmissions.” CMAJ.

    Weinberger, M., E. Z. Oddone, et al. (1996). “Does increased access to primary care reduce hospital readmissions? Veterans Affairs Cooperative Study Group on Primary Care and Hospital Readmission.” N Engl J Med 334(22): 1441-1447.

    Weissman, J. S., R. S. Stern, et al. (1994). “The impact of patient socioeconomic status and other social factors on readmission: a prospective study in four Massachusetts hospitals.” Inquiry 31(2): 163-172.

    Share
    Comments closed