By KIP SULLIVAN, JD

The notion that hospitals can reduce readmissions, and that punishing them for “excess” readmissions will get them to do that, became conventional wisdom during the 2000s on the basis of very little evidence. The Medicare Payment Advisory Commission (MedPAC) urged Congress to enact the Hospital Readmissions Reduction Program (HRRP) beginning in 2007, and in 2010 Congress did so. State Medicaid programs and private insurers quickly adopted similar programs.

The rapid adoption of readmission-penalty programs without evidence confirming they can work has created widespread concern that these programs are inducing hospitals to increase utilization of emergency rooms and observation units to reduce readmissions within 30 days of discharge (the measure adopted by the Centers for Medicare and Medicaid Services [CMS] in its final rule on the HRRP), and this in turn may be harming sicker patients. Determining whether hospitals are gaming the HRRP and other readmission-penalty schemes by diverting patients to ERs and observation units (and perhaps by other means) should be a high priority for policy-makers. [1]

In Part I of this series I proposed to address the question of whether hospitals are gaming the HRRP by asking (a) does research exist describing methods by which hospitals can reduce readmissions under the HRRP and, in the event the answer is yes, (b) does that research demonstrate that those methods cost no more than hospitals save. If the answer to the first question is no, that would lend credence to the argument that the HRRP and other readmission-penalty schemes are contributing to rising rates of emergency visits and observation stays. If the answer to second question is also no, that would lend even more credence to the argument that hospitals are gaming the HRRP.

In Part I, I noted that proponents of readmission penalties, including MedPAC and the Yale New Haven Health Services Corporation (hereafter the “Yale group”), have claimed or implied that hospitals have no excuse for not reducing readmission rates because research has already revealed numerous methods of reducing readmissions without gaming. I also noted many experts disagree, and quoted a 2019 statement by the Agency for Healthcare Research and Quality that “there is no consensus” on what it is hospitals are supposed to do to reduce readmissions.

In this article, I review the research MedPAC cited in its June 2007 report to Congress, the report that the authors of the Affordable Care Act (ACA) cited in Section 3025 (the section that instructed CMS to establish the HRRP). In Part III of this series I will review the studies cited by the Yale group in their 2011 report to CMS recommending the algorithm by which CMS calculates “excess” readmissions under the HRRP. We will see that the research these two groups relied upon did not justify support for the HRRP, and did not describe interventions hospitals could use to reduce readmissions as the HRRP defines “readmission.” The few studies cited by these groups that did describe an intervention that could reduce readmissions:

• were based on carefully selected patient samples that were much narrower than the broadly defined patient population covered by the HRRP;
• focused on diseases of the heart, primarily heart failure (HF), and shed little or no light on how hospitals were supposed to reduce readmissions of patients diagnosed with pneumonia or COPD, or patients who had knee or hip replacements, all of whom are now covered by the HRRP;
• offered limited evidence that some interventions could pay for themselves if hospitals were allowed to select the patients to whom the intervention would apply, but no evidence that any intervention could do that when applied to the broad pool of patients defined by the HRRP. 

MedPAC’s faith-based diagnosis

Congress authorized CMS (via Section 3025 of the ACA) to punish readmissions of patients with one of five diagnoses, or who received two procedures, mentioned by MedPAC in their June 2007 report to Congress. Those conditions/procedures were: HF, chronic obstructive pulmonary disease (COPD), acute myocardial infarction (AMI), pneumonia, coronary artery bypass graft (CABG) surgery, percutaneous transluminal coronary angioplasty (PTCA), and “other vascular.” [2]

Obviously, MedPAC’s June 2007 report played a critical role in convincing Senator Max Baucus and other authors of the ACA the HRRP would work. But that report was extremely misleading (1) because its diagnosis of the “excess readmission” problem was wrong, and (2) because its solution was based on a few studies that did not support the proposed solution, i.e., the HRRP.

In that 2007 report, and in subsequent reports, MedPAC took the position that research exists showing hospitals how to reduce readmissions, but hospitals refuse to deploy these proven methods because hospitals are paid fee-for-service, and because consolidation of the industry into chains of hospitals, nursing homes, and clinics has not progressed far enough. Here is how MedPAC explained the failure of hospitals to implement the numerous (allegedly proven) methods of reducing readmissions in their 2007 report to Congress: “Research shows that specific hospital-based initiatives … can avert many readmissions…. Policy changes could encourage more hospitals to adopt successful strategies…. [H]ospitals and other providers have not broadly invested in their role in managing the [post-discharge] transition. Two related factors account for this. First, providers often operate independently of one another (in ‘silos’)…. A second related factor is Medicare’s (and other insurers’) fee-for-service payment policy.” (p. 115)

MedPAC cited not a single study in support of this diagnosis. The reason was, of course, that there was no research to cite.

MedPAC’s off-label solution

Having offered its evidence-free diagnosis, MedPAC went on to offer a solution based on research that did not support their prescribed solution. If you go to MedPAC’s website and open up Chapter 5 of their 2007 report, you will come upon a section with the encouraging (and very misleading) title, “How can hospitals reduce readmissions?” (pp 111-114) In this section, MedPAC asserted hospitals can reduce readmissions four ways: “Provide better, safer care during the inpatient stay”; “Attend to patient’s medication needs at discharge”; “Improve communication with patients before and after discharge”; and “Improve communication with other providers.” The commission cited 17 studies in support of these statements.

The table below offers brief descriptions of the 17 studies. I have grouped them into two categories: Those that did not study an intervention, and those that did. To call your attention to how many of the studies in the latter group examined heart failure, I have put “heart failure” in red.

Studies cited by MedPAC in their June 2007 report to Congress, pp. 111-114*

Studies that did not examine an intervention (10)

Bernard and Encinosa, 2004: “likelihood of readmission doubled (from 14 percent to 28 percent) with an adverse patient safety event during the initial hospitalization.”

Pennsylvania Health Care Cost Containment Council, 2006: “rates of readmission after CABG with hospital-acquired infections are more than double those of uninfected CABG patients.”

 Pennsylvania Health Care Cost Containment Council, 2005: “blood clots are among the top three reasons for readmission” of knee and hip replacement patients.

Hannen et al., 2003: “early extubation or use of beta blockers and aspirin on discharge for CABG patients could … contribute to lower readmission rates.”

Forster et al., 2003: “19 percent of all patients discharged from the hospital experienced an associated adverse event within three weeks; 66 percent of them were adverse drug events.”

Coleman et al., 2005: “elderly patients who had medication discrepancies at discharge were more than twice as likely to be rehospitalized within 30 days.”

Kripalani et al., 2007: Discharge summaries are often incomplete and often don’t get to outpatient doctors.

HMO Group, 2004: Lahey Clinic in MA holds quarterly meetings with “SNF physicians” to reduce readmissions. No data on what happens at these meetings or what impact they have on readmissions.

Kind et al., 2007: “Stroke patients … readmitted after suffering a stroke were much more likely than other patients to be dehydrated and have electrolyte imbalances.”

Trisolini et al., 2006: The participants in the Physician Group Practice (PGP) Demonstration “have put in place better systems to reduce readmissions.” (Note: The final evaluation of the PGP published in 2012 reported the ten PGPs failed to reduce readmissions.)

Studies that examined an intervention (7)

Lappe et al., 2004: Medication protocol for patients with heart failure, coronary artery disease, AMI or atrial fibrillation “significantly improved mortality rates after discharge and 30-day readmission rates, particularly for patients with HF.”

Naylor et al., 1999: Two Philadelphia hospitals “were able to … reduce readmissions by 45 percent over the 24 weeks of the study by having nurses repeatedly meet with patients at high risk for poor outcomes after discharge, both during hospitalization and at home after discharge.” Patients were diagnosed with heart failure, and seven other diagnoses.

Coleman et al., 2006: “Transition coaches” lowered readmissions at 30, 60 and 90 days for patients with one of 11 chronic disease diagnoses (including heart failure, and coronary artery disease, but not pneumonia).

Institute for Healthcare Improvement, 2004a: Phone calls to heart failure, patients reduced readmissions, but no description of study available.

Institute for Healthcare Improvement, 2004b: Phone calls to heart failure, patients reduced readmissions, but no description of study available.

Berenson, 2006: Describes interventions used for heart failure, patients by Billings Clinic (MT) and Park Nicollet Clinic (MN), but makes no mention of their effect on readmissions.

VanSuch et al., 2006: Heart failure, patients were less likely to be readmitted if they received complete discharge instructions. Readmission reduction was not significant after adjustment for patient characteristics and distance from the hospital.

*  The descriptions of all studies are based on MedPAC’s  June 2007 Report to Congress, Chapter 5. All quoted statements are from MedPAC’s report. The descriptions of the last four studies are also based on my reading of those studies. Complete references for the articles cited here may be found at p. 120 of Chapter 5 in the 2007 report.

Three features of this table are notable. First, of the 17 studies, ten did not examine an intervention.

Second, of the remaining seven studies that did report on an intervention, all but one focused on cardiovascular patients (primarily HF patients). Third and most significant, only three of those seven studies – those by Lappe et al., Naylor et al., and Coleman et al. (2006) – described a successful intervention as well as the methodology used to evaluate the intervention. [3]

It gets worse. Two of those three studies – the Coleman (2006) and Naylor studies – restricted their samples so severely they bore little resemblance to the populations that MedPAC recommended should be exposed to readmission penalties, and which the HRRP later covered. As I noted in Part I, the population to which the HRRP applies is largely unrestricted; there are no requirements, for example, that patients give consent, live near the hospital, be able to understand their case manager, have a caregiver at home, take x number of drugs per day, etc. in order to be exposed to the HRRP.

I discussed the 2006 study by Coleman et al. in Part I. We saw that the authors restricted their patient sample so severely (with exclusion criteria such as patients had to consent, live near the hospital, have a phone, not be demented, etc.) that they wound up excluding 60 percent of the patients they screened. Hospitals laboring under the threat of HRRP penalties do not have the luxury of excluding 60 percent of, say, their pneumonia patients, and reporting on the readmission rate of only the remaining 40 percent.

Naylor et al. restricted their participants just as rigorously. First, patients had to have been admitted through one of two hospitals owned by the University of Pennsylvania. Here is how Naylor et al. described their exclusion criteria for patients admitted to those two hospitals: “Eligible patients had to speak English, be alert and oriented when admitted, be able to be contacted by telephone after discharge, and reside in the geographic service area. Patients also had to meet at least one of the following criteria….: age 80 years or older; inadequate support system; multiple, active, chronic health problems; history of depression; moderate-to-severe functional impairment; multiple hospitalizations during prior six months; hospitalization in the past 30 days; fair or poor self-rating of health; or history of nonadherence to the therapeutic regimen.” The patient sample was further restricted by the requirement that patients consent to the study. A whopping 61 percent refused to consent. [4]

Turning now to Lappe et al. – the third of the three studies listed in the table above that actually reported a successful intervention: That study examined a patient sample that was relatively unrestricted compared with the subjects in the Coleman and Naylor studies, but was still quite restricted compared with the population swept up by the HRRP. Lappe et al. did not count readmissions to hospitals outside of the ten Intermountain Healthcare System hospitals that participated in the study (all hospital readmissions count under the HRRP); and they did not count readmissions unrelated to one of the four heart-related diagnoses that were the focus of the study  — HF, AMI, coronary artery disease, and atrial fibrillation (under the HRRP, with the exception of planned admissions, all admissions within 30 days of discharge count, regardless of whether they relate to the diagnosis of the index admission).

To sum up, of the 17 studies MedPAC cited in the course of explaining how hospitals “can reduce readmissions,” only seven described an intervention; of those seven, only three reported that the intervention reduced readmissions; and of those three, only one (Lappe et al.) studied a population somewhat similar to the broadly defined population covered by the HRRP, and that one was limited to diseases of the heart. [5]

Having made their evidence-free diagnosis of the “excess readmission” problem, and having completed their discussion of these 17 studies, MedPAC then recommended that Congress consider punishing “hospitals with high risk-adjusted rates of readmission,” and that Congress start such a program with a “subset [that] would include conditions for which some hospitals have successfully reduced readmissions.” (p. 115) MedPAC alleged that these conditions include HF, four other cardiac conditions or procedures, COPD, and pneumonia. MedPAC, to its credit, did recommend that punishable readmissions should consist of only those “related” to the index admission. But MedPAC, to its discredit, did not warn Congress about the aggressive exclusion criteria used by the three studies that demonstrated successful interventions, nor in any other way convey to Congress that the research they cited provided no basis for their recommended solution. Instead, MedPAC recommended that all patients be included in the pool regardless of their ability to give consent, how far away they lived from the hospital, etc.

Egregious off-label prescribing

The MedPAC commissioners and staff were by no means the only people flogging the readmission-penalty notion in the 2000s. But they were among the most influential. Their support for a readmission-penalty program imposed upon broadly defined swaths of Medicare enrollees was instrumental in persuading Congress to authorize the HRRP, and the authorization of the HRRP by Congress in turn encouraged copycat behavior by the insurance industry. MedPAC has an obligation to all of us – taxpayers, other policy-makers, doctors, hospital executives, and above all patients – to base their recommendations to Congress on rigorous evidence, not hunches and groupthink.

But MedPAC did not do that. The studies cited by MedPAC, listed in the table above, did not warrant endorsement of the HRRP. Nor did those studies warrant the assumption made by MedPAC and others that numerous methods of reducing readmissions (as MedPAC and the HRRP defined them) had already been proven, and all hospitals had to do was pull them off the shelf and implement them. Hospital executives who read every one of the 17 studies listed in the table would not have found a single intervention that could reduce readmissions (as MedPAC and the HRRP defined “readmissions”) for even one of the seven broadly defined categories of patients recommended by MedPAC, never mind all seven of them. Of the 17 studies, only the one by Lappe et al. studied a patient sample somewhat resembling the population exposed to the HRRP, and that study was limited to HF and three other heart conditions.

In Part I of this three-part series, I compared MedPAC’s behavior to “off-label prescribing” – the prescribing of a drug for a purpose that has not been approved by the FDA because there is little or no evidence for using that drug for the proposed purpose. MedPAC’s endorsement of the HRRP on the basis of the research listed in the table above was egregiously “off label” in two respects: The patients tested in those studies were very different from the patients exposed to the HRRP; and those studies focused on HF even though MedPAC urged Congress to impose the HRRP on non-HF patients. Recommending a readmission penalty for pneumonia or COPD patients on the basis of research conducted on carefully selected patients who don’t have pneumonia or COPD is neither rational nor ethical.

[1] Determining whether gaming has been induced by readmission-penalty programs is far more difficult today than it would have been if the readmission-penalty notion had been subjected to rigorous research before it was inflicted on the populace. Because research like that is a public good, and because it’s expensive to run controlled experiments, we can’t expect the private sector to pay for it. The research should have been done by CMS before Congress authorized CMS to establish the HRRP. A well-designed trial would have shown, among other things, whether the HRRP induced gaming (just as the Physician Group Practice Demonstration revealed that the experimental groups gamed that demo by upcoding at twice the rate of the control groups). But the research was not done, and now we’re trying to fix the plane while it’s in flight.

[2] In 2012, CMS selected a list of conditions and procedures for the HRRP that was slightly different from that recommended by MedPAC in their 2007 report. When the HRRP program began on October 1, 2012, three conditions were targeted – HF, AMI, and pneumonia. On October 1, 2014, hip and knee replacement and COPD were added. On October 1, 2016, CABG was added. To sum up, to MedPAC’s list CMS added hip and knee replacement, and deleted PTCA or “other vascular.”

[3] As the table in the text indicates, MedPAC listed two studies by the Institute for Health Improvement (IHI), one labeled 2004a and the other 2004b. In both cases, MedPAC provided only the title of the study and the IHI’s web address. Plugging the title of the 2004a study into IHI’s search engine produced a short “improvement story” that reads like a press release. It contains no information about how the study, if one was done, was conducted. Plugging the title of the 2004b study into IHI’s search engine produced no matches.

[4] Naylor et al. reported screening 1,296 patients. Of these, only 28 percent were enrolled in the study. Naylor et al.’s explanation for the attrition is confusing. They report that 376 of the 1,296 patients were “discharged before screening” (which left 920), which is inconsistent with the statement that all 1,296 were screened. Of the remaining 920, 61 percent (557) refused to give their consent.

[5] MedPAC also discussed the readmission-penalty notion in their June 2008 report to Congress (this was in addition to their chapter on readmissions in their June 2007 report). In a chapter on “bundling payments” in the 2008 report, MedPAC formally recommended Congress adopt a program like the HRRP. But with the exception of an article about Geisinger’s program to improve the quality of care of CABG patients, and an article about the Catholic Health Partnership’s effort to reduce readmissions of HF patients, MedPAC cited no research on specific interventions to reduce readmissions beyond what they had cited in their 2007 report. The article about the Geisinger program did not report a statistically significant drop in CABG readmission rates. The article about the Catholic hospitals reported enormous reductions in readmission rates for HF patients, but the author provided no methodology for the study.  

Kip Sullivan is a member of the Health Care for All MN advisory board, and of MN Physicians for a National Health Program.