Invited Commentary
May 1, 2023
The Disappointing Impact of Interventions to Prevent Hospital Readmissions
Robert M. Wachter
JAMA Intern Med. Published online May 1, 2023. doi:10.1001/jamainternmed.2023.0804
When it comes to intervention trials in health services research, we are often left with the challenge of interpreting the generalizability of a positive study, particularly when the intervention involved humans interacting with other humans. Yes, the study found that the intervention (such as patient counseling or team building) worked, but would it work in my clinic or hospital? Or were the results skewed by the special circumstances created by the trial itself.
The study by Donzé et al1 in this issue of JAMA Internal Medicine presents us with the opposite problem: how to interpret a negative health services research intervention study. The authors conducted a single-blind randomized clinical trial in 4 Swiss teaching hospitals, enrolling patients at high risk of readmission as determined by a validated prediction tool. Patients randomized to the intervention group received several interventions designed to ease the posthospital transition and prevent readmissions: inpatient medication reconciliation, a predischarge education session with accompanying written educational material, an early postdischarge primary care visit, and 2 postdischarge follow-up telephone calls. Patients in the control group received usual care from their hospital physician, along with a short information sheet.
The findings were unambiguously negative. The interventions had no impact on the 2 main outcomes: unplanned readmissions and 30-day mortality. Nor was there any measurable impact on other outcomes such as postdischarge health care use or patient satisfaction with the transition process. There was also a small but significant increase in mortality in the intervention arm, a finding that is sufficiently counterintuitive that it raises the possibility of unmeasured confounding.
Transition interventions have a long and storied history in health services research. Between 1994 and 2009, several studies2-4 described the results of 3 different multimodal interventions (each with relatively subtle differences, mostly in the role of a health professional, such as a transitions coach or a nurse practitioner) in preventing readmissions. All 3 interventions—the Naylor et al2 Transitional Care Model (TCM), the Coleman et al3 Care Transitions Intervention (CTI), and the Jack et al4 Re-Engineered Discharge (RED) – were associated with significant reductions in readmissions.
In addition to the literature reporting that many readmissions could be prevented by evidence-based interventions, in 2009 a seminal study by Jencks et al5 found that patients with Medicare who were hospitalized had a nearly 1 in 5 chance of readmission within a month. The publication of the Jencks study, just at the time that the Patient Protection and Affordable Care Act (informally known as Obamacare) was being crafted by the Obama Administration and debated in Congress, set the stage for a major policy intervention. After all, if readmissions are common and partly preventable through implementation of feasible albeit expensive interventions, why not make hospitals accountable for their readmission rates? The policy, the Hospital Readmissions Reduction Program (HRRP), was embedded in Obamacare in 2010 and formally launched in 2012. It mandated public reporting of readmission rates for 3 common diseases: myocardial infarction, heart failure, and pneumonia. Hospitals with unusually high rates of readmissions would incur financial penalties, often quite substantial.
Recently, Cram et al6 advanced the case that the HRRP should be modified, and its financial penalties removed. The authors (including me) argued that the policy has led to substantial gamesmanship, was associated with reductions in readmission rates no greater than those seen in countries without such policies, and was costing—and possibly wasting—billions of dollars of scarce hospital quality improvement dollars. Moreover, we observed that several studies conducted since the implementation of HRRP have demonstrated that most readmissions are not preventable,7 which makes it unfair to penalize hospitals for them. The study by Donzé et al1 adds to this body of research.
What has changed in the past decade to render these types of transition interventions, so strikingly effective in the preceding 15 years, ineffective in reducing readmission rates? It’s hard to know for sure. Perhaps these interventions work best for less ill patients – as the threshold for hospitalization has grown higher, there are fewer such patients in the hospital.
It is also possible that “usual care” has improved, making it harder to demonstrate the impact of more intensive interventions. As someone who has practiced hospital medicine for over 30 years, this explanation certainly feels credible. Most hospitals now perform at least a version of medication reconciliation, and many conduct postdischarge follow-up telephone calls, as part of routine practice. When I began as an inpatient attending in the early 1990s, we discussed discharge planning in an interdisciplinary meeting with social workers and case managers at a short meeting that occurred every Thursday. Yes, once a week. Today in my hospital, such multidisciplinary meetings take place every morning, with an abbreviated afternoon teatime meeting serving as an additional check-in on most days. As most hospitals have improved their routine discharge processes, demonstrating additional benefit from a more intensive intervention becomes that much harder.
It is difficult for me to know what constitutes usual care (which is not described in the study) in Swiss hospitals when it comes to facilitating a safe discharge . But it is safe to say that, as a US program without a Swiss equivalent, the Medicare HRRP is an unlikely explanation for any improvement in routine discharge practices there.
Transitions of care remain a period of heightened risk. The answer cannot be to abolish programs aimed at facilitating a safe discharge. Instead, it is worth continuing to hardwire in practices proven to be effective and to explore whether new models—including better analytics to determine which patients are at greatest risk for posthospital problems and tailored follow-up based on individualized patient needs and preferences—will result in a safer postdischarge period. Until we find new and better interventions, health care organizations should consider the possibility that their expensive labor-intensive programs designed to improve transitions and reduce readmissions may not be having their desired effects. And policy makers should absorb the growing literature pointing to the ineffectiveness of complex and expensive transition intervention programs. They may conclude, as I have, that the HRRP can no longer be considered an evidence-based policy intervention.
