Primary Care Corner with Geoffrey Modest MD: 30-day hospital readmission rates, ?? an appropriate QI marker

By Dr. Geoffrey Modest

A recent NIH study looked at the effect of the Medicare Hospital Readmissions Reduction Program (HRRP) on 30-day readmission rates after hospitalizations for acute myocardial infarction, congestive heart failure, or pneumonia, and in particular looking at whether the previously lowest performing hospitals improved more than the higher performing ones after the introduction of HRRP  (see doi:10.7326/M16-0185).

Details:

• 15,170,008 Medicare patients discharged alive from US acute care hospitals between 2000 and 2013.
• Mean age 79.5, 54% female, 85% white/10% black/4.7% other race, 19% admitted with acute MI/45% CHF/36% pneumonia, 52% discharged to home/18% to home with care/23% to nursing home, average length of stay 6 days, 25% rural hospitals/65% private nonprofit/9% major teaching hospitals, overall observed readmission rate 23%.
• HRRP penalties for 30-day readmission rates were: 0% for the highest performing hospitals, 0-0.5% for average performing hospitals, 0.5-0.99% for low performing hospitals, and >1% for the lowest performing ones.
• Of 2868 hospitals serving 1,109,530 Medicare discharges annually, 30.1% were highest performers, 44.0% were average performers, 16.8% were low performers, and 9.0% were lowest performers

Results:

• Overall risk-standardized readmissions increased by an estimated 0.5 per 10,000 discharges per year prior to the passage of HRRP, then decreased by 76.6 per 10,000 discharges per year after passage.
  • For acute MI, risk-standardized readmissions decreased by 23.7 per 10,000 discharges per year before passage, then by 99.3 per 10,000 discharges per year after passage
  • For CHF, risk-standardized readmissions increased by 5.1 per 10,000 discharges per year before the passage and then decreased by 84.7 per 10,000 after.
  • For pneumonia, risk-standardized admissions increased by 3.1 per 10,000 discharges per year before passage and then decreased by 48.2 per 10,000 after.
• After controlling for pre-HRRP trends, readmissions per 10,000 discharges that were averted and attributable to the law were:
  • 6 for the highest performing hospitals
  • 8 for the average performers
  • 4 for the low performers
  • 1 for the lowest performers

Commentary:

• It does seem that after passage of the law, there was a pretty dramatic decrease in the 30-day readmission rate among all hospitals, but especially among those that had been the lowest performers.
• It should be pointed out that the data overall on the utility of financial incentives in changing the “quality of care” metrics is pretty mixed. This study found that the lowest performing hospitals were able to change the most; however, other studies of financial incentives have not shown this to be true, often attributed to the fact that these hospitals had insufficient infrastructure to implement change. Also, this does raise the ideological concern that using financial incentives to make change in fact reinforces the conception and actuality in the US that health care is just a business and not a fundamental human right that should be managed as in most other industrialized (and may less resource-rich) countries: an essential governmental social program such as education.
• However, and the reason I bring up the study, is that it really brings up to me some concerns about quality goals, especially when looking at surrogate markers.
  • For A1c: as mentioned in several blogs, using a target A1c goal is fraught with potential downsides
    • For patients who have really erratic blood sugars (often because of lack of dietary consistency), just increasing meds to lower the A1c (and get “credit” for better care) may well lead to significantly poorer real-world outcomes from hypoglycemia (e., over treating patients at times when their blood sugars are already low).
    • Some of the meds that decrease A1c may actually increase clinical morbidity (rosiglitazone increasing cardiac disease, —  see the many blogs on other new but concerning meds at: https://stg-blogs.bmj.com/bmjebmspotlight/category/diabetes/ )
    • And those who use A1c as a metric do not include actual clinical outcomes as part of their assessment.
  • In terms of hospital 30-day readmission rates, there certainly should be a mechanism to make sure that hospitals don’t just discharge and readmit patients as a means to increase their earnings (e., getting paid for 2 admissions instead of 1), but there is also a real down-side to focusing on decreasing the readmission rate:
    • Hospitals are dangerous places to be because:
      • They are crawling with resistant bacteria
      • There is a tendency/imperative to do lots of testing for things that we in outpatient medicine might just observe and workup later as needed (this is due to several issues: specialists are often involved in the hospitalized patient’s care, and studies have shown that specialists order more tests than generalists; and, even if there is a lowish probability of a problem, it does in some ways make sense to get more tests in hospitalized patients to see what is going on, since putting off the tests might prolong the hospitalization. But the net result of more testing is the likelihood of more adverse events (either because of the test itself, or because of the downstream further testing/procedures for false positive findings)
      • And those who use hospital readmission rates as a metric do not include actual clinical outcomes as part of their assessment.​ interestingly, when there have been doctors’ strikes and dramatic decreases in hospital admissions, as in 1976 in California as well as others, there has been an attendant lower mortality (and, that is a clinical outcome…..)
    • So, I think it makes sense to avoid unnecessary hospitalizations, and, I would think, to keep the length-of-stay as short as possible
    • My practice until 2 years ago (when our health center was still doing in-patient rounding) was to discharge patients as soon as I felt they were stable (often during my rounding early the morning after their admission) when I felt they had a roughly 80% chance of doing well at home, but with aggressive outpatient follow-up (home visits, or seeing them in clinic the next day, ). And my experience was that it was really uncommon for patients to be readmitted. But with the incentives being strongly to avoid readmissions, I am afraid that might translate into longer in-hospital observation and lengths-of stay (at least it was clear that I discharged patients much sooner than the house staff would have, in large part because I could assure timely and appropriate outpatient follow-up). The point here is that we should be developing coherent integrated systems of care that would allow decreased hospitalizations overall, and lower lengths-of-stay if possible when hospitalizations happen. and, not simply using a single marker of “quality” for the complex and often highly individual decisions on how long to keep a patient in the hospital (for example, the same patient who is homeless or does not have adequate home supports may need to stay in the hospital longer appropriately.)
    • ​And, the other side of the issue: if a patient is really sick with end-stage heart failure, , they are likely to be readmitted within 30 days perhaps no matter what happens (though, of course, we should do as aggressive outpatient management as we can). And their being home as much as possible may have important value to them: being with family, in a friendly and supportive environment, etc., even if they are aware they might be back in the hospital soon

So, the real issue is how does one blend the need for some quality control issues (better care for diabetics or decreasing hospitalizations, in the above examples) but avoid using a blunt instrument (a1c levels, 30-day readmissions) which may well decrease real quality care????  This is certainly not easy to do by large-scale data-mining, looking just at numbers (a1c’s) or billing (readmission rates), but I think really requires looking at individual patients to see what an appropriate a1c might be for them, or whether they were really discharged too early and needed readmission because of poor clinical judgment. If you send me their emails, I can add them to the list