Treat the patient or the data point?
Study examines 30-day mortality rate
There’s a thing in science called the observation effect, where the very act of observing something changes the outcome. Is it possible that something similar is happening with cardiac patients? A study in Health Services Research1 has found a bump in 30-day mortality rates at 30 days. There’s no reason why explored in this paper — that research is yet to come, says Bryan Maxwell, MD, MPH, assistant professor at Johns Hopkins Medical Institutions in Baltimore and one of the study authors. But there are some hypotheses, including that physicians, knowing that there is a key benchmark related to 30-day mortality rates, change their care of the patient in order to get that patient to that key 30-day mark, and that patients who would otherwise be sent to hospice are more aggressively treated at great financial cost and with little benefit in terms of quality or length of life.
Maxwell and his colleagues closely analyzed a subtle pattern in the rate at which people die after surgery. "We have speculated about why, but we have only proven the pattern. Subsequent work will look further at the story of what’s happening. But hypothetically, consider this scenario," he says. "A patient goes through a big surgery, with 10 hours in the OR, then the ICU for a few days, and a few days in the hospital — maybe a week or 10 days total. The riskiest time for most patients is right around the time of surgery. Other complications can arise later, but the risk is smaller."
For more complex patients, they might be in the ICU for weeks. That’s the subgroup that is probably responsible for this 30-day mortality bump, Maxwell says. "Does the timing of death with them have something to do with those patients or with the decision making about their care?" For example, a patient doesn’t die suddenly, but has a series of problems — dialysis, ventilator, nothing going in the right direction. At some point, Maxwell says, the doctor sits down with the family and says, "This isn’t working, we can’t fix all the problems and don’t think there is a good chance of meaningful recovery." There are discussions about what the patient would want in terms of life support and transitioning to hospice care. "But what is not always known is when are those decisions made? You might think it just happens with what clinically is happening. But what we found is it might be that there is some influence that’s about all the pressure of benchmarks."
If the patient is at day 25, and one physician thinks things look grim, another might want to give it a few more days. It might be a front-of-mind thought about the 30-day benchmark, Maxwell says, or it might be back of mind. Either way, it’s not malicious. "The reason they exist is to incent physicians to act a certain way, to improve quality, improve survival. But just like any process of measuring, people know what’s being measured and what isn’t," he says.
Teaching to the test?
He relates it to standardized tests: They exist to measure how well students learn, but what increasingly happens is that teachers teach to the tests.
Maxwell doesn’t believe this is necessarily specific to heart surgeons. They picked cardiac surgery because postoperative mortality is more common, and it would be easier to see if a pattern existed there. "It’s just easier to see when more of your patients are likely to die in that specialty."
He worried that heart surgeons would be upset, but they’ve largely been supportive. "My sense is that they want to do the right thing for patients, and what they largely feel about this study is they want better quality metrics and benchmarks, too."
He doesn’t think we should just get rid of what we have. It’s easy to measure 30-day mortality. Some groups and hospitals have started to tweak the measure by looking at in-hospital post-operative mortality regardless of when it occurs. But Maxwell also thinks that measuring things other than death is important: "If I was a family member or a patient wondering where to have heart surgery, I would want to know things like, how many patients return to a regular life versus go to a rehab facility. How long does it take to get back to normal activities? An expansion of what is being measured is called for."
He understands that the more complex and nuanced the metric, the more burdensome it is for those who collect and keep track of the data — and the harder for patients and the wider public to understand. "Risk-adjusted median length of stay is harder to understand than mortality. There is a trade-off. We have to provide what matters without creating too much of a burden on the organizations that provide care. I don’t know how to balance that."
He thinks that as we get more efficient at collecting data on programs and patients, it may be easier to include information such as what he mentioned above, without having to put more resources into it. For now, Maxwell says he believes that QI managers should have some honest conversations with surgeons about what things matter in terms of outcomes other than mortality — such as how long it takes to recover and how fully patients recover.
"That’s the kind of outcome information we should strive for, and how do we improve those data points is what’s important. If you can, I’d collect and measure that."
For more information on this topic, contact Bryan Maxwell, MD, MPH, Assistant Professor, Johns Hopkins Medical Institutions, Baltimore, MD. Email: [email protected].
- Maxwell BG, Wong JK, Miller DC, Lobato RL. Temporal Changes in Survival after Cardiac Surgery Are Associated with the Thirty-Day Mortality Benchmark. Health Serv Res. 2014 Apr 9. doi: 10.1111/1475-6773.12174
