Some Surrogate Endpoints Are Fine

The Study of the Week considers a trial that tested an intervention for helping a major crisis in American healthcare

A true crisis in humanity goes unnoticed in nearly every US hospital. Routine is the image of people receiving brutal non-beneficial care at the end of their life.

Patients who have no chance of meaningful survival get stuck on ventilators and dialysis machines. Frail patients made more frail with each chemo infusion.

Clinicians choose to work at the bedside because they like helping people. Inhumane care at the end of the life is the opposite of help. It is pure harm. Worse: it’s really hard to stop the inhumanity once it’s started.

JAMA has published a randomized controlled trial of an intervention meant to reduce this blemish of American healthcare.

The idea is modest: promote conversations. Specialists call it a goals of care discussion.

No matter the name, the notion is that before a seriously ill person becomes incapacitated in an ICU, doctors should discuss their goals.

Another name for this is advance care planning. It includes a) clarifying a patient’s understanding of their illness and its treatments, b) understanding their goals of care and c) learning what is important to them.

Previous studies have shown that having these discussions may reduce the odds of suffering at the end of life. Previous studies also show that these discussions don’t often occur. And that is why so many people end up suffering from non-beneficial care at end of life.

The trial tested a simple one-page prompt in the electronic health record. See screenshot:

The endpoint was the proportion of patients whose chart had a documented goals of care care discussion.

The setting of the trial was 3 hospitals in one healthcare system. Patients were 72 years on average and the most common diseases were coronary disease, heart failure, lung failure, kidney failure, cancer and peripheral vascular disease.

The results

The proportion of patients with electronic health record–documented goals-of-care discussions within 30 days was 34.5% (433 of 1255 patients) in the intervention group vs 30.4% (382 of 1257 patients) in the control arm.

The absolute risk difference of 4.1% reached statistical significance.

Comments:

JAMA published this study yesterday. My initial impression when we talked about it on our podcast was that goals of care discussions were so important that there was no need to submit prompts to have them to randomization.

Then I slept on it and this morning I have come to a different conclusion. Yes, of course, we need to study policy interventions. The one-page prompt to physicians seems small, but the electronic health record overflows with warnings and advisories and prompts.

These are all policy choices. Some one or some group felt that placing that distraction in the record was an obvious positive. The grand idea being that diverting the doctor’s attention to this or that reminder will help improve care. Each diversion seems small. But taken together the effects of EHR diversions can be a negative.

What’s cool about this study is that the authors’ tested whether their diversion of attention made a difference.

But there is another reason this study was important to conduct. Look at the results.

In a university hospital system known for its strength in palliative care, only a third of seriously ill patients in the intervention arm have a documented goals of care discussion. If it is that low at the University of Washington, in Seattle, it’s surely lower in less enlightened centers.

---

Some will quibble about the softness of the endpoint—making a statement in the chart that a discussion was had is definitely a surrogate endpoint. The real endpoint would be whether there were fewer patients suffering from non-beneficial care at the end-of-life.

This is a strong argument. Surrogate endpoints are often problematic.

My counter would be that no trial is perfect. Good trials raise as many questions as they answer. The next step could be a larger trial looking at harder endpoints.

My other counter comes from an emotional place. I, like any bedside clinician, hate to see people suffer. I look at that one-page intervention with awe.

The language suggestions are brilliant. “What have other doctors told you about how serious your illness is and what to expect?” You would be shocked at how few patients have been told about the seriousness of their illness or its treatments.

Immediately after that question, a treating doctor or nurse will know what a patient knows. This sets the stage for the next two questions.

Patients and families may not have a quick answer to the “if you were to get sicker, what would be most important to you? question.

But the effect of these questions is vital. They make everyone think. Think about possible outcomes—both good and bad.

That is a start. A crucial start.

The intervention is low-cost, low-burden and immensely elegant. It is a first step to improving our crisis in end of life care.

That it was tested in RCT form is good news. That it passed the test of randomization is great news.