Almost a year ago, I wrote a column at the Upshot on how doing more for patients often does no good. One of the examples I used was advanced life support (ALS) versus basic life support (BLS). The basis for my piece was a recent study that showed that patients with a cardiac arrest who had received ALS before coming to the hospital were less likely to survive to discharge than patients who received BLS. They were also less likely to survive to 90 days after discharge, and they had worse neurological outcomes.
Many of the comments and emails I received after it was published were incredulous. People couldn't believe that this was possible. After all, paramedics administering ALS must be better than just getting BLS. A number of the more charged missives I received challenged me to demand only BLS if my life were in danger.
The more serious charges, though, picked at the methods. Since this wasn't a randomized controlled trial, then we can't establish causality. It's possible that the sicker patients were the ones who got ALS, and that's why their outcomes were worse. It's true the authors tried a number of sensitivity analyses to check for this, but it's still a legitimate complaint.
Of course, a randomized controlled trial is likely never going to happen. One of the biggest reasons is that paramedics refuse to do it. It's unethical, many believe, to randomize people to get BLS, because ALS must be better. In fact, the Ontario Prehospital Advanced Life Support Trial, another cohort study of ALS versus BLS, was intended to be an RCT, but the emergency medical services people refused to make it so for this very reason.
When confronted with such a case, researchers must turn to the tools they have. A more sophisticated analysis, using instrumental variables and propensity scores (which Austin has written about extensively) can help. In a recent study published in Annals of Internal Medicine, researchers did just that:
Background: Most Medicare patients seeking emergency medical transport are treated by ambulance providers trained in advanced life support (ALS). Evidence supporting the superiority of ALS over basic life support (BLS) is limited, but some studies suggest ALS may harm patients.Objective: To compare outcomes after ALS and BLS in out-of-hospital medical emergencies.
Design: Observational study with adjustment for propensity score weights and instrumental variable analyses based on county-level variations in ALS use.
Setting: Traditional Medicare.
Patients: 20% random sample of Medicare beneficiaries from nonrural counties between 2006 and 2011 with major trauma, stroke, acute myocardial infarction (AMI), or respiratory failure.
Measurements: Neurologic functioning and survival to 30 days, 90 days, 1 year, and 2 years.
The older study also just looked at cardiac arrest. This study, however, also looked at trauma, stroke, acute myocardial infarction, and respiratory failure. It also added in an instrument variable, in the form of geographic penetration of ALS across counties.
They did two main analyses. In the first, the researchers used propensity scores to match patients within counties to test how BLS performed against ALS. It's still possible that dispatchers sent out ALS to worse cases than BLS, but they called EMS systems to ask them directly, and found this didn't occur. In that analysis, BLS outperformed ALS again.
In the second instumental variable analysis, with respect to 90-day survival, those who received BLS were 4.1% more likely (absolute) to survive major trauma, 4.3% more likely to survive stroke, 5.9% more likely to survive an acute MI. There was no significant difference with respect to respiratory failure.
There will always be limitations to research, and there are with this study. It's possible that selection bias is still occurring. But these researchers were super careful. They did more than one analysis. The propensity score analysis could be biased if ALS went to sicker patients. But with most conditions, it would be difficult to know what the "severity" was before sending out the ambulance. With respect to trauma, they further controlled for severity. Finally, the patients that got BLS were older and had more comorbidities than those who got ALS. It's hard to imagine this would bias results against ALS.
The instrumental variable analysis can be confounded is counties with more ALS had worse quality hospital care. Again, the researchers were careful. They did falsification tests, and found that ALS penetration wasn't associated with factors like non-emergent surgical mortality or intensive care mortality. They even did sensitivity analyses to ensure that more BLS patients didn't die at the scene or en route to the hospital.
Still, critics dismiss the results:
"Their premise is flawed," said Howard Mell, a spokesman for the American College of Emergency Physicians and director of emergency services in Iredell County, N.C. He said advanced ambulances tend to carry much sicker patients. "That's why they have much worse outcomes."
Nothing is a guarantee. But this analysis accounted for this critique. It doesn't seem to make a difference.
For a long time, the tobacco companies were able to get away with saying we never proved that "smoking causes cancer" since no RCT was available. One will also never be done. At some point, when we aren't going to get an RCT, we have to start to accept the best evidence we can get. We could still do an RCT of ALS versus BLS. But if people refuse, then they should help design a study they'd accept and allow, as enough evidence to consider the idea that ALS might be harmful. Otherwise, they should consider that they might be unwilling to change their minds, no matter what new data come to light.