Knowing the actual probability that a patient will have a bad outcome can help you communicate risk to a patient or family, and allow them to share in the decision-making process.
Emergency physicians have transparency issues when it comes to risk. We know that we deal in risk every day, but it’s a nebulous sort of nonspecific sense, a shadow that’s out there somewhere, the fear of some bad future outcome. But when you’re armed with the right data, communicating risk becomes less of a vague and scary proposition.
As we talk about risk, we need to define our terms. Risk is something bad in the future that could possibly happen. In general, that’s very different from, say, legal risks. We’re talking here about the possibility of certain negative medical outcomes, not the possibility that some money grubbing low-life will finally catch up to an ambulance with your patient in it.
Also, as we talk about risk, remember that these are people who haven’t had a bad outcome. They are well and they look good. The risks we’ll be talking about are outcomes that are kind of major and patient-oriented. We’re dealing here with what you say to a patient about what their medical risk for a given outcome. We’ll start with chest pain patients.
We see this class of patient under 40 years old every day. They make us nervous because if you miss it on a 35 year old it’s a very big deal; there’s a lot of future life on the line. We also know that there is a minimal risk that something bad is going to happen. But what do the numbers tell us?
First, let’s narrow in on well patients who have had a negative ED work up – the EKG doesn’t show an MI and the troponin, if you sent one, was negative. For these patients, under 40 years old, showing up to an emergency department, the 30 day event rate is somewhere in the range of 0.2%. That’s 2 in 1,000. In this particular data set, four patients had MIs over the 30 days after their emergency department visit, and one patient died.
The one patient who died had metastatic cancer, but because it was a death, it got classed as a potential miss. With the patient with metastatic cancer, I have no proof that the death is really related to their chest pain, but to be conservative we add that outcome in there. Which means that 1 in 500 is the bad news rate. The reason I tell everybody this is because I want them to think about what it means to have that number at your fingertips. Think about what it means to know that number rather than feeling a nebulous chance that something dark and disturbed is waiting to happen.
This nebulous sense that something dark and disturbed is going to happen can easily throw us off track, and it can certainly throw our patients off track. When we communicate the risks we tend to exaggerate the numbers, at least in the tone that we take. We exaggerate them because we’re trying to be safe, and in many ways that is OK. We’re trying to be conservative because we are making this decision for somebody else. That is just human nature; it is expected and normal. It’s not wrong to over reach and it’s not wrong to over guess and overestimate risk when you’re guessing for somebody else. However, once you know these numbers you don’t have to guess; you can shine some light into the shadowy places. It’s no longer nebulous, and you can talk plainly, if necessary, about your patient’s 1 in 500 chance that something bad will happen.
Now there is another category of chest pain, and another round of data we need to look at. These data are from two very high quality studies in which the only way to get into the study was to be a chest pain patient in the emergency department and be classified by the emergency physician as low risk. The average age was close to 50. On average, patients had one Framingham risk factor. They could have hypertension, they could have diabetes. In a handful of cases they actually had a couple of Framingham risk factors and in some cases they had zero. Sixty percent of them had a normal EKG and about one-third had a nonspecific EKG. Another 10% had a variety of findings, none of which were clearly ischemic. This is a group that is very, very common for us to see. This is the average chest pain patient that we see, in fact: about 50 years old, hypertensive or diabetic or maybe hypertensive and diabetic, with an EKG that is nondiagnostic or normal and a negative troponin.
Now, the 30-day outcome rate in this group was 5 MIs in 1,400 patients with zero deaths. That’s pretty amazing, and decent proof that emergency physicians are really good at this. When an emergency physician says, “I think this is a low risk patient” it turns out to be a low risk patient. Because 5 MIs – and zero deaths – in 1,400 subjects is an incredibly low outcome rate. It’s almost unbelievable. What that means is that with a negative ED workup, 0.4% is the number needed for badness – about 1 in 250. And that is something you can say to your patients in the emergency department. Once again, the numbers take away the ambiguity, changing it from a shadow game guesswork to a numbers game. People get to make decisions based on their actual risk.
The next category is another chest pain cohort, and it’s one of my favorite recent studies. This time they’re in their mid-50s and on average they have two Framingham risk factors, are diabetic and hypertensive or maybe they’re a smoker and they’re diabetic. On average, two Framingham risk factors, and EKGs are either normal or nondiagnostic. In 1,000 subjects, 8 MIs with zero deaths. Plus, they actually have a six month outcome instead of a 30-day outcome, making it even more conservative. Impressed by the study, I talked to the lead researcher and asked how these patients were classified by the emergency physicians. Well the only patients who got into this study were patients who were being admitted to the hospital overnight. They had to be admitted. It couldn’t even be a chest pain observation unit. And they had to be getting nuclear stress tests. They didn’t get into the study unless they were getting myocardial perfusion imaging. They were a group that emergency physicians thought were medium to moderate risk. They were not the low risk population. And yet there were still only 8 MIs, and zero deaths. That’s 0.8%, or a number needed for badness of 125. I think that’s pretty amazing.
I know what you’re probably thinking. These numbers are way lower than you’ve always heard. I’ll tell you why: these studies restricted the outcomes being counted to heart attack and deaths – you don’t include somebody who got a stent put in, or somebody who got maximal medical therapy, or some sort of an intervention. If you don’t include those patients in the outcome, the outcomes become much smaller. More than half of the outcomes in a lot of these patients in these studies are actually that somebody put a stent in there.
Now I do not count that as an outcome. That’s a caveat that you have to take home with you and understand when you’re trying to understand these numbers. The reason I don’t include them is because there is all kinds of great data now, courage trial and many others, showing that stents do not reduce heart attacks or deaths. They may have a small impact on symptoms, but they do not reduce heart attacks and deaths. And this is now well proven and well known by cardiologists everywhere. It’s a little bit of a shock to us when we talk about it but actually it’s true that if stents don’t reduce heart attacks and deaths, I can’t really count those as an important outcome to a patient.
We’re talking about you and patient, communicating risk. You’re talking to them about whether or not they’re going to have a bad outcome. This is not about whether or not some doctor is going to treat them in some way in the future. This is whether they’re going to have the outcome that they came to the hospital worried about. The reason they showed up was chest pains. They were worried that they might have a heart attack or die. They didn’t show up because they were worried that some cath jockey might put a stent in them for no good reason. That’s different.
Those are the chest pain risks. That’s a lot easier to remember than I thought it was going to be when I first went through the literature. And I didn’t engineer these numbers in any way. I wasn’t trying to play math, it’s just worked out this way. It starts at 500, it goes down to 250, then it halves again to 125.
Next we’re going to talk about adult minor head injury. The first thing to remember is that these are just studies, and they’re not going to be the same as every patient you see. The first data set has to do with patients who fall into a specific, lower-risk category of head injury. There had to have been witnesses loss of consciousness or amnesia. We’re only including the people from these studies who had GCS of 15 and no neurological deficits. You couldn’t have any kind of seizure. You couldn’t have had any kind of funny neurological finding on your exam in the emergency department. If you don’t have a GCS of 15 or you have some sort of neurological finding then you’re in a higher risk category. Don’t apply these numbers to those patients. Lastly, injury had to be within the last day.
Within this cohort we see some amazing numbers. Among thousands and thousands of subjects enrolled – all of whom get CAT scans – what is totally startling for me is that it turns out that between 3% and 10% of the patients had a CT finding, and 0.4% had neurosurgery. And that 0.4%, amazingly, is exactly the same across all four of these studies. Mortality in these studies was also a surprise: nobody died. I suppose that means that these patients all got to an OR quickly enough and that a handful of patients with obvious neurological derangement didn’t make it into the study. But if you showed up to an emergency department with a GCS of 15, blunt trauma, your neurosurgical rate was 0.4%, and mortality was zero. In other words, with a clinically important traumatic brain injury, 0.4% – or 1 in every 250 – needed neurosurgery.
That’s not nothing. It actually scares me a little bit to think that 1 out of every 250 subjects I see with a totally normal neurological exam and a GCS of 15 is going to go on to a neurological procedure of some sort. But those are the numbers.
Or are they? If we’re just looking at the study population, the number is indeed 1 in 250. However, if you are negative for the Canadian CT Head Rule, it’s near zero. So what is the Canadian Head CT Rule? It’s not difficult to remember. I’ll give you a little mnemonic that is a shout-out to our friends above the border: “Get Sloshed and Vomit.” I love Canadians, but they have a reputation that they are proud of, and if it helps you remember the Canadian Head CT Rule, all the better. If you want to substitute another “S” word for “sloshed” go right ahead. Get creative.
So “get” is the G and that’s just a GCS of 15. You knew it had to be GCS of 15 anyway. The S in sloshed is for skull fracture. This means you have to do a good scalp exam. And I mean really look over their scalp. You have to make sure they don’t have a battle sign behind their ear. You have to look in their ears and make sure they don’t have any signs of a basilar skull fracture or a hemotympanum in there. And they can’t have any depressed skull fracture when you feel their scalp. “And” is for age: They cannot be older than 65 years old. And vomit is just for vomit. They can’t have vomited more than once. So “Get sloshed and vomit” is the Canadian CT rule. GCS of 15, skull fracture of any kind, age over 65 and vomiting more than once. If you remember those four things, according to the best data we have, the patient in front of you can be said to have a zero probability of a neurosurgically important bleed.
A side note about “zero probability.” If you’re CT negative or Canadian CT head rule negative, your chances of a neurosurgical outcome are down below 1 in 5,000. I said zero but keep in mind there is always a confidence interval and if we calculate this margin for error, the best guess is less than 1 in 5,000. Which is pretty darn good.
A couple more bits about that data that I like. What if your patient is negative on the “G” “S” and “V” of the Canadian rule, but is over 65? All of a sudden the percentage who will need neurosurgical intervention is about 1%, which is significant. And yet, when I tell this number to patients, many of them still opt to go home, rather than wait in the hospital, and return if the situation worsens. When I tell people that number and we have this discussion I’m frequently surprised at how much more risk they’re willing to take than I was willing to give them.
So it’s 1% of you are over the age of 65 but you haven’t broken any other piece of the rule. I like that because I don’t want to automatically have to get a CAT scan just because someone is over the age of 65. I want to be able to have a discussion about it and talk about the risks without it being nebulous.
Why are all of these numbers so important? Because we need to let patients in on the decision making process. It needs to be a collaborative process. And what’s more, when we do let patients in on the numbers, we might just have less stress. I sleep better at night. When I have a 1 in 100 or 1 in 250 patient, I let them decide whether they want to be admitted to the hospital and whether they want to have the CAT scan. I actually talk it over with them and let them decide. And when I do that, I’m often surprised at their willingness to take more risk than I would have allowed them.
The next big question is this: What do you write on the chart? I’ve created a macro that I have in my computer charts. It says, “Miss Smith and I discussed the possibility of her being admitted to the hospital for testing and observation and she understands the estimated risk of a heart attack or death is about 0.4% or 1 in 250. Based on that number, she has chosen to follow up with primary care doctor instead. She is confident. She is reasonable. She verbalized and understands these risks.”
Is this legally secure? About four years ago I had that patient who, after I had put this entry in, came back with an MI the next day. She did well but she had an MI. So I showed the chart note to my legal department and they said this is as safe as you can be. This is as good as you can do, and this is a reasonable chart entry that will protect you.
We all know that the doctor who admits every patient for chest pain. We all know the doctor who sends almost every patient home and we are sort of blown away and amazed that anybody would do that. I think we can all agree it’s a bit unfair that which doctor the patient gets is the arbiter of whether they get admitted. A better idea is to know the risk numbers, have an intelligent conversation about them with your patient, and let them help drive the bus.
Dr. Newman is the author of Hippocrates’ Shadow: Secrets from the House of Medicine (Scribner).