This little letter, tucked away in the Correspondence section of Annals delves into the Pulmonary Embolism Rule-Out Criteria – a decision instrument of some controversy in Emergency Medicine. Specifically, this letter addresses a case report from a previous issue of Annals of, essentially, a large pulmonary embolus diagnosed in a young patient who was otherwise PERC negative.
The authors from Carolinas Medical Center have a registry of 1,880 PE+ patients with which to evaluation, and they performed a retrospective application of the PERC rule. Overall, 6% of this cohort was PE positive and PERC-negative. When compared with the patients with PE who were PERC-positive, there are a few statistically significant differences – pleuritic chest pain was more common in PERC-negative patients with PE, along with pregnancy or post-partum status. Unfortunately, these statistically significant relative differences reflect only small absolute differences of essentially clinically irrelevant magnitude. The only mildly interesting tidbit from the letter is the statistic that none of PERC-negative PEs died within 30 days, compared with 5.7% of the PERC-positive cohort.
The authors suggest a couple weak clinical implications from the data, but these are limited by the retrospective nature of the analysis. It is enough to remember that PERC-negative does not actually “rule-out” PE – it is simply a collection of negative likelihood ratios working against a pretest probability, resulting in clinical equipoise regarding the expect benefits vs. harms of CT pulmonary angiogram and the resultant harms of treatment in physiologically uninteresting PE.
“Clinical Features of Patients With Pulmonary Embolism and a Negative PERC Rule Result”
www.ncbi.nlm.nih.gov/pubmed/23260692
Category: Clinical Decision Rules
The Latest Prognostication for Stroke
We have a fairly robust vascular neurology program at my institution, and – unsurprisingly – they’re rather pro-thrombolysis. While our disagreements over the efficacy of thrombolysis for acute strokes are generally set aside in a truce stemming from academic and research interests, the main philosophical difference between our services remains this: the difference between eligible and indicated.
Vascular neurology tends to treat these terms as synonymous regarding thrombolysis and acute stroke, while it’s clear from the literature that not every patient benefits from thrombolysis. The most recent issue of Neurology features another prognostic tool, the SPAN-100, which is the simplest by far: NIHSS + age. If this score is >100, fewer patients will benefit from tPA than will be harmed. There’s a quality-of-life discussion to be had regarding individualized treatment decisions in SPAN-positive patients, and this is derived from a very small cohort, but it’s consistent with the remaining literature.
The accompanying editorial is also pro-thrombolysis, but does recognize these scoring systems are important clinical tools in educating patients and families regarding the potential for benefits and harms. Most importantly, this table from the editorial summarizes the growing body of literature available to assist the decision-making process:
I look forward to seeing these develop such that clinicians have better tools with which to separate eligible from indicated.
“Stroke Prognostication using Age and NIH Stroke Scale: SPAN-100″
www.ncbi.nlm.nih.gov/pubmed/23175723
www.ncbi.nlm.nih.gov/pubmed/23175723
ALTEs That Need Admission Need Admission
Coming from the west-coast PEM powerhouses of Harbor-UCLA, CHLA, and USC, this prospective observational study attempts to distill the clinical characteristics of “apparent life threatening events” requiring hospitalization. Traditional teaching has always errs on the side of admission for ALTEs, despite the typical low-yield nature of the admission.
They collected data on 832 ALTEs, 191 (23%) of which they felt truly necessitated admission for a set of predefined criteria. Based on this data, they came up with a simple decision rule to identify ALTEs for admission:
– They obviously need to be admitted.
– Concerning medical history/prematurity/congenital comorbidities.
– >1 ALTE in 24 hours.
This captured 89% of necessary hospitalizations with a specificity of 61%, with an AUC of 0.71.
It’s a bit of an odd rule that includes “obvious need for admission”, but, I suppose it’s rather pragmatic. However, the adoption of a rule such as this – after prospective validation – would depend on the “acceptable miss rate” in an infant with a possible life-threatening condition. A sensitivity of 89% probably isn’t going to cut it, so, in the end, what this study is only good for is perusing the interesting data they’ve collected along the way.
“Apparent Life-Threatening Event: Multicenter Prospective Cohort Study to Develop a Clinical Decision Rule for Admission to the Hospital”
New & Improved Glasgow Blatchford Score
Clinical decision instruments that predict short-term interventions and outcomes are fabulous things – precisely the sort of instruments that help Emergency Department physicians decide who will benefit from hospitalization.
This is the Glasgow Blatchford Score, a decision instrument applicable to presentations for upper gastrointestinal bleeding. It has been “improved” by physicians from Kaiser, who have performed addition by subtraction – eliminating variables without a significant change in performance. The original score has eight clinical features – the “improved” version has five, dropping chronic disease, melena, and syncope from the criteria. Despite this, the AUC for therapeutic endpoints as well as for rebleeding and mortality is no different than the original score – at 0.85 and 0.83, respectively.
It is only a 200 patient cohort, and they don’t break down exactly how many patients were in each quartile of possible modified GBS, but essentially, a score of 0 or 1 means <5% chance of needing a clinical intervention, scores of 2-6 about 20% need an intervention, and anything above that is 70% chance of intervention. The incremental improvement over regular physician judgement is not examined, but, more objective evidence is always better.
“A modified Glasgow Blatchford Score improves risk stratification in upper gastrointestinal bleed: a prospective comparison of scoring systems”
When Cardiologists Risk-Stratify Heart Failure
This is the “MARKED” score, a multi-marker prognostic risk score derived from emergency department patients presenting with acute dyspnea. The authors state they’ve presented a “simple, straightforward” score that “may help the treating physician at the ED to decide on urge of intervention, admission, and timing of re-evaluation.”
Sounds perfect! A valuable tool to determine which patients are at high-risk for short-term mortality, possibly to predict which patients may have unanticipated poor outcomes if discharged home?
Ah, sadly, no.
When these cardiologists risk-stratify heart failure, they’re using 90-day mortality – an endpoint almost certainly irrelevant to acute evaluation of dyspnea. The authors recognize the “single-center” aspect of their study as a limitation – but, considering this comes from a dedicated “cardiology ED” in Holland, the external validity is extraordinarily limited. The authors also do not offer any practical suggestion regarding how this score might be used in practice – or how decision-making using this score effectively changes outcomes compared with usual care.
Finally, this “simple” score features the commonly used laboratory tests such as NT-proBNP, high-sensitivty cardiac troponin T, Cystatin-C, high-sensitivity C-reactive protein, and Galectin-3. It should be no surprise a few authors receive compensation from Roche Diagnostics and ACS Biomarker B.V.
“Multimarker Strategy for Short-Term Risk Assessment in Patients With Dyspnea in the Emergency Department”
www.ncbi.nlm.nih.gov/pubmed/23021334
The Ehrlanger HEARTS3 Score
I hate using the TIMI score to risk-stratify patients in the Emergency Department. It wasn’t derived from a question asked in the Emergency Department, but has been co-opted by hundreds of studies as it has some value as part of our common language with inpatient medicine and cardiology teams. We’re familiar enough with it’s shoehorning into our environment that we can use it to assist in some rough decisions about prognosis, but, clearly a better tool must exist.
A couple years back, the HEART score came out of the Netherlands. In a small derivation and validation cohort, it did a reasonable job of predicting outcomes, using language and variables more relevant to the Emergency Department. However, these authors from Ehrlanger in Chattanooga recognized one of the limitations of the HEART score was the somewhat arbitrary “expert” weighting of the various elements. They therefore undertook a study with the goal of using logistic regression and likelihood ratios of the various included elements to expand the score and modify the weighting.
The good news: they improved the AUC of the scoring system from 0.827 and 0.816 for acute MI and 30-day ACS, respectively, to 0.959 and 0.902. At the reasonable cut-off, the HEARTS3 score gets up close to ~98% sensitivity with ~60% specificity for 30-day ACS.
The bad news: a complex clinical situation requires a complex clinical decision instrument. No one will be able to hold this in their head like the NEXUS criteria, the TIMI score, or Wells criteria – if we were even bothering to hold all these hundreds of decision instruments in our heads to start. Luckily, smartphones, the Internet, and decision-support built-in to electronic health records is making progress towards readily available peripheral brains with which to quickly reference risk-stratification instruments such as this.
It still needs external validation, but this is one of the tools seeming to have the greatest potential I’ve recently seen
“Improving risk stratification in patients with chest pain: the Erlanger HEARTS3 score”
http://www.ncbi.nlm.nih.gov/pubmed/22626816
A Chest Pain Disposition Decision Instrument
This article has three things I like – information graphics, informed patients, and an attempt to reduce low-yield chest pain admissions. Unfortunately, in the end, I’m not sure about the strategy.
This is a prospective study in which the authors developed an information graphic attempting to illustrate the outcome risks for low-risk chest pain presentations. They use this information graphic as the intervention in their study population to help educate patients regarding the decision whether to be observed in the hospital with potential provocative stress testing, or whether they would like to be discharged from the Emergency Department to follow-up for an outpatient provocative test. They were attempting to show that use of this decision aid would lead to increased patient knowledge and satisfaction, as well as reduce observation admissions for low-risk chest pain.
The good news: it definitely works. Patients reported increased knowledge, most were happy with the decision instrument, and a significantly increased proportion elected to be discharged from the Emergency Department – 58% of the decision aid group wanted to stay vs. 77% of the “usual care” arm wanted to stay.
My only problem: this study truly exposes the invalidity of our current management of chest pain. If these patients are low-risk and they’re judged safe enough for the outpatient strategy in this study – why are any of them being offered admission? Of course, it’s probably because they don’t have timely follow-up, and AHA guidelines dictate stress testing urgently following the index visit. But, truly, in an ideal world, few (if any) of these low-risk patients – such as the one who ruled in by enzymes – should be offered admission.
But, other than that, I’m all for information graphics and patient education techniques to include them in a shared decision-making process!
“The Chest Pain Choice Decision Aid : A Randomized Trial”
www.ncbi.nlm.nih.gov/pubmed/22496116
Rational Clinical Examination: GI Bleeding
This series of articles, “The Rational Clinical Examination” in JAMA is by far one of my favorite approaches to medicine. They ask simple clinical questions, and they do literature searches to find evidence to apply. Additionally, the form in which they distill the evidence tends to be likelihood ratios – a far more useful statistical construct in estimating how a particular finding contributes to ruling-in or ruling-out disease.
This most recent literature review covers gastrointestinal bleeding – and it covers a few worthwhile points. Most encouragingly, the authors are exceedingly skeptical about the utility of NG tube placement – reasonable positive LR for UGIB, but, as the authors note, a suspected source is usually well-established prior to NG tube placement. Additionally, they note that the NG lavage does not tend to influence final patient-oriented outcomes – and lean towards not recommending its use. Secondly, they also cover the Blatchford and Rockall scores, which are decision instruments that might have value in helping triage patients for outpatient management.
“Does This Patient Have a Severe Upper Gastrointestinal Bleed?”
www.ncbi.nlm.nih.gov/pubmed/22416103
ABCD2 For Cerebrovascular Dizziness
This is a bit of an interesting idea – a repurposing of the ABCD2 prediction instrument for TIAs as a risk-stratification instrument for cerebrovascular causes of “dizziness.”
Every ED physician loves the complaint of “dizziness.” It’s either giddiness, unsteadiness, lightheadedness, vertigo, and it’s frequently difficult to elicit any pertinent neurologic symptoms to clarify one of the benign causes of vertigo or a cerebrovascular cause.
This is a retrospective chart review in which they evaluated the charts of 907 “dizzy patients”, 37 of which had a cerebrovascular cause – 4.1%. It’s a small sample size – so the confidence intervals for their odds ratios are very wide – but for multivariable adjusted odds, age > 60 had an increased OR of 5.1, BP >140/90 had an increased OR of 2.9, speech disturbance had an OR of 6.2, and unilateral weakness had an OR of 10.9. Essentially, it’s interesting to see – and it makes sense – that the same features that generally portend stroke after TIA also might help predict which of your dizzy patients will be higher yield for a more intensive evaluation.
“Application of the ABCD2 Score to Identify Cerebrovascular Causes of Dizziness in the Emergency Department”
http://www.ncbi.nlm.nih.gov/pubmed/22442167
It’s Another Chest Pain Prediction Rule!
Yet again, the insanity of the race to a zero-miss culture funds another chest pain discharge prediction rule. In fact, the most telling part of this paper is in the very end when they compare the chest pain admission rates of the Canadian hospitals in this article to the U.S. hospital – 18% and 20% in Canada compared to 96% in the U.S. (combined ED observation status and inpatient). The difference in those numbers is insane – and I’m sure people could easily debate which is the preferred side of those numbers to be on.
In any event, the study is a prospective, observational data-gathering study of 64 variables related to the presentation of chest pain – some of which are objective and some of which are historical. It’s an interesting read – in part because the inter-observer kappa for a lot of the historical variables is so terrible they weren’t even usable. After collecting all their data, they did 30-day telephone follow-up or vital records review to evaluate the combined endpoint of death, myocardial infarction, or revascularization.
Via the magic of recursive partitioning, a patient without new EKG changes, a negative initial troponin, no history of CAD, atypical pain, and age less than 40 years separated out 7.1% of their study population that had zero 30-day outcomes. Adding a second negative troponin six hours later for the 41-50 year group gives another 11.2% of patients that had zero outcomes. So, a facility that admits 96% of their patients could potentially reduce admissions – but it might have less utility in Canada.
I’d rather see a two-hour second troponin than a six-hour one; it might reduce sensitivity, but it’s wholly impractical to tie up a bed in the ED for 6 hours for a patient you want to send home. And, like most of these articles, the combined endpoint of death, MI, and revascularization is irritating. Considering there were twice as many revascularizations as myocardial infarctions, there really ought to be more granularity in these sorts of studies with regard to the actual coronary lesions identified rather than simply lumping them into a combined endpoint.
“Development of a Clinical Prediction Rule for 30-Day Cardiac Events in Emergency Department Patients With Chest Pain and Possible Acute Coronary Syndrome”
www.ncbi.nlm.nih.gov/pubmed/21885156