Monday, October 12, 2015

STONE Score Close, But Not Quite

The STONE score is a lovely idea for a common problem – ureteral stones are a frequent source of low-value imaging in the Emergency Department.  An adequately sensitive or specific decision-instrument would improve diagnostic accuracy while reducing imaging.  Unicorns would bloom from the frozen tundra.

STONE, however, has a couple issues.  Firstly, it was derived and validated at a single institution in similar cohorts.  Secondly, it has a few variables that probably lack face validity, i.e., three points for non-black ethnicity?

This report is from the large, multi-center trial comparing CT vs. ultrasound for the diagnosis of ureterolithiasis.  Of the 2,759 patients randomized from the original trial, 845 underwent CT as gold standard and had the data available for evaluation of the STONE score.  This constituted their retrospective validation cohort.

So, STONE is OK.  Not great, but OK.  The AUC of STONE was 0.78, which is a step up from 0.68 of physician gestalt.  However, it was only 87% specific at the high end in the validation to rule-in stone, and, then, 96% sensitive at the low end as stone rule-out.  Phrased otherwise, there were about 10 to 20% fewer cases of ureterolithiasis identified in the moderate and high cohorts in this validation, compared with the original.  There were also, unfortunately, a greater number of significant alternative diagnoses in the low and moderate cohorts, as well.

The authors come to very reasonable conclusions.  First, the STONE score isn’t perfect.  It can be a useful adjunct to risk-stratification and shared decision-making, but the positive and negative likelihood ratios are inadequate as a standalone rule.  As a corollary, the score definitely requires refinement or reinvention, based on their analysis of the predictive contribution of different variables in the original score.  It is probably still fair to use this score to supplement gestalt in the context of pursuing judicious use of resources, but it must be considered in the context of other predictive features to determine the appropriate imaging strategy.

“External Validation of the STONE Score, a Clinical Prediction Rule for Ureteral Stone: An Observational Multi-institutional Study”

Friday, October 9, 2015

Again, One Troponin is Enough

As we’ve seen suggested in prior work, an undetectable high-sensitivity, high-precision troponin at presentation is a powerful predictor of minimal risk in the short-term.

In this prospective study based at four hospitals in Scotland, 6,304 eligible patients with suspected acute coronary syndrome were captured for analysis across three prospectively gathered cohorts.  These patients were split across a derivation cohort to establish a threshold with an appropriate negative predictive value, an internal validation cohort, and an external validation cohort.  Their target for a 30-day MACE of MI or cardiac death was a negative predictive value of 99.5%.

Making essentially a long story short, they found a threshold of 5 ng/L established this minimal-risk population, with only 9 patients meeting the primary outcome out of of 2,302 patients with an initial troponin less than the threshold.  And, if you want to be even more airtight, 6 of these presented within 2 hours of the onset of chest pain, and 3 had apparent ischemic changes on their initial EKG – which may have been picked up by a rapid repeat testing protocol in the ED for some patients.

There are a few holes in this paper, of course.  Not all patients were hospitalized and had such repeat testing, so some small nSTEMI or vasospasm events could have been missed.  Patients in this study required a median of 54 minutes for blood sampling after presentation to the ED, so some caution should be exercised regarding repeat troponin testing if your department is efficient regarding phlebotomy at presentation and onset of symptoms was just prior to arrival.  But, on the whole, this greatly adds to the body of evidence we’ve been building – that cutting edge troponin assays alone can provide powerful prognostic information.

Now, what to do with all the “intermediate” cases ….

“High-sensitivity cardiac troponin I at presentation in patients with suspected acute coronary syndrome: a cohort study”

Wednesday, October 7, 2015

Is NODESAT Overhyped?

In the last few years, we’ve had a little bit of a sea-change in oxygenation during intubation.  We’ve stopped relying solely on pre-oxygenation to bridge our patients through apnea, and started providing passive oxygenation during intubation.  Usually supplied by high-flow nasal cannula, this takes advantage of physiology and diffusion to distribute oxygen into circulation.

But, as these authors state, the evidence for this practice is spotty – mostly observational evidence from controlled intubation settings.  Our critically-ill patients hardly have the same physiology as those undergoing elective airway procedures, and are generally less responsive to oxygenation adjuncts.  So, this is the FELLOW trial, a pragmatic, open-label randomized trial comparing apneic oxygenation vs. “usual care” – which was none.

With 150 patients in their intention-to-treat analysis, this cartoon sums up the results sufficiently:

Not much difference!

Their two groups were relatively well-balanced in terms of physiology and airway comorbidities.  The intubating operators were reasonably experienced (median >50 intubations), and 2/3rds of the patients were intubated on the first attempt.  There were probably no important differences in pre-oxygenation or procedural factors.

But, it is quite a small trial.  There are small differences here favoring the apneic oxygenation arm that simply might not reach statistical significance.  The exclusion criteria included “if the treating clinicians felt a specific approach to intra-procedural oxygenation or a specific laryngoscopy device was mandated for the safe performance of the procedure”, which could have introduced a selection bias.  The open-label effects may or may not be confounding.  The ICU environment and exclusion criteria also affect generalizability to the Emergency Department.

In the end, the answer is: apneic oxygenation still probably helps, particularly considering the pre-study evidence favored the intervention, and this one study does not move the needle much.  However, the observation here of a clinically unimportant effect size is not unreasonable.  If the effect size is small, the cost of an intervention becomes important.  However, in this case, the cost is fairly minimal – a small addition to set-up time and procedural complexity.  Considering the low cost and the post-test odds still favoring the intervention, it would be erroneous to stop providing apneic oxygenation based on this trial, and further study is indicated.

“Randomized Trial of Apneic Oxygenation during Endotracheal Intubation of the Critically Ill”

Monday, October 5, 2015

When Peripheral Vertigo Isn't

But, such misdiagnosis is rather rare.

These authors conduct a retrospective, administrative database study of patients discharged from an Ontario, Canada, Emergency Department between 2006 and 2011 with a diagnosis of peripheral vertigo.  They identify 41,794 patients with appropriate data and an explicit diagnosis of peripheral vertigo, and follow them for adverse outcomes.  Most importantly for this paper, looking at subsequent stroke diagnoses, they find 57 (0.14%) received a diagnosis of stroke within 7 days.  To these authors, this number represents a miss rate for actual cerebrovascular causes of vertigo on initial presentation.

It’s a little hard to put great confidence in this precise number, however, because there were actually 270,865 visits discharged with diagnoses of dizziness & vertigo.  The largest cohort of these, 71% of them, were discarded from analysis because their diagnosis was “Dizziness, giddiness, lightheadedness, vertigo NOS”, which was not specific enough for these authors' purposes.  It would have been enlightening to have included this less-specific cohort in this analysis, at least as a parallel comparison group, for additional insight.

However, the authors do something I quite like in this paper, and something I think would improve the robustness of many other studies:  a propensity-matched analysis attempting to establish a baseline level of risk.  For these authors, they choose renal colic – a generally benign, yet frequent, condition, that ought confer no special additional risk of stroke.  These authors have no difficulty finding 34,872 of patients to pool into a cohort relatively well-matched on important risk factors.  These renal colic patients have very similar rates of falls and fractures in the short-term after discharge – but they have many fewer strokes.  Most of the difference between cohorts arises in the first 7 days, with 50 occurring in the vertigo cohort versus zero in renal colic.  Between day 7 and and 365, however, the rate difference narrows, with 74 strokes in the vertigo cohort and 49 in renal colic.

Despite the massive limitations of retrospective review and relying on diagnosis codes, I think the authors’ general observation is correct.  Dizziness can be a quite challenging diagnosis to evaluate in the Emergency Department, and clearly some patients are being erroneously given diagnoses of peripheral etiologies.  However, the rate of misdiagnosis in this cohort is likely somewhere better than 1 in 700, which I find generally clinically acceptable considering the impossibility of utilizing advanced imaging in such a vast population.

We can certainly endeavor to do better, but we have a lot of work cut out for ourselves in improving the specificity of our clinical evaluation first.

“Outcomes among patients discharged from the emergency department with a diagnosis of peripheral vertigo”

Friday, October 2, 2015

A Call to Retire Routine C-Spine Immobilization

Remember the heartwarming children’s tale of the young woman with the ribbon around her neck?  Once removed, the head falls from the body, and the unfortunate woman expires.  Thus, the inspiration and evidential basis for modern trauma care utilizing immobilization of the cervical spine.

This article asks a very simple question: is there evidence to support the notion of incidental cerebrospinal movement resulting in subsequent paralysis?  Their answer: probably not.

These authors review several thousand abstracts to extract twelve publications describing a mere 41 cases of patients who were not completely immobilized, and thence suffered subsequent neurologic deterioration.  The individual cases reported upon provide, generally, rather spotty detail regarding the circumstances.  Some patients had additional falls or trauma in the ED, others were intoxicated and combative, while many others seemed to have gradual worsening without a specific event.

These authors propose this gradual worsening represents the primary time scale of neurologic deterioration – and suggest the suspected precipitating events documented by these cases represent contextual red herrings.  Rather than becoming tetraplegic as a consequence of repositioning in the Emergency Department, it is more likely clinical manifestations result from ligamentous disruption, bleeding, and edema related to the primary injury, and immobilization would not have prevented their progression.

Of course, this paucity of documented examples cannot represent an exhaustive report of all known secondary deterioration after an initial, non-immobilized injury.  However, just as erroneous is the presumption that immobilization prevents such secondary deterioration – particularly when coupled with the known inconveniences and harms of mass immobilization during transport and evaluation.

It’s time for routine cervical spine immobilization to go!  However, such discontinuation need be undertaken in such a setting as capable of detecting any adverse events resulting from such.

“Early Secondary Neurologic Deterioration After Blunt Spinal Trauma: A Review of the Literature”

Wednesday, September 30, 2015

A Very Odd Look at CT In the ED

Why do we perform CTs in the Emergency Department?  It’s fair to say the primary indication is diagnostic certainty: the ruling-in or ruling-out of a disease process of substantial clinical relevance.  However, this study begs the question: have we lost touch with this concept of "substantial clinical relevance"?

This is a qualitative study evaluating physician decision-making in the context of CT ordering.  These authors provided physicians, approximately 2/3rds attending physicians, a questionnaire pre- and post-CT for 1,280 patients in the Emergency Department.  The main gist: what are you worried about?  How confident are you in the diagnosis?  And, then, after CT, how about now?

The bullet-point summary:
  • Physician confidence in their diagnosis grew after CT.  Splendid.
  • CT excluded or confirmed alternative diagnoses in 95+% of cases.  Excellent.
  • Increasing pre-CT confidence in a leading diagnosis was associated with lesser changes in leading diagnosis post-CT.  OK.
  • Many pre-CT leading diagnoses were benign, but with low physician confidence.  Except for CT head.
  • Nearly 3/4ths of CT scans performed of the head had a leading diagnosis of “Benign headache” or other, had no change in diagnosis following CT, and confidence was generally pretty high.  This is awful.
  • Finally, if you were hoping a CT would prevent bouncebacks: no.  15% of abdominal pain returned within a month for related reasons, as well as 14% of chest pain/dyspnea, and 11% of headache.
CT is an important tool.  It certainly makes the life of the risk-averse physician much, much easier.  However, the instances in which CT identified an important diagnosis in this study are certainly in the minority – most final diagnoses were either benign or could have been achieved through other means.  Unfortunately, very few specific actionable items can be taken away from this study – excepting CT for headache (ugh) – but it certainly shows there is fertile ground for a culture change to take root and decrease low-yield CT utilization.

“CT in the emergency Department: A Real-Time Study of Changes in Physician Decision Making”

Monday, September 28, 2015

The “Routine” Chest X-Ray

Many presenting complaints in the Emergency Department call for cardiothoracic imaging.  Some can be assessed by point-of-care ultrasound, but, for the most part, plain radiography is the established routine.  Whether the pretest probability of disease warrants such widespread use is one matter.  This article documents yet another – duplication of imaging.

These authors review four years of radiology from their institution and document 3,627 patients for whom both CXR and chest CT were ordered.  Their main analysis breaks down the use of radiology mostly looking at the order of which these studies were requested, and whether results from one were available prior to the completion of the other.

For the most part, the CXR was ordered first, and the images were available for review before the subsequent CT chest.  However, in 354 (9.8%) cases, the CXR images hadn’t even yet been acquired when the CT chest was ordered.  This probably generally overlaps the 134 (3.7%) cases where the CT chest was ordered simultaneously or prior to the CXR.  Regardless – if the results were clinically irrelevant, why order the test?

I think it’s fair to say many of the CXRs included in this study were pointlessly redundant – especially when the decision for CT was obviously made prior to their acquisition.  No doubt the CXR is included in most ED protocols for certain chief complaints, and is ordered reflexively without thought.

Looking for waste to target in the system?  Here you go.

“Inefficient Resource Use for Patients Who Receive Both a Chest Radiograph and Chest CT in a Single Emergency Department Visit”

Friday, September 25, 2015

Central Line Showdown

Major complications related to insertion of central line catheters remain: infection, thrombosis, and “mechanical complications”(read: pneumothorax, arterial puncture).  And, central lines have three typical locations:  subclavian, jugular, and femoral.

Wouldn’t it be cute if each of these catheter insertion sites each had their own specific realm of superiority?

And, based on this randomized trial of 3,471 catheter insertion events, they just about do:

  • The femoral site has the fewest mechanical complications, but the most thrombosis (1.4%).
  • The subclavian site had the fewest infections (0.5%) and thrombotic events (0.5%), but the most mechanical complications (2.1%).
  • And the jugular site was essentially a middle ground between the two, although, had the greatest infection rate (1.4%).

However, many of these differences need be taken with a grain of salt.  Ultrasound was not mandated, which probably led to the unusual incidence of mechanical complications for the jugular approach.  Specific antimicrobial dressings were not used regularly.  There was a high rate of failure and crossover in the subclavian arm (14.7%).  Then, the number of complications for each were measured typically in the near-single digits, while 469 patients died before catheter removal – a large enough number of potentially unmeasured events to significantly affect the primary outcome.  Only symptomatic patients were screened for thrombosis – again, leaving many patients with potentially missed outcomes.

So, in the end, what’s likely best?  A skilled subclavian – or one guided by an ultrasound approach – is probably ideal, but should be avoided if clinician comfort is low.  Following that, an ultrasound-guided jugular approach is likely best.  It is not reasonable to suggest the femoral site would routinely be superior to either approach, but these very low rates of complications indicate it need not be shunned when clinically appropriate.

“Intravascular Complications of Central Venous Catheterization by Insertion Site”

Wednesday, September 23, 2015

The “No Objective Testing” Rule for Chest Pain

As “zero-miss” philosophies for chest pain proliferated, so did the “early objective testing” phenomenon.  Chest pain presentations of all colors and ages, following an initial evaluation in the Emergency Department, have been invariably referred for inpatient observation and stress testing.  The problem?  Most of these tests are performed in patients for whom the pretest likelihood of coronary artery disease is low, and the incidence of false-positives outweighs that of true-positives.

This rehash of prior observational data aims to address this issues – specifically deriving a prediction rule for those who are so low risk any follow-up testing ought be obviated.  Based on cohorts enrolled for two previous studies of potential acute coronary syndrome presentations in New Zealand and Australia, these authors track 2,396 patients for 30 days.  Following an initial negative ED evaluation, 5.6% were diagnosed with acute coronary syndrome within 30 days – mostly “unstable angina”, and mostly based on follow-up objective testing.  Using standard statistical bootstrapping techniques, the authors partition out 31.1% of this cohort as having a <1% risk of ACS within 30-days: age <50, <3 risk factors, and no prior CAD or MI.  The authors suggest patients meeting these criteria, following a negative ED evaluation, need not be referred for any objective testing.

The puzzle here is in the outcomes measured: 30-day AMI, cardiovascular death, unstable angina, or revascularization.  The leap, then, involves associating a lack of these outcomes with the disutility of provocative or anatomic testing.  It is probably true there is colinearity between clinically important coronary artery disease and incidence of major adverse cardiac outcomes, but this is a bit tenuous of a foundation for their conclusions.  A lack of 30-day adverse event does not exclude benefit from cardiac imaging, and, vice versa, the presence of a 30-day adverse event does not imply a benefit from cardiac imaging.  However, given the lack of other useful evidence, this is as valid a proxy as we have to inform practice.

Finally, what tends to disappoint me most in these sorts of articles is something of a disconnect between the study design and the clinical questions relevant in practice.  Chest pain presentations lay somewhere on a bell curve.  At the low end, a subset of young and healthy folks are mostly harmed by overdiagnosis and additional resource utilization.  Then, at the high end, patients with multiple risk factors and known disease can no longer be substantially helped by additional invasive interventions.  Each of these cohorts is appropriate for early discharge from the Emergency Department without additional cardiac imaging, yet the focus is persistently on only the very low-risk end of the spectrum.  I hope future study is able to provide a more sophisticated analysis of the features of moderate- and high-risk patients to identify the individuals within these cohorts who do benefit from cardiac imaging, and which specific testing best informs downstream care.

“A Clinical Decision Rule to Identify Emergency Department Patients at Low Risk for Acute Coronary Syndrome Who Do Not Need Objective Coronary Artery Disease Testing: The No Objective Testing Rule”

Monday, September 21, 2015

The Internet Knows If You’ll Be Dead

As another Clinical Informatics “window into the future” – a window into the future.

These authors used three years of electronic health record data to derive a predictive Bayesian network for patient status.  Its scope: home, hospitalized, or dead.  There are many simple models for predicting such things, but this one is interesting because it attempts to utilize multiple patient features, vital signs, and laboratory results in a continuously updating algorithm.  Ultimately, their model was capable of predicting outcomes up through one week from the initial hospitalization event.

Some fun tidbits:

  • What mattered most on Day 1?  Neutrophils, Hct, and Lactate.
  • As time goes by, the network thinks knowing whether you’re on the Ward at Day 3 is prognostic.
  • By Day 5, variables like a simple count of the total number of tests received, the presence of cancer, and albumin levels start to gain importance.

Their Bayesian prediction network was best at predicting death, with an average accuracy of 93% and an AUROC of 0.84.  Similarly, the prediction engine was most accurate on Day 1, with an average accuracy for each outcome of 86% and an AUROC of 0.83.  Overall, for the entire week and all three outcomes, the AUROC was 0.82.

What was also quite interesting was the model, while also predicting outcomes during the index hospitalization, also detected readmission events within the time period scope.  The authors provide a few validation examples as demonstrations, and include a patient whose probability of hospitalization was trending upwards at the time of discharge – and subsequently was readmitted.

Minority Report, medicine style.

“Real-time prediction of mortality, readmission, and length of stay using electronic health record data”