Emergency Medicine Literature of Note
Musings on Emergency Medicine, Clinical Informatics, & High-Value Care.
The BMJ article highlighting conflict-of-interest in medical guidelines has a thriving Rapid Response section – including responses from Michael Bracken, Peter Sandercock, Fred Geisler, and David Newman. A fascinating microcosm of sorts of some of the recent controversies in Emergency Medicine!
“Why we can’t trust clinical guidelines – Rapid Responses”
http://www.bmj.com/content/346/bmj.f3830?tab=responses
There’s been a little hullabaloo recently regarding the influence of financial conflicts-of-interest on guidelines – the result of a recent BMJ investigative report. But, what effect do these conflicts truly have? Is there any way to compare, side-by-side, a conflicted guideline with a non-conflicted guideline?
Why – yes!
In the very popular American Journal of Medical Quality comes this tiny gem, a comparison between two guidelines written just over a year apart. Both guidelines describe treatment options for Primary Immune Thrombocytopenia, and were both published in the same journal. One guideline was written by a financially untarnished societal group, while the other guideline was written by sponsored experts. In addition, the sponsored guideline had supplemental assistance by a professional scientific writing group funded by pharma.
Table 4 is a lovely, side by side comparison of the major treatment recommendations. Unsurprisingly, various thrombopoietin-receptor agonists and anti-D immunoglobulin received top billing in the sponsored guideline, while more conventional therapies were recommended in the non-sponsored guideline.
This article was, however, written by members of the non-conflicted guideline group – so, perhaps there’s some ulterior motive at work. Regardless, at least, it’s a fascinating look at the tangible effects of financial conflicts-of-interest.
“Conflicts of Interest and Clinical Recommendations: Comparison of Two Concurrent Clinical Practice Guidelines for Primary Immune Thrombocytopenia Developed by Different Methods”
www.ncbi.nlm.nih.gov/pubmed/23550214
Lidocaine as adjunctive treatment following cardiac arrest in the context of ventricular fibrillation/ventricular tachycardia has generally fallen out of favor and been replaced with amiodarone, a result of historical data and more recent trials. However, the quality of the evidence is generally poor, and confounders are frequent.
So, here’s some more chaotic, retrospective evidence gathered over the course of 16 years. This is the King County registry of OHCA, of which 1,721 patients with VF/VT and ROSC were identified. Of these, 1296 patients received prophylactic lidocaine after ROSC, in theory, to prevent re-occurrence of VF/VT. And, in both their unadjusted, adjusted, and propensity matched cohorts, there was a reduction in recurrence of VT/VF. However, in their propensity-matched cohort – which may or may not be a better tool for comparing two groups than their multivariate adjustment – there was no difference in admission rate or survivors to hospital discharge.
At the least, as these authors suggest, this data provides the clinical equipoise needed to justify prospective OHCA trials exempt from informed consent.
“Prophylactic lidocaine for post resuscitation care of patients with out-of-hospital ventricular fibrillation cardiac arrest”
www.ncbi.nlm.nih.gov/pubmed/23743237
Answer: Not isolated subarachnoid hemorrhage.
Caveats abound, of course. In this retrospective review of 404 patients transferred to a facility with neurosurgical capabilities for traumatic intracranial hemorrhage, only 48 suffered subsequent neurologic deterioration. Of this cohort, 75 were isolated traumatic SAH, and only 1 deteriorated – and that was due to an aspiration event in a 98 year-old male. Therefore, these authors, citing prior literature, feel this population appropriate for a prospectively-evaluated more-restrictive transfer protocol.
Their study is, of course, retrospective. Their sample size is rather small. There’s no granular data on characteristics of individual head injuries; not all traumatic SAH or SDH is created equal. Few of their odds ratios for deterioration were helpful in making predictions; isolated SAH had an OR for deterioration of 0.078, while SDH and epidural injuries were twice as likely to deteriorate as baseline (particularly since they tended to co-occur with other intracranial injuries).
But, in our era of cost-conscious medicine, these authors clearly describe the presence of a spectrum of intracranial disease that does not benefit from transfer or hospitalization.
“Patients with traumatic subarachnoid hemorrhage are at low risk for deterioration or neurosurgical intervention”
www.ncbi.nlm.nih.gov/pubmed/23694879
Modern trauma care has essentially, based on several articles published in the Journal of Trauma, migrated to 1:1 ratios of FFP to PRBC in massive transfusion. This was motivated in part by the outcomes observed in battlefield trauma associated with the use of whole blood. However, multiple individuals feel the true answer might lay somewhere in between – and that the observed association between equivalent transfusion ratios and survival is related to “survival bias”. In essence, trauma patients arriving the ED receive blood first – and the FFP only arrives if they survive long enough to receive it. Therefore, on retrospective analysis, the data is automatically skewed in favor of the folks who receive FFP, simply because they survived long enough to do so.
This is another retrospective review, but one that attempts to control for time-dependent exposure in the transfusion ratio. These authors generate hour-by-hour transfusion ratios throughout the first 48 hours of a patient’s hospital stay – and use these sort of time-dependent ratios to make their associations with outcomes. Using the “conventional” model – as other studies have done – there is a 2.50-fold (1.54 – 4.05) risk of death among patients with a low FFP:RBC ratio. With their “time-dependent” model, the adjusted hazard rate is only 1.25-fold (0.78 – 2.00).
There are, of course, a few problems with methods. Emergency transfusions where RBCs were in the Emergency Department had time imputed from retrospective analysis, rather than from blood bank issue time. In the 10 randomly selected patients the authors performed chart review to validate their analysis, they saw a median difference of 0.14 hours between issue time and transfusion time – but with an interquartile range of 0.0 – 2.5 hours. This means some of their analysis is based off extraordinarily imprecise data. Then, the reviewing anesthesiologist performing the outcome review was not consistently blinded to the transfusion ratio of the patient – which potentially biases the results in whatever manner suits the authors.
In the end, it’s another inconclusive data point in the massive transfusion literature. Luckily, there is a prospective trial already in progress.
“Effect of Plasma-to-RBC Ratios in Trauma Patients: A Cohort Study With Time-Dependent Data”
http://www.ncbi.nlm.nih.gov/pubmed/23782963
Anyone and everyone who works in the Emergency Department is aware there is a spectrum of individual practice. Certainly, any resident currently in training frustratingly knows this all too well, and can easily predict the prevailing culture of their shift in advance depending on their supervising attending.
So, it comes as no surprise this review of 389,120 Emergency Department visits in a three hospital, single-system review shows substantial variation in admission rate. There was, obviously, variation between hospitals – expected as different hospitals may have variable patient demographics. But, within hospitals – and across all three sites – there was up to a 2.3-fold (21% to 49%) variation in admission rate. As with all retrospective, observational studies, there are limitations inherent in the data set. However, the authors attempted adjustments based on several factors without substantially altering the outcomes.
There is no data on patient outcomes – particularly as relevant to those discharged by physicians with the lowest admission rates. Considering our culture of over-diagnosis and over-treatment, I expect, with further prospective or cross-sectional study, we would find the physicians with the lowest admission rates to have indistinguishable health outcomes from their peers. The factors that contribute to this variation – as well as interventions to reduce the variation – require further study.
“Emergency Department Physician-Level and Hospital-Level Variation in Admission Rates”
www.ncbi.nlm.nih.gov/pubmed/23415741
Urinalysis is frequently performed in elderly females presenting to the Emergency Department with non-specific symptoms – owing to the general trend of non-focal constitutional complaints failing to localize disease as age increases. Understandably, then, this population is thoroughly subjected to increased diagnostic testing.
And, of course, if you run enough tests, you’ll find some “answers”. The question, of course, is whether these “answers” are in fact true positives, accurately reflecting underlying clinical disease. In this retrospective cohort of 153 elderly patients for whom a urine culture was sent along with a urinalysis, nearly half of urinalyses were false positives. Adding in the not-insignificant incidence of non-pathogenic asymptomatic bacteriuria, some of these positive culture results were likely false positives as well. The authors of this article feel this reflects substantial overdiagnosis and overtreatment.
This assertion is almost certainly correct. However, these authors are short of suggestions regarding the improvement of diagnostic accuracy. They suggest, perhaps, urine samples ought only be taken by catheter, and treatment be initiated only on positive culture results. However, more practical and expedient diagnostic methods are lacking.
“Overtreatment of Presumed Urinary Tract Infection in Older Women Presenting to the Emergency Department”
www.ncbi.nlm.nih.gov/pubmed/23590846
A little hullabaloo today regarding a new report in the BMJ regarding problems with the creation of clinical practice guidelines, including a substantial portion of the article devoted to the conflicts of interest swirling about the use of tPA in acute ischemic stroke.
There have already been personal attacks – even from ACEP’s official twitter account – against the author. These seem to be motivated by the author’s coverage of the new ACEP/AAN clinical practice guidelines for tPA – and miss the overall point that patients are best protected and served by guidelines that are formulated in the absence of conflict-of-interest. tPA for acute stroke, erythropoetin endorsement by the National Kidney Foundation, the errant one-man crusade for steroids in spinal cord trauma, and cholesterol treatment guidelines are all discussed in the context of cautionary tales regarding the influence of conflict-of-interest.
Whichever side of the expand/limit tPA in acute stroke debate you fall upon, the issues of sponsorship bias, one-sided panelists on a still-controversial practice, and lack of open peer review for the ACEP/AAN guidelines ought to be unacceptable. Patients and practicing clinicians benefit from healthy debate and recognition of the limitations of the science, which seems clearly to have been lacking in the creation of these guidelines.
“Why we can’t trust clinical guidelines”
http://www.bmj.com/content/346/bmj.f3830
The holy grail of tPA proponents, the time-to-revascularization theory, is the subject of this most recent article in JAMA. This is a data mining exercise from the Get With The Guidelines-Stroke Registry – and, actually, there’s not a lot to say. They evaluate 55,000 patients from the registry, and there are significant differences between the cohorts receiving tPA between 0-90, 91-180, and 181-270 minutes – so all their outcomes are dependent on multiple statistical adjustments.
And, when I say multiple, I mean overwhelming:
The variables used in the risk models were patient-level and hospital-level risk adjustors that were expected to be predictive of outcome, based on empirical analysis, prior literature, and clinical judgment.
Patient-level factors included age, race/ethnicity, sex, medical history (including atrial fibrillation, prosthetic heart valve, previous stroke or TIA, coronary heart disease or prior myocardial infarction, carotid stenosis, peripheral vascular disease, hypertension, dyslipidemia, diabetes, and current smoking), stroke severity (NIHSS), an age-by-NIHSS interaction term, arrival time during regular work hours (7 AM-PM Monday-Friday), arrival mode (ambulance, private vehicle), and select classes of vascular risk prevention medications prior to admission.
Hospital-level factors included hospital size, region, teaching status, rural location, certified primary stroke center status, average number of patients treated with tPA annually, and average number of annual stroke dis- charges. All variables were included in the predictive models without use of a stepwise or other formal variable selection process.
The pharmaceutical industry conflict-of-interest disclosure is even longer.
These folks could have made these data say whatever they desired with their statistical weighting. They report a positive association with time-to-treatment and improved outcomes; the astute reader may interpret this as they are wont.
It is also worth mentioning the earlier time-to-treatment populations are probably more likely to include stroke mimics and TIAs – both of which tend towards excellent outcomes, with or without tPA. The percentage of stroke mimics treated with tPA ranges between 6.5% and 15.5% at academic centers using MRI as imaging confirmation, and has been estimated to be as high as 25% to 29% in community settings. The GWTG-Stroke registry specifically fails to account for stroke mimics in their coding instructions – a patient that receives tPA and rapidly improves is to be coded as “aborted stroke”, even though contemporary evidence throws this concept into doubt.
Regardless, the percentage of stroke mimics confounding the results likely dwarfs the magnitude of effect of the proposed time-to-treatment association reported by these authors.
“Time to Treatment With Intravenous Tissue Plasminogen Activator and Outcome From Acute Ischemic Stroke”
http://jama.jamanetwork.com/article.aspx?articleid=1697967
Not a day goes by, it seems, the New York Times or some other equally prominent journalistic source publishes a scathing invective regarding the irresponsible cost of healthcare in the U.S. In this context, it is equal parts highly entertaining and appalling to measure how ignorant Emergency Physicians are regarding the cost of the care they provide.
This single-center study performed in Philadelphia gives results similar to prior work in the same vein. 23 attending physicians and 21 residents were surveyed regarding estimates of cost of care for 102 of their patients discharged from the Emergency Physician. Each estimate for the total cost of care was compared with the actual final charges billed to the patient or their insurance carrier.
Median estimated charge: $1,268
Median actual charge: $2,175
There was no difference between attending and resident performance.
Not an encouraging result – particularly as patients are likely going to be burdened with ever-increasing portions of their healthcare costs, and we ought to be able to communicate with them the cost of care as part of shared decision-making in the Emergency Department.
“Emergency physicians’ knowledge of the total charges of medical care”
www.ncbi.nlm.nih.gov/pubmed/23685055