Improving Outcomes, Mythbusting, Neurology

Early vs late meningitis diagnosis: capturing the needle in the haystack

Needle in the haystack, infectious pathway, take 6.

This is a retrospective study looking at early vs late diagnosis of bacterial meningitis from three hospitals in Denmark (one looking at data from 1998-2014; the other two from 2003-2014). To be eligible, patients had to be >15 years of age, and, obviously, had to be hospitalized with a clinical presentation consistent with possible community acquired meningitis (any combination of headache, neck stiffness, fever, altered mental status, petechiae) with no alternative diagnoses made during or after admission. Furthermore, all patients also had to have a proven bacterial etiology by either: positive CSF culture, positive blood culture and CSF with >10 wbcs, bacteria seen on CSF gram stain, or bacteria in CSF by PCR or antigen analysis.

So what is early and what is late diagnosis? They define “early diagnosis” as being recognized in the ED (1.3 hours to antibiotics median), and “late diagnosis” as, well, not diagnosed in the ED (ie, diagnosed on the wards- 13 hours to antibiotics median). Over roughly 15 years, they saw 358 cases of bacterial meningitis, (~8 cases per year per institute – seems a bit high? They do not mention total number of annual ED visits), with 32% being classified as diagnosed “late.” … so, probably 2-3 cases a year of “late” diagnosis – a true needle in the haystack.

Why the late diagnosis? They tended to be older (65 years of age vs 56), less likely presenting with headache (58% vs 82%), less likely with neck stiffness (36% vs 78%), less likely with fever (59% vs 78%), with the classic triage of AMS, fever, and neck stiffness was only present 20% of the time in the late diagnosis group vs 50% in the early diagnosis…. So, it wasn’t an easy catch.

Why does this matter?  Welp, with early antibiotics having a positive effect on mortality (18% vs 36%) as well as unfavourable outcome (which they do not actually define, 37% vs 66%, in favor of early antibiotics).  This is a HUGE difference in mortality and unfavourable outcomes if you do not catch it early!  … Then again, do we do more harm by giving 1-2g of ceftriaxone to everyone who is a bit altered?  Would the risk of cdiff then outweigh the 2-3 annual misses? I’m not so sure.  What about the recurrent headaches and repeat visits for post-LP headaches?

If you really want to tease out the data a bit, 53% of late diagnosis patients vs 26% or earlier diagnosis patients had a head CT before the LP. 72% of “late diagnosis” patients tentatively had a non-infectious etiology- so let’s explore some of the tentative diagnoses:

loss of consciousness (19 patients)

stroke (12 patients)

intracranial / subarachnoid hemorrhage (7 patients)

impaired mental status (6 patients)

headache (5 patients)

back pain (5 patients)

seizures (5 patients)

loss of vision (2 patients)

(among others)


What I’m seeing here is a a trend towards a neurologic issue (a CT scan, a diagnosis of syncope / seizures, AMS, etc) – which may indicate that the thought of meningitis (or even endocarditis) may not have been entertained. Cant make the diagnosis if you dont think about it. In a similar vein, this diagnosis is rare and runs across a spectrum – on one end, the febrile, meningeal and altered, on the other, the vaguely unwell.  And that, surprisingly, even a 12 hour delay to antibiotics can wreck havoc on the patient.

The take home points?  Be vigilant, entertain the spectrum of disease for meningitis, but remember that every decision you make has consequences, including the decision to, and not to, perform an LP, not to mention the decision to indiscriminately give antibiotics for those “altered”.  Choose wisely, and remember there is no such thing as zero risk.


Dedicated post-arrest services? meh.

Should we regionalize post-arrest care?  Well, if your facility does not have a cath lab, then the answer is yes.  But Academic Hospital A, which sees >100 arrests a year, just started advertising a fancy post-cardiac-arrest service.  Academic Hospital B also sees >100 arrests a year, but does not have a post-arrest service aside from their MICU.

You, being at hospital C without a cath lab, have just achieved ROSC in a witnessed arrest. Who do you transfer to?

This study looks at 987 post-arrest patients that survived to admission at 7 hospitals in and around Southwestern Pennsylvania. One of them is a regional referral center with post cardiac arrest services consulted on OHCA with ROSC, accepts sudden cardiac arrests from outside facilities, and is consulted on in-hospital arrests with ROSC. There are two additional tertiary care centers that see >100 SCA annually, and 4 “low volume” centers. They look at multiple variables, and evaluate discharge disposition, discharge CPC, and length of survival post-discharge.

They improved numbers of discharge CPC – the post arrest service center with a discharge CPC of 1 or 2 32% of the time vs 37% of the time for the other 6 facilities. More patients were discharged to home (41% vs 32%) from the post-arrest service center and survived for longer if they were treated with the post-arrest service.


While the authors claim similar patient characteristics between the post-arrest service center and the other 6 hospitals…. 46% of patients were transferred to the post-arrest service center vs 16% at hospitals 2-7 – perhaps skimming a healthier patient that made it through the transfer (remember- you had to survive to discharge to be counted) – the authors even acknowledge that their transferred patients did better than their other arrests.

Add in that the initial rhythm 51% of the time was VT/VF for the post-arrest service center vs 41% in the other six hospitals, and you’ve got plenty of confounders. Frankly, given all of this, it’s a bit strange that the proportion of patients surviving to discharge did not differ at all. One would think if you have a post arrest service and the scales are tipped in your favor to begin with, that you’d have a higher percentage of patients surviving.

Ultimately, patients lived longer post-arrest when treated at a facility with a post-arrest service, and the authors are touting this as reason to (further) regionalize post-arrest care.  Sure, there are slightly better neurologic outcomes, but the scales were tipped in their favor to begin with. I don’t trust this conclusion, especially when the post-arrest service had an advantageous patient population to begin with that should have led to a measurable increase in improved survival, in addition to an increase in length of survival.

Cardiology, Cardiology, Critical Care, Improving Outcomes, Improving Throughput, Mythbusting, Pulmonary, Radiology, Radiology

Probing the dyspneic patient.

For undifferentiated dyspnea, how would you like to have an accurate diagnosis in 24 minutes?

I love this study.

Basically, for all dyspneic patients (not trauma related, and over age 18), 10 EP’s were given an H&P, vital signs, and an EKG, as well as access to a Chest X-Ray, Chest CT, cardiologist performed echo, and labs including an ABG.

These same 2,683 patients, in tandem, had point of care ultrasound testing (lung, IVC, echo). Here’s the catch – the ultrasonographers were only provided the H&P, vital signs, and EKG then asked to make a diagnosis. The treating provider was blinded to POCUS diagnosis.

These numbers for diagnostic accuracy of POCUS are astounding.

+LR for acute HF? 22 (-LR 0.12)

+LR for ACS? 105 !!!

+LR for pneumonia? 10.5 (-LR 0.13)

+LR for pleural effusion? 95 (-LR 0.23)

+LR for pericardial effusion? 325!!! (-LR 0.14)

+LR for COPD/asthma? 22 (-LR 0.14)

+LR for PE? 345!!!

+LR for pneumothorax? 4635!!! (-LR 0.12)

+LR for ARDS? 90

Yes, for certain things like pneumonia, the difference in p-values between tradition means and POCUS diagnosis was not significantly different, but what about volume status? I cant imagine blindly giving 30 cc/kg would benefit the patient with a plethoric IVC and pleural effusion. There is some elegance a play here.

Additionally, sure, ED diagnosis for ACS had a higher LR, but they also had a cardiologist performing and interpreting echos in the ED (a rather rare siting in a US ED I would imagine) – without much improvement in their -LR (0.53 vs 0.48). For PE, the -LR of POCUS was predictably mediocre if not outright bad (0.6), while the -LR for ED diagnosis of PE, with the benefit of chest CT, was -0.10.

Now look, I get that these EP’s were quite sono-savvy. They all had 2+ years of experience, over 80 hours of ultrasound lessons & training, with at least 150 lung and 150 ED echo’s under their belt. The diagnosis was made in 24 minutes with POCUS in comparison to 186 minutes for traditional means. And while most of us can not do a year+ ultrasound fellowship, and neither can we all be as savvy with the probe as these authors (or Matt, Mike, Jacob, Resa, Laleh, etc) – it does not mean we shouldnt try. You can still greatly increase your yield just by practicing. To boot, the cognitive offload you experience by saving yourself a few hours by (correctly!) knowing which direction you are heading with a patient is an immense boon to both your mental heath & your patients well being.

Improving Outcomes, Mythbusting, Pulmonary

Sono-guided ACLS

This study demonstrates what many of us have probably suspected – the absence of cardiac activity on ultrasound portends a grave diagnosis; but this study really is so much more.

Utilizing 20 sites across the US and Canada from May 2011-Nov 2014 looked at all nontraumatic in-ED and out of hospital cardiac arrests that arrived to the ED in either PEA or asystole, and whether or not POCUS demonstrated a potential role in resuscitation.

953 patients, 793 used for final analysis (106 not included due to resuscitation under 5 minutes, 8 patients DNR, 1 uninterpretable sono, 3 with incomplete timing data, 42 for no ACLS meds given) – had a cardiac sono at the “beginning and end of ACLS.” The primary outcome was percentage of patients that survived to hospital admission, with secondary outcomes of survival to discharge and ROSC. Unfortunately, neurologic intact survival was not evaluated. The treating EP’s were credentialed in POCUS at their local institutions and unblinded. Digital clips were reviewed by a single reviewer in a blinded fashion for agreement (which was deemed to be “substantial agreement”).

The data (numbers are percentage, such that “28.9” = the percentage of patients with cardiac activity on POCUS during the resuscitation who survived to admission):



Cutting to the chase, this study brings up a number of key points:

-PEA on the monitor may not necessarily be PEA, with a whopping 54% of patients having cardiac activity on POCUS

-asystole on the monitor may not be cardiac standstill, as 10% had cardiac activity on POCUS

– survival to admission with cardiac activity on POCUS is MUCH higher – 28.9% vs 7.2%, but….

– cardiac activity on POCUS for PEA/asystole portends only a 3.8% survival to discharge

-no cardiac activity = poor prognosis, 0.6% of patients survived (3 out of 530). With two of the three patients were Vfib at some point during EMS working on them.

-pericardial effusion was seen in 34 patients (4.3% of those in the final analysis). 15.3% of patients whom had a pericardiocentesis performed survived to discharge.

– only 15 patients received lytics for suspected PE, with only one (6.7%) surviving to discharge. (which was almost the MORTALITY rate of PEAPETT)


Whew. This is a lot to digest. Let’s just say that ultrasound helps you tease out a spectrum of disease and further characterizes what you are dealing with. I’m looking at POCUS in codes as a risk stratification tool. Is there a prolonged time without cardiac activity without a potentially reversible causes? Might want to consider calling it earlier since survival to discharge is abysmally low. And sheesh… 1 out of 25 cardiac arrests had a pericardial effusion??? Wow. Time to brush up on those pericardiocentesis skills.

Caveats- this was done by EP’s credentialed for POCUS, so they’re likely more talented than the rest of us.  Dont let that scare you though, rather, this.  Perhaps seeing cardiac movement on ultrasound lends a “bridge to hope” and the team puts in a more-heroic-than-usual effort.

And of course, this also leads to more questions- of those 28.9% with cardiac activity that survive to admission, what if they are brought straight to the cath lab? Or started on ECMO? Would this potentially alter survival rates and neurologically intact survival in meaningful ways? Time shall tell.  Until then, cut that KT window, pick up the probe, and have your TPA & long pericardiocentesis needle ready.

Cardiology, Critical Care, Improving Outcomes

Brady arrest? Go Full Bore.

Do you believe in full bore medicine?

(hat tip to SMACCdub for that line)

This paper puts those thoughts under the microscope a bit, and challenges us to think ahead and be prepared.

They looked at all OHCA from 2006 to 2012 with initial brady/asystolic arrests to determine if they may benefit from pre-hospital pacing, and to look at survival rates associated with various rhythms. Clear non-cardiac causes (trauma, drowning, respiratory, neurologic, suicide) were excluded.

7925 OHCA in the Netherlands

less non-cardiac (6681 patients)

less those without EKGs (~500 patients)

less ~3000 patients with VF/VT (now at 2643 patients)

less those with normo/tachycardia and those with pacers previously placed (~300 patients)


This leaves us with 2333 brady/asystolic (idioventricular, junctional, sinus brady, 3rd degree with/without escape, asystole) patients – or, about 30% of their OHCA.

Unwitnessed arrest still protends a poor outcome, with survival about 0.5%. However, for witnessed arrests, they report 4% survival for idioventricular / junctional arrests and 6.8% for sinus brady arrests. This seems consistent with prior studies. However, for a study trying to determine whether or not pacing is beneficial, their pace rates were quite low. They paced 11 of 220 sinus brady patients and 41 of 452 idioventricular / junctional patients, with a delay of 30.1 and 16.5 minutes to pacing respectively – with an electrical capture rate of 55% and 70% to boot.  Esssentially, they can’t answer the question “Does pacing help” with such a care gap.

So, why is this? For sinus brady, maybe patients are hanging in the 40’s-50’s and felt to be quasi-stable. Maybe its the angst of floating a pacer. Perhaps the lengthy delay for sinus brady is giving atropine, then giving it again… and maybe again- akin to pressor-angst for sepsis (giving a 4th, 5th, and 6th liter rather than starting pressors or a central line).  I imagine there is a mental barrier – whether it be not thinking about pacing or passing the buck (“I’ll let the ICU figure it out.”). The evolution of the ED-ICU model (and perhaps UPMC’s cardiac arrest unit) may be the best place to look at this type of “full bore” medicine and whether or not it would be beneficial.

But for now, there is a large gap in care. Bradycardic arrests represent about 10% of arrests, have a reasonable survival rate, and are (potentially) suboptimally managed – and you have the tools to potentially improve an outcome.  We can not say whether or not pacing is futile care for this condition.

Until then, go full bore. Your patients & their families deserve it until pacing is demonstrably shown to not be beneficial in bradycardic arrests.

Critical Care, Improving Outcomes, Mythbusting, Neurology

Compazine… for infectious disease?

Today’s article’s (1, 23 ) are a break from the usual trials that are typically discussed and a bit more “benchside medicine” than bedside medicine.  In fact, let’s look at this as an early request for one of the 12 trials of Christmas.

Phenothiazines have demonstrated in vitro (as well as some in vivo) activity for gram positive cocci, mycobacteria, amoeba (4; 5), and some gram negative rods.

It should be noted that Klebsiellae, pseudomonads and acenetobacters were highly resistant to almost all of these drugs.

The MIC for phenothiazines are usually not reached with conventionally used doses, but these compounds do enhance the activity of various antibiotics to which various bacteria are susceptible (including vancomycin), and even decrease the MIC of resistant organisms.

So where am I going with all of this? For starters, lets look at some common causes of meningitis, in no specific order:

Strep pneumo (gram positive); group B strep (gram positive); staph aureus (gram positive); Listeria (gram positive); Neisseria meningitidis (gram neg diplococci); H flu (gram neg)

All things phenothiazines are thought to have activity against.

You’re likely to be giving patients with potential meningitis something for pain (I hope?), so why not go with compazine?  Likewise, patients whom you may suspect bacteremia from a cellulitis, why not give compazine to, ummm, “counteract the nausea” associated with the opiates you gave for pain control?

I think this falls into the unlikely to harm, might help category, and is seemingly a ripe area for research.  Is this practice changing?  Nope, not at all.  Food for thought, but until compazine is proven unsafe in an infectious process, I will continue my love affair with compazine for headaches, nausea, and vomiting (regardless of suspected etiology).

Critical Care, Improving Outcomes, Mythbusting

1 in 10 EGLS saves a life.

Do current sepsis guidelines go far enough?

That was my first thought when I read today’s article. This single center ICU study looked at 220 patients divided into two categories- one category in which patients that were managed in adherence with the 2012 surviving sepsis guidelines – 20-49 ml / kg initial IV fluid bolus, continued fluid challenges until CVP of 8-12, with more given based on treating team. Noradrenaline until MAP of 65, and dobutamine for cvSO2 <70% in combination with either lactate >2 or urine output <0.5 ml / kg / hr). The other 110 patients had treatment guided by limited echo:

Treatment options looked like this:

1) IVC fluctuation <15% & normal LV function= give pressors only (discontinue IV fluid)

2) IVC fluctuation >15% & normal LV function = 20-40 ml /kg IV fluid given

3) IVC fluctuation >15% & mod/severe LV function = 10-20ml/kg IV fluid given AND initiate dobutamine 5ug/kg/min

4) IVC fluctuation <15% & mod/severe LV function = discontinue IV fluid and initiate dobutamine 5 ug/kg/min


These patients were pretty sick- all patients were mechanically ventilated and on noradrenaline. Limited echo was performed within 24 hours of presentation to ICU and within 36 hours of presentation to the ED (actual times were within 7-15 hours in the ICU, on average, 11 hours). Patient characteristics were pretty similar in terms of age, APACHE scores, and labs (similar ESRD/CHF percentage as well ~20% of both patient arms). Surprisingly, patients received a ridiculous amount of IV fluid from the ED – 68 (55-70) ml / kg in the echo group vs 65 (55-72) ml / kg in the standard of care arm. Yes, even with 20% of patients having ESRD / CHF – the least amount of IVF given was 55 ml / kg !


Despite all of this IV fluid given in the ED, 35% of patients still have >15% IVC collapse (!). 65% of patients had their fluid restricted, and 22% in the echo arm vs 12% in the standard of care were started on dobutamine. On Day 1 in the ICU, patients received less IVF in the echo arm (49 (33-74) ml / kg, vs 66 (42=100) ml / kg) – but still a significant amount if IVF.

28 day survival was 56% vs 66% in favor of the echo arm, with significantly less acute kidney injury (65% vs 88% for all AKI, and 19% vs 36% for stage 3 AKI).

So your NNT to save a life is 10, and 4 to reduce incidence of any AKI.

So, is this really an ED paper? Well, it depends on your area of practice. The local flavor of the authors is such that their local policy was to initiate dobutamine in the ICU and not in the ED. Are you boarding ICU players? Are your hospitalists ultrasono savvy? How involved are your intensivists in patient care while patients are awaiting an ICU bed? Are you okay with administering at least 40 ml /kg IV fluid and starting pressors on your septic shock patients? If the answer is no or “not really” to any of these questions, then the answer is yes.