The New Issue of CJSM

Under the lights, on the gridiron:
the site of a lot of concussions.

I was very excited to find the eTOC (electronic table of contents) for the September 2013 CJSM in my email inbox this morning.

As an associate editor, I am privy to this sort of information prior to launch, so to speak, but it’s usually all I can do on these blog pages not to talk about the contents prior to publication.  The offerings are always so varied, so interesting.  Well, the ‘horse is out of the barn now,’ the new CJSM is travelling over the ether, and I can at last discuss an article or two!

Before I forget, let me remind you that you, too, can get the CJSM eTOC and find out the lineup when it’s hot off the press, so I encourage you to go to the home page and sign up, the link is near the top of the masthead.

I wanted to write briefly this morning about one of the new studies in the September CJSM, “The Prevalence of Undiagnosed Concussions in Athletes,” Meehan et al.

William P. Meehan III, M.D. is a friend of mine.  To paraphrase a line that has had a lot of currency recently here in the states, the arc of Bill’s career bends towards success.  It is really hard to keep track of all the work he is doing in the field of concussion.  In fact, in an attempt to do so, I have asked him to participate in a “Five Questions with CJSM” blog post interview, like the one I did with Dr. Jason Mihalik last month.   Keep your eyes posted here, because I’m hoping Bill can sit for a few minutes between attending to his multiple, other commitments and I’ll write up his thoughts in a blog post  before the end of the month, I hope.

The new study was conducted at Boston Children’s Hospital and the University of Pittsburgh.  The authors designed a cross-sectional study to enroll all patients presenting for treatment at the two sports concussion clinics in a calendar year, with exclusion criteria including concussions sustained by motor vehicle accident and other more significant trauma.  731 patients met inclusion criteria, with fully 31% not participating, resulting in 486 patients included for final analysis. Outcome measures included current score on the “Post Concussion Symptom Scale,” loss of consciousness (LOC) with current injury, and mean symptom duration in days.  The independent variable was the answer to this question, “Have you ever sustained a blow to the head which was not diagnosed as a concussion but was followed by one or more of the signs and symptoms listed in the Post Concussion Symptom Scale,” with patients answering ‘yes’ defined as having a previously undiagnosed concussion. Data analysis was conducted to see if there were differences in the three outcome measures in the patients who had a history of concussion and those who had not.

The authors report that there was a statistically significant (P < 0.038) difference in rates of LOC in the two groups, with patients having a previously undiagnosed concussion having LOC with their current concussion at a rate of 30.6% (compared with 21.8% in the groups with no previously undiagnosed concussion).  The other outcome measure of statistical significance was the mean PCSS score at initial visit, which was reported as 33 (group with previously undiagnosed concussion)  vs. 25 (those without) p < 0.004.  In terms of length of symptom duration with the current injury, there was no statistically significant difference.

I enjoyed this study.  It adds to the growing body of literature that supports the notion that concussion is an undereported injury. The mean age in the two groups was 15.4/15.5, resembling the population I take care of at Nationwide Children’s Hospital,   The numbers in the study were large, even if they may have been underpowered to achieve some potential secondary outcome measures, as the authors note in their ‘Discussion.’

My critical thoughts mostly focus on the PCSS outcome measure.  The PCSS the authors used, part of the SCAT2, has been incorporated again in the 4th Zurich International Consensus Statement on Concussion, with its updated SCAT3.  The PCSS in both tools is a 22 item Likert scale, scored 0 to 6, asking about symptoms, ranging from headache, to irritability to blurred vision; the score can, therefore, range from 0 to 132 (6 x 22) on any administration.

One big critique is that the PCSS is an ordinal scale, but not an interval or ratio scale.  Rating one’s headache on a 0 to 6 Likert scale can give us a notion that 5 headache is worse than a 1, but we can’t say it’s ‘five times worse’ as we could with a ratio scale; we can’t even say the difference between a 1 and a 2 headache is the same as that between a 4 and a 5 headache (interval scale).  We must be very careful, therefore, how we use ordinal scales in quantitative analyses.  It is common practice to take the total PCSS and use that one number to infer that, say, a PCSS of 33 means the patient is more symptomatic than one with a 25.  Despite it’s being a common practice, it is a questionable practice.

Furthermore, though a statistically significant difference, the reported mean PCSS in the study’s two groups are different by only 8 points (33 and 25).  There are a couple of ways of looking at this difference; ultimately, we want to know if this difference is clinically significant–if it might be detectable in an individual patient we are treating, or whether that magnitude of change might have prognostic significance.

One way to begin this process would be to say the patients with a previously undiagnosed concussion had a 32% higher PCSS than those without [ ((33 – 25)/25)x100% = 32% ].  That sounds fairly high.  Another way might be to reflect that the PCSS scale ranges from 0 to 132, so an 8 point increase on that scale would represent ‘only’ a 6% change.

A better way–a more statistically sound way– would be to calculate the effect size, which is a basic measure that has more use (unfortunately) in the social sciences than in the clinical sciences.

I would direct the readership to an excellent, if slightly sassy, essay, “It’s the Effect Size, Stupid.”  There are different ways of measuring the effect size, which is a standardized indicator of the ability of scores on a measure to distinguish between two groups (are you in the ‘previously undiagnosed with concussion’ group or are you out?)  One way is to calculate something called Cohen’s d, and there are useful online calculators for this as typically you will not find it on SPSS.   The effect size measures are simply standardized mean differences between the two groups in question; Cohen’s d for instance takes into account the reported means and standard deviations to calculate an ‘effect size,’ which is a unitless measure and can be useful, for instance, in therapeutic interventions:  determining whether the effect of one intervention in a measurable outcome is superior to another.  Effect sizes less than 0.5 are small (possibly inconsequential), whereas those above 0.8 are large, and a Cohen’s d = 2 would be whopping.  Qualitatively, all this is saying is, for instance, if a reported effect size were 1.0, this is something notable and noticeable.

I can’t calculate the effect size in the two groups in the Meehan study as we have the reported means but not the standard deviations. Hmm…..maybe I can get them from Bill!

I’m not touching you, am I bothering you?

Well, I fear for a Sunday I may have already introduced an early nap for some of the readers out there.  If you’re interested in issues such as ‘minimal clinically significant difference’ and the like, by all means, contact me on the twitter feed @cjsmonline or comment here on the blog.  In the meantime, most definitely head to the new CJSM journal, and check out Dr. Meehan’s study and more!!!

The sporting week ahead is an exciting one:  US Open tennis finals, a new IOC president to be announced, and the first NFL Sunday of the season, just to name a few of the highlights I’ll be tracking.  May your week be a good one!