Blog | Monday, May 29, 2017

Interpreting and evaluating the Centor score


In response to a twitter request, this post will in detailed fashion discuss the score, how I recommend using it and how to evaluate it. I will go into more detail than I generally do, because the questions require that detail.

The original study, published in 1981 based on data collected in 1980, used logistic regression to evaluate predictors of positive group A beta hemolytic streptococcal cultures taken from adults (aged 16 and above) coming to an emergency department for a chief complaint of sore throat. We collected candidate symptoms and physical examination signs.

The idea was always to use the resulting scores to stratify patients. We never favored dichotomizing the scores. Dichotomizing means transforming the scores into two parts (e.g., negative if the score is 2 or less and positive if the score is 3 or greater). Using the score in this way loses information. We found (as do the majority of validation studies) that the higher the score, the higher the probability of a positive throat culture.

In one of our early papers, we suggested reassuring patients with a score of 0 or 1, testing patients with a score of 2 and empirically treating 3s and 4s. Almost all authors and guidelines agree with the reassurance for 0s and 1s. Some guidelines urge testing for 2s, 3s and 4s. Others favor empiric antibiotics for 4s.

Sensitivity and specificity arguments miss the point of the score. We should not consider the score as a positive or negative test, but rather as providing a probability estimate.

Recent data make this even more complex. Group C/G pharyngitis and Fusobacterium necrophorum pharyngitis have the same signs and symptoms as GAS pharyngitis. Thus if one believes that all 3 bacteria are clinically important (some argue that we do not have sufficient data to include these additional bacteria in our treatment considerations), then the probability of positive bacterial pharyngitis increases dramatically for these higher scores.

That consideration is not important in pre-adolescents, as they rarely have Group C/G or Fusobacterium pharyngitis. To answer a follow-up question, I do not understand why adolescents (and older) have different pharyngitis pathogens, but I do believe that our approach to adolescent/young adult pharyngitis should focus more on preventing suppurative complications and considering specific viral infections (EBV and acute HIV). This recommendation should not translate to pre-adolescents.

In regards to using the score to evaluate pre-adolescents, the principles of validation would involve 2 steps. The first part is discrimination. I have always used the area under the ROC curve. Generally the score shows an ROC area of approximately of 0.7 for GAS. Thus it has reasonable but not outstanding discrimination. Second, I would look at prediction – i.e., for each score what is the likelihood ratio (this measure is sometimes called the interval likelihood ratios) – and is it similar to previous reports.

One cannot look at the actual probabilities, because they will depend on the prevalence of GAS in the population tested. Over the years I have seen prevalence as high as 25% in some ER populations and as low as 5% in some student health clinics. One can adjust the probability predictions for the pre-test probability. Practically, we can use the score to guide our thoughts anyway.

Another way to consider the score is a way to standardize our history and physical examination. We should not ignore other clinical clues in making our further testing or treatment decisions.

The score is meant for initial visits. The mistakes that I see are several:
1. Not adjusting for pretest probability – a 27 year old man comes to see you, and both of his children have documented GAS. Even a rapid test is a mediocre test here.
2. Not considering a wider differential and different approach to the patient who has worsening disease
3. Not knowing that seemingly straightforward sore throats can progress
4. Insisting that the infection is viral, when it is acting like a bacterial infection

I am happy to expound on any questions that this post raises.

db is the nickname for Robert M. Centor, MD, FACP. db stands both for Dr. Bob and da boss. He is an academic general internist at the University of Alabama School of Medicine, and is the Regional Associate Dean for the Huntsville Regional Medical Campus of UASOM. He still makes inpatient rounds over 100 days each year. This post originally appeared at his blog, db's Medical Rants.