Imagine this: you hear about a new test of an Important Swallow Measure and check out the article for its normative data. You see this table:

The next day, you try out the test and your 25-year-old patient scores 1.57 seconds.

Is their Important Swallow Measure within normal limits?
 
Hold that thought while we consider what within normal limits even means.
 
Whenever we make a clinical judgement, we’re doing some kind of assessment of whether the patient’s behavior is normal, functional, or disordered. Often, we collect quantitative data from our patients and then want to see how their performance compares with “normal limits.” To define those limits, we typically ask: What are the lower and upper values that include the vast majority (commonly 95%) of the results from a healthy sample?
 
In some areas of our field, finding these values might be relatively straightforward, like a score on a published norm-referenced cognitive assessment. But a lot of the normative data for up-and-coming quantitative measures of swallow function are coming straight from research articles that aren’t trying to provide a normal range (like this one). These articles might look like they present the data we want—even when they don’t. So we need to be careful not to accidentally use the wrong values to make clinical judgements, just because they are the only values available. 
 
There are a number of appropriate ways to define a normal range. For example, we might use +/- 2 standard deviations (SD) from the mean to capture ~95% of the normal values (if we know the data is normally distributed). Or, we could define normal limits based on percentiles (especially if the data is skewed). For example, 90% of US women’s heights fall between the 5th percentile of 4’11” and the 95th percentile of 5’8” [1]. If you don’t remember much about SD and percentiles, don’t worry. Just know that these are both meaningful ways to describe a range within which most of the values fall. And, therefore, they are good measures to use when defining normal limits or thresholds.
 
Now let’s go back to that table. Is your 25-year-old patient’s score of 1.57s within the normal range?

So, what is a CI around the mean? The CI of a mean is a measure of how precisely we estimated the mean. Oversimplified, there is a 95% chance that a particular 95% CI includes the true population mean. For example, if we took a survey of women’s heights, we might find a mean of 5’3” and a 95% CI of 5’2”–5’4”. This tells us that the true average height among all women is highly likely to be between 5’2” and 5’4”. (If you want to learn more about CIs, here is a less technical explanation or here is a more technical one.)
 
There’s three important points to take away about a 95% CI:

• The 95% CI does not represent the range that includes 95% of the values. In fact, the 95% CI gets narrower as the sample size increases, even when the distribution of the values stays the same. Check out this website for some helpful visuals on this point.
• The 95% CI is often going to be much narrower than an appropriate normal range. This means that basing a reference range on 95% CI values will likely lead to overidentifying normal results as being abnormal. Using the height example above, it would be like labeling a woman who is 5’5” (165 cm) as being outside the normal limits of height because the 95% CI values are 5’2”–5’4”.
• CIs are an entirely different sort of metric from a range of the distribution. Using CI values to define a normal range is a little like trying to dribble an orange because it has the same shape as a basketball. They may look similar but they serve different purposes.

By now, you’ve probably gotten the point that we shouldn’t use CI values to inform clinical decisions. But if you’ve ever looked at CIs in a table or forest plot and thought they were values of a normal range, you wouldn’t be the first. It’s a common mistake. So before you use “healthy reference ranges” for quantitative swallowing measures—even if you see them used in a peer-reviewed article--it’s worth checking that those values are based on SD or percentiles, and not CIs. The source of those normative values should make this information clear—it’s even sometimes in the abstract.
 
We’re all trying to do best by our patients. When we incorporate more standardized, quantitative data on swallowing into clinical practice, we’re typically spending extra time in the hopes of making more accurate judgements. So let’s make sure that we are using meaningful normal ranges and interpreting our data appropriately. Because no one wants to be accidentally labeling normal swallow function as impaired.

 

[1] Fryar CD, Carroll MD, Gu Q, Afful J, Ogden C. Anthropometric reference data for children and adults: United States, 2015-2018. Vital Health Stat. (2021) https://www.cdc.gov/nchs/data/series/sr_03/sr03-046-508.pdf
​
[2] Technically, you could calculate a SD from a CI if you know the sample size and that the data is normally distributed. But some swallowing measures are skewed and journal articles with normative values do not always indicate whether the findings are normally distributed. (See here for more details.)