The p value is commonly used to determine whether a change in a patient-reported measure of comfort and function (MCF) is statistically significant. By contrast the minimum clinically important difference (MCID) is a threshold value that can used to try to determine whether the change in a patient-reported measure of comfort and function (MCF) is clinically significant.
While the application of the MCID has gained wide popularity there are are some substantial problems with this metric.
1. There are different methods of calculating the MCID:
The anchor based method, which is determined by asking a group of patients to indicate at the time of follow-up how their current health state compares to their baseline assessment. The problems here include (a) the MCID depends on patient recall of their baseline (risk of recall bias), (b) variable approaches for determining what is a "meaningful change, (c) the set of patients on which the MCID is determined may not be the same as the set of patients to which the calculated MCID is applied.
The MCID can also be determined using the distribution method. Again there are varying approaches to this calculation: (a) evaluating the probability that a difference in scores occurred by chance, (b) using sample variation based on the effect size, the standardized response mean (the mean change divided by the standard deviation of score differences), or an internal responsiveness statistic, and (c) calculating the standard error of the mean.
As a result, there is substantial variation in the values used for the MCID for each patient-reported measure of comfort and function.
2. Regardless of how it is determined, the calculated value for an MCID depends on a host of variables, including the characteristics of the patients included in the calculation group (age, sex, co-morbidities, socioeconomic status, length of followup, etc (see this link)). It is also affected by the diagnoses being treated and by the type of treatment being investigated. Thus an MCID calculated for a reverse total shoulder for patients treated in Boston is likely to be different from an MCID calculated for anatomic total shoulder arthroplasty treated in Denver. A given value of the MCID may not be generalizable.
3. The use of MCID dichotomizes patients into those who do and those who do not improve by the threshold. For example if one applies an MCID of 15 to the ASES score, a patient that improved by 16 points would be considered a success and one that improved by 14 points would be a failure, even though the amount of improvement was essentially the same.
4. Improvements in the MCID along the continuum of scores are not the same: consider two patients that improved by an MCID of 16: (a) patient #1 that improved from an ASES score of 80 to 96 - a really good result and (b) patient #2 that improved from an ASES score of 10 to 26 - still has a bad shoulder.
We have found that the percent of maximal possible improvement (%MPI) (see this link) avoids many of the problems with the MCID (see MCID vs %MPI at this link).
The %MPI is determined using the simple formula;
(post treatment score minus pre treatment score)
divided by
(maximum possible score minus preoperative score)
In comparison to the problems with the MCID, the %MPI has the following attributes:
1. There is no reliance on a calculated threshold value
2. The %MPI can be applied to any patient population without concern for its relevance
3. The results are not dichotomized into success or failure, but are expressed as a continuum that expresses the relative amount of improvement
4. It does not equate an improvement at the low end of the scare to the same amount of improvement at the high end of the scale.
5. The concept of percent maximum possible improvement is easier to explain to patients than the MCID (see this link).
6. It is relatively independent of which patient-reported measure of comfort and function is used (see this link); thus it enables comparison of outcomes among studies using different outcome measures.
So, considering two hypothetical patients: (a) patient #1 that improved from an ASES score of 80 to 96 - a really good result and (b) patient #2 that improved from an ASES score to 10 to 26 - still has a bad shoulder. Yet both improved by the ASES MCID of16.
However,
patient #1 had a %MPI of 80%
(96-80)/(100-80)
patient #2 had a %MPI of 17%
((26-10)/(100-10)
The results are quite different. Even though both patients exceeded the MCID of 15, patient #1 is likely to be more pleased with the outcome than patient #2.
As we consider how to assess the success of shoulder arthroplasty from the patient's standpoint, we should look at the real data, as presented in what we refer to as a Codman Graph. Take a look at each patient and decide whether the MCID or the %MPI best presents the outcome.
Of course, it is not an "either/or" situation. Authors can employ both the MCID and the %MPI. In fact, comparing and contrasting the two is of interest. Future studies should assess whether the MCID or the %MPI is more strongly associated with patient satisfaction.
Follow on facebook: https://www.facebook.com/frederick.matsen
Follow on LinkedIn: https://www.linkedin.com/in/rick-matsen-88b1a8133/
How you can support research in shoulder surgery Click on this link.
Here are some videos that are of shoulder interest
