Losing patients to followup can artificially inflate outcomes.

Long term followup carries the risk of progressively losing patients that have poor results so that the results appear to get better with time. 

A Call for a Standardized Approach to Reporting Patient-Reported Outcome Measures: Clinical Relevance Ratio

These authors make a case for a more rigorous reporting of outcome data to account for the proportion of patients who do not reach the thresholds for clinically important improvement. They point out that reporting p values for the average improvement in patient-reported outcome measures (PROMs) can overestimate the benefit of treatment. 

Instead the clinical significance of the outcome should be reported using clinical thresholds

such as 

-the minimum clinically important difference (MCID), a threshold indicating the smallest change in the treatment outcome that a patient perceives as beneficial, and 

-the patient acceptable symptom state (PASS), a threshold indicating the highest level of symptoms beyond which a patient considers himself or herself well. 

Use of these values enables the investigator to dichotomize outcomes by identifying which patients did  and did not exceed the threshold. 

Finally, they propose a  “clinical relevance ratio,” which reports the number of patients achieving clinical importance at a given time point divided by the total number of patients included in the study at its outset. 

Unlike other common PROM-reporting approaches, the clinical relevance ratio is not skewed by patients who are lost to follow-up with increased time after the procedure; it is recognized that the more-satisfied patients are reported to be more likely to follow up, introducing response bias with an increasing loss to follow-up (see this link and this link).

Consider the example below. Preoperatively (T0) all 120 patients scored below the PASS. 

At T1, 5 patients had been lost to followup - 52% (60/115) of the remainder exceeded the PASS.

At T2, 10 patients had been lost to followup - 82% (90/110) of the remainder exceeded the PASS.

At T3, 30 patients had been lost to followup (including all those failing to exceed the PASS at T2) - 100% (90/90) of the remainder exceeded the PASS. This could be reported as a 100% success rate at final followup.

But wait - by dividing the number of patients exceeding the PASS threshold at each time point by the number of patients entering the study (120), we see that only 75% of the patients were known to be successful at both the T2 and T3 time points. 

Comment: Previous posts (see this link for example) have discussed non-response bias. This is why it is important to use followup methods that minimize the risk of non-response (i.e. loss to followup). The prevailing wisdom is that loss to followup is minimized when mail-in forms are used and when the non-responders are pursued by followup mailings and phone calls. This obviates the biasing effects of requiring patients to return to the office or of requiring patients to use a computer interface.

Follow on twitter: https://twitter.com/shoulderarth

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

Shoulder arthritis - what you need to know (see this link)

Shoulder arthritis - x-ray appearance (see this link)

The smooth and move for irreparable cuff tears (see this link)

The total shoulder arthroplasty (see this link).

The ream and run technique is shown in this link.

The cuff tear arthropathy arthroplasty (see this link).

The reverse total shoulder arthroplasty (see this link).

Shoulder rehabilitation exercises (see this link).

This is a non-commercial site, the purpose of which is education, consistent with "Fair Use" as defined in Title 17 of the U.S. Code.          

Note that author has no financial relationships with any orthopaedic companies.