Is retrospective application of the Oxford Shoulder Score valid?

Application of quantitative outcome scores before and after an intervention can enhance research design and statistical power. For unplanned events such as trauma, prospective scoring is not possible. This study aims to investigate the reliability of shoulder symptom recall in patients undergoing a surgical procedure on their shoulder.

Fifty patients undergoing shoulder surgery completed an Oxford Shoulder Score (OSS) before surgery. OSS questionnaires were mailed to the individuals at a mean of 50 days postoperatively. They completed the OSS according to how they remembered their symptoms before their operation.

On recall, the OSS increased from 39.4 to 40.5 showing no statistically significant difference (P = 0.09). However, individual score agreement, assessed by a Bland Altman plot, revealed 95% confidence intervals from -7.7 to 9.9.

This study shows that individual patient recall of symptoms was variable. There is considerable disagreement between their original and recalled scores. However, when applied to a large group, recollection of shoulder symptoms as assessed by the OSS was not subject to recall bias. It is, therefore, possible to use the mean value of recollected scores, within a population, to assess the impact of an intervention. For example, following shoulder trauma where a prospective measurement is not possible.