A Critical Appraisal of the Statistical Approaches Used in Within-Individual Observations in Hand Surgery.

Abstract

Hand surgery studies often include data from multiple hands, digits, or joints from 1 individual without using appropriate statistical approaches to assess within-individual observations, allowing for potential bias regarding treatment effects. We critically appraised the statistical methods used among studies, including dependent observations in hand surgery literature. All publications from the year 2020 to 2022 were retrieved from PubMed for 5 hand surgery journals. Studies containing ≥5 participants who performed a hand intervention in the operating theater were included. The proportion of patients with nonindependent observations and the proportion of nonindependent observations were calculated. A total of 10 128 articles were screened, of which a total of 465 studies were identified. Of these, 124 studies (27%) included multiple hands, joints, or digits from 1 individual. Only 79 (64%) studies provided data on the number of the digits, hands, and joints from a given patient. Of these, the proportion of patients with nonindependent observations was 14%. The proportion of nonindependent observations was 26%. Sixty-seven percent of articles did statistical comparisons between groups, but only 14.5% used methodological adjustments for within-patient relationships. Of the 71 studies that did not do proper statistical adjustments, 63 (88.7%) reported at least one significant result. In conclusion, there is a significant amount of nonindependent observations from single individuals and limited studies accounting for multiple observations in hand surgery literature. Most studies that did not do statistical adjustments for nonindependent observations still reported a significant finding, which raises the risk of bias.