David W Roberson1, Erna Kentala, Peter Forbes. 1. Department of Otolaryngology and Communication Disorders, Childrens Hospital Boston, Boston, Massachusetts 02111, USA. david.roberson@childrens.harvard.edu
Abstract
OBJECTIVES: The goals of this project were 1) to develop and validate an objective instrument to measure surgical performance at tonsillectomy, 2) to assess its interobserver and interobservation reliability and construct validity, and 3) to select those items with best reliability and most independent information to design a simplified form suitable for routine use in otolaryngology surgical evaluation. DESIGN: Prospective, observational data collection for an educational quality improvement project. METHODS: The evaluation instrument was based on previous instruments developed in general surgery with input from attending otolaryngologic surgeons and experts in medical education. It was pilot tested and subjected to iterative improvements. After the instrument was finalized, a total of 55 tonsillectomies were observed and scored during academic year 2002 to 2003: 45 cases by residents at different points during their rotation, 5 by fellows, and 5 by faculty. Results were assessed for interobserver reliability, interobservation reliability, and construct validity. Factor analysis was used to identify items with independent information. RESULTS: Interobserver and interobservation reliability was high. On technical items, faculty substantially outperformed fellows, who in turn outperformed residents (P < .0001 for both comparisons). On the "global" scale (overall assessment), residents improved an average of 1 full point (on a 5 point scale) during a 3 month rotation (P = .01). In the subscale of "patient care," results were less clear cut: fellows outperformed residents, who in turn outperformed faculty, but only the fellows to faculty comparison was statistically significant (P = .04), and residents did not clearly improve over time (P = .36). Factor analysis demonstrated that technical items and patient care items factor separately and thus represent separate skill domains in surgery. CONCLUSIONS: It is possible to objectively measure surgical skill at tonsillectomy with high reliability and good construct validity. Factor analysis demonstrated that patient care is a distinct domain in surgical skill. Although the interobserver reliability for some patient care items reached statistical significance, it was not high enough for "high stakes testing" purposes. Using reliability and factor analysis results, we propose a simplified instrument for use in evaluating trainees in otolaryngologic surgery.
OBJECTIVES: The goals of this project were 1) to develop and validate an objective instrument to measure surgical performance at tonsillectomy, 2) to assess its interobserver and interobservation reliability and construct validity, and 3) to select those items with best reliability and most independent information to design a simplified form suitable for routine use in otolaryngology surgical evaluation. DESIGN: Prospective, observational data collection for an educational quality improvement project. METHODS: The evaluation instrument was based on previous instruments developed in general surgery with input from attending otolaryngologic surgeons and experts in medical education. It was pilot tested and subjected to iterative improvements. After the instrument was finalized, a total of 55 tonsillectomies were observed and scored during academic year 2002 to 2003: 45 cases by residents at different points during their rotation, 5 by fellows, and 5 by faculty. Results were assessed for interobserver reliability, interobservation reliability, and construct validity. Factor analysis was used to identify items with independent information. RESULTS: Interobserver and interobservation reliability was high. On technical items, faculty substantially outperformed fellows, who in turn outperformed residents (P < .0001 for both comparisons). On the "global" scale (overall assessment), residents improved an average of 1 full point (on a 5 point scale) during a 3 month rotation (P = .01). In the subscale of "patient care," results were less clear cut: fellows outperformed residents, who in turn outperformed faculty, but only the fellows to faculty comparison was statistically significant (P = .04), and residents did not clearly improve over time (P = .36). Factor analysis demonstrated that technical items and patient care items factor separately and thus represent separate skill domains in surgery. CONCLUSIONS: It is possible to objectively measure surgical skill at tonsillectomy with high reliability and good construct validity. Factor analysis demonstrated that patient care is a distinct domain in surgical skill. Although the interobserver reliability for some patient care items reached statistical significance, it was not high enough for "high stakes testing" purposes. Using reliability and factor analysis results, we propose a simplified instrument for use in evaluating trainees in otolaryngologic surgery.