Appendix 2: Mixed-effects Model for Pain Scores
For analysis of the pain scores, a mixed-effects regression model was fit. Patients were treated as a random effect to account for multiple observations per patient. Fixed effects covariates included treatment group (morphine vs. placebo) and time (1, 2, 6, 12, 24, and 48 h) indicators, and group by time interaction terms. Initially a linear model was fit, and results were confirmed using an ordinal logistic model. The parameterization of the model was of the general form
where Yij is the pain level for patient i at timepoint j, G is an indicator variable for treatment group (1 = morphine, 0 = placebo), vi1, vi2, …vi48 are indicator variables for the time at which the measurement was obtained, ai is the random patient effect, and eij is residual error. Also, b1T, b2T, b6T, b12T, b24T, and b48T are treatment group differences at the six timepoints. An overall joint test (b1T = b2T = b6T = b12T = b24T = b48T = 0) based on 6 degrees of freedom was performed to test the significance of the treatment effect. In addition, an overall treatment effect was calculated as the average of the treatment group differences at the six timepoints. The results are provided in Table A2.1.
Naveen Nathan, M.D., Department of Anesthesiology, Northwestern University Feinberg School of Medicine, Chicago, Illinois, contributed to figure 1.
Support was provided solely by the Department of Anesthesia and Critical Care at the University of Chicago Medical Center, Chicago, Illinois.
Dr. Balkhy has received honoraria from Intuitive Surgical (Sunnyvale, California). The other authors declare no competing interests.
Anesthesiology. 2021;135(5):864-876. © 2021 American Society of Anesthesiologists | Lippincott Williams & Wilkins