Statistical Analysis
Summary statistics were calculated using standard methods. Multivariable
mixed effects ordered logistic models were employed using self-reported
frequencies across symptom categories. The primary modeling framework
included an interaction of peri-surgical time point and pain management
protocol, adjusted for age, race, gender, body-mass index (BMI), history
of intravenous (IV) drug use, and use of pain medications at home around
the surgical window. Graphs were drawn using marginal estimates from
fully adjusted models. Due to limited data in some response categories,
results were presented for participants reporting “none” or “some”
symptom occurrences at each time point. Furthermore, for some outcomes,
absence of variation in the data would not allow for model convergence,
so these outcomes are omitted from reported results. There was no reason
to assume missing data were not missing at random (NMAR). The missing at
random (MAR) assumptions of the mixed models utilized in analyses were
deemed valid. All statistical analyses were completed using Stata v16.1.