8 143 Impact of Socioeconomic Status in Patients with Achilles Tendinopathy Statistical analysis The dataset of included subjects was examined using scatterplots to identify any outliers or disparities. Missing data were recorded and the reasons behind missing data were carefully evaluated. For the primary outcome measure there was 26%, 30% and 42% of missing data at 6, 12 and 24 weeks respectively. As a substantial proportion (>10%)32 of the primary outcome measure was missing (at 24 weeks follow-up), we performed the Little MCAR test.33 The missingness of this data was found to be plausible for missing completely at random (MCAR), according to the non-significant (p=0.242) Little MCAR test. We performed a complete case analysis (CCA) utilizing pairwise deletion for both the primary and secondary outcome measures. We compared the VISA-A scores between Q1 and Q5 at baseline, 6, 12 and 24 weeks using a general linear model while adjusting for the following pre-defined set of variables that we also used in a similar prospective study: age, sex, BMI, AAS and duration of symptoms.34 We did not detect any significant collinearity among the variables (supplementary file 5). Additionally, we adjusted for baseline VISA-A score as it significantly (p < 0.001 in univariate analysis) influenced the VISA-A scores at 6, 12 and 24 weeks follow-up. For the secondary outcome measure (patient satisfaction), we used a modified Poisson regression analysis adjusting for the same set of variables. These methods ensured unbiased analyses.35-38 When it is plausible that data is MCAR, conducting a complete case analysis does not introduce bias since the incomplete datasets can be considered representative of the entire dataset.35,38,39 However, it is important to note that a CCA may lead to increased standard errors due to the reduced sample size resulting from missing data.39 Additionally, as a substantial (>40%) amount of data is missing for the outcome measures, the results obtained from the analysis should be interpreted as hypothesis-generating rather than definitive.39 To explore the robustness of our findings, a sensitivity analysis was performed using a (generalized) linear mixed-effects model for the primary and secondary outcome measures separately. IBM SPSS Statistics (version 28.0.1.0) were used for the CCA and the sensitivity analysis was conducted using R software, version 4.2.1. Used packages included ‘nlme’, ‘GLMMadaptive‘, ’emmeans’. Residuals were checked for all models.
RkJQdWJsaXNoZXIy MTk4NDMw