In these models, baseline measurements were included since the SHS exposure occurred after the baseline measurements were made.
To keep the results comparable over time, specifications that use variables which are available in all waves of the LSAC and contain the least missing information (see Table 1 and Section 4 for a list of variables included in our baseline models) are used.
Baseline outcomes were included within the models to control for previous level of functioning; further non-EE variables that were associated with outcomes at follow-up were identified as additional potential predictor variables.
However, when the baseline values were included in the model, the associations did not remain significant (tables 1, 2).
Increase in baseline HR was related to lower LAD but not LAD indices, when only baseline covariates were included in the model.
When diagnostic group status and baseline cognition were included in the model simultaneously, the main effects of group, F 1, 72) = 33.3, p < .001, and baseline cognition, F 1, 72) = 24.9, p < .001, remained significant.
Only baseline characteristics were included in the model; the addition of whether a participant had experienced a surgical procedure during the 12 months was excepted as this was likely to reduce TTR.
As a result, exacerbation and baseline PaO2 were included in this model as explanatory variables after stepwise variable selection (R = 0.20, p = 0.007, and R = 0.09, p = 0.03, respectively).
In addition to the Elixhauser comorbidity index used as a continuous variable, measures of overall healthcare utilization, such as the total insurance-covered medical costs and the total number of outpatient visits for all causes during the 6-month baseline period, were included in the model to adjust for patient baseline health status.
Baseline values were included in the regression model.
