Skip to main content
. Author manuscript; available in PMC: 2019 Sep 26.
Published in final edited form as: Respiration. 2018 Sep 26;96(5):434–445. doi: 10.1159/000490258

Table 4.

The Impact of Selection Bias on Estimates of Bias

Prediction of FEV1 % of Predicted
Method of
Prediction		 Unadjusted Model
Cohort A
(N=79)		 Adjusted Prediction Model
Cohort B
(N=114)
Intercept
(95% CI)		 Slope
(95% CI)		 Intercept
(95% CI)		 Slope
(95% CI)
Quantitative
Perfusions Scan		 23.54
(14.67, 32.42)		 −0.38
(−0.51, −0.25)		 13.94
(6.43, 21.45)		 −0.27
−0.38, −0.17)
Segment counting
with 18 segments		 22.96
(14.95, 30.97)		 −0.37
(−0.49, −0.26)		 15.69
(8.55, 22.82)		 −0.29
(−0.39, −0.19)
Segment counting
with 19 segments		 21.98
(14.15, 29.81)		 −0.36
(−0.47, −0.25)		 14.64
(7.50, 21.77)		 −0.28
(−0.38, −0.18)
Prediction of DLCO % of Predicted
Unadjusted Prediction Model
Cohort A
(N=79)		 Adjusted Prediction Model
Cohort B
(N=114)
Intercept
(95% CI)		 Slope
(95% CI)		 Intercept
(95% CI)		 Slope
(95% CI)
Quantitative
Perfusions Scan		 31.97
(24.16, 39.79)		 −0.53
(−0.65, −0.41)		 19.97
(13.07, 26.87)		 −0.39
(−0.50, −0.27)
Segment counting
with 18 segments		 29.09
(21.49, 36.69)		 −0.48
(−0.60, −0.36)		 20.19
(13.36, 27.02)		 −0.38
(−0.49, −0.27)
Segment counting
with 19 segments		 28.23
(20.50, 35.96)		 −0.47
(−0.59, −0.35)		 19.38
(12.45, 26.32)		 −0.37
(−0.48, −0.26)

Cohort A consists of all surgical patients, Cohort B includes patients that did not have surgery (n=35) as well as those that did have surgery to adjust for selection bias.