Skip to main content
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: J Biomed Inform. 2020 Dec 3;113:103652. doi: 10.1016/j.jbi.2020.103652

TABLE 4a.

Model estimates, predictive diagnostics, and relative contribution of PheRS in the presence of covariates in UKB.

t Model
Covariates only Covariates + risk factors PRS only PRS + covariates + risk factors PheRS only PheRS + covariates + risk factors PheRS + PRS only PheRS, PRS, covariates + risk factors
0 PheRS Odds Ratio - - - - 1.69 1.67 1.69 1.67
95% CI - - - - 1.65, 1.73 1.63, 1.71 1.65, 1.72 1.63, 1.71
PRS Odds Ratio - - 2.15 2.17 - - 2.11 2.16
95% CI - - 1.76, 2.62 1.77, 2.65 - - 1.72, 2.59 1.76, 2.65
Diagnostics AUC 0.692 0.694 0.581 0.711 0.701 0.805 0.741 0.812
AUC 95% CI 0.673, 0.711 0.676, 0.713 0.560, 0.603 0.693, 0.729 0.680, 0.721 0.786, 0.824 0.718, 0.765 0.794, 0.830
HL Stat (P-value) 37.50 (9.22E-06) 43.10 (8.44E-07) 10.90 (2.05E-01) 69.8 (5.39E-12) 1010 (0.00E+00) 821 (0.00E+00) 607 (0.00E+00) 781 (0.00E+00)
Brier Score 0.00166 0.00166 0.00166 0.00166 0.00166 0.00167 0.00167 0.00168

1 PheRS Odds Ratio - - - - 1.55 1.52 1.55 1.52
95% CI - - - - 1.51, 1.59 1.48, 1.56 1.51, 1.58 1.48, 1.56
PRS Odds Ratio - - 2.15 2.17 - - 2.12 2.15
95% CI - - 1.76, 2.62 1.77, 2.65 - - 1.73, 2.59 1.76, 2.64
Diagnostics AUC 0.692 0.694 0.581 0.711 0.660 0.772 0.695 0.779
AUC 95% CI 0.673, 0.711 0.676, 0.713 0.560, 0.603 0.693, 0.729 0.638, 0.682 0.751, 0.793 0.669, 0.720 0.759, 0.800
HL Stat (P-value) 37.50 (9.22E-06) 43.10 (8.44E-07) 10.90 (2.05E-01) 69.8 (5.39E-12) 1570 (0.00E+00) 966 (0.00E+00) 653 (0.00E+00) 889 (0.00E+00)
Brier Score 0.00166 0.00166 0.00166 0.00166 0.00166 0.00167 0.00166 0.00167

2 PheRS Odds Ratio - - - - 1.64 1.59 1.65 1.59
95% CI - - - - 1.58, 1.71 1.52, 1.65 1.58, 1.71 1.53, 1.65
PRS Odds Ratio - - 2.15 2.17 - - 2.16 2.18
95% CI - - 1.76, 2.62 1.77, 2.65 - - 1.77, 2.64 1.78, 2.66
Diagnostics AUC 0.692 0.694 0.581 0.711 0.612 0.737 0.647 0.748
AUC 95% CI 0.673, 0.711 0.676, 0.713 0.560, 0.603 0.693, 0.729 0.589, 0.635 0.715, 0.758 0.621, 0.674 0.727, 0.769
HL Stat (P-value) 37.50 (9.22E-06) 43.10 (8.44E-07) 10.90 (2.05E-01) 69.8 (5.39E-12) 1260 (0.00E+00) 599 (0.00E+00) 483 (0.00E+00) 617 (0.00E+00)
Brier Score 0.00166 0.00166 0.00166 0.00166 0.00167 0.00167 0.00167 0.00167

5 PheRS Odds Ratio - - - - 1.62 1.57 1.62 1.57
95% CI - - - - 1.56, 1.68 1.50, 1.63 1.56, 1.68 1.51, 1.63
PRS Odds Ratio - - 2.15 2.17 - - 2.16 2.18
95% CI - - 1.76, 2.62 1.77, 2.65 - - 1.76, 2.63 1.78, 2.67
Diagnostics AUC 0.692 0.694 0.581 0.711 0.605 0.732 0.651 0.742
AUC 95% CI 0.673, 0.711 0.676, 0.713 0.560, 0.603 0.693, 0.729 0.585, 0.626 0.710, 0.754 0.626, 0.676 0.720, 0.763
HL Stat (P-value) 37.50 (9.22E-06) 43.10 (8.44E-07) 10.90 (2.05E-01) 69.8 (5.39E-12) 613 (0.00E+00) 509 (0.00E+00) 301 (0.00E+00) 437 (0.00E+00)
Brier Score 0.00166 0.00166 0.00166 0.00166 0.00167 0.00167 0.00167 0.00167

Abbreviations: AUC, area under the receiver-operator characteristics curve; CI, confidence interval; HL, Hosmer-Lemeshow Chi-Square test; PheRS, phenotype risk score; PRS, polygenic risk score.

Notes: Variable estimates come from a logistic GLM model. PheRS constructed using multivariate betas obtained in MGI and applied to UKB data. Diagnostic results are obtained on using the fitted values from the corresponding model. The covariates used are sex, birthyear, genotyping batch, and the first four principal components from the genotype data. The risk factors are BMI (continuous), drinker (ever/never), and smoker (ever/never).