Table 5 Univariate and multivariate covariate-specific HR for radiation vs. active monitoring for prostate cancer.

From: Uncovering interpretable potential confounders in electronic medical records

 

Univariate analysis

Multivariate analysis

Covariates

HR

95% CI

P value

HR

95% CI

P value

W.radiation

1.22

[0.63, 2.4]

0.551

0.62

[0.24, 1.6]

0.316

struct:patient_age*

265.19

[9.1, 7.8e+03]

0.001

275.03

[3.7, 2.1e+04]

0.011

struct:race_white

0.43

[0.15, 1.3]

0.129

0.39

[0.099, 1.5]

0.170

struct:race_api

1.38

[0.29, 6.7]

0.687

0.62

[0.09, 4.3]

0.626

struct:race_black

1.69

[0.18, 16]

0.646

1.90

[0.19, 19]

0.582

struct:hispanic

0.38

[0.014, 11]

0.572

0.39

[0.016, 9.7]

0.567

struct:clinical_stage

2.42

[0.2, 30]

0.491

1.12

[0.056, 22]

0.939

struct:tumor_grade

0.89

[0.034, 23]

0.942

533.75

[0.015, 1.9e+07]

0.240

struct:grade_unknown

0.12

[0.0017, 8.9]

0.337

0.00

[3e-07, 32]

0.220

struct:diagnosis_year

0.69

[0.053, 9.1]

0.781

4.96

[0.1, 2.4e+02]

0.417

text:carotid*

44.60

[4.1, 4.9e+02]

0.002

9.63

[2.1, 43]

0.003

text:resident*

185839.25

[80, 4.3e+08]

0.002

1288062.91

[1e+03, 1.6e+09]

<0.001

  1. HR hazard ratio, CI confidence interval, * intersection terms.
  2. The * denotes intersection terms identified by our method. The lower block of covariates represents terms extracted from clinical notes. For each covariate, we show the effect size (HR), the 95% confidence interval (CI), and the statistical significance (P value) from a Wald statistics test.
Back to article page