Skip to main content

Table 1 Proportional hazards models predicting 45-year all-cause mortality (1447 deaths among 1591 men) by three methods

From: Artificial neural networks versus proportional hazards Cox models to predict 45-year all-cause mortality in the Italian Rural Areas of the Seven Countries Study

 

Forced Cox model

Forward stepwise Cox model

Backward stepwise Cox model

(N = 1591)

(N = 1591)

(N = 1591)

β

t

HR(±95%CI)

β

t

HR(±95%CI)

β

t

HR(±95%CI)

Age (years)

0.1024

16.81

1.68(1.58-1.79)

0.1017

16.78

1.67(1.58-1.78)

0.1029

17.01

1.69(1.59-1.79)

Father status (codes 0–1)

0.1422

2.19

1.15(1.01-1.31)

0.1421

2.19

1.15(1.01-1.31)

0.1436

2.21

1.15(1.02-1.31)

Mother status (codes 0–1)

0.2063

3.11

1.23(1.08-1.40)

0.2121

3.22

1.24(1.09-1.41)

0.2072

3.16

1.23(1.08-1.40)

Family history of CVD (codes 0–1)

0.0326

0.60

1.03(0.93-1.15)

=

=

=

=

=

=

Corneal arcus (codes 0–1)

0.2092

2.70

1.23(1.06-1.43)

0.1988

2.57

1.22(1.05-1.42)

0.2037

2.63

1.23(1.05-1.43)

Physical activity (codes 1-2-3)

−0.0830

−1.92

0.92(0.85-1.00)

=

=

=

−0.1042

−2.51

0.90(0.83-0.98)

Cigarettes smoked per day (N)

0.0174

6.26

1.18(1.12-1.24)

0.0177

6.37

1.18(1.12-1.25)

0.0177

6.36

1.18(1.12-1.25)

Body mass index (Kg/m2)

−0.1840

−2.77

0.51(0.31-0.82)

−0.1711

−2.62

0.53(0.33-0.85)

−0.2097

−3.30

0.46(0.29-0.73)

Body mass index2 [(Kg/m2)2]

0.0033

2.70

1.90(1.19-3.04)

0.0031

2.62

1.85(1.17-2.93)

0.0036

3.03

2.02(1.28-3.18)

Arm circumference (cm)

−0.0026

−1.64

0.94(0.88-1.01)

−0.0034

−2.21

0.92(0.86-0.99)

=

=

=

Mean blood pressure (mmHg)

0.0180

7.76

1.28(1.20-1.36)

0.0186

8.24

1.29(1.21-1.36)

0.0189

8.43

1.29(1.22-1.37)

Heart rate (beats/min)

0.0018

0.78

1.02(0.97-1.09)

=

=

=

=

=

=

Serum cholesterol (mmol/l)

0.0722

2.81

1.08(1.02-1.14)

0.0758

2.97

1.08(1.03-1.14)

0.0751

2.94

1.08(1.03-1.14)

Forced expiratory volume (l/m2)

−0.4691

−3.93

0.89(0.84-0.94)

−0.4815

−4.05

0.89(0.84-0.94)

−0.4742

−3.99

0.89(0.84-0.94)

ECGm abnormalities (codes 0–1)

0.0518

0.47

1.05(0.85-1.31)

=

=

=

=

=

=

Prevalence CVD (codes 0–1)

0.3658

2.85

1.44(1.12-1.85)

0.3520

2.75

1.42(1.11-1.83)

0.3905

3.05

1.48(1.15-1.90)

Prevalence K (codes 0–1)

2.1479

4.72

8.57(3.51-20.89)

2.2056

4.87

9.08(3.73-22.07)

2.1644

4.76

8.71(3.58-21.22)

Prevalence DIAB (codes 0–1)

0.2545

2.06

1.29(1.01-1.64)

0.2502

2.02

1.28(1.01-1.64)

0.2674

2.16

1.31(1.03-1.66)

ROC ± standard error (±95%CI)

 

0.829 ± 0.0137 (0.810-0.847)

  

0.830 ± 0.0136 (0.811-0.849)

  

0.830 ± 0.0137 (0.810-0.848)

 
  1. β = coefficient; t = t value of coefficient (when t > |1.96| p<0.05); HR = hazard ratio; CI = confidence intervals. The differences for hazards ratio are expressed as 0–1 for dichotomic variables, as 1 unit for physical activity, and as standard deviations for continuous variables.
  2. CVD = cardiovascular disease; DIAB = diabetes; ECGm = minor ECG; K = cancer.
Back to article page

BMC Medical Research Methodology

ISSN: 1471-2288

Contact us