Skip to main content

Table 3 Performance measures of survival models, with the cutoff value for predicted probability = 10%,* for the 2003 sample

From: Outcomes of prolonged mechanic ventilation: a discrimination model based on longitudinal health insurance and death certificate data

Measure

Sensitivity†

Specificity†

PPV†

NPV†

 

n

%

(se)

n

%

(se)

n

%

(se)

n

%

(se)

3-month survival

2974/2984

99.7

(0.1)

68/2639

2.6

(0.3)

2974/5545

53.6

(0.7)

68/78

87.2

(3.8)

6-month survival

2329/2360

98.7

(0.2)

207/3263

6.3

(0.4)

2329/5385

43.3

(0.7)

207/238

87.0

(2.2)

1-year survival

1762/1823

96.7

(0.4)

543/3800

14.3

(0.6)

1762/5019

35.1

(0.7)

543/604

89.9

(1.2)

2-year survival

1235/1329

92.9

(0.7)

1099/4294

25.6

(0.7)

1235/4430

27.9

(0.7)

1099/1193

92.1

(0.8)

  1. PPV: positive predictive value
  2. NPV: negative predictive value
  3. se: standard error
  4. * As 10% was chosen as the cutoff value for predicted probability, patients with a predicted probability of survival < 10% were classified as a group that would die before the end of observation period, and those with a predicted probability of survival > = 10% were classified as a group that would survive the whole observation period. By comparison of predicted outcomes with actual outcomes, we further calculated the four measures for reflecting the quality of predictions in different perspectives.
  5. † The formulae for these measures are below:
  6. sensitivity = (the number of patients with a predicted probability of survival > = 10%)/(the number of patients who actually survive the whole observation period);
  7. specificity = (the number of patients with a predicted probability of survival < 10%)/(the number of patients who actually died before the end of observation period);
  8. PPV = (the number of patients who actually survive the whole observation period)/(the number of patients with a predicted probability of survival > = 10%);
  9. NPV = (the number of patients who actually died before the end of observation period)/(the number of patients with a predicted probability of survival < 10%).