Skip to main content

Table 3 Sensitivity, Specificity, and Predictive Value Positive (PVP) and Predictive Value Negative (PVN) Rates, and Predictive Summary Index (PSI) for Different MMRI-R "Cut-Points" - Validation Data

From: The MDS Mortality Risk Index: The evolution of a method for predicting 6-month mortality in nursing home residents

MMRI-R Cut-Point

Sensitivity

Specificity

PVP

PVN

PSI

0

100.00

 

23.15

  

6

99.75

1.81

23.43

96.01

0.19

11

98.55

10.42

24.89

95.99

0.21

16

94.69

25.09

27.57

94.01

0.22

21

85.91

45.60

32.23

91.48

0.24

26

70.64

67.47

39.54

88.41

0.28

31

51.44

83.24

48.03

85.06

0.33

36

34.38

92.23

57.14

82.35

0.39

41

20.65

96.82

66.13

80.20

0.46

46

11.28

98.89

75.30

78.73

0.56

51

5.60

99.67

83.73

77.81

0.62

56

2.58

99.91

89.93

77.30

0.67

61

1.01

99.97

90.99

77.03

0.68

66

0.31

99.99

93.94

76.91

0.71

71

0.05

100.00

100.00

76.86

0.77

  1. Table notes:
  2. • Sensitivity is the probability of observing an MMRI-R score greater than or equal to the cut-point given mortality within six months.
  3. • Specificity is the probability of observing an MMRI-R score less than the cut-point given survival beyond six months.
  4. • Predictive value positive (PVP) and is the probability of dying within six months given an MMRI-R score is greater than or equal to the cut point.
  5. • Predictive value negative (PVN) and is the probability of surviving beyond six months given an MMRI-R score less than the cut-point.
  6. • The appropriateness of PVP and PVN depends on having a good estimate of population prevalence, in this case the prevalence of six month mortality.
  7. • For ease of interpretation all probabilities were converted to percentages.
  8. • Predictive Summary Index (PSI), calculated as PVP + PVN -100, is a overall measure of the gain in certainty associated with using the MMRI-R to make decisions. The practical range of PSI values is 0 to 1 with 0 reflecting a useless test and PSI = 1 for a test yielding perfect predictions.