Skip to main content

Table 2 Performance characteristics of the different surveillance models

From: The augmented value of using clinical notes in semi-automated surveillance of deep surgical site infections after colorectal surgery

 

Sensitivity,

% (95%CI)

Specificity,

% (95%CI)

PPV,

% (95%CI)

NPV,

% (95%CI)

WR,

%

Model 1

97.6 (87.1–100.0)

69.6 (62.4–76.1)

41.7 (31.7–52.2)

99.2 (95.7–100.0)

57.3

Model 2

87.8 (73.8–95.9)

79.9 (73.4–85.4)

49.3 (37.4–61.3)

96.7 (92.5–98.9)

67.5

Model 3

95.1 (83.5–99.4)

73.4 (66.4–79.6)

44.3 (33.7–55.3)

98.5 (94.8–99.8)

60.9

Model 4

95.1 (83.5–99.4)

73.4 (66.4–79.6)

44.3 (33.7–55.3)

98.5 (94.8–99.8)

60.9

Model 5

92.7 (80.0–98.5)

77.7 (71.0–83.5)

48.1 (36.7–59.6)

97.9 (94.1–99.6)

64.9

Model 6

95.1 (83.5–99.4)

70.6 (63.5–77.1)

41.9 (31.8–52.6)

98.5 (94.6–99.8)

58.7

Model 7

92.7 (80.1–98.5)

79.3 (72.8–84.9)

50.0 (38.3–61.7)

97.9 (94.2–99.6)

66.2

Model 8

85.4 (70.8–94.4)

82.1 (75.8–87.3)

51.5 (39.0–63.8)

96.2 (91.8–98.6)

69.8

  1. PPV: positive predictive value; NPV: negative predictive value; WR: workload reduction; 95%CI: 95% confidence interval; NLP: natural language processing.
  2. Model 1: Original semi-automated algorithm with structured data only.
  3. Model 2: Model 1 augmented with NLP component using decision tree and raw counts.
  4. Model 3: model 1 augmented with NLP component using decision tree and discretized counts.
  5. Model 4: model 1 augmented with NLP component using decision tree and binary counts.
  6. Model 5: model 1 augmented with NLP component using random forest and raw counts.
  7. Model 6: model 1 augmented with NLP component using random forest and discretized counts.
  8. Model 7: model 1 augmented with NLP component using random forest and binary counts.
  9. Model 8: Model 1 augmented with a rule-based component.