Columbia Implementation

Algorithm Type: 
Case, Control
Institution: 
Columbia University
Implementation Details: 

- The case definition is unclear in the algorithm.

  - Here we reported PPV for the case defined as results from chest x-array report (NLP). 

  - If the case is defined by both ICD and x-array report, then the PPV is 100%. The number of case is 156.

  - If the case is defined by ICD, x-array report and ABX, then the PPV is 100%. The number of case is 2.

- The control defined in this study is unclear as well. There are control A, control B, and controlAB. Here we reported PPV for all controls (three groups). 

- The overall impression is the negation NLP is not accurate which leads to the high error rate.

 

Files: 
review of 25 cases
review of 25 controls
Phenotype: 
Pneumonia- VUMC eMERGE v5.1
Dictionaries
Implementations/Datasets/
Cases (156) Actual Class (Expectation)
Predicted Class
(Observation)
True Positive: 
100.00
Correct result 		Unexpected result
Control (8015) Actual Class (Expectation)
Predicted Class
(Observation)
True Positive: 
0.96
Correct result
False Positive: 
0.04
Unexpected result
Control Positive Predictive Value:
 96%

Comments