This work is licensed under a Table 3.
Results of the manual coding and Automated Phenotyping Tool for identifying Developmental Language Disorder cases in health systems data (APT-DLD) classification of electronic health records (EHRs) in the discovery sample, in terms of true positive rate (TPR), true negative rate (TNR), concordance, positive predictive value (PPV), and negative predictive value (NPV).
| Group | Category | Discovery sample (N = 973) |
||||
|---|---|---|---|---|---|---|
| Manual coder classification n (%) |
APT-DLD classification n (%) |
TPR or TNR | Concordance | Clinician gold standard | ||
| Inclusion | 1 | 200 (21) | 205 (21) | 100% | 98% | PPV = 95% |
| 2 | 269 (28) | 272 (28) | 98% | |||
| Exclusion | 3 | 393 (40) | 399 (41) | 99% | 100% | NPV = 85% |
| 4 | 111 (11) | 97 (10) | 87% | |||
Note. Columns 3 and 4 indicate the numbers of records and percentages of the discovery sample (in parentheses) classified in each category. In Columns 5 and 7, TPR and PPV refer to the inclusion group, whereas TNR and NPV refer to the exclusion group. TPR, TNR, and concordance are calculated relative to the trained research assistant coding of the entire discovery sample. PPV and NPV are calculated with reference to the 97 EHRs (10% of the discovery sample) manually reviewed by our speech-language pathologist clinician coders, C. T. B. and S. M. C.