Table 1.
Bipolar disorder algorithms | h2g (SE) | Sample size | ||||
---|---|---|---|---|---|---|
Liability scale | Observed scale | P-valueb | PPV | cases | controls | |
95-NLP | 0.24 (0.10) | 0.25 (0.10) | 0.015 | 0.86 | 862 | 3952 |
Coded-strict | 0.09 (0.05) | 0.15 (0.08) | 0.064 | 0.84 | 1968 | 3952 |
Coded-broad | 0.13 (0.04) | 0.22 (0.08) | 0.003 | 0.80 | 2581 | 3952 |
Coded-broad-SV | 0.00 (0.11) | 0.00 (0.18) | 0.591 | 0.50 | 408 | 3952 |
All except coded-broad-SV | 0.12 (0.04) | 0.21 (0.07) | 0.004 | 0.83 | 3013 | 3952 |
ICCBD + PGCBDa | 0.23 (0.01) | 0.41 (0.02) | 3.17 × 10−80 | NA | 13902 | 19279 |
SNP-based heritability on liability scale was converted from observed scale based on population prevalence of 1%
aICCBD + PGCBD: Bipolar disorder genome-wide association study from the ICCBD and PGC1 with cases ascertained by traditional methods (Charney et al.7)
bTest for h2g different from 0. PPV: positive predictive values from clinical validation (Castro et al.10). 95-NLP: probabilistic algorithm with 95% specificity based on natural language processing. Coded-strict, Coded-broad, Coded-broad-SV: coded rule-based algorithms with decreasing stringency. SV: single visit. SE: standard error