Table IV.
Posterior predictive checking results from the medical records and medicare claims, a subsample of CanCORS data.
| Medical records | Medicare claims | |||||
|---|---|---|---|---|---|---|
| Study site |
Observed capturing rate |
Median from predictions |
ppp | Observed capturing rate |
Median from predictions |
ppp |
| GHC | .38 | .31 | .78 | .92 | .89 | .64 |
| HPHC | 1 | 1 | .41 | 1 | 1 | .16 |
| HFHS | .78 | .67 | .73 | .70 | .83 | .18 |
| KPHI | .20 | .38 | .48 | 1 | 1 | .20 |
| KPNW | .67 | .67 | .48 | .85 | .85 | .51 |
| NCCC | .25 | .26 | .43 | .78 | .83 | .25 |
| UAB | .20 | .22 | .30 | 1 | .89 | .92 |
| UCLA | .44 | .41 | .73 | .87 | .83 | .79 |
| IOWA | .73 | .71 | .66 | .90 | .88 | .72 |
| UNC | .22 | .23 | .46 | 1 | 1 | .44 |
| Var | .08 | .07 | .21 | .01 | .01 | .54 |
Note: The subsample consists of 2669 CanCORS patients who died within 15 months of diagnosis and had at least one observed hospice-use record from two sources. The posterior predictive checks are based on 105 replicates. ppp, posterior predictive p-values. For example in the GHC site, the capturing rate is 23/(23 + 38) = .38 for medical records and 23/(23 + 2) = .92 for Medicare claims. The corresponding numerators and denominators can be found in Table AI.