As in the model predicting index acute admission length stay, severity measures were important factors in predicting the use of any PAC services. The odds of using any post-acute care services increased with increasing severity level across each set of severity measures. In the case of APR-DRGs (Set 1, Table 3-24), for beneficiaries in APR-DRG level 4 (extreme severity), the odds of using any PAC were 7.7 times the odds for beneficiaries in the lowest severity level. The odds of using any PAC were also higher for beneficiaries with MS-DRG with MCC (OR=2.7) and with CC (OR=1.7) compared with beneficiaries without CCs (Set 2, Table 3-25).
In Set 3 (Table 3-27), the magnitude of the odds ratios were smaller on the MS-DRG variables compared to Set 2 (w MMC OR=1.7; w CC OR=1.3), but the direction was the same indicating increasing odds of PAC use with increasing severity. In this model, many of the HCC measures also increased the odds of using any PAC services. For example, beneficiaries with HCC96 Ischemic or Unspecified Stroke had 3.5 times the odds of using PAC compared beneficiaries without this HCC. In the models without the HCC indicators (Set 1, Table 3-25, and Set 2, Table 3-26), the DRG indicator variables demonstrated similar results. The odds of using any PAC service were significantly higher for beneficiaries with index acute hospital admission for DRGs 209, 210, and 014.
Age was another significant factor in these models. Increasing age was associated with higher odds of PAC use. For example, in the model run using APR-DRG severity measures (Set 1, Table 3-25), the odds of using any PAC for beneficiaries aged 75-84 were 2.8 times the odds for beneficiaries less than age 65 and the odds increased to 5.1 for beneficiaries aged 85 and older. Similar patterns were observed across the independent variable sets.
The census region variables were each significant across the models indicating that the odds of using any PAC services were lower in each region compared to odds of PAC use in New England. The only supply variable that was significant across the models was the supply of LTCH beds/1,000 beneficiaries. Higher numbers of LTCH beds per beneficiary was associated with a slight increase in the odds of any PAC use.
The odds of PAC use were approximately 20.0 percent higher for beneficiaries discharged from hospitals located in urban areas compared with nonurban areas (OR=1.21 in Set 3, Table 3-26). Beneficiaries discharged from government-run hospitals were slightly less likely to use PAC compared with other beneficiaries (OR=0.93 in Set 3, Table 3-26).
The organizational relationship variables in this model were not significant indicating that the presence of any colocated post-acute providers or post-acute subproviders does not affect the odds of using PAC.