Table-2.
Estimation of logit, conditional-fixed effects logit (FE), random-effects logit (RE) and average partial effects (APE), stratified by facility ownership & eligibility for bed-hold payment
| Medicaid | Medicaid × Payer- mix+ | Payer-mix+ | ||
|---|---|---|---|---|
| Pooled model (include all facilities) | Logit model 1 | 0.239***(0.027) | 0.109***(0.012) | |
| FE model1 | 0.162***(0.030) | 0.004 (0.029) | ||
| RE model2 | 0.167***(0.030) | 0.016 (0.028) | 0.186***(0.036) | |
| APE3 | 0.028*** (0.005) | |||
| FP & Bed-hold eligible | Logit model 1 | 0.320*** (0.055) | 0.056***(0.220) | |
| FE model1 | 0.296*** (0.065) | 0.113** (0.054) | ||
| RE model2 | 0.302*** (0.065) | 0.118*** (0.053) | 0.077(0.066) | |
| APE3 | 0.054*** (0.012) | |||
| FP & Bed-hold ineligible | Logit model | 0.065 (0.061) | 0.112*** (0.026) | |
| FE model | 0.009 (0.064) | -0.028 (0.068) | ||
| RE model | 0.012 (0.064) | -0.029 (0.066) | 0.252*** (0.082) | |
| APE | 0.001(0.012) | |||
| NFP & Bed-hold eligible | Logit model | 0.252*** (0.041) | 0.151*** (0.020) | |
| FE model | 0. 160*** (0.045) | -0.023 (0.046) | ||
| RE model | 0.166*** (0.045) | -0.005 (0.046) | 0.188*** (0.062) | |
| APE | 0.026*** (0.007) | |||
| NFP & Bed-hold ineligible | Logit model | 0.291*** (0.080) | 0.105*** (0.035) | |
| FE model | 0.223*** (0.086) | -0.059(0.085) | ||
| RE model | 0.224*** (0.086) | -0.049(0.083) | 0.205*** (0.097) | |
| APE | 0.037*** (0.013) | |||
The numbers in the cell indicates the coefficients; SE value in the parenthesis.
P<0.01,
P<0.05,
*P<0.1. All tests are based on 1 -tail test. Shaded cells indicate that the estimates from logit model are statistically different from the estimates from FE model (p<0.01 for the pooled model ; P =0.06, <0.01and =0.09 FP bed-hold ineligible, NFP bed-hold eligible and NFP bed-hold ineligible NHs respectively).
The original payer-mix variable, which ranges from 0-100%, is divided by 10. Thus one unit increase represents 10% increase in payer-mix level. This variable is centered at the state mean value.
The following individual level variables were controlled: age, male, white, do-not-hospitalization/hospice enrollment, hospitalization within past 30 days, unstable condition, case-mix index, antipsychotic medications, physical restraints, ADL(long form), cognitive performance scale, pressure sore, diabetes, congestive heart failure, hypertension, other heart diseases, cancer, chronic obstructive pulmonary disease/asthma, Alzheimer /dementia, renal failure fell within last 30d, fell within past 31-180d; obese (>=30); low BMI (<=19) , CHESS (CHESS scale>0), linear, squared and cubed terms of the interval between observational start date till the end of the year ,and infection.
The following variables were controlled for RE models, additional to the above individual level variables: facility payer-mix, facility size, onsite physician extender, onsite registered nurse, resources (onsite x-ray, onsite-lab, onsite pharmacy), registered nurse hourly wage, Medicaid payment rate, facility average case-mix level, average occupancy rate, % Medicare in the facility, rural location, downstate location, the number of hospital beds per person for the population aged 75+ years (county level), and the Herfindahl index (county level).
Estimated by random-effects probit model; SE was obtained by bootstrapping.