Medicaid in Residential Care. Multivariate Analysis: Predicting Residential Care Facilities' Participation in Medicaid

09/01/2013

To better understand the predictors of facility participation in Medicaid, we conducted multivariate analyses of factors that were hypothesized to be related to serving Medicaid residents. The dependent variable for this analysis is the bivariate variable indicating facility Medicaid participation, which is defined as a facility serving at least one resident whose LTC services are paid by Medicaid. Medicaid participation is coded zero if the facility reports that it does not have residents whose LTC services are paid by Medicaid and one if the facility reports that it does. The multiple logistic regression model used in our analysis is:

Facility Medicaid Participation = intercept +state RCF Medicaid coverage category + the number of residential care beds + the total direct staff hours per resident day (HPRD) ratio + MSA + facility chain status + private/for-profit ownership + high privacy facility + percentage of residents with short-term memory problems or disoriented all or most of the time during the last 7 days + error term

Because the multivariate analysis is conducted at the facility level, the direct staff ratio used in the model is at the facility level.

Table 5 presents descriptive data for the variables used in the multivariate analysis, by Medicaid payment status. States are categorized by Medicaid coverage of RCFs: whether the state Medicaid program covers residential care with HCBS waivers, state plan personal care, both, or neither, with neither as the reference category. Most facilities are located in states where Medicaid HCBS cover services in RCFs.8 Facilities have an average of slightly more than 30 beds, an average direct care ratio of 4.2 hours per resident per day. In addition, most facilities are located in Metropolitan Statistical Areas (MSAs), are not part of a chain, and have for-profit ownership. Facilities are about evenly split between high-privacy and low-privacy living arrangements. In the average facility, almost half of residents have some signs of cognitive impairment. Facilities serving Medicaid residents statistically significantly differ from facilities not serving Medicaid residents on all of the variables used in the regression model, except for staffing levels and ownership type.

TABLE 5. Means and Proportions for Variables Used in the Model Predicting Medicaid Participation
Model Predictive Variables Total RCF FacilitiesN=2,302 (% or mean2); Non-MedicaidFacilitiesN=1.292 (% or mean2) Facilities Serving Any Residents on MedicaidN=998 (% or mean2) Significance Test p-value1

SOURCE: RTI analysis of the 2010 NSRCF.

NOTES:

  1. P-values are based on chi-square tests for categorical variables and t-test for continuous variables.
  2. Percent for categorical variables; means for continuous variables.
  3. Direct care hours include RN/LPN/LVN, personal care aide and direct care administrator hours per patient per day.
  4. To determine whether a facility is part of a chain, the survey asks if the facility is owned by a chain, group, or multifacility system.

**p<0.05, ***p<0.001, ns: not significant, p>0.1.

State Medicaid RCF coverage category
   HCBS waivers only 59.9% 66.9% 50.8% ***
   State plan personal care only 13.4% 10.0% 17.6% ***
   Both HCBS waivers and state plan personal care 15.6% 9.0% 24.2% ***
   Neither HCBS waivers nor state plan personal care 11.1% 14.1% 7.4% ***
Number of RCF beds 31.3 34.5 26.8 ***
Direct care staff ratio3 4.2 4.2 4.1 ns
Located in a MSA 80.5% 85.8% 73.7% ***
Part of a chain4 37.7% 38.9% 36.1% ns
Private, for-profit 82.4% 83.3% 81.2% ns
High-privacy (>80% of units are single rooms or apartments) 46.7% 49.1% 43.8% **
Percent of residents with short-term memory problems or are disoriented all or most of the time during the last 7 days   43.8% 48.0% 38.3% ***

In our logistic regression model, we hypothesize that type of Medicaid coverage of RCFs services will be significant predictors of facility participation in Medicaid. In general, our expectations are that Medicaid participation will be associated with lower cost, including smaller facilities, lower direct care staffing ratio, location outside of a MSA, not being part of a chain, for-profit ownership, lower privacy in living arrangements, and lower prevalence of people with Alzheimer's disease and other dementias.

Table 6 presents the odds ratios and confidence intervals for the full model. The R2 for the model is 10.3 percent, which suggests that most of the variation is accounted for by variables not in the model. The Medicaid state plan personal care and the category representing states with both HCBS waiver and personal care coverage were statistically significant predictors of Medicaid participation. Somewhat surprisingly, coverage of RCF services only through HCBS waivers was not a statistically significant predictor of Medicaid participation, suggesting that the need to serve persons with a nursing home level of care may be a deterrent to participation.

TABLE 6. Logistic Regression Model Predicting Facility Medicaid Participation
Model Predictive Variables Odds Ratios (95% confidence interval) Significance Test p-value1

SOURCE: RTI analysis of the 2010 NSRCF.

NOTES:

  1. P-value for the t-test that the beta of the predictive variable equals zero (ß=0).
  2. Direct care hours include RN/LPN/LVN, personal care aide and direct care administrator hours per patient per day.
  3. To determine whether a facility is part of a chain, the survey asks if the facility is owned by a chain, group, or multifacility system.

*p<0.1, **p<0.05, ***p<0.001, ns: not significant, p>0.1.

Intercept 1.61 (0.95, 2.73) *
Medicaid coverage category
   HCBS waivers only 1.32 (0.88, 1.98) ns
   State plan personal care only 2.57 (1.65, 4.00) ***
   Both HCBS waivers and state plan personal care 5.12 (3.17, 8.28) ***
   Neither HCBS waivers nor state plan personal care ref ---
Number of residential beds 0.99 (0.99, 1.00) ***
Direct care staff ratio2 0.98 (0.95, 1.02) ns
Located in a MSA 0.48 (0.38, 0.61) ***
Part of a chain3 1.17 (0.93, 1.47) ns
Private, for-profit ownership (vs. private non-profit or state/county/local government) 0.88 (0.68, 1.16) ns
High-privacy (>80% of units are private, i.e., for 1 person) 0.86 (0.69, 1.08) ns
Percent of residents with short-term memory problems or are disoriented all or most of the time during the last 7 days   0.99 (0.99, 1.00) ***
Model Fit R2 = 0.103990 ---

The odds that a facility will participate in Medicaid are lower if the facility is located in an urban area. Higher incomes in urban areas may contribute to residents or their families being better able to pay higher rates, making facilities less willing to participate in Medicaid where the payment rates are believed to be lower. The odds of participating in Medicaid decrease as the percentage of residents with short-term memory problems or who are confused or disoriented most of the time in the last seven days increases. In other words, the higher the number of residents with short-term memory problems, most likely associated with Alzheimer's disease, the less likely the facility is to participate in Medicaid. Moreover, the odds of a facility participating decrease as the average number of residential care beds in the facility increases (i.e., the larger the facility, the less likely it is to accept Medicaid residents). The direct care staff ratio, for-profit ownership, being part of a chain, and having a high percentage of high-privacy units were not statistically significant predictors of Medicaid participation.

View full report

Preview
Download

"FacBenDif.pdf" (pdf, 1.23Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®