R-5: Does VR Effectively Support Community Living?
Sample, Data Collection and Measurement, Data Analysis
This project will utilize data from the state’s VCM database to study the various factors that may impact consumer employment/rehabilitation outcomes. The VCM database system in which VR staff and administrators generate regular electronic case-management reports to monitor and track consumer-level VR service activities and employment or other rehabilitation outcomes. It includes all pertinent information regarding each VR case from referral to final closure. The unit of analysis of this database is individuals.
The overall sample size includes 145,684 cases referred for services over the 6-year target period; of those, 110,070 cases were accepted for application, 85,528 accepted for services, and 56,831 remain in active plan status. We will examine the characteristics of those that were rejected at the early stages of the process to determine if there were any patterns in demographic characteristics such as race, disability type, or place of residence.
A major strength of the database is that it contains all of the VR cases that made contact with the system over the 6-year period. The sample should be sufficient for testing our research questions and predictive model for the hierarchical linear regression. In particular, we know from our preliminary analysis that we have sufficient sample size for two of the comparison groups: individuals living in the community (N = 87,477 or 92% of the sample) and individuals living in institutions or nursing homes (N = 5,515 or 5.8% of the sample). We do not know the number of cases of individuals who moved from a residential facility to a private residence over July, 2004 through October, 2010 in the sample. These individuals are included in the community living sample and we plan to conduct a careful examination of residence changes in that sample to identify individuals who made the transfer.
A major weakness of the database is that there are data entry errors which forced the deletion of 962 inaccurate and invalid cases. For instance, some of the deleted cases were customers who had been reported as being both male and female, and/or having more than 20 dependents.
Data Collection and Measurement Techniques
The data for these preliminary analyses were extracted from the VCM mainframe dataset from the state VR. The database covers the cases from July, 2004 through October, 2010, and contains all personal characteristics, case history, types of services, and employment outcome information on consumer receiving VR services in the state. This large dataset is used daily by VR staff to enter customer-level service activities across its VR offices in the state. These data were made available to University of Illinois at Chicago (UIC) researchers as a result of a contractual data sharing agreement with the state VR agency. We will maintain a log of the decision rules regarding extraction and cleaning of the database to include in the final report. We will conduct separate analyses to examine the characteristics of the cases that were rejected for services.
For this project we will estimate the effect of socio-demographic characteristics, case-level characteristics, and VR-service characteristics on the employment/rehabilitation outcomes of VR consumers residing in or transferring out of institutions and those residing in the community using Hierarchical Logistical Regression Modeling (HLM). The UIC team will clean and transpose the VCM data to conduct baseline analysis of demographic and outcome-related findings of the VCM cases.
Our analyses will identify the trends and outcomes for VR consumers based on a variety of factors, starting with descriptive statistics of all demographic, VR service and outcome variables. We will also conduct Pearson correlations of demographic variables (e.g., race, disability, age, gender, place of residence, etc.) with both service (e.g., type and number of services received, expenditures), and outcome variables (e.g., employment, education, independent living) using SPSS. Finally, we will use HLM to estimate the predictive value of all demographic, case level and VR-level factors on rehabilitation outcomes (i.e. employment/no employment). We will conduct the process in three steps.
In step 1, we will enter demographic variables like race, gender, place of residence and disability type (using a particular group as a reference category for each variable). In step 2 we will analyze case level indicators like referral source, source of income, and actual income. In step 3 we will enter VR services such as the number and type of services received by consumers, the amount of money spent per case, the service provider, etc. The output test statistic for this process is the Negelkerke R2 value for each step and an estimate of the model’s predictive ability (accuracy of predicting) of the rehabilitation/non-rehabilitation outcomes.