Analysis of the California In-Home Supportive Services (IHSS) Plus Waiver Demonstration Program. Implications for Modeling Recipient Outcomes


The preceding sections presented information about the living arrangements, functional limitations, and chronic health conditions of IHSS recipients and how these were distributed by age and provider type. Comparisons were also made between those entering the IHSS program in 2005 versus those continuing from 2004. Several conclusions can be drawn from these analyses relative to the recipient and other attributes that need to be adjusted in comparing recipient outcomes by provider type.29 First, it is apparent that the factors associated with Parent, Spouse, and others providers are, in part, a function of the family and other resources available. For example, among those without parents, spouses or other relatives, the options reduce to using non-relatives. This influence is most apparent among minor children, where the vast majority of those with available parents have paid Parent providers; and among the few adults with spouses. Additionally, there are preferences and other influences that are not measured by CMIPS assessments. Typically, a two-stage model would be used to estimate the “predicted” provider type in the first stage, and estimate the predicted outcomes associated with the provider type in the second stage. Ideally such a process adjusts for “selection” effects on provider choice, with the outcome of these models compared against the observed outcomes of waiver vs. non-waiver recipients. However, the absence of complete information in CMIPS about the availability of relatives (including legally responsible relatives) and recipient-provider preferences severely limits the applicability of such two-stage models here. Given the data limitations constraining the estimation of such models, the outcomes analysis reported in the subsequent sections uses observed provider type as one of the predictors of service use and expenditure outcomes. Provider type will be based on the notion of “intention to treat” described in the Methods section. If a legally responsible relative is ever used in the study year, this provider type is the presumed preference regardless of changes in provider type made during the year. Similar assumptions are made contrasting other relatives with non-relatives.

A third conclusion is suggested by the differences among race/ethnicity groups in their association with provider type. These differences are present across all age groups after adjustments for physical and cognitive limitations, household size, and IHSS wage effects. This suggests the appropriateness of using race/ethnicity as a proxy for cultural preferences or predispositions to assume caregiving roles.

Per capita income, one of several county-level measures tested, represents the cost of living in the counties, and has a significant, if modest association with provider type. This measure is retained in the outcome models.

Finally, the differences among some of the provider types in the association with managed care membership may have an effect on comparisons in analyses of Medicaid expenditures and health care events. Medicaid claims-records are generally not available for those in managed care because monthly payments are made to the health plan based on member characteristics, not on reimbursement for the use of specific services. Groups with a greater propensity toward managed care participation may have fewer chronic health conditions and lower Medicaid expenditures, but this cannot be determined with the data available. Analyses within age group, adjusting for other risk factors will help minimize this differential reporting, but it cannot fully eliminate any systematic difference if healthier (or sicker) persons enroll in managed care compared to those in fee for service. For this reason, the analyses when using health conditions as a control variable exclude recipients who are in Medicaid managed care. Payment for community care services, including IHSS, is not included in the managed care capitation payments. Consequently, analysis of this outcome is done both including adjustments for medical conditions (obtained from claims data and limited to those in fee for service), and all IHSS recipients without adjustment for medical conditions.

29. Appendix C extends the descriptive findings using logistic regression to adjust for recipient differences within a provider group. Separate analyses were conducted by recipient age group to assess the adjusted association of recipients and the “selection” of provider type. These analyses also evaluated the relative value of using IHSS wage rate as a proxy for county IHSS policy. Conclusions coming from these analyses were that the comparison of provider effects on recipient outcomes could be accommodated by using models which compare effects associated with provider type rather than using separate models by provider type of those using predicted provider types as covariates. IHSS modal wage rates were used with all comparisons being made to Los Angeles and Fresno Counties which reflect 45% of all IHSS recipients statewide and the statewide median IHSS wage rate. Among minor children, the comparison of recipients in counties across all modal IHSS wage levels found few statistically significant provider choice differences from the reference counties. The exception was that in counties with modal hourly wages of $10 or more, the likelihood of a parent being a paid provider reduced relative to the likelihood of recipients in the reference counties. No differences were found for the other provider groups. Recipients age 18-64 offer a somewhat similar pattern. Parents in counties with modal IHSS wages above $9 per hour were less likely to be paid providers, and there was a modest tendency for Non-Relatives to assume the provider role. The choice of Spouse provider was positive across wage rate levels, suggesting that choice of spouses was not related to IHSS wage rates. Among aged recipients, the prior pattern for Spouse providers holds, accept in the highest wage rate counties, which do not differ from the reference counties. Across all but the highest wage rates, counties show a tendency toward more Other Relative providers and somewhat less likelihood of Non-Relative providers than in the reference counties.

View full report


"IHSSPlus.pdf" (pdf, 4.9Mb)

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®