Analysis of the California In-Home Supportive Services (IHSS) Plus Waiver Demonstration Program. Recipient Characteristics Measures


Recipient Characteristics measures were obtained largely from CMIPS. These files are compiled monthly and include recipient eligibility and assessment files, provider eligibility, and payment files. Recipient assessment data in CMIPS are generally updated every two years or after a substantial change in status. IHSS Plus Waiver recipients receive annual assessments. To obtain reasonable comparability between waiver and non-waiver recipients we averaged each recipient’s assessment measures drawn from each IHSS participation month in 2005. These items can vary from month to month with changes in status or periodic reassessments. If the individual was not a recipient in January, then the first assessment in 2005 was used as the starting assessment. Following is a description of the measures drawn from CMIPS for this analysis.10

  • Recipient Age. Used to classify recipients into the target age cohorts.

  • Recipient Gender.

  • Recipient Race/Ethnicity. There are 16 race/ethnicity categories used in CMIPS, we consolidated these into four groups: White, Hispanic, African-American, Asian/and all Others for the analysis.11

  • Recipient Household Size and Living Arrangement. There are several measures potentially available for these items. One is a count of persons in the household, excluding non-IHSS recipients <age 14. A second is a measure indicating for those under 18, if a parent is present and whether the parent is able and available to be a care provider. There is a similar measure relative to spouses, but this measure was found to be unreliable and was not used.12

  • Provider Relationship. This information is obtained from the provider eligibility files. Three types of provider relationships are used to classify the IHSS recipient into a waiver and non-waiver status: Parent/Spouse, Other Relative, Non-Relatives. (See the sampling discussion for the decision rules used when there were changes in providers, and multiple providers in a month or year.)

  • Recipient Income and Share of Cost. Recipient income was not used in the analysis as all study subjects were Medicaid eligible. Fewer than 4% among all provider groups had any IHSS “Share of Cost” payment requirements.

  • Eligible Months. Number of months eligible for IHSS in 2005 is used as the denominator to standardize IHSS authorized hours, and total IHSS and Medicaid expenditures into averages per IHSS participation months in 2005.

  • Authorized IHSS Hours. This measure is represented in the analysis by the monthly authorized hours in a calendar year, averaged over the number of active IHSS participation months in the year. More precise time varying measurement was not used because of data limitations. Authorized hours can change with a change in recipient status or living arrangement, but the items in the assessment files are commonly not updated until the next assessment -- which may be more than a year away.13

  • Recipient Cognitive Limitations. For consistency in classifying recipients across all age groups and payer sources, recipient cognitive limitations are limited to three items available in CMIPS: Cognition is defined by: Memory, Orientation, and Judgment. Each scored 1 independent; 2 able to perform, but needs verbal assistance such as reminders, guidance, or encouragement; 5 cannot perform without human assistance.

  • Recipient Limitations in ADLs. ADLs include bathing and grooming; dressing; transferring; bowel, bladder and menstrual; eating. Each task is scored on a five point scale: 1 and 2 as per above, 3 Can perform with some human direct physical assistance from the provider, 4 Can perform with a lot of human assistance, 5 cannot perform without human assistance.14

  • Recipient Breathing problems (which includes assistance with self-administration of oxygen and the cleaning of this equipment) are scored in CMIPS as 1 independent, 5 cannot perform without human assistance, 6 Paramedical Services needed. The measure used in the analysis is the presence/absence of a Breathing item with a score of five or more.

Health Conditions and Diagnoses were compiled from Medicaid claims to supplement the CMIPS recipient characteristics.15 These records include up to two diagnoses, coded using the International Classification of Disease or ICD-9-CM (CDC, 2007), for each individual service claim. Analyses using claims and other administrative data have adopted a variety of approaches for identifying and adjusting for patient diagnoses. Important areas of conceptual consensus are that the diagnostic categories be: (a) clinically meaningful and related to well-specified disease or medical conditions; (b) the categories predict medical expenditures or other specified outcomes of interest (e.g., mortality); and (c) have sufficient prevalence to permit stable estimates. Two of the most prominent approaches are used in this analysis, one for recipient characteristics (described below), the other as a health care outcome indicator (described in the Outcome Measures section).

The CMS hierarchical condition categories (HCC) are used as health condition predictor variables in our analysis.16 HCC classification uses both inpatient and outpatient data. The HCC provides a standardized protocol for combining over 15,000 ICD-9 categories into 189 condition categories or CCs (Pope, Ellis, Ash, et al., 2000). Most CCs describe a broad set of similar diseases, generally organized into 23 body systems, but CCs 185-189 are assigned by beneficiary utilization of selected types of durable medical equipment. The CCs can be organized into hierarchies, designed so that a person is coded only for the most severe manifestation among the related diseases defining the CC. Within the same HCC a person is classified once. This avoids the problem of duplicative counting of related conditions. For unrelated diseases (i.e., diseases in other CCs), the number of HCC’s accumulate.

HCCs are assigned using any mention of the eligible diagnosis from any of five sources.17 Information or the frequency of mentions are not differentially weighted among these sources (Pope, et al., 2004):

  • Principal hospital inpatient;
  • Secondary hospital inpatient;
  • Hospital outpatient;
  • Physician claims-record; and
  • Clinically-trained non-physicians (e.g., psychologist, podiatrist, nurse practitioner).

Additional information required for the identification of a qualifying diagnosis is a date on the eligible record establishing that the diagnosis was made (or was present) during the relevant reporting period.18

  1. Provider attributes such as race/ethnicity, age, and gender are available in CMIPS, but other than relationship to the recipient, these data were not used in the analysis.

  2. Appendix BTable B-1 shows the distribution of the study samples’ race/ethnicity groups, by recipient age group and provider type by new recipients in 2005 and those continuing from 2004.

  3. Measures of living arrangement, such as housing type, having a live-in provider, and various shared housing arrangements were incompletely coded in assessments and correlated with household size. Consequently, only household size was used.

  4. Within CMIPS there is a calculated unmet need, defined to be the difference in total need hours and authorized hours. This measure was not used because it is confounded by ceilings on the maximum number of authorized hours (283 hours/month) used by IHSS, and non-transparent adjustments made for household composition or unmeasured changes in status. An alternative unmet need measure derived from the difference between authorized hours and paid hours was considered. This measure proved to be problematic as the distribution of hours per week is not determinable from the monthly payment data. Consequently, there may be unmet hours in particular days or weeks that are masked by accumulated monthly billings.

  5. IADLs included in CMIPS are housework, laundry, shopping and errands, meal preparation and clean-up, mobility inside one’s home. Each task is scored on a five point scale: 1 and 2 as per above, 3 Can perform with some human direct physical assistance fro the provider, 4 Can perform with a lot of human assistance, 5 Cannot perform without human assistance. This measure is not included in the analysis because of the absence of variance. Across all provider groups, 85% or more of the recipients have four or more limitations with a score of three or higher.

  6. The project considered using the CDER and the POS file -- both from DDS. Together, these provide recipient assessment information and service use data. However, given the proportionately small number of IHSS recipients in these data sets, and the incomparability of the assessment measures with those in CMIPS, the redundancy with salient POS items with those in Medicaid claims, the decision was made to limit age and provider analysis to the uniform common data available from CMIPS and Medicaid claims.

  7. The CMS-HCC model was developed for Medicare using claims data to provide risk adjustment for Medicare capitation payment rates (Pope, Kautter, Ellis, et al., 2004). This method has been extensively tested for predictive validity among aged and disabled persons; and with both community and institution-based populations.

  8. Diagnoses from other claims records (including home health providers, durable medical equipment providers, skilled nursing homes, ambulatory surgery centers, hospice, clinical laboratories, radiology/imaging) are excluded. The basis for these exclusions are practical. This is due to poor predictive power found in the development of the HCC model, and concern about the reliability of the diagnoses from non-physicians, or confusion arising from the coding of “rule-out” diagnoses that sometimes appears on laboratory or imaging records.

  9. Applications of HCCs for prospective payment protocols require that the diagnoses be obtained from the baseline (i.e., prior) year. These classifications are used as the basis for reimbursement in the subsequent year. This model evolved from multiple studies over two decades (e.g., Ash, Porell, Gruenberg, et al., 1989; Ellis, Pope, Iezzoni, et al., 1996). Clinical applications of HCC or other condition groups, such as for assignment of members/patients into special clinics or care management panels, have found improved prediction of service use and expenditures if concurrent diagnoses are incorporated into the classification (Dudley, Medlin, Hammann, et al., 2003). Because of this and evidence that using a single year to identify diagnoses for an individual may lead to an under counting of conditions and a bias toward classifying beneficiaries who have higher cost (e.g., those with hospital stays or frequent or specialty physician visits) (Newcomer, Clay, Luxenberg, Miller, 1999), we have elected to use concurrent year claims in HCC assignment. Even with this adjustment there is still a concern that chronic condition prevalence and service use are under reported in the IHSS recipient population. This occurs for several reasons. First, Medicaid reimbursed service use is reliably reported only for those in fee for service. Services covered under managed care capitation agreements (such as hospitals, skilled nursing facilities, physicians, and other health care providers) do not usually generate a billing or reimbursement claim. Managed care enrollees are omitted from any analysis involving diagnostic classifications or counts of conditions. Secondly, recipients dually eligible for Medicare or other payers such as the Veterans Administration may have services exclusively or substantially paid for by these sources. In such circumstances, there will be no or fewer Medicaid claims and diagnoses reported. A third factor is that Medicaid claims have fields for recording only two diagnoses. When a patient has (or their service claim involves) more than two conditions, then the number of diagnoses will be under reported on the claim. This may result in some conditions not being recorded on the claims records. These factors are not thought to be differentially distributed within recipient age groups or their provider types.


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®