Comparison of Fee-for-Service, Mainstream Managed Care and HIV Special Needs Plans (SNPs) Shows Better Quality in Managed Care
There are currently three delivery system options that New York City Medicaid beneficiaries with HIV may choose: the fee-for-service system, mainstream managed care plans or HIV SNPs. The Department of Health designed a study to determine how the three delivery systems compared on a variety of quality of care and service utilization measures. Preliminary results follow.
For this initial comparison of quality and utilization, it was determined to use administrative measures, which can be calculated with claims and encounter data. Most of the measures used in the study were adapted from the National Commission for Quality Assurance's (NCQA) HEDIS® measurement set and modified to comply with HIV clinical guidelines. Measures were chosen that related to HIV care (e.g. antiretroviral therapy and viral loads) as well as adult preventive services, (e.g. dental visits and breast cancer screening). Other measures were chosen to assess how well the systems controlled undesirable events such as emergency room visits and hospital admissions.
Staff from the Department's Office of Health Insurance Programs and the AIDS Institute met to identify relevant measures and refine denominator and numerator specifications. The resulting measurement set is the table below:
|1) Antiretroviral Therapy (ARV)||The percentage of beneficiaries with HIV, age 2 and older, who had 12+ 30-day prescriptions for antiretroviral medications.|
|2) Viral Loads||The percentage of beneficiaries with HIV, age 2 and older, who had 12+ 30-day prescriptions for antiretroviral medications and who had at least two viral load measurements, at least six months apart, during the measurement year.|
|3) Dental Visits||The percentage of beneficiaries with HIV, age 2 and older, who had at least one preventive dental visit during the measurement year.|
|4) Inpatient Admissions||The percentage of beneficiaries with HIV, age 2 years and older, who had at least one non-maternity inpatient admission during the measurement year.|
|5) Inpatient Admissions for Pneumonia||The percentage of beneficiaries with HIV, age 2 years and older, who had at least one inpatient admission during the measurement year and was diagnosed with pneumonia.|
|6) Emergency Room||The percentage of beneficiaries with HIV, age 2 years and older, who had at least one emergency room visit during the measurement year.|
|7) Breast Cancer Screening||The percentage of women with HIV, ages 42 to 69, who had a mammogram during the measurement year or the year prior.|
|8) Cholesterol Screening||The percentage of beneficiaries, with HIV, age 2 years and older, with at least three prescriptions for ARV who had a cholesterol screen at least 30 days after their first ARV prescription during the measurement year.|
Numerator specifications for each measure were programmed in SAS, and both Medicaid claims and encounter data were pulled to look for evidence that a service was delivered within requisite timeframes. All coding was validated by programmers at IPRO, the Department's external quality review organization.
To be eligible for the study, a beneficiary had to have a diagnosis of HIV in calendar year 2005 and still be enrolled in Medicaid in 2006 for nine months or more.
The denominator pull was conducted using criteria developed by the AIDS Institute and included rate codes, diagnosis and procedure codes, hospital DRGs (diagnostic-related groups) and pharmacy data. This initial pull of 67,000 beneficiaries was reduced to 52,951 by eliminating dually eligible beneficiaries and those less than two years of age and by only considering beneficiaries with nine or more months in the service delivery system in 2006. To eliminate false positives (most prevalent in the Medicaid managed care cohort) the cohort was run through the 3M Clinical Risk Grouper (CRG). The CRG tool uses both diagnosis and procedure codes and pharmacy data to assign beneficiaries based on severity of illness. By using the CRG tool, and requiring two or more diagnoses of 042, the eligible group was reduced to 21,745. This was the denominator used for the study and was stratified by delivery system
Demographic variables (age, gender, race/ethnicity, and aid category) differed across delivery systems. The observed rates may therefore be confounded by these variables, as well as other factors such as the presence of co-morbidities among persons with HIV. In order to control for these factors, and make rate comparisons more meaningful, a risk adjusted rate was calculated for each measure.
The table below provides preliminary risk-adjusted rates of performance across the three delivery systems:
|Measures:||Fee for Service||Managed Care||SNPs||Statewide Average|
|1) Antiretroviral Therapy||73.9||69.2||76.3||73.7|
|2) Viral Loads||39.8||42.4||55**||40.6|
|3) Dental Visits||46.0||32.2*||48.9||45.2|
|4) Inpatient Admissions||34.9**||25.7*||27.2*||34.0|
|5) Inpatient Admissions for Pneumonia||9.6||5.7*||6.5*||9.3|
|6) Emergency Room||43.4||42.2||40.6||43.2|
|7) Breast Cancer Screening||24.8*||52.6**||55**||27.7|
|8) Cholesterol Screening||15.9*||54.7**||67.3**||20.4|
- * significantly below statewide average
- **significantly above statewide average
There are several issues related to data used in this study that must be considered when interpreting the results. In order to limit burden on providers, reduce study expenses and expedite study results, the Department used administrative data from the Medicaid data system. This could influence study results as follows:
Despite our best efforts to refine the denominator criteria, we may still have beneficiaries in the study who do not have HIV. Without confirming diagnosis through medical record review we are dependent on physician coding of claims and encounters to identify our eligible group. Through the use of CRGs, we have attempted to minimize this problem; however, we believe that false positives still exist to some extent in the mainstream managed care cohort.
Claims in the form of billed rate codes may not have the necessary detail to substantiate that a service was rendered. If an accompanying CPT code indicating a service was performed was not on the claim, the FFS delivery system was not credited for the measure.
Incomplete Encounter Data
Underreporting of encounter data is an issue for some managed care plans. This problem is particularly evident if the plan is reliant on a vendor to submit data, such as dental.
These rates are preliminary and are subject to change pending completion of a data validation project involving the AIDS registry. Additional measures to be included in the study include CD4 counts and cervical cancer screening. Going forward, the Department will require all managed care plans (mainstream and SNPs) to annually report on a set of HIV-specific measures as part of the annual collection of Quality Assurance Reporting Requirement (QARR) data. In addition, a study to examine key components of HIV care will be conducted, focusing on providers who have not taken part in any Department-sponsored HIV measurement and improvement activities. Ongoing monitoring of service utilization, medications and member satisfaction is also planned.