摘要：Objectives. We sought to identify people living with HIV/AIDS from Medicare and Medicaid claims data to estimate Medicaid costs for treating HIV/AIDS in California. We also examined how alternate methods of identifying the relevant sample affect estimates of per capita costs. Methods. We analyzed data on Californians enrolled in Medicaid with an HIV/AIDS diagnosis reported in 2007 Medicare or Medicaid claims data. We compared alternative selection criteria by examining use of antiretroviral drugs, HIV-specific monitoring tests, and medical costs. We compared the final sample and average costs with other estimates of the size of California’s HIV/AIDS population covered by Medicaid in 2007 and their average treatment costs. Results. Eighty-seven percent (18 290) of potentially identifiable HIV-positive individuals satisfied at least 1 confirmation criterion. Nearly 80% of confirmed observations had claims for HIV-specific tests, compared with only 3% of excluded cases. Female Medicaid recipients were particularly likely to be miscoded as having HIV. Medicaid treatment spending for Californians with HIV averaged $33 720 in 2007. Conclusions. The proposed algorithm displays good internal and external validity. Accurately identifying HIV cases in claims data is important to avoid drawing biased conclusions and is necessary in setting appropriate HIV managed-care capitation rates. In 2010, the White House Office of National AIDS Policy outlined an ambitious National HIV/AIDS Strategy for the United States that called for evaluation strategies that would “obtain data (core indicators) that capture the care experiences of people living with HIV without substantial new investments.” 1 Surveillance systems already in place in each state provide the Centers for Disease Control and Prevention with comprehensive data on incident HIV and AIDS cases. 2 However, much less is known about the medical treatments received by people living with HIV/AIDS and the cost of those treatments. Much of the cost of HIV/AIDS treatment is borne by public insurance programs, principally Medicaid and Medicare. These 2 programs provide health insurance for more than half of people living with HIV/AIDS who are receiving care. 3,4 The importance of Medicaid as a source of funding for HIV/AIDS treatment of low-income persons will grow substantially after full implementation of the Affordable Care Act, which eliminates the additional disability requirement for Medicaid eligibility in states accepting the Medicaid expansion, thereby extending coverage to nondisabled, low-income people living with HIV/AIDS in those states. Because of its prominent role in insuring low-income people living with HIV/AIDS, Medicaid can provide a rich source of data on the types and costs of treatments delivered to some of the most vulnerable individuals with HIV/AIDS. Insurance claims data can potentially allow us to monitor HIV/AIDS treatment without substantial new investments because most claims data are stored as computerized records. Claims data provide a comprehensive picture of medical care received from a variety of providers in multiple settings (outpatient, inpatient, laboratory, pharmacy), contain procedure codes that detail the services provided, and include cost of the treatment. By contrast, medical records tend to have smaller scope, in terms of both numbers of patients and services covered. Furthermore, medical records most often lack payment information. Insurance claims data can provide information on a large number of individuals, even among those with relatively low-prevalence conditions, which is valuable in reducing the variability of estimates of per capita expenditures. However, the greater precision afforded by large administrative data sets is of little value if estimates are based on an inappropriate sample. Claims data are primarily designed for billing purposes; thus, they generally lack clinical detail important for selecting cases with a particular disease. 3,5 For example, claims data will document whether a laboratory test was performed, but not the test result. Therefore, analysts must rely on the diagnosis information on insurance claims. 6 Professional medical records specialists code diagnoses on inpatient claims, leading to greater accuracy and reliability of diagnosis information coming from inpatient stays. However, diagnosis coding is more error-prone in the outpatient sector, which has accounted for an increasing percentage of HIV/AIDS care since 1996 when antiretroviral medication (ARV) began to dramatically reduce hospitalization for HIV/AIDS. 7 This has increased the challenges of identifying people living with HIV/AIDS from insurance claims data. We applied a practical algorithm for identifying people living with HIV/AIDS in insurance claims data to estimate Medicaid costs for treating HIV/AIDS in California. We also examined how alternate methods of identifying the relevant sample affect estimates of per capita costs.
