Patients suffering from serious illnesses such as heart disease, diabetes, and cancer who use newer drugs are likely to live longer than patients using drugs approved by the Food and Drug Administration in earlier years.
In The Effect of Drug Vintage On Survival: Micro Evidence from Puerto Rico's Medicaid Program (NBER Working Paper No.10884), NBER Research Associate Frank Lichtenberg finds that patients suffering from serious illnesses such as heart disease, diabetes, and cancer who use newer drugs are likely to live longer than patients using drugs approved by the Food and Drug Administration in earlier years.
Lichtenberg's data mainly come from information recorded by Puerto Rico's Health Insurance Administration (ASES), which contracts with private managed care organizations to provide services to Medicaid patients, which account for 40 percent of Puerto Rico's population. His sample includes data on over 500,000 Medicaid patients. To qualify for Medicaid in 2002 a family of four could not have an annual income larger than $16,440.
Lichtenberg analyzes the impact of the vintage of drugs used to treat patients on their probability of survival, conditional on demographic characteristics (age, sex, and region), their use of medical services, and the nature and complexity of their illness. Because each pharmacy claim includes the National Drug Code, Lichtenberg can determine the earliest date of the FDA's approval for each prescription's active ingredients. He calculates values for drugs introduced post-1970, post-1980, and post-1990. He further organizes his data by disease group, and examines mortality data provided by Puerto Rico's Department of Health during the period 2000-2. Lichtenberg factors into his models such variables as the number of medical, pharmacy, and hospital claims, 15 different diagnosis categories, and dummy variables for region and for single-year-of-age-by-sex.
The data reveal that ASES beneficiaries using newer drugs during January-June 2000 were less likely to die by the end of 2002, conditional on the covariates. The estimated mortality rates strictly decline with respect to drug vintage. For pre-1970 drugs, the estimated mortality rate was 4.4 percent. The mortality rates for users of the 1970s, 1980s, and 1990s drugs were 3.6 percent, 3.0 percent, and 2.5 percent respectively. These differences in mortality rates are highly significant statistically, Lichtenberg concludes.
Moreover, he notes that people who used more medical services during January-June 2000 were more likely to die by the end of 2002. For example, one additional medical claim (physician visit) is associated with a .0031 (about 8 percent) increase in the probability of death. Although use of medical services presumably reduces mortality, given initial (pre-treatment) health status, people in the worst initial health group use the most medical services.
Medicaid patients in Puerto Rico receive older medications than U.S. Medicaid patients. Lichtenberg believes this is so at least in part because in Puerto Rico, physicians bear the costs of the drugs they prescribe, with those costs deducted from the physician's health fund reimbursements. The estimates suggest that if the ASES vintage distribution were the same as U.S. Medicaid's, the ASES mortality rates would have been 5.3 percent lower (3.5 percent versus 3.7 percent).
In addition to estimating the model for the entire ASES population, Lichtenberg estimates the model separately for patients with diseases of the circulatory system, patients with endocrine, nutritional, and metabolic diseases and immunity disorder (primarily diabetes), and patients with neoplasms. With only one exception, within each group the coefficients of all three drug-vintage variables were negative and highly significant.
"In this study," Lichtenberg cautions, "we did not control for the effect of the vintage of medical products and services other than drugs on survival, and this may have affected our estimates of the effect of drug vintage. We plan to address this issue in future research."
-- Matt Nesvisky