摘要：Objectives. We implemented active surveillance for Guillain–Barré syndrome (GBS) following seasonal or H1N1 influenza vaccination among the Medicare population during the 2009-2010 influenza season. Methods. We used weekly Medicare claims data to monitor vaccinations and subsequent hospitalizations with principal diagnosis code for GBS within 42 days. Group sequential testing assessed whether the observed GBS rate exceeded a critical limit based on the expected rate from 5 previous years adjusted for claims delay. We evaluated the lag between date of service and date of claims availability and used it for adjustment. Results. By July 30, 2010 (after 26 interim surveillance tests), 14.0 million seasonal and 3.3 million H1N1 vaccinations had accrued. Taking into account claims delay appropriately lowered the critical limit during early monitoring. The observed GBS rate was below the critical limit throughout the surveillance. Conclusions. Medicare data contributed rapid safety monitoring among millions of 2009–2010 influenza vaccine recipients. Adjustment for claims delay facilitates early detection of potential safety issues. Although limited by lack of medical record review to confirm cases, this claims-based surveillance did not indicate a statistically significant elevated GBS rate following seasonal or H1N1 influenza vaccination. In 2009, public health efforts in the United States to address the pandemic of a novel influenza A (H1N1) virus included a federal effort to facilitate development and distribution of vaccines. 1,2 Five monovalent vaccines against the novel strain (A/California/7/09-like virus) were approved as strain change formulations to each manufacturer’s licensed seasonal influenza vaccine. 3–9 Safety monitoring of licensed vaccines has traditionally relied largely on passive surveillance, 10 which is national and timely but subject to multiple limitations (e.g., underreporting, reporting biases, and lack of denominator data on vaccinated persons 11 ), and epidemiological studies, 12 which may require years to complete. Early detection of safety issues is especially important for influenza vaccine monitoring because most vaccinations are administered within a short period of a few months. Rapid active surveillance using data from managed care networks has emerged more recently. 13–15 However, these databases may underrepresent the elderly and may not have sufficient statistical power to evaluate very infrequent adverse events. To address these issues and prepare for a potential influenza pandemic, a pilot project was initiated in 2006 by the Food and Drug Administration (FDA) and Centers for Medicare and Medicaid Services (CMS) that aimed to monitor influenza vaccine safety as rapidly as possible (i.e., near real time) among the Medicare population. Medicare insures approximately 38.8 million elderly (aged ≥ 65 years) and 7.8 million younger persons with disability or end-stage renal disease. 16 The pilot phase (2006–2009) aimed to develop the technical and methodological capability to use Medicare administrative data for safety monitoring as soon as data accrued each week. It focused on surveillance for specific conditions after vaccination such as Guillain–Barré syndrome (GBS), an autoimmune disorder that produces weakness or paralysis, which was associated with the 1976–1977 swine influenza (A/New Jersey/1976/H1N1) vaccine. 17–20 Some studies of the 1976–1977 vaccine indicated relative risks of 4 to 9, corresponding to 5 to 10 excess cases per million persons vaccinated, with most or all of the elevated risk within 6 weeks after vaccination. Higher relative risks (approximately 11–18 in some analyses) were found for onset in weeks 2 and 3. 18–20 A challenge in using claims and potentially other types of health care databases for active surveillance is the lag between date of service and date when information is available in the system. In the Medicare program, a vaccination or hospitalization is observed after a health care provider seeks payment for services by submitting a claim to Medicare administrative contractors that process it and then submit it to CMS. Not accounting for this delay can potentially lead to biased results during early monitoring. We implemented methods to account for claims delay in sequential monitoring and describe results of the first prospective application of these methods, specifically to monitor for GBS after seasonal or H1N1 influenza vaccination among the Medicare population during the 2009-2010 influenza season.
