The contribution of surveillance systems in providing valuable information for injury prevention and control is widely recognized; for example, surveillance data can be used to highlight the burden of injury, set priorities for prevention, and evaluate preventive strategies [1, 2]. Estimates of the population burden of injuries differ, though, depending on how information is obtained. Detailed trauma registries and special surveillance systems [e.g., ] contain rich contextual information on particular subsets of injuries, but since such databases are generally not population-based, they cannot be used to estimate the incidence of injury. Although population-based surveys can yield estimates of the total burden of non-fatal injuries across a broad spectrum of injury severity, they often include insufficient sample sizes for studying small population subgroups , and are subject to recall errors .
Administrative health care databases, due to their presumed near complete coverage of injuries requiring medical care and their lack of reliance on self-reports, may be particularly useful for injury surveillance. Such databases allow for local or regional estimates of the burden of injury, which has been identified as an important goal [2, 6, 7], and since they are pre-existing, they are cost-efficient. Administrative data also provide an opportunity to examine health care use for injury. Administrative databases only capture injuries that receive medical care, however, and since surveillance using administrative data is often based on hospitalization data alone, only relatively severe injuries are included. Decisions regarding whether to seek medical care and where to seek care for an injury may be influenced by outside factors (such as access to care, care-seeking, and practice patterns), which may lead to selection biases [8–10].
In Ontario, Canada, administrative health care databases that may provide information on the incidence of non-fatal injuries include hospital discharge and physician billing data. Although hospital discharge data have been widely used in Canada to study injury, the feasibility of using physician billing information for injury surveillance has rarely been investigated . These data, if valid, may help to expand the coverage of administrative databases to include more minor injuries, capturing care delivered in physicians' offices and emergency departments. Although minor injuries have less impact on individuals and are less costly to the health care system on a per-injury basis, minor injuries have a large impact in terms of total population morbidity due to their frequent occurrence [12, 13]. Expanding coverage by including physician billing data would thus serve to provide a more comprehensive picture of the total health care burden of injury, and may also reduce selection biases in the surveillance data.
It is not clear how the injury information provided by administrative databases compares with that obtained from population-based surveys. A study of adolescent injuries we conducted using data from the Ontario Health Survey (OHS)  presented a unique opportunity to explore such a comparison; a subset of the 1996–1997 OHS data was linked by respondent to Ontario administrative health care databases, including both hospital discharge and physician billing data.
The overall purpose of this study was to explore the feasibility and value of using physician billing data for Ontario, Canada, along with hospitalization data, for the surveillance of adolescent injuries. The first objective was to document the burden of adolescent injury based on administrative health care data, focusing on the relative contribution of physician billing information and comparing overall estimates with surveillance information from survey data. The second objective was to examine data quality issues, by directly comparing adolescent injuries identified using administrative health care databases ("administratively-defined injuries") with those identified using self-report survey data ("self-reported injuries").