Stroke tracked by administrative coding data: is it fair?

C ommunity-based registries are considered the best source of information on stroke incidence, risk factors, characteristics , and outcome. They can be used for developing adequate strategies for stroke care and prevention. Criteria to perform an ideal stroke incidence study had been defined and refined in the past 2 decades and include standards for definitions, methods , and data presentation. 1–4 Complying with those criteria makes the obtained results reliable and the studies comparable to one another. However, community-based registries are challenging and are extremely time, labor, and resource demanding. Cross-sectional, cohort, and case-control studies, not at the community level, are routinely performed to obtain reliable data on several aspects concerning stroke characteristics and care. They can be as challenging as community-based registries , particularly when high numbers of subjects need to be included to ensure statistical power. Furthermore, those same studies do not reflect disparities across a nation because they are usually performed in geographic regions that more likely involve centers with active teams promoting better stroke care. Alternatively, administrative coding data (also called claims data or billing) have been introduced as a tool for stroke research because they are associated with less resource wasting. Nearly all Western countries currently collect hospital discharge data in some form. They are anonymous, plentiful , inexpensive, and widely available in electronic format. 7 In addition, they are quickly retrievable, thereby offering a remarkable advantage in terms of cost and affordability. Furthermore, because stroke is prominently a hospital-managed condition, administrative data might be suitable for monitoring its epidemiology. The advantage is greater when multiple years of data are needed for the analysis, such as when monitoring time trends or assessing the impact of different treatment strategies by using observations before and after the change. Administrative files typically contain infor-The application of administrative coding data to surveillance purposes can, therefore, often be accomplished with relatively modest additional efforts and costs. Despite the advantages of administrative coding data, they are not flawless, and the degree of their weakness raises the question of the validity of some of the studies that have used them. Users of administrative coding data should recognize that data, particularly diagnosis-related group codes, were not designed for surveillance purposes but rather to justify reimbursement. Each admission is assigned to a diagnosis-related group, and depending on that diagnosis-related group the hospital is paid a predetermined lump sum, regardless of the costs associated with …