目標：瞭解地理區位差異在公共衛生研究扮演重要的角色，但在全民健保資料庫中，並不包含居住地資訊，侷限探索地理與健康關連的可行性。本研究之目的在於提出估計居住地變項的原則及其適用性。方法：利用2005年承保抽樣歸人檔，依不同被保險人特質及就醫地，建立三種不同的居住地估計方法，並與內政部各鄉鎮市區現住人口數為基準，比較其相關係數，再以肺癌與肝癌就醫人數的鄉鎮分佈進行驗證。結果：以被保險人投保地、呼吸道感染門診就醫地、被保險人身分、投保類別與被保險人年齡相互搭配之居住地估計方法，在不同都市化程度的鄉鎮與不同年齡層都有較佳的表現。僅以投保地為居住地或僅以呼吸道感染就醫地為居住地的估計方法，則僅適用於特定鄉鎮市區與特定年齡層人口。結論：本研究比較不同都市化程度與年齡結構的分層結果，確認居住地估計方法的可行性，並瞭解不同居住地估計方法的適用性。本研究結果可提供研究者選擇適合之居住地估計原則。

Objectives: Understanding the geographic patterns and regional differences in health status plays an important role in public health research; however, the place-of-residence (township level) of an insured is not available in National Health Insurance Research Databases (NHIRD). The objective of this study was to propose principles for estimating the place-of-residence (township level) in NHIRD. Methods: Based on demographic characteristics, insurance classification, location of hospital visit, and insurance registration of the insured, this study compared three methods of estimating the place-of-residence (township level) from the Longitudinal Health Insurance Database of NHIRD in 2005. Official statistics of the usual resident population in each township from the Department of the Interior were used as reference data for comparisons among the three methods. The study further verified these methods by comparing the estimated numbers and official statistics for the medical treatment of lung and liver cancer patients in 2005. Results: This study found that the method which combined insurance classification, location of hospital visit, and insurance registration provided an optimal estimate of place-of-residence in each area by different levels of urbanization and age-group. Consideration of either location of hospital visit or insurance registration may be appropriate for specific townships and age groups. Conclusions: This study demonstrated the feasibility of estimating place-of-residence in NHIRD and the applicability of these proposed methods.