臺灣地區糖尿病健康結果與危險因子解構之地理差異：一項區域性健康不平等的全面研究

Abstract

背景：糖尿病（Diabetes Mellitus, DM）是一種全球盛行的代謝性疾病，對公共衛生構成長期且日益嚴峻的挑戰。糖尿病經常伴隨多種潛在併發症，並對個人與整體社會造成顯著的醫療與經濟負擔。在臺灣，雖然近年（2017–2021年）糖尿病的發生率與盛行率趨於穩定，且死亡率並未顯著上升，但相關危險因子與糖尿病健康結果（如發生率與死因別死亡率）在各縣市間仍呈現顯著的地理差異，反映出明確的區域性健康不平等現象。 目的：本研究旨在建構完整的疾病歷程架構，包含「危險因子暴露」→「疾病發生」→「醫療照護（診斷與治療）」→「死亡」的進程，並探討其中健康不平等最為顯著、可被解釋與最具介入潛力的環節。本研究更進一步解析臺灣各縣市間糖尿病健康結果及其相關危險因子分布的地理差異，並量化三項主要可修正之危險因子──高身體質量指數（High BMI）、身體活動量不足（Physical Inactivity）與吸菸行為（Smoking）──對糖尿病發生率不平等的貢獻程度。 方法：本研究涵蓋2017至2021年，並整合臺灣健保資料庫（NHIRD）、國民健康訪問調查（NHIS）、營養健康狀況變遷調查（NAHSIT）與國人死因統計檔，分析糖尿病發生率、死因別死亡率、死亡對發病比（MI ratio）、治療覆蓋率及血糖控制率等指標，並針對三項主要危險因子估算其單獨與聯合之族群可歸因分率（Population Attributable Fraction, PAF）。本研究進一步運用變異係數（Coefficient of Variation, CV）作為地區不平等分析工具，比較縣市別糖尿病發生率的目前情境與使用PAF估計移除某危險因子後之縣市別糖尿病發生率反事實情境，兩情境下變異係數之改變，以量化各危險因子對地理空間差異的解釋力。 結果：研究顯示，臺灣各縣市間在糖尿病相關健康指標上存在明顯的區域性差異。東部地區（如花蓮縣、臺東縣）具有較高的糖尿病發生率；花蓮縣與屏東縣之糖尿病死亡率亦居各縣市前列；部分中南部縣市則是在治療覆蓋率及血糖控制成功率方面表現相對落後。在三項主要危險因子當中，高BMI之族群可歸因分率最高（男性介於33.24%至45.72%；女性介於17.55%至42.67%），對糖尿病發生率的整體貢獻最大；相較之下，身體活動量不足（男性介於14.90%至17.92%；女性介於13.80%至18.83%）與吸菸（男性介於12.75%至21.98%；女性介於0.23%至2.84%）之貢獻度明顯較低。進一步以目前與反事實情境下變異係數改變量之分析後發現，男性高BMI對縣市間糖尿病發生率差異之解釋力最高（3.45%）；女性則以身體活動量不足為最（1.18%）。然而，這三項因子之解釋力皆不足以全面說明地理不平等，顯示尚有其他潛在因素驅動臺灣地區糖尿病發生的不平等分布。整體而言，研究結果突顯出目前既有策略在應對地區性健康不平等方面的侷限，亦顯示需針對具代表性的危險因子採取更差異化的政策。 結論：本研究強調糖尿病防治策略應具備地區化與性別導向特性，並呼籲將肥胖防治納入政策核心，作為減少糖尿病負擔與健康不平等的首要目標。研究亦透過CV方法量化並分解不平等來源，提供精準公共衛生於規劃中的資源配置與介入優先順序之實證依據。

Background: Diabetes Mellitus (DM) is a globally prevalent metabolic disorder, representing a persistent public health challenge. It is frequently accompanied by a wide range of complications and imposes a substantial medical and economic burden on individuals and society. In Taiwan, while the incidence and prevalence of diabetes remained relatively stable during the study period from 2017 to 2021, and the diabetes-specific mortality did not exhibit a significant upward trend, there are still notable geographical disparities persisting across counties in both diabetes-related outcomes as well as in the distribution of associated risk factors. These variations in both outcomes and risk factor distributions reflect pronounced regional health inequalities. Objectives: This study aims to conceptualize the full disease pathway framework, including “risk exposure” → “disease onset” → “healthcare access (diagnosis and treatment)” → “diabetes-attributable mortality”, seeking to identify the stages where disparities are most pronounced, most attributable, and most amenable to intervention. This study also aims to analyze the geographical disparities in diabetes-related health outcomes and associated risk factor distributions across Taiwan’s counties, and to further quantify and decompose the contributions of three major modifiable risk factors—high body mass index (BMI), physical inactivity, and smoking—to the observed inequalities in diabetes incidence. Methodology: This study analyzes data from 2017 to 2021, using nationally representative sources: National Health Insurance Research Database (NHIRD), National Health Interview Survey (NHIS), Nutrition and Health Survey in Taiwan (NAHSIT), and Cause of Death Data. Key diabetes-related indicators, including incidence, diabetes-attributable mortality, mortality-to-incidence ratio (MI-Ratio), treatment coverage, and glycemic control, are assessed across regions. Population Attributable Fractions (PAFs) are estimated for high body mass index (BMI), physical inactivity, and smoking, being calculated both individually and jointly, which are also disaggregated by sex and geographic unit. This study further employed the Coefficient of Variation (CV) as a tool to analyze reginal inequality by comparing county-level diabetes incidence under the current scenario with a counterfactual scenario in which the population-attributable fraction (PAF) for each risk factor was removed. The change in CV between the two scenarios is used to qualify the explanatory power of individual risk factors on spatial disparities. Results: The findings reveal persistent geographical inequalities in diabetes-related health indicators across Taiwan. Eastern regions such as 花蓮縣and臺東縣 exhibit higher diabetes incidence; 花蓮縣 and 屏東縣exhibit elevated diabetes-specific mortality; several central and southern counties demonstrate relatively low treatment coverage and suboptimal glycemic control rates. Among the three risk factors, high BMI demonstrates the highest average population attributable fraction (PAF) across counties (males range from 33.24% to 45.72%; females range from 17.55% to 42.67%), indicating its dominant contribution to diabetes incidence. In comparison, the PAFs for physical inactivity (males range from 14.90% to17.92%; females range from 13.80%% to 18.83%) and for smoking (males range from12.75% to 21.98%; females range from 0.23% to 2.84%) have lower attribution. Further analysis of the difference in CV between the current and counterfactual scenarios indicates that, for males, high BMI have the strongest explanatory power for inter-county disparities in diabetes incidence (3.45%); among females, physical inactivity accounts for the greatest proportion (1.18%). However, the explanatory power of all the three risk factors remains limited, suggesting the presence of other unmeasured factors potentially driving the unequal geographic distribution of diabetes incidence in Taiwan. Conclusion: This study underscores the importance of implementing geographically and gender-sensitive diabetes prevention strategies in Taiwan, with an emphasis on obesity reduction as a core public health policy priority to reduce the burden of diabetes. By applying Coefficient of Variation (CV) as decomposition methods to quantify the sources of disparities, the study provides an empirical foundation for precision public health planning, including targeted resource allocation and prioritized interventions.