影響成人心血管疾病患者合併中風之個人因素與地區變異分析 (2000-2007年)

Abstract

目的：首先藉由空間統計分析瞭解臺灣地區20-64歲心血管疾病與其合併中風之地區變異，包含偵測心血管疾病與其合併中風高盛行率聚集之地區、以及比較不同年盛行率群聚之時空變遷。其次探討心血管疾病患者之個人層次預測因子、社會經濟地位、鄉鎮層次地區變項與跨層次交互作用對其有無合併中風的影響。 方法：採橫斷式研究設計、次級資料分析。資料主要以2001年國民健康訪問調查、2005年國民健康訪問暨藥物濫用調查及其全民健康保險資料串連資料檔為主，對象為心血管疾病患者。空間統計分析部分，以最近鄰法(k-Nearest Neighbors)之最近5個鄉鎮(k = 5)為鄰區的權重矩陣，並以全域型空間自相關(Global Spatial Autocorrelation)指標Moran’s I 與Getis-Ord Genral G以及區域型空間自相關(Local Spatial Autocorrelation)指標LISA(Anselin Local Moran’s I)與Getis-Ord Gi*進行群聚分析比較。多層次分析中的階層線性迴規模型部分，包含隨機係數模型、截距預測模型；個人層次自變項包括人口特質、健康狀態、健康行為、飲食狀態、時間、社會經濟地位變項，鄉鎮(市區)層次變項有四個，分別為Log平均所得、Log平均所得LISA指標、地區劣勢因素分數、地區劣勢因素分數LISA指標，資料來源為財政部財稅資料中心的綜合所得稅申報核定統計專冊、內政部統計處、以及全民健康保險資料庫。 結果：空間統計分析部分，2001年、2000-2002年心血管疾病盛行率在臺灣地區整體上有明顯群聚現象，分佈模式相似且皆位於部分台北縣市、高雄縣市地區。2001年至2005年之心血管疾病盛行率由群聚現象轉變為隨機，且2005年心血管疾病高盛行率的顯著群聚區域已不是2001年分佈於南北都會區(部分台北縣市、高雄縣市)等地，而有往臺灣中西部(即彰化縣、雲林縣、嘉義縣、南投縣等)包含沿海地區移動之趨勢。階層線性迴規模型部分，刪除遺漏值後，納入本研究之心血管疾病患者樣本共計1,360人，197個鄉鎮(市區)。研究顯示，年齡、性別、有無工作狀況與心血管疾病患者是否合併中風有相關；而在控制個人層次後，鄉鎮(市區)層次有達統計上顯著差異的地區變項與有無合併中風皆呈現顯著負相關。

結論：隨著年齡增加、男性、以及沒有工作的心血管疾病患者其合併中風之風險會提升許多；另外，當居住在平均所得較高、鄰區皆為高平均所得的群聚地區，其合併中風的風險會較低。由此可知，不同的個人預測因子、社會經濟地位與地區因素對心血管疾病患者是否會合併中風造成不同程度的效果與影響。

Objectives: To find out geographical variations of cardiovascular disease (CVD) and cardiovascular disease with stroke aged 20-64 in Taiwan by applying spatial analysis, including the detection of cluster(s) of high CVD prevalence and high CVD with stroke prevalence, and the comparison of the space-time transitions of high prevalence cluster(s) among different years. Then, assess the effects of individual factors, individual socioeconomic status, town-level variables, and cross-level effects from a sample of CVD adult patients with stroke. Methods: Data were obtained from National Health Interview Survey in 2001 (2001 NHIS), National Health Interview Survey in 2005 (2005 NHIS), and each National Health Insurance Research Databases. Distanced-based k-Nearest Neighbors (k=5) were used as spatial weights matrix for spatial analysis, and the measures of Global Spatial Autocorrelation (Moran’s I and Getis-Ord Genral G) and Local Spatial Autocorrelation (LISA and Getis-Ord Gi*) were also used to exhibit the spatial clustering of CVD and CVD with stroke. Individual-level variables were included individual characteristics, health outcome, health behavior, diet condition, time, and individual socioeconomic status. Town-level data were derived from Ministry of Finance (Financial Data Center), Ministry of the Interior (Department of Statistics), and National Health Insurance Research Database, including 4 variables. Multilevel models including random coefficient model and intercept-as-outcome, were used in the analyses.

Results: 2001 and 2000-2002 CVD prevalence rates demonstrated spatial clustering characteristics, which distribution patterns of high prevalence were not only similar but also clustered in part of Taipei and Kaohsiung. The spatial clustering of high CVD prevalence was transformed from clustered to random among 2001 and 2005, moreover, the significant spatial clustering areas of high CVD prevalence were tended to shift to the Midwest of Taiwan from 2001 to 2005. For multilevel analysis, we excluded missing data on individual information and the remaining study sample included 1,360 CVD patients nested within 197 towns. As shown in the results, age and gender were positively related to CVD patients with stroke, and work condition were negatively related to CVD patients with stroke. Town-level variables were negatively related to CVD patients with stroke after controlling individual-level variables. Conclusions: Elder, male, and unemployed CVD adult patients had higher risk with stoke, compared to those younger, female, working patients. Furthermore, CVD adult patients who live in the higher average income areas and clustered districts had lower risk with stroke, compared to those live in the lower average income areas and clustered districts. Therefore, different individual-level and town-level variables were associated with CVD adult patients with stroke.