台灣癌症病患自殺模式之時空分析

Abstract

目的：相較於其他歐美國家，台灣癌症病患之自殺率高出許多。然而目前相關研究多著重在個人層面，為了解台灣癌症病患自殺之整體情況，此研究從時空分佈的角度切入，描繪台灣癌症病患之自殺模式。空間分析可界定未知健康事件的位置(location)，對疾病的預防、健康服務和改善資源分配極為重要。 方法：此研究將使用國民健康局癌症登記串死因檔，對台灣1995-2007年間之癌症病患自殺之情形進行了解和探索。納入之危險因子包括性別、年齡、罹患癌症類型、確診至自殺間隔時間、自殺季節以及自殺方式；社區環境因子則由2001和2005年國民健康局之國民健康訪問調查中，計算各鄉鎮市區吸菸率、嚼食檳榔率；該鄉鎮市區是否有提供安寧照護服務之醫療機構則由健保局特約醫事機構查詢得知。以羅吉斯迴歸(logistic regression)進行癌症患者自殺危險因子的探討，使用之工具為SAS9.2；並以Scan statistics對台灣癌症病患自殺情形進行時空分析(space and space-time analysis)，以了解不同時間點，癌症病患自殺情形在空間分佈上的改變，使用工具為SaTScan，並以ArcGIS繪製地圖。 結果：研究結果發現，若是男性、40-64歲之中壯年或者罹患之癌症會造成外觀改變者，會有較高的自殺風險，且皆有隨著時間上升的趨勢；若是罹患預後較好的癌症種類，自殺風險反而較高，與先前研究有所呼應，先前的研究發現台灣自殺之癌症病患中，比例最多為肺癌以及口腔癌患者，與其他國家的研究結果相仿，但是接著為預後良好的乳癌和血液類腫瘤；但若是罹患預後良好的癌症，確診後短期內自殺的風險顯著較低。癌症病患之自殺情形有空間群聚，高風險群聚區在雲嘉南、台南、高雄、屏東、台東和花蓮等地區，且群聚區內的鄉鎮有顯著較高的吸菸率和嚼食檳榔率。時空分析發現高風險群聚區也分佈在類似的地區，群聚時間為1999-2003年；縮小掃描半徑後(cluster size<10%)，發現兩個次群聚區，第一個次群聚區涵蓋8個台南、高雄地區之鄉鎮，群聚時間為2002年；第二個次群聚區涵蓋55個雲嘉南地區之鄉鎮，群聚時間為1999-2003年。 結論：整體而言，性別、年齡、罹患癌症類型和確診後至自殺時間間隔皆對癌症病患自殺風險有顯著的影響。癌症病患的自殺情形在1995-2007年間有顯著的時空群聚，高風險群聚區和高風險次群聚區內，年齡和性別對癌症患者自殺之影響不同於全台灣的分析結果。將來可針對高風險群聚區之人口特性、文化、社會和環境等，進行自殺防治策略之設計。

Cancer patients' suicide rate in Taiwan is much more than other European countries. This research aims to understand the whole picture of cancer patients’ suicide by using time-space analysis. Space analysis can locate health events which are unknown. It's important to disease prevention, health services and resources distribution. The dataset used in this study taken from the 1995-2007 Cancer Registry linking Death Certificate File published by Taiwan’s Department of Health, which include gender, age, cancer type, time since diagnosis, death month and suicide methods. We also use 2001 and 2005 National Health Interview Survey published by Taiwan’s Department of Health to calculate rates of betel quid chewing and cigarette smoking by township. The dataset about terminal care services were taken from the website of Bureau of National Health Insurance. The statistical methods used in this study are logistic regression by SAS9.2, and Scan Statistics by SaTScan. We found that if cancer patient is male, 40-64 years old or his/her cancer type would change appearance will have higher risk to commit suicide, and the risk is rising by time. Cancer patients with good prognosis have lower suicide rate in short time interval since diagnosis. Suicide behaviors among cancer patients have spatial clusters. The most likely clusters are in Yunlin, Chiayi, Nantou, Tainan, Kaohsiung, Pingtung, Taitung and Hualien, those townships have higher rate of smoking and chewing betel nuts besides. Time-space analysis also found clusters in those townships, in 1999-2003. There are two subclusters founded after we lessen the scan radius (cluster size<10%), the primary subcluster locate in Tainan and Kaohsiung in 2002; the secondary subcluster locate in Yunlin, Chiayi and Nantou between 1999-2003. On the whole, risk factors such gender, age, cancer type and time since diagnosis have significant influences to suicide behavior among cancer patients. We found out that there have time-space clusters in 1995-2007, and those clusters have different characteristics. We have to focus on the environment of higher risk clusters and those patients who live in there to make the policy of suicide prevention.