日本在COVID-19疫情期間自殺率增加的空間群聚特徵

Abstract

背景：在COVID-19全球流行期間，世界多數國家地區自殺率降低或維持不變，而日本是少數自殺率增加的地區。現有文獻大多討論自殺率改變的時間序列特徵，而討論疫情對於自殺率影響的空間群聚位置與特徵較為缺乏。

研究目的：本研究欲偵測因為重大影響社會的事件（COVID-19），短期間「自殺率的變化」的空間群聚及環境特徵。並與疫情期間「自殺率」的空間群聚位置與特徵比較，闡述偵測「變化量」的空間群聚的在疫情期間的角色。

方法：本研究關注於2020年4月至2021年6月，日本疫情爆發至疫苗施打之前的四波疫情，以日本作為研究區，市區町村作為空間單元，針對「自殺率」與「自殺率的改變」兩者，偵測空間群聚。再利用多階層羅吉斯回歸模型捕捉熱區相對於冷區，分別找出可能的環境特徵，進行比較以及詮釋。

結果：自殺率增加的空間群聚與自殺率高的空間群聚，其位置與環境特徵均具差異。在位置上，自殺率增加傾向群聚於城市的邊緣，而高自殺率傾向群聚於鄉村與山區。在環境特徵上，前者群聚於COVID-19感染率高、人口密度高、獨居比例低的區域，而後者群聚於人口密度低的地方，且與COVID-19感染率無關。

詮釋：這項研究顯示了相較於偵測自殺率群聚，發生短期極端事件（疫情）時，偵測「自殺率改變」的空間群聚更能偵測到極端事件的影響。推測其能排除固有的地區環境因子，去偵測事件對於自殺率影響的空間群聚。

Background: During the COVID-19 pandemic, Japan experienced an excess suicide rate compared to the pre-pandemic period, whereas most other countries experienced the opposite. Most studies have focused on the change in suicide rate in the timeline, but few have discussed the geographic variation in the impact of COVID-19 on the suicide rate.

Objectives: This study aimed to detect spatial clusters of increased suicide rates during an acute significant event (COVID-19 in Japan) and their environmental characteristics. Subsequently, the spatial clusters of high suicide rates were used as comparisons to interpret the role of detecting spatial clusters of increased suicide rates.

Methods: This study focused on four waves of virus outbreaks in Japan from April 2020 to June 2021, before the vaccine was implemented. This study considers non-isolated municipalities in Japan as the study area. After detecting the spatial clusters of increased suicide rates and high suicide rates, this study used multi-level logistic regression models to capture environmental characteristics.

Results: Spatial clusters of increased suicide rates differed from those of high suicide rates. In terms of location, increased suicide rates tend to be clustered on the fringes of urban areas, whereas high suicide rates tend to be clustered in rural areas. In terms of environmental characteristics, increased suicide rates tended to cluster in places with higher COVID-19 infection rates, higher population densities, and lower single-household ratios. However, spatial clusters with high suicide rates tend to occur in areas with lower population densities and are unrelated to COVID-19 infection rates.

Interpretation: This study shows that the detection of spatial clusters of “changes in suicide rates” during short-term extreme events (epidemics) might be more sensitive than the detection of clusters of suicide rates in detecting the impact of extreme events. This is because it can exclude inherent regional environmental factors and show geographic variation in the impact of events on suicide rates.