目標：本研究旨在探討全人工膝關節置換術（total knee arthroplasty，簡稱TKA）病人術前共病分群與術後感染之關聯。方法：採回溯性世代研究並以英國生物資料庫住院檔（採計1980到2020年）判定初次TKA患者（n=20,112），追蹤結果變項術後感染直至2021年止；術前共病以術前所有醫療紀錄之疾病診斷碼定義。以潛在類別分析（latent class analysis，簡稱LCA）法進行共病分群，並用Cox比例風險迴歸模式評估TKA患者之共病分群的短期與長期術後感染之風險。結果：LCA分群最合適的結果為四群，分別命名為低共病（n=10,573, 52.6%）、三高及心血管疾病（n=4,597, 22.9%）、皮膚及感染性疾病（n=3,390, 16.8%）及多共病群（n=1,552, 7.7%）。與低共病群相比，皮膚及感染性疾病群短期術後感染風險較高（風險比（hazard ratio, HR）=1.86, 95%信賴區間（confidence interval, CI: 1.14-3.05）；至於長期術後感染風險，皮膚及感染性疾病群（HR=2.25, 95% CI: 1.52-3.35）、三高及心血管疾病群（HR=1.90, 95%CI: 1.25- 2.89）與多共病群（HR=2.78, 95% CI: 1.52-5.08）皆比低共病群高。結論：TKA病人術前共病可分成四大分群且與低共病群相比，其他三群都具有長期術後感染風險，其中皮膚及感染性疾病群也有較高的短期術後感染風險，研究發現有助臨床醫師辨別術後感染較高風險病人。

Objectives: This study was conducted to investigate the association between preoperative comorbidity clusters and surgical site infection after total knee arthroplasty (TKA). Methods: This retrospective cohort study was conducted using the data (from the UK Biobank database) of 20,112 adults who had undergone primary TKA between 1980 and 2020. The primary outcome was postoperative infection. The patients were followed up until the end of 2021. Information on the patients’ preoperative comorbidities was obtained from their pre-TKA medical records. A latent class analysis was performed to cluster patients with TKA according to preoperative comorbidities. A Cox proportional-hazards model was used to estimate the risks of short- and long-term postoperative infection in patients with different comorbidity clusters. Results: The latent class analysis revealed four comorbidity clusters: low comorbidity (10,573 patients [52.6%]); hypertension, hyperglycemia, hyperlipidemia, and cardiovascular disease (4,597 [22.9%]); skin and infectious diseases (3,390 [16.8%]), and multiple comorbidities (1,552 [7.7%]). The risk of short-term postoperative infection was higher in the skin and infectious diseases cluster than that in the low-comorbidity cluster (hazard ratio [HR]: 1.86; 95% confidence interval [CI]: 1.14–3.05). The risk of long-term postoperative infection was elevated in the skin and infectious diseases cluster (HR: 2.25; 95% CI: 1.52–3.35); hypertension, hyperglycemia, hyperlipidemia, and cardiovascular disease cluster (HR: 1.90; 95% CI: 1.25–2.89); and multiple comorbidities cluster (HR: 2.78; 95% CI: 1.52–5.08). Conclusions: We identified four comorbidity clusters in patients who underwent TKA. Compared with the low-comorbidity cluster, the other three clusters had increased risks of long-term postoperative infection; the skin and infectious diseases cluster also had an elevated risk of short-term postoperative infection. Our findings can help clinicians identify patients at an elevated risk of post-TKA infection.