口腔癌篩檢轉介及診斷延遲分析及相關因子研究

Abstract

研究背景 口腔癌篩檢為次段預防手段，旨在病灶發生前期施以早期治療或針對癌前病變提早介入，以期降低口腔癌死亡率，甚或發生率。然而，在篩檢到治療及最後追蹤會牽涉到不同階段，包括針對陽性的轉介、轉介後至確診，及確診到治療，任一時期的延遲均可能對口腔癌預後有不良的影響。 研究目的 本論文旨在利用近五年臺灣針對吸菸及嚼食檳榔的30歲以上成人口腔黏膜篩檢探討整個口腔癌篩檢多階段進程，及各階段的影響因素。 材料與方法 本研究針對台灣2017-2021年參與口腔黏膜篩檢計3,115,239人次，探討其陽性、轉介、診斷、治療及預後。在各個階段先各別以傳統羅吉斯迴歸分析及Cox比例風險迴歸分析探討接受篩檢個案人口學變項及提供篩檢機構特質對於個案在各個階段的影響。隨後發展口腔癌篩檢健康照護多階段模式，利用貝氏蒙地卡羅Metropolis Hasting迴歸抽樣機器學習估計同時估計從篩檢、陽性、轉介、確診、治療至死亡之相關參數。 結果 2017-2021年參與口腔黏膜篩檢民眾的口腔黏膜下纖維化及均質性薄白斑比率約為4.87%，其他陽性發現比率約為2.88%，陽性個案中整體轉介率約為70%；年齡較長、單純吸菸或已戒菸戒檳、醫療院所篩檢、醫療層級較高、耳鼻喉科醫師或牙醫師、較嚴重陽性發現等均有較高的轉介率，影響的因子不論是利用羅吉斯迴歸分析或Cox比例風險迴歸均相似，在疫情前(2017-2019年)與疫情後(2020-2021年)各影響因子的顯著性及影響大小亦差異不大。 轉介個案中有1502人有口腔癌診斷，分析轉介至診斷間隔，發現年齡愈高，診斷愈慢 (相較於30-39歲，60-69歲者aOR=0.50 [95% CI: 0.27-0.94]，70歲以上者aOR=0.41 [95% CI: 0.20-0.84])。而教育程度為無或國小者(aOR: 1.56, 95% CI: 1.13-2.17)、由牙科醫師篩檢者(aOR: 2.06, 95% CI: 1.40-3.04)、篩檢結果為疑似口腔癌者(aOR: 5.35, 95% CI: 1.99-14.36)，診斷速度較快。針對確診者，男性(aHR=1.78, 95% CI: 1.29-2.46)及專科、大學教育程度以上(aHR=1.27, 95% CI: 1.03-1.58)接受治療速率明顯較快。 在口腔癌篩檢健康照護多階段模式中，疫情後出現陽性個案略多(aOR=1.04, 95% CI: 1.03-1.06)，但轉介速率較疫情前快 (aRR=2.36, 95% CI; 2.32-2.40)，男性轉介速率(aRR=1.03, 95% CI: 1.00-1.07)及接受治療速率(aRR=1.32, 95% CI: 1.26-1.40)均高於女性，其他延遲轉介速率的因素尚有年齡較輕、教育程度較低、嚼檳且吸菸、較基層篩檢單位、非牙科及耳鼻喉科醫師及山區離島縣市等。而延遲治療的因素則包括教育程度較低、較基層篩檢單位、牙科醫師、非醫療不足縣市。 結論 本論文針對口腔癌篩檢不同階段進展分析與其相關因素找出較有延遲轉介、診斷及治療的高風險對象，研究結果有助於針對高風險對象加強健康促進。

Background: Oral cancer screening, the secondary prevention, is anticipated to reduce mortality from or even incidence of oral cancer with detection and treatment of early cancer and oral potentially malignant disorders. However, delays in the different stage of screening process from screen, referral, confirmation, and treatment may deter the effectiveness of screening. Objectives: To develop an Oral Cancer Health Care Multi-state Network Model and explore the delays in Taiwan oral cancer screening, which targets at betel chewers and cigarette smokers aged 30 years and elder with empirical data, and elucidate the associated factors. Materials and Methods: We collect 3,115,239 subjects who participated in the oral cancer screening in Taiwan from 2017 to 2021. We explored the compliance of referral after positive finding, diagnosis, treatment and prognosis. Logistic regression models and Cox proportional hazards regression models were used to explore the different factors including demographic variables and the type of screening institutions at each step. We also developed an Oral Cancer Health Care Multi-state Network Model to incorporate the sequential process following screening. The Bayesian Monte Carlo Metropolis Hasting regression sampling of machine learning was applied for the posterior distribution of hyperparameters. Results: Between 2017 and 2021, the proportion of oral submucosal fibrosis (OSF) and homogeneous thin leukoplakia (LP) among 3,115,239 participants in oral mucosal screening was about 4.87%, and the proportion of other positive findings was about 2.88%. Among the later, the overall referral rate was about 70%. Elderly people, no chewing or quit chewing but smoking, quit chewing and quit smoking, screening in hospitals, higher medical level, otolaryngologists or dentists, more severe positive findings was associated with higher referral rates. The results of logistic regression or Cox proportional hazards model are similar. The effect size in factors before the pandemic (2017-2019) and after the pandemic of COVID-19 (2020-2021) is also similar. A total of 1,502 referred cases had oral cancer diagnosis. Among them, elderly people tend to delay in diagnosis (aged in 60-69: aOR=0.50 (95% CI: 0.27-0.94) (aged over 70: aOR=0.41 [95% CI: 0.20-0.84]) compared with those in 30-39. And those with no education or elementary school (aOR: 1.56, 95% CI: 1.13-2.17), screened by a dentist (aOR: 2.06, 95% CI: 1.40-3.04), suspected oral cancer screening results (aOR: 5.35, 95% CI: 1.99-14.36), the diagnosis was faster. Among patients of oral cancer, males (aHR=1.78, 95% CI: 1.29-2.46) and those with college or university education (aHR=1.27, 95% CI: 1.03-1.58) had significantly faster treatment rates. In the Oral Cancer Health Care Multi-state model, there were slightly more positive cases after the pandemic of COVID-19 (aOR=1.05, 95% CI: 1.03-1.06), but the referral rate was faster than that before the pandemic (aRR=3.78, 95% CI; 3.22 -4.18). Male had higher referral rate (aRR=1.04, 95% CI: 1.00-1.07) and treatment rate (aRR=1.50, 95% CI: 1.27-1.77) than female. Other factors associated with lower referral rate included younger age, lower education level, chewing and smoking, basic level medical institution, non-dental nor otolaryngologists, and residents in remote areas. Factors that delay in treatment included lower education, basic level medical institution, dentists, and not areas with insufficient medical resources. Conclusions: We developed an Oral Cancer Health Care Multi-state Network Model and elucidated the relevant factors to different steps involved in this network. The findings from this study is anticipated to identify the high-risk group for delays in referral, diagnosis and treatment. Such a findings may aids the precision promotion program in the compliance of oral cancer screening.