彰化肺結核防治效益評估

研究背景結核病為全球最重要之傳染病之一，自1994年起，世界衛生組織陸續推出針對結核病防治的不同面向包含臨床診斷與治療、持續性追蹤，以及國家層級的持續性策略等全球結核病防治建議與策略規劃，包括：1994年的DOTS strategy以及2006年的Global Plan to Stop TB。2015年WHO推行End TB Strategy，主要目標為相較於2015年，至2035年結核病發生率降低90％，死亡降低95%。彰化縣長期投入結核病防治措施包含實驗室診斷代檢網及都治計畫。本論文目的在於 (1) 建立貝氏多階層變動點卜瓦松模型階層結核病罹病風險評估模型， (2) 運用彰化結核病監測實證資料結合(1)評估此兩項結核病防治措施(實驗室診斷代檢網及都治計畫)推行後對於降低結核病罹病風險之成效， (3) 基於(1)與(2)預測並推估彰化縣是否可於20235年達到WHO End TB Strategy目標。材料與方法本研究使用前期與後期比較之類實驗設計(Before-and-After Quasi-experimental Design)方法，評估結核病發生率在代檢網及都治計畫介入前後對於結核病罹病風險降低之效益。運用彰化縣由1990年至2016年各年結核病新發個案數與當年人口結構分佈之實證資料進行效益評估。所收集之實證資料包含性別、年齡層(分為0-29歲、30-49歲、50-69歲及70-84歲)、本國籍與非本國籍別，以及26個鄉鎮(區分為6類都市化程度)。本研究建構貝氏多階層變動點卜瓦松模型，將前述實證資料中年齡、性別、國籍別以階層迴歸方法納入考量，並且將鄉鎮異質性以及都市化程度之對於結合罹病風險之影響納入於模型建構中。本研究除將前述年齡、性別、國籍之人口特性以及鄉鎮差異在結核病罹病風險(基礎發生率)與年代效應之影響納入考量，並將以兩項結核防治介入措施：實驗室診斷代檢網及都治計畫開始推行之年代建立此貝氏模型中的兩個變動點進行效益評估及結核病罹病風險未來趨勢之預測。在此貝氏多階層變動點卜瓦松模型結合彰化實證資料，評估前述兩項主要結核病防治措施對於結核病罹病風險之影響，做為族群介入措施效益評估，以及預測2035年是否可達到WHO結核防治目標之依據。研究結果彰化縣1990年至 2016年之間共有22,240例新診斷結核病例，發生率為每 100,000人年63.5人。自1990年至2016年在代檢網及都治兩個重要防治措施介入之後的結核病發生率實際趨勢以及利用模型進行1990年至2035年的發生率預測。在介入前階段(1999-2001)，整體發病率呈上升趨勢，從1990年的36.5人/100,000上升至2001年76.2人/100,000。在實施代檢網之後(第一個變動點)，結核病診斷日期提前，因此在2002 年發生率提高至106人/100,000。隨之在2006年整體發生率下降至75.3人/100,000。在實施都治之後(第二個變動點)，整體發生率更下降到55.0人/100,000，以貝氏變動點模型預測發生率2035年結核病發生率約為26.3/100,000。結核病發生率在不同年齡之間的差異大。預測結核病發生率在2035年5.2/100,000下降到1.8/100,000、14.5/100,000到3.7/100,000、53.8/100,000到13.1/100,000和331.2/100,000到114.9/100,000，下降趨勢在30-49歲以及50-69歲最為明顯。除此之外，加入外籍勞工結核病個案，0-29歲發生率在近期呈現上升的趨勢，從2015年11.8/100,000上升至2035年12.4/100,000。本研究進一步利用貝氏變動點卜瓦松模型評估代檢網及都治介入措施的效益，代檢網及都治介入措施效益為66.8% (95%可信區間: 60.6%, 72.0%)。在 0-29歲、30-49歲、50-69歲、70 歲以上分別介入措施效益達65.1% (95%可信區間: 39.9%, 81.7%)、72.6% (95%可信區間: 59.4%, 82.3%)、73.8% (95%可信區間: 66.0%, 80.3%) 和 63.3% (95%可信區間: 54.8%, 70.5%)。然而在不同都市化程度的地區差異上，從都市化效益高至低介入效益分別為73.1 (95%可信區間: 59.3, 83.1)、63.2 (95%可信區間: 45.4, 76.3)、78.8 (95%可信區間: 69.6, 85.9)、67.6 (95%可信區間: 46.5, 81.7)、65.4 (95%可信區間: 24.9, 87.0)及62.7 (95%可信區間: 43.9, 76.5)，顯現出都市化程度不同對於鄉鎮間結核病罹病風險之差異性與影響。結論本論文利用貝氏多階層變動點模型結合彰化長期結核病族群資料評估結果，顯示結核病實驗室診斷代檢網以及都治策略兩項主要介入措施可降低族群結合風險達66.8%，此介入效益隨年齡層以及都市化程度有所不同。雖然結核病發生率在結核防治措施介入後呈現下降趨勢，但所有年齡層族群皆無法達到在2035年相較於2015年結核病罹病風險下降發生率90%之目標。而外籍勞工結核病發生率逐漸上升之趨勢，對於彰化縣0-29歲年齡層結核病發生率影響甚大，對於年長族群介入造成之風險下降亦不及於其他族群。結核病防治措施之規畫應針對不同目標族群罹病風險特性發展包含篩檢、診斷，以及治療之介入措施。

關鍵字

結核病防治；都治計畫；代檢網；貝氏多階層變動點卜瓦松模型

並列摘要

Background As one of the chronic contagious diseases that threaten the population health of the word, tuberculosis (TB) is the second leading cause of death among infectious diseases affecting a total of 10 million people worldwide in 2020. In response to this threat, Word Health Organization (WHO) has initiated a series of mitigating strategies including DOTS (1994), Global Plain to Stop TB (2006), and End TB Strategy (2015) covering the continuum of diagnosis, treatment, and follow-up and monitoring strategies at nationwide level with emphasis on the sustained support and implementation of policy makers and stakeholders. The End TB Strategy proposed in 2015 set the goal of decreasing TB incidence and mortality by 90% and 95%, respectively. Following the nationwide strategies implemented in Taiwan, Changhua county have engaged in TB prevention including the establishment of National Reference Laboratory Network (NRLN) for accurate laboratory diagnosis and sentinel laboratory-based surveillance since 2001 and DOTS for the provision of sustained and monitored treatment for TB patient since 2006. The aims of this study is thus (1) to establish the Bayesian hierarchical change point Poisson regression model for assessing the risk of TB; (2) to apply the model established in (1) to the empirical data on TB risk with long-term follow-up in Changhua county to evaluate the effectiveness of the two prevention strategies, NRLN and DOTS, in reducing the risk of TB; and (3) to predict the risk of TB in Changhua in 2035 guided by the goal of WHO End TB Strategy on the basis of the estimated results derived from (1) and (2) Material and methods This study applied a before-and-after quasi experimental design to the empirical data on TB collected in Changhua covering 27 years (1990 to 2016) to assess the effectiveness in reducing TB risk by comparing the risk before and after the implementation of the two prevention strategies, NRLN and DOTS. The information on the demographics including sex and age (categorized into the groups of 0-29 years, 30-49 years, 50-69 years, and 70-84 years), nationality (Taiwanese and foreign workers), and township (categorized into six urbanization levels) were collected. The Bayesian hierarchical change point Poisson regression model were developed to take into account the factors associated with TB risk by using the empirical information mentioned above and also the heterogeneity across the townships of Changhua county. The two interventions, NRLN and DOTS, were incorporated into the model by setting two change points for year 2002 and 2006, respectively. The effectiveness of the two interventions and the risk of TB in Changhua in 2035 were then derived on the basis of the estimated results of the Bayesian hierarchical change point regression model. Results A total of 22,420 newly diagnosed TB cases were reported in Changhua county between 1990 and 2016, which gives the average TB incidence rate of 63.5 per 100,000 person-year. The empirical time trend of TB incidence in parallel with two important nationwide intervention policies from 1990 to 2016 was observed and the predicted tine trend was projected from 2017 until 2035. During the pre-intervention phase (1999-2001), the overall incidence had shown an escalating trend from 36.5 per 100,000 person-year in 1990 until 76.2 100,000 in 2001. After the implementation of supra national reference laboratory network (NRLN, the first change point) that advanced the date of TB diagnosis since 2002, there was a jump to 106 per 100,000 person-year. Since then, the overall incidence regressed to 75.3 per 100,000 person-year until 2006. After implementation of DOTS (the second change-point), the overall incidence further dropped to 55 per 100,000 person-year (Period 3). Based on Bayesian change point model to predict the incidence trend in the next two decades, the predicted incidence of TB will be 26.3 per 100,000 person-year. There was a significant difference in the change in TB incidence between year 2015 and 2035 across the age groups in Changhua. The risks of TB in year 2015 were estimated as 5.2, 14.5, 53.8, and 331.2 for the age group of 0-29, 30-49, 50-69, and above 70 years, respectively. The corresponding TB risks predicted form the Bayesian hierarchical change point Poisson model in year 2035 were 1.8, 3.7, 13.1, and 114.9, respectively. The reduction in TB incidence among those aged 0-29 and aged 30-49 were larger than other two groups. Regarding the population of young foreign workers aged 0-29 years, there was an increasing trend in the recent period, from 11.8 per 100,000 in 2015 to 12.4 per 100,000 in 2035, whereas a decreasing trend were estimated for the corresponding Taiwanese population (5.2 per 100,000 in 2015 to 1.8 per 100,000 in 2035). On the basis of the estimated results derived from the Bayesian hierarchical change point Poisson regression model, the effectiveness of the two nationwide intervention, NRLN and DOTS, were evaluated. The overall effectiveness of NRLN and DOTS are estimated as 66.8% (95% CI: 60.6%, 72.0%). Among the different age group, the effectiveness for aged 0-29, aged 30-49, aged 50-69 and aged above 70, are estimated as 65.1% (95% CI: 60.6%, 72.0%),72.6% (95% CI: 59.4%, 82.3%), 73.8% (95% CI: 66.0%, 80.3%) and 63.3% (95% CI: 54.8%, 70.5%), respectively. However, in terms of the heterogeneity among different counties with six levels of urbanization, the effectiveness of intervention from high to low urbanization level are estimated as 73.1 (95% CI: 59.3, 83.1), 63.2 (95% CI: 45.4, 76.3), 78.8 ( 95% CI: 69.6, 85.9), 67.6 (95% CI: 46.5, 81.7), 65.4 (95% CI: 24.9, 87.0) and 62.7 (95% CI: 43.9, 76.5). The heterogeneity among different counties with urbanization level largely affects the baseline risk and intervention impact of TB. Conclusion On the basis of the empirical data of TB with long-term follow-up for the population in Changhua, the effectiveness of the implementation of NRLN and DOTS in reducing TB risk was estimated as 66.8%. The heterogeneity in the effectiveness of these population-wide strategies across the townships with different level of urbanization were revealed. Although these two interventions were effective in reducing TB risk, the goal of WHO End TB Strategy in reducing TB incidence by 90% compared with that of 2015 cannot be reached by 2035. An increasing trend in TB risk were identified among the foreign labors aged 0-29 years. The effectiveness of the intervention strategies is also attenuated for the elder group. These results suggest the necessity for a prevention strategy covering the aspects of screening, diagnosis, and treatment tailored by the characteristics of each group.