某醫學中心血液檢驗白血球分類自動驗證閾值的適當性評估

Evaluation the Autovalidation Criteria of White Blood Cell Differential Count Flagging of Hematology Examination in the Medical Center

洪維彤(Wei-Tung Hung) ； 潘品靜(Pin-Jing Pan) ； 溫士玄(Xei-Shi Wen) ； 湯勝輝(Sheng-Hui Tang) ； 張錦標(Jin-Biou Chang) ； 張皓婷(Hao-Ting Chang)

《中華職業醫學雜誌》 29卷3期 (2022/07) Pp. 205-210

摘要

目的：檢驗室血液常規分析中的白血球分類計數，因人工驗證耗時長且過程繁瑣。本研究為進一步提升檢驗的敏感度及特異性，降低檢驗報告時間（turnaround time; TAT）及大量檢體處理效率，及減少人員錯誤，能有效增加檢驗室進行檢驗報告的自動驗證（autovalidation）的比例，可符合檢驗室認證之要求。材料與方法：Sysmex XN-10血球計數分析儀，原廠建議IG%（immature granulocyte）數值大於2需人工閱片，Left shift、Atypical lymphocyte、Blasts/Abn lymphocyte之Q-flag閾值分別為110、120、100。收集2021年1月至6月共55446筆報告，評估IG%最佳閱片閾值。IG%大於2，區分2.1-5.0、5.1-10.0、10.1-15.0三組進行分析。利用MedCalc繪出ROC曲線，針對白血球相關且具有Q-flag之閱片閾值進行評估。結果：IG%的三組統計分析結果：（1）靈敏度分別為：59.4%、27.8 %、7.5%。（2）特異性分別為：100%、99.7%、98.6%。（3）偽陽性分別為：0%、0.3%、1.2%。（4）偽陰性分別為：6.6%、11.7%、15.0%。檢驗室閱片閾值，Q-flag的分析結果；其ROC Curve在原本閾值的特異性分別為：85.4%、96.9%、80.3 %。偽陽性分別為：14.3%、3.0%、19.5%。為減少不必要的人工閱片時間，找出Q-flag最佳調整點，分別為：240、180、180。其ROC Curve在特異性分別為：96.2%、99.9%、91.6%；偽陽性分別為：3.7%、0.1%、8.3%。結論：經本研究可驗證原廠建議與檢驗室作業相同。IG%大於2需同時觀察是否有其它白血球相關的flag出現才具有臨床意義。而Q-flag閾值調整後，Left shift、Atypical lymphocyte及Blasts/Abn lymphocyte的特異性都增加，分別為：11.2%、3%, 11.3%；偽陽性都下降，分別為：10.6%、2.9%、11.2%；需人工確認的檢體數減少，分別為：72.2%、90.7%、55.2%，提升TAT之要求。

關鍵字

自動驗證 ； 檢驗報告時間（TAT） ； Q-Flag ； Sysmex XN-10

並列摘要

Purpose: The analysis process of differential count of WBC is time-costing and complex. The aim of our study is to elevate efficiency, specificity and sensitivity of examination, and to decrease turnaround time and error rate of reports. To meet medical center certification and laboratory accreditation requirements. Materials and methods: According to Sysmex XN-10 manufacturer's suggestion, the immature granulocyte % (IG %)＞ 2 needs artificial confirm through blood smear by medical staff. We collected 55546 samples from 2021 January to 2021 June. The threshold of Left shift、Atypical lymphocyte、Blasts/Abn lymphocyte of Q-flag is 110, 120, 100 separately. To verify the review criteria for IG %, we divided sample into three groups 2.1-5.0, 5.1-10.0 and 10.1-15.0. Using MedCalc software to draw ROC curve for evaluation. Results: The three group of IG % (1) Sensitivity: 59.4%, 27.8 %, 7.5% separately. (2) Specificity: 100%, 99.7%, 98.6% separately. (3) False positivity: 0%, 0.3%, 1.2% separately. (4) False negativity: 6.6%、11.7%、15.0% separately. Specificity of original criteria of ROC Curve: 85.4%, 96.9%, 80.3 % separately, and False positivity：14.3%, 3.0%, 19.5% separately. To decrease microscopic review rate of our laboratory. The best adjustment cutoff value of Q-flag: 240,180,180 separately. Specificity of original criteria of ROC Curve: 96.2%, 99.9%, 91.6 % separately, and False positivity: 3.7%, 0.1%, 8.3% separately. Conclusion: Through our study can prove our clinical result has the same review criteria with manufacturer's suggestion. When IG %＞ 2 and other white blood cell related flag are both observed at the same time is meaningful. After adjusted the threshold of Q-flag, increasing specificity of Left shift、Atypical lymphocyte、Blasts/Abn lymphocyte:11.2%, 3%, 11.3%; decreasing False positivity:10.6%, 2.9%, 11.2%; The percentage for artificial review: 72.2%, 90.7%, 55.2% separately. Improving the efficiency of autovalidation for differential count in the laboratory and TAT.

並列關鍵字

Auto validation ； Turnaround time ； Q-Flag ； Sysmex XN-10