Cembrowski G S, Hackney J R, Carey N
Park Nicollet Medical Center, Minneapolis, Minn. 55416.
Arch Pathol Lab Med. 1993 Apr;117(4):437-43.
The Clinical Laboratory Improvement Act of 1988 (CLIA 88) has dramatically changed proficiency testing (PT) practices having mandated (1) satisfactory PT for certain analytes as a condition of laboratory operation, (2) fixed PT limits for many of these "regulated" analytes, and (3) an increased number of PT specimens (n = 5) for each testing cycle. For many of these analytes, the fixed limits are much broader than the previously employed Standard Deviation Index (SDI) criteria. Paradoxically, there may be less incentive to identify and evaluate analytically significant outliers to improve the analytical process. Previously described "control rules" to evaluate these PT results are unworkable as they consider only two or three results. We used Monte Carlo simulations of Kodak Ektachem analyzers participating in PT to determine optimal control rules for the identification of PT results that are inconsistent with those from other laboratories using the same methods. The analysis of three representative analytes, potassium, creatine kinase, and iron was simulated with varying intrainstrument and interinstrument standard deviations (si and sg, respectively) obtained from the College of American Pathologists (Northfield, Ill) Quality Assurance Services data and Proficiency Test data, respectively. Analytical errors were simulated in each of the analytes and evaluated in terms of multiples of the interlaboratory SDI. Simple control rules for detecting systematic and random error were evaluated with power function graphs, graphs of probability of error detected vs magnitude of error. Based on the simulation results, we recommend screening all analytes for the occurrence of two or more observations exceeding the same +/- 1 SDI limit. For any analyte satisfying this condition, the mean of the observations should be calculated. For analytes with sg/si ratios between 1.0 and 1.5, a significant systematic error is signaled by the mean exceeding 1.0 SDI. Significant random error is signaled by one observation exceeding the +/- 3-SDI limit or the range of the observations exceeding 4 SDIs. For analytes with higher sg/si, significant systematic or random error is signaled by violation of the screening rule (having at least two observations exceeding the same +/- 1 SDI limit). Random error can also be signaled by one observation exceeding the +/- 1.5-SDI limit or the range of the observations exceeding 3 SDIs. We present a practical approach to the workup of apparent PT errors.
1988年的《临床实验室改进法案》（CLIA 88）极大地改变了能力验证（PT）的做法，它规定：（1）对某些分析物进行令人满意的PT作为实验室运营的条件；（2）为许多这些“受监管”分析物设定固定的PT限值；（3）每个测试周期增加PT样本数量（n = 5）。对于许多这些分析物，固定限值比以前采用的标准差指数（SDI）标准宽得多。矛盾的是，识别和评估具有分析意义的异常值以改进分析过程的动力可能会减少。以前描述的用于评估这些PT结果的“控制规则”不可行，因为它们只考虑两三个结果。我们使用参与PT的柯达Ektachem分析仪的蒙特卡罗模拟来确定用于识别与使用相同方法的其他实验室结果不一致的PT结果的最佳控制规则。使用分别从美国病理学家学会（伊利诺伊州诺斯菲尔德）质量保证服务数据和能力验证测试数据中获得的不同仪器内和仪器间标准差（分别为si和sg），对三种代表性分析物（钾、肌酸激酶和铁）的分析进行了模拟。在每种分析物中模拟分析误差，并根据实验室间SDI的倍数进行评估。使用功效函数图（检测到的误差概率与误差大小的关系图）评估用于检测系统误差和随机误差的简单控制规则。根据模拟结果，我们建议筛查所有分析物中是否出现两个或更多超过相同±1 SDI限值的观测值。对于任何满足此条件的分析物，应计算观测值的平均值。对于sg/si比值在1.0至1.5之间的分析物，平均值超过1.0 SDI表示存在显著的系统误差。一个观测值超过±3 - SDI限值或观测值范围超过4个SDI表示存在显著的随机误差。对于sg/si比值更高的分析物，违反筛查规则（至少有两个观测值超过相同的±1 SDI限值）表示存在显著的系统或随机误差。一个观测值超过±1.5 - SDI限值或观测值范围超过3个SDI也可表示存在随机误差。我们提出了一种处理明显PT误差的实用方法。
