Center for Injury Epidemiology, Liberty Mutual Research Institute for Safety, 71 Frankland Road, Hopkinton, Massachusetts 01748, USA.
Inj Prev. 2011 Dec;17(6):407-14. doi: 10.1136/ip.2010.030593. Epub 2011 Apr 11.
Bayesian methods show promise for classifying injury narratives from large administrative datasets into cause groups. This study examined a combined approach where two Bayesian models (Fuzzy and Naïve) were used to either classify a narrative or select it for manual review.
Injury narratives were extracted from claims filed with a worker's compensation insurance provider between January 2002 and December 2004. Narratives were separated into a training set (n=11,000) and prediction set (n=3,000). Expert coders assigned two-digit Bureau of Labor Statistics Occupational Injury and Illness Classification event codes to each narrative. Fuzzy and Naïve Bayesian models were developed using manually classified cases in the training set. Two semi-automatic machine coding strategies were evaluated. The first strategy assigned cases for manual review if the Fuzzy and Naïve models disagreed on the classification. The second strategy selected additional cases for manual review from the Agree dataset using prediction strength to reach a level of 50% computer coding and 50% manual coding.
When agreement alone was used as the filtering strategy, the majority were coded by the computer (n=1,928, 64%) leaving 36% for manual review. The overall combined (human plus computer) sensitivity was 0.90 and positive predictive value (PPV) was >0.90 for 11 of 18 2-digit event categories. Implementing the 2nd strategy improved results with an overall sensitivity of 0.95 and PPV >0.90 for 17 of 18 categories.
A combined Naïve-Fuzzy Bayesian approach can classify some narratives with high accuracy and identify others most beneficial for manual review, reducing the burden on human coders.
贝叶斯方法在将大型行政数据集的伤害叙述分类为原因组方面显示出了前景。本研究考察了一种联合方法，即使用两种贝叶斯模型（模糊和朴素）来对叙述进行分类或选择进行手动审查。
从 2002 年 1 月至 2004 年 12 月期间向工人赔偿保险公司提交的索赔中提取伤害叙述。将叙述分为训练集（n=11,000）和预测集（n=3,000）。专家编码员为每个叙述分配了两位数字的美国劳工统计局职业伤害和疾病分类事件代码。使用训练集中手动分类的病例开发了模糊和朴素贝叶斯模型。评估了两种半自动机器编码策略。第一种策略是如果模糊和朴素模型对分类意见不一致，则将病例分配进行手动审查。第二种策略使用预测强度从同意数据集选择更多病例进行手动审查，以达到 50%的计算机编码和 50%的手动编码水平。
当仅使用一致作为过滤策略时，大多数由计算机编码（n=1,928，64%），留下 36%供手动审查。整体（人工加计算机）的综合敏感性为 0.90，对于 18 个 2 位数事件类别中的 11 个，阳性预测值（PPV）>0.90。实施第二种策略可提高结果，整体敏感性为 0.95，对于 18 个类别中的 17 个，PPV>0.90。
朴素-模糊贝叶斯方法的组合可以以较高的准确率对某些叙述进行分类，并识别最有利于手动审查的其他叙述，从而减轻人工编码员的负担。
