RE: "ESTIMATING THE EFFECT OF PREVENTIVE SERVICES WITH DATABASES OF ADMINISTRATIVE CLAIMS: REASONS TO BE CONCERNED".
RE: "ESTIMATING THE EFFECT OF PREVENTIVE SERVICES WITH DATABASES OF ADMINISTRATIVE CLAIMS: REASONS TO BE CONCERNED".
复制标题
回复：“利用行政索赔数据库评估预防服务的效果：需要关注的理由”。
DOI：
10.1093/aje/kwz050
复制
发表时间：
2019
影响因子：
5
通讯作者：
Weiss,NoelS
中科院分区：
文献类型：
--
作者：
Weiss,NoelS
由于结肠镜检查期间可以识别并切除结肠和直肠息肉，因此进行该测试有可能降低后期结直肠癌 (CRC) 的发病率。 García-Albéniz 等人(2) 的队列研究旨在衡量筛查结肠镜检查在降低 CRC 发病率方面的成功程度。García-Albéniz 等人(2) 列举了从筛查时开始一直持续长达 13 年的 Medicare 参保者中的病例。他们将这一数字与根据当时未接受筛查的医疗保险参保者的比率（无论他们是否在以后接受筛查）进行的预期进行了比较。多年来，我一直主张采取措施提高流行病学研究的敏感性，即增加在确实存在关联时识别关联的可能性 (3)。我在本期杂志 (4) 中的评论是由于担心 García-Albéniz 等人 (2) 使用的方法会导致对与结肠镜筛查相关的任何真实发病率降低的迟钝估计：• 在初次筛查检查中发现的已经存在的 CRC 病例，无法通过在检查期间识别和切除息肉来预防。• 在某种程度上，医疗保险记录未能区分针对体征或症状进行的结肠镜检查和针对体征或症状进行的结肠镜检查如果进行筛查，由于这种错误分类，“已筛查”队列中将会出现大量虚假病例。• 如果最初未筛查的人员在随访期间继续进行筛查，则筛查后其 CRC 发病率将“污染”真正未筛查人员的发病率。García-Albéniz 等人。选择在随访开始后不评估队列成员的筛查接受情况，因此他们不得不接受这段时间筛查状态错误分类对结果的影响（2）。然而，在我的评论中，我提出了一项建议（呼应几年前提出的一项建议（5）），旨在加强筛选群体和未筛选群体之间的对比。具体来说，我建议将筛查者的随访推迟一段时间，该时间间隔与隐匿性癌症的估计持续时间相对应 (4)。对筛查发现的病例进行识别并从分析中删除，将导致仅在这段时间内癌症筛查呈阴性的人中的发病率较低，之后可以对筛查组和未筛查组的其余成员之间的癌症发病率进行公平比较。（在他们的信（1）的第二段中，García-Albéniz等人提到了对CRC 8年累积发病率的分析（6），其中对筛查组的“早期随访”
Because colon and rectal polyps can be identified and removed during colonoscopy, performance of this test has the potential to reduce the later incidence of colorectal cancer (CRC). In their cohort study, which sought to measure the extent to which screening colonoscopy succeeded in reducing CRC incidence, García-Albéniz et al.(2) enumerated cases among Medicare enrollees beginning at the time of screening and continuing for up to 13 years thereafter. They compared this number with that expected on the basis of rates among Medicare enrollees who had not been screened at that time (irrespective of whether they had gone on to be screened at a later date). For some years, I have advocated taking steps to increase the sensitivity of epidemiologic studies—that is, to increase the likelihood of identifying an association when one truly is present (3). My commentary in this issue of the Journal (4) was stimulated by a concern that the approach used by García-Albéniz et al.(2) would lead to a blunted estimate of any true incidence reduction associated with colonoscopy screening:• CRC cases identified at the initial screening examination, already being present, could not have been prevented by polyp identification and removal during that examination.• To the extent that Medicare records failed to discriminate between a colonoscopy performed in response to signs or symptoms and one performed for screening, there would be a spuriously large number of cases in the “screened” cohort resulting from this misclassification.• To the extent that initially unscreened persons went on to be screened during follow-up, their CRC incidence rates following screening would “contaminate” the rate for truly unscreened persons.García-Albéniz et al. chose not to evaluate receipt of screening in cohort members after follow-up began, so they had to accept the impact on the results of the misclassification of screening status during this period of time (2). In my commentary, however, I made a suggestion (echoing one made several years earlier (5)) that sought to sharpen the contrast between screened and unscreened groups. Specifically, I recommended delaying the initiation of follow-up of screened persons for an interval of time corresponding to the estimated duration of occult cancer (4). The identification and removal from the analysis of screen-detected cases would give rise to a low incidence among persons who screened as negative for cancer only for this duration of time, after which a fair comparison of cancer incidence could be made between the remaining members of the screened and unscreened groups.(In the second paragraph of their letter (1), García-Albéniz et al. refer to an analysis of the 8-year cumulative incidence of CRC (6) in which the “early follow-up” of screened