Assessment of the reliability of data collected for the department of Veterans Affairs National Surgical Quality Improvement Program
Assessment of the reliability of data collected for the department of Veterans Affairs National Surgical Quality Improvement Program
复制标题
DOI：
10.1016/j.jamcollsurg.2007.01.012
复制
发表时间：
2007-04-01
影响因子：
5.2
通讯作者：
Booker, Brigitte
中科院分区：
文献类型：
--
作者：
Davis, Chester L.;Pierce, John R.;Booker, Brigitte
背景：退伍军人事务部（VA）医疗检查员办公室研究了VA国家外科质量改进计划（NSQIP）收集的数据的可靠性。该研究的重点是病例选择的偏见，准确性的患者死亡的报告，和患者的风险变量和outcome.Study设计：从15个VA医疗中心的样本的手术数据进行了分析。对于病例选择偏倚，审查人员应用NSQIP标准从数据库中纳入或排除2，460名患者，将其结果与NSQIP工作人员的结果进行比较。为了准确报告死亡患者，审查人员将10，444份NSQIP记录的社会安全号码与VA受益人识别和记录子系统，VA患者治疗文件和社会安全管理局死亡文件中的社会安全号码进行了比较。为了测量评价者间的可靠性，评价者在550例患者的病历中重新提取了59个变量，这些病历也记录在NSQIP database.RESULTS中：在病例选择偏倚方面，评价者同意NSQIP在2,460例病例中的2,418例（98%）的决定。计算机记录匹配确定的死亡人数比NSQIP的198人多4人，相差约2%。对于52个分类变量，未经机会校正的一致性为96%。对于48的52个分类变量，kappas范围从0.61到1.0（实质上几乎完美的协议），没有变量的kappas小于0.20（轻微到差的协议）。结论：这个样本的医疗中心显示，坚持在选择病例的NSQIP数据库，报告死亡，并收集患者的风险变量的标准。（美国科尔外科杂志2007; 204：550-560。(C)2007年由美国外科医生学会出版）。
BACKGROUND:The Office of the Medical Inspector of the Department of Veterans Affairs (VA) studied the reliability of data collected by the VA's National Surgical Quality Improvement Program (NSQIP). The study focused on case selection bias, accuracy of reports on patients who died, and interrater reliability measurements of patient risk variables and outcomes.STUDY DESIGN:Surgical data from a sample of 15 VA medical centers were analyzed. For case selection bias, reviewers applied NSQIP criteria to include or exclude 2,460 patients from the database, comparing their results with those of NSQIP staff. For accurate reporting of patients who died, reviewers compared Social Security numbers of 10,444 NSQIP records with those found in the VA Beneficiary Identification and Records Locator Subsystem, VA Patient Treatment Files, and Social Security Administration death files. For measurement of interrater reliability, reviewers reabstracted 59 variables in each of 550 patient medical records that also were recorded in the NSQIP database.RESULTS:On case selection bias, the reviewers agreed with NSQIP decisions on 2,418 (98%) of the 2,460 cases. Computer record matching identified 4 more deaths than the NSQIP total of 198, a difference of about 2%. For 52 of the categorical variables, agreement, uncorrected for chance, was 96%. For 48 of 52 categorical variables, kappas ranged from 0.61 to 1.0 (substantial to almost perfect agreement); none of the variables had kappas of less than 0.20 (slight to poor agreement).CONCLUSIONS:This sample of medical centers shows adherence to criteria in selecting cases for the NSQIP database, for reporting deaths, and for collecting patient risk variables. (J Am Coll Surg 2007; 204:550-560. (C) 2007 by the American College of Surgeons).