Center for Pharmacoepidemiology Research and Training, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA; Center for Clinical Epidemiology and Biostatistics, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA; Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA.
Department of Clinical Pharmacology and Pharmacometrics, Bristol-Myers Squibb, Philadelphia, PA.
J Am Coll Surg. 2019 May;228(5):752-759.e3. doi: 10.1016/j.jamcollsurg.2019.02.037. Epub 2019 Feb 15.
Venous thromboembolism (VTE) events are tracked in trauma registries and by administrative data sets. Both databases are used to assess outcomes, despite having varying processes for data capture.
This study was performed at an urban, university-based, Level I trauma center from 2004 to 2014. Retrospective review of the trauma registry and the hospital's administrative database was performed querying for all VTEs. Each VTE was then validated through manual chart review. Confirmed events were those with radiographic evidence of VTE by ultrasound, CT, and/or ventilation-perfusion scan. Sensitivity, specificity, and predictive values were calculated and compared between databases.
There were 19,353 trauma patients admitted during the study period; 656 VTEs were identified in the registry and 890 were identified via administrative data; 527 potential events were identified by both databases; 129 events were only in registry; and 363 were only found in the administrative database. We confirmed 636 of 656 events in registry (positive predictive value, 97%; 95% CI, 95.6% to 98.3%) vs 815 of 890 events in administrative data (positive predictive value, 91.6%; 95% CI, 89.75% to 93.4%; p < 0.001). Sensitivity was higher for administrative (87.2% vs 68.0%; p < 0.001), as 299 confirmed VTE events were not in the registry. Differences between the 2 databases were diminished when the analysis excluded untreated events and those present on admission. Twenty-three percent of validated deep vein thrombosis events in the registry were upper extremity events.
The trauma registry showed higher specificity and lower sensitivity compared with administrative data. The low false-positive rate of the trauma registry supports its validity in VTE outcomes research. Additional investigation is needed to evaluate the relevance of the variable sensitivity, likely due to definitional differences. Supplementation of trauma registry data with administrative data can strengthen its completeness.
静脉血栓栓塞症（VTE）事件在创伤登记处和行政数据集进行追踪。尽管数据采集过程不同，但这两个数据库都用于评估结果。
这项研究于 2004 年至 2014 年在一所城市、大学附属的一级创伤中心进行。对创伤登记处和医院行政数据库进行回顾性检索，以查询所有 VTE。然后通过手动图表审查验证每个 VTE。通过超声、CT 和/或通气灌注扫描确认 VTE 的放射学证据的事件被确认为确诊事件。在两个数据库之间计算并比较了敏感性、特异性和预测值。
研究期间共有 19353 名创伤患者入院；登记处发现 656 例 VTE，行政数据中发现 890 例；两个数据库共发现 527 例潜在事件；129 例仅在登记处；363 例仅在行政数据库中发现。我们确认了登记处 656 例中的 636 例（阳性预测值，97%；95%置信区间，95.6%至 98.3%）与行政数据中的 890 例中的 815 例（阳性预测值，91.6%；95%置信区间，89.75%至 93.4%；p<0.001）。行政数据的敏感性更高（87.2%比 68.0%；p<0.001），因为 299 例确认的 VTE 事件未在登记处。当分析排除未治疗的事件和入院时存在的事件时，两个数据库之间的差异减小。登记处 23%的已确认深静脉血栓形成事件为上肢事件。
与行政数据相比，创伤登记处显示出更高的特异性和更低的敏感性。创伤登记处的低假阳性率支持其在 VTE 结果研究中的有效性。需要进一步调查以评估可变敏感性的相关性，这可能归因于定义上的差异。用行政数据补充创伤登记处的数据可以提高其完整性。
