When is ‘Urgent’ Really Urgent and Does it Matter? Misclassification of Procedural Status and Implications for Risk Assessment in Cardiac Surgery

详细信息    查看全文

• 作者：Md N Karim ; MBBS ; M Clin Epi^a ; Christopher M Reid ; PhD^b ; ^c ; Andrew Cochrane ; MD ; FRCS^d ; Lavinia Tran ; PhD^b ; Baki Billah ; PhD^a ; ^{baki.billah@monash.edu" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Clinical status ; Misclassification ; Global model ; Risk prediction ; Cardiac surgery ; 30-day-mortality
• 刊名：Heart, Lung and Circulation
• 出版年：2016
• 出版时间：February 2016
• 年：2016
• 卷：25
• 期：2
• 页码：196-203
• 全文大小：1120 K

文摘

Many patients classified as “urgent” in Australia New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS) registry contradict the prescribed definition (surgery within 72 hours of angiogram or unplanned admission). The aim was to examine the impacts of this misclassification on the prediction of 30-day mortality following cardiac surgery.

Methods

The ‘reported clinical status’ was compared with a ‘corrected clinical status’ following reclassification based on the standard definition calculated from raw data. Observed-to-predicted risk ratios (OPRs) of 30-day mortality were calculated for the model using reported status and corrected status and compared. A Bland-Altman plot was generated to examine the level of agreement between the two OPRs.

Results

Of 18496 cases reported as urgent, 49.9% were operated after 72 hours, leading to misclassification of 14.6% in the registry. Misclassified patients had significantly higher mortality (3.5%) than true urgent patients (2.9%). Underweight (OR:1.6,CI:1.2-2.1), dialysis (OR:1.4,CI:1.1-1.7), endocarditis (OR:2.1,CI:1.7-2.5), shock (OR:1.6,CI:1.3-2.0) and poor ejection fraction (OR:1.2,CI:1.1-1.4) were significant predictors of misclassification. Bland- Altman plot demonstrates significant disagreement between two risk estimates (P<0.001). Misclassification results in overestimation of risk by 9.1%. Observed-to-predicted risk increased with corrected definition (0.8975 vs 0.9875), suggesting poorer calibration with reported status.

Conclusions

In the ANZSCTS database, misclassification prevalence is 14.6%. Misclassification compromises the discrimination capacity and calibration of the model and results in overestimation of mortality risk.