Lack of Association Between Intraoperative Handoff of Care and Postoperative Complications

A Retrospective Observational Study

Vikas N. O'Reilly-Shah; Victoria G. Melanson; Cinnamon L. Sullivan; Craig S. Jabaley; Grant C. Lynde


BMC Anesthesiol. 2019;19(182) 

In This Article


This study was approved by the Emory University Institutional Review Board (IRB00108382) prior to data acquisition and analysis, including a waiver of written informed consent. Our data analysis plan, including variables to include as confounders, was included with our IRB submission. This manuscript was prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement and guidelines for improved reporting of observational studies and propensity score analyses.[11]


Data for the present work were drawn from two academic hospitals (Emory University Hospital and Emory University Hospital Midtown) within Emory Healthcare (EHC), which share a common anesthesia information management system (AIMS). AIMS and other electronic medical record data are consolidated nightly into the EHC Clinical Data Warehouse (CDW). At both sites, anesthetic care is delivered via the care-team model. For each surgical case, a combination of attending physician anesthesiologists, trainee physicians (i.e. residents and fellows), and non-physician anesthetists (i.e. certified registered nurse anesthetists and anesthesiologist assistants) provide care in concert. The attending physician anesthesiologist maintains responsibility for one or more concurrent patients, supervising a trainee physician or anesthetist who maintains continuous presence in the operating room. Attendings prescribe and medically direct the anesthetic plan. Only one attending anesthesiologist is responsible for the care of a patient at a time; handoffs may occur amongst attendings and/or non-attendings indicating a transfer of responsibility.

There is no mandatory handoff communication tool that is required from attending to attending. Anesthetists and residents at both hospitals are requested to complete a paper communications tool that is used for both provider and recovery room transfers of care. However, compliance is not mandatory, and no data exists to evaluate the frequency this communications tool is utilized.

Data Collection

Both studied EHC hospitals are participants in NSQIP, which entails collection and reporting of structured, high quality patient-level demographic and outcomes data by abstractors hired and trained specifically for this purpose. Data from January 2014 through December 2017 were combined with patient-level data from the EHC CDW for analysis. All cases reported to NSQIP utilizing intraoperative EMR were included in this analysis.

NSQIP Reporting

NSQIP reporting and sampling procedures are defined and monitored by the American College of Surgeons.[12] EHC reports both targeted and sampled data to NSQIP. Specific targeted surgeries, with 100% reporting of data from cases performed include: breast, colon, hepatobiliary, thyroid, and vascular. All other NSQIP reported cases are sampled utilizing an 8-day sampling technique where the first 40 cases performed in a rolling 8-day period are reported.


This study is a retrospective analysis of routinely collected clinical data in the anesthesia record combined with the aforementioned NSQIP data, which is also collected routinely (using the aforementioned case sampling strategy) as part of surgical quality improvement monitoring at our institution. Due to the retrospective nature of the planned work, no a priori power analysis or sample size calculation was performed; the research protocol called for inclusion of all NSQIP data available within the study period.

Primary Outcome

The primary outcome (dependent variables) for this study was a composite outcome of NSQIP postoperative occurrences. Specifically, any patient flagged as having any of the following occurrences was considered to have experienced the composite outcome: progressive or acute renal failure, cardiac arrest requiring cardiopulmonary resuscitation, stroke, any type of surgical site infection or sepsis, myocardial infarction, unplanned intubation, mechanical ventilation greater than 48 h, pneumonia, deep vein thrombosis, venous thromboembolism, urinary tract infection, or readmission within 30 days.

Primary and Secondary Exposures

The primary exposure (independent variable) of interest was handoff at the attending level. This was defined as the presence of more than one attending recorded on the anesthetic record via the AIMS.

The secondary exposure of interest was a complete handover of care. This was defined as the presence of more than one attending and more than one non-attending documented on the electronic anesthetic record. At the studied institutions, breaks for non-attending providers are documented through a separate mechanism; as such, routine breaks were not considered when determining complete handovers.

Independent Variables

The independent variables included in the present analysis were patient age, sex, body mass index (BMI), ASA-PS classification, case length, surgical case complexity, and evening/weekend start time. Total case length was recorded in NSQIP; missing values were calculated as the difference between the routinely recorded anesthesia start and stop times. Surgical case complexity was defined as the quantity of American Society of Anesthesiologists Relative Value Guide base units ascribed to the case.[13] Evening/weekend start time was a binary determination based on surgery start time occurring (on Monday through Friday) prior to 7 A.M. or after 7 P.M., or occurring any time on a Saturday or Sunday. Normality was assessed via visual inspection of normalized quantile-quantile plots; case length and surgical complexity were logarithmically transformed to improve normality.

Statistical Analysis

Statistical analysis was performed in R v3.3.2 (R Core Team, Vienna, Austria) using the RStudio platform v1.1.423 (R Studio Team, Boston, MA).[14,15] Baseline characteristics were compared on a single variable basis between those exposed to a handoff and those not exposed. Chi-square test was used for sex, ASA-PS classification, evening/weekend case, surgical specialty, and the presence of a composite NSQIP adverse event. Wilcoxon rank-sum test was used for age, BMI, and case length.

The association of the composite outcome with the primary and the secondary exposures of interest was tested in multiple ways. In addition to chi-square analysis described above, a single variable logistic regression model (binomial logistic regression with one Y/dependent variable with two levels, one X/independent variable) was constructed to test the unadjusted association of the outcome with the exposure. Additionally, two multiple variable logistic regression models were constructed (binomial logistic regression with one Y/dependent variable with two levels, multiple X/independent variables). The simpler model contained the exposure of interest plus age, sex, and ASA-PS classification. The final model contained these independent variables plus case length, surgical case complexity, and evening/weekend start time.