Assessing the impact of a semi-structured intraoperative anaesthesia handoff cognitive aid on surgical patient outcomes: study protocol for a cluster randomised trial

Abstract

Introduction

Intraoperative anaesthesia handoffs represent a risk point in the care of surgical patients. Although often necessary to prevent fatigue, improve vigilance and optimise operational efficiency, critical information can be lost, potentially leading to postoperative complications. Structured handoffs can increase the transfer of knowledge during intraoperative anaesthesia handoffs, improving their quality. We therefore propose to test the primary hypothesis that a semi-structured intraoperative anaesthesia handoff cognitive aid reduces the number of serious 30-day complications in surgical patients.

Methods and analysis

We will enrol adults having non-cardiac surgery who are scheduled to have an intraoperative anaesthesia handoff for operational reasons. We plan a cluster randomised trial (enrolling over 18 months, anticipated sample size approximately 4500 patients) that will compare the Epic Electronic Health Record intraoperative anaesthesia handoff cognitive aid to routine handoffs. Our primary outcome will be the number of serious postoperative complications within 30 days. Our secondary outcomes will be: (1) the number of minor complications; and (2) the duration of postoperative hospitalisation. Bayesian analysis with generalised linear multilevel modelling will be used to estimate the effect of structured handoffs on the primary and secondary outcomes.

Ethics and dissemination

This study has been approved by the local institutional review board with a waiver of informed consent. Results will be disseminated in the medical literature with de-identified data available on request.

Trial registration number

NCT06533111.

Strengths and limitations of this study.

• The large sample size will increase the precision of the estimate of the effect.

• The broad inclusion criteria and limited exclusion criteria will increase the generalisability of the findings.

• The intervention being tested is a widely available electronic handoff tool, increasing the ability for other institutions to replicate our trial or implement the tool based on our results.

• The inability to directly assess how clinicians are using the handoff cognitive aid means that we do not have a direct understanding of the differences in exposure between the intervention and control groups.

• There is a possibility of a learning effect over the trial duration as clinicians change their usual care handoffs due to intermittent exposure to the cognitive aid.

Background

Postoperative complications are surprisingly common, and about 2% of patients who have major inpatient surgery die within a month.¹ Information lost during intraoperative handoffs may worsen patient outcomes, including postoperative complications and mortality.²,⁷ For example, failure to transmit key care elements such as glucose management in diabetics or blood pressure management in patients at cardiovascular risk may contribute to serious complications. Nonetheless, a randomised trial suggests that intraoperative anaesthesia handoffs do not worsen patient outcomes,⁸ a finding supported by another retrospective study.⁹ Divergence in these findings may result from limited generalisability due to the imposed structure of trials and confounding in retrospective analyses, especially by case duration.

The quality and quantity of information conveyed during verbal handoffs may be improved by the use of semi-structured handoff cognitive aids.¹⁰,¹² For example, electronic handoff checklists increase transfer of clinical content, including discussion of antibiotics from 63% to 97%, vasopressors from 44% to 85% and anti-emetics from 15% to 46% of handoffs.¹¹ Structured cognitive aids are routinely used in other safety-critical industries such as the military and aviation.¹³

We therefore propose to test the primary hypothesis that a semi-structured handoff cognitive aid reduces the number of serious postoperative complications within 30 days. Secondarily, we will test the hypotheses that a handoff cognitive aid: (1) reduces the number of minor complications; and (2) shortens the duration of postoperative hospitalisation. Our protocol adheres to the Standard Protocol Items: Recommendations for Interventional Trials standards.¹⁴

Methods

We will conduct a cluster crossover trial at Memorial Hermann Hospital-Texas Medical Center, an academic training centre with approximately 1000 beds that performs more than 2000 adult non-cardiac operations per month. Two groups of operating rooms on separate floors will participate, each with about 15 rooms. One group will be cluster randomised every 4 weeks to use a semi-structured handoff cognitive aid or handoff per clinical routine using a computer-generated random sequence. The alternate group of operating rooms will use the opposite method. Thus, about half the operating rooms will use the semi-structured handoff cognitive aid at any given time. We plan to enrol 18 clusters in each set of operating rooms, for a total of approximately 4500 patients. Operating room assignments will be announced at the start of each 4-week block, with the planned sequence known only to the principal investigator.

To encourage proper use of the handoff cognitive aid, we used implementation mapping¹⁵ to identify and overcome barriers to its use. Based on the results, we developed an implementation plan that includes communication strategies, educational activities, feedback and incentives to optimise use of the cognitive aid and improve trial compliance.¹⁶

Patients assigned to semi-structured handoffs will use the cognitive aid embedded into Epic Electronic Health Records¹⁷ which is structured to approximate the IPASS format: illness severity, patient summary, action list, situation awareness and synthesis by receiver.¹⁸ The Epic cognitive aid facilitates complete information transfer while allowing flexibility for clinicians to emphasise situational details.

Patients assigned to current care will have handoffs per routine without a cognitive aid. There is no standard for intraoperative handoffs at the study site and methodology varies among clinicians.

Handoff cognitive aid use will be documented in the electronic health record. Additionally, we will observe intraoperative handoffs on a subset of cases in both trial arms, targeting observations of 50–100 cases total (~1%–2% of the overall sample size), to better understand the differences between trial arms in terms of the style and content of the intraoperative handoffs. These will be focused ethnographic observations,¹⁹ with prospectively defined clinical content, teamwork behaviours and human factors items to identify, as well as space for open notes and observations. The clinical content will be derived from prior studies on intraoperative handoffs.^{11 20} The human factors and teamwork behaviours, such as asking questions, closed loop communication and interaction with the cognitive aids, will be derived from prior work in describing the context of intraoperative anaesthesia handoffs²⁰ as well as how clinicians interact with cognitive aids.²¹

Our primary outcome will be the unweighted number of serious postoperative complications plausibly related to information transfer during handoffs. We selected complications that could be linked to hand-off adequacy,²² specifically elevated troponin, acute kidney injury, pneumonia, bleeding requiring transfusion, prolonged intubation, unplanned reintubation, rapid-response team activation, hospital readmission and mortality within 30 days. Secondarily, we will evaluate the number of less serious complications, specifically postoperative hypotension (defined as treatment with a vasopressor in the first 24 hours postoperatively), hypoxia (defined as an oxygen saturation to fraction inspired oxygen ratio less than 315 averaged over the first 2 hours postoperatively) and uncontrolled pain (defined as a numeric pain score average of 5 or greater over the first 2 hours postoperatively) and intraoperative hypotension (mean arterial pressure <65 mm Hg for at least 15 min) subsequent to handoffs. Table 1 lists the complications we will consider and their detailed definitions. An additional secondary outcome will be the duration of hospitalisation.

Table 1. Definitions of components of primary and secondary outcomes.

Complication	Operational definition
Primary outcome components
Myocardial injury after non-cardiac surgery/myocardial infarction	High sensitivity troponin I ≥75 ng/L
Acute kidney injury	At least kidney disease: improving Global Outcomes Grade 1 (creatinine ≥1.5× baseline within 30 days)
Post-op blood transfusion	Transfusion of whole blood, packed red blood cells or fresh frozen plasma in first 24 hours postoperatively
Prolonged intubation	Patient does not emerge and extubate at end of case, whether planned or unplanned
Pneumonia	International Classification of Disease, 10th Revision codes within 30 days
Reintubation	Reintubation within 24 hours postoperatively (intubation note, bolus dose propofol, etomidate, rocuronium, cisatracurium, succinylcholine, vecuronium)
Rapid response team activation	Rapid response called within 24 hours postoperatively
Readmission	Readmission within 30 days for any cause
Mortality	All-cause 30-day mortality
Secondary outcome components
Postoperative hypotension	Documentation of pressor administration (phenylephrine, vasopressin, norepinephrine, epinephrine, ephedrine) in first 24 hours postoperatively
Hypoxia	Oxygen saturation to fraction inspired oxygen ratio <315 on average in first 2 hours postoperatively
Uncontrolled pain	Pain score average ≥5 on a 10-point scale over first 2 hours postoperatively
Intraoperative hypotension	Mean arterial pressure <65 mm Hg for at least 15 min intraoperatively following first intraoperative anaesthesia handoff

Complications will be abstracted electronically from medical records. Troponin I will be measured using the Siemens Atellica IM High-Sensitivity Troponin I Assay with myocardial injury after non-cardiac surgery defined as a level greater than or equal to 75 ng/L.²³ Troponin I and creatinine values will be extracted from the laboratory data. Pneumonia diagnoses will be captured from the International Classification of Disease, Revision 10 codes (all codes considered are listed in online supplemental material). Clinical note types will be abstracted for procedure notes for intubation outside of the operating room, rapid response activation notes and readmission history and physical notes. Prolonged intubation will be abstracted from the events documented in the anaesthesia record. Blood transfusion requirements, vital signs, pain scores and oxygen requirements will be retrieved from nursing flowsheets. Medications indicative of intubation or hypotension treatment will be abstracted from the medication administration record. Hospital length of stay will be abstracted from admission and discharge dates in administrative data.

Evaluation of a random selection of charts for abstraction by the principal investigator will ensure data quality. In the event of a discrepancy between the data pulled from the electronic health record and the judgement of the principal investigator, a second member of the research team, with clinical expertise, blinded to the treatment group, will resolve the discrepancy.

Data analysis

Our prespecified goal is to produce a precise, unbiased estimate for the effect of the handoff intervention on postoperative complications. Our grant supports 18 months of enrolment, and we anticipate having about 4500 patients in that time period. We do not plan interim analyses.

Analysis will be restricted to adults ≥18 years who have non-cardiac surgery in a participating operating room starting between 07:00 and 17:00 Monday through Friday. We will consider only the initial operation within any 30-day period. We define handoffs as a permanent intraoperative care transition, excluding lunch and other short breaks. And finally, we will exclude patients designated American Society of Anesthesiologists (ASA) Physical Status 6 (organ donors).

Geographic cluster randomisation of patients will result in additional clustering within surgical teams (ie, initial anaesthesiologist, second anaesthesiologist and surgeon). Given that some clinicians will participate in more than one surgical team, we will use generalised multilevel models with four random intercept terms (ie, geographic floor, initial anaesthesiologist, second anaesthesiologist, surgeon), adding a fifth as needed if the patient has multiple intraoperative handoffs, to account for clustering of patients within floors and surgical teams of shifting composition (R V.4.3.2 and Stan V.2.30; packages rstanarm and brms) for both continuous and discrete outcomes.²⁴,²⁶

We will rely on leave-one-out cross-validation^{27 28} and posterior predictive checking for model-fitting. Primary analyses will use an intention-to-treat approach. Priors for intercepts and regression coefficients will be ~Normal (µ=0, σ=10) and ~Normal (µ=0, σ=0.561) in the log-form respectively, and (for continuous outcomes) ~Normal (µ=0, σ=100) for both intercepts and regression coefficients in the identity link (with some adjustment for scaling) and level one and two error variances will be specified as ~Half T (df=3, mean=0, SD=10). The priors for regression coefficients in the log-form assume that any risk ratio is centred at the null hypothesis (R.R.=1) and that the 95% credible interval falls between 0.33 and 3.0.

Hypothesising that the use of the semi-structured handoff cognitive aid will decrease serious postoperative complications and mortality, less serious postoperative complications and length of stay compared with usual care handoffs, multilevel logistic modelling will evaluate the probability of postoperative complications as a function of treatment, while Poisson/Negative-Binomial multilevel models will evaluate length of stay as a function of treatment.

We hypothesise, in an exploratory manner, that the treatment effect shows heterogeneity, with a larger impact in higher risk groups—(1) patients with an ASA Physical Status 4 and 5, (2) patients cared for by supervised trainees versus solo attending anaesthesiologists, (3) patients undergoing a procedure with an Operative Stress Score²⁹ 4 and 5—representing a more physiologically stressful surgery and (4) patients receiving more than one intraoperative anaesthesia handoff.

Treatment-effect heterogeneity will be examined for prespecified subgroups as interactions using the same multilevel models discussed before. These results should be interpreted as hypothesis-generating. Assessment of learning (eg, clinicians improving usual care handoffs as the trial progresses) will additionally be done by considering time as an interaction.

The primary analysis will be intent-to-treat with a secondary treatment-received analysis, where treated subjects will be defined as subjects for whom at least half of the clinicians giving intraoperative handoff used the handoff cognitive aid. The data analyst will be blinded to the treatment arm.

Patient and public involvement

There was no formal patient or public involvement in the design of this research trial.

Discussion

Much information is lost during conventional care transitions. Semi-structured handoff cognitive aids may diminish information loss during care transitions, which may in turn decrease postoperative complications. Our goal is therefore to determine whether use of semi-structured cognitive handoff aid reduces postoperative complications using a cluster randomised cross-over trial design.

We plan to evaluate the handoff cognitive aid embedded within Epic Electronic Health Records because it is widely available and included in the base version of the Epic Anesthesia Information Management System. The cognitive aid was designed through a consensus process that involved diverse anaesthesia clinicians, which improves its acceptability across a range of practice settings.¹⁷

Our primary outcome is the number of serious postoperative complications that are plausibly linked to inadequate information transfer. Fortunately, serious complications are sparse. However, that makes it challenging to enrol sufficient patients to generate a robust treatment-effect estimate. We therefore plan a cluster randomised trial that will allow us to enrol a large number of patients, probably about 4500 which will give us considerable power to detect even a small benefit. Given that cognitive handoff aids are readily available, cost-free and that all our primary outcomes are serious, we will consider any reduction to be clinically meaningful.

We will use Bayesian analyses with weakly informative neutral priors that will result in posterior efficacy estimates and associated precision. Bayesian analyses permit probabilistic inference even in the context of traditionally underpowered trials³⁰ and further map directly onto clinicians’ typical reasoning where clinicians frequently consider their estimates of pretest priors in interpreting test results, thus using inductive inference around probabilities.³¹

Cluster randomised designs encourage generalisability by including a broad range of clinicians. Specifically, we will include a wide range of anaesthesia clinicians, including anaesthesia faculty at all levels, anaesthesiologist assistants and residents. Planned subgroup analyses will estimate the treatment effect in various provider groups.

A critical feature of any trial is controlling the experimental exposure. That is usually easy if the intervention is under clear control of the investigators, such as giving a drug versus a placebo. Our situation differs. We cannot force clinicians to use the Epic cognitive aid, much less to use it correctly. Conversely, we cannot prevent clinicians from adopting elements of the Epic handoff tool into their routine practice and using those procedures even during periods designed for routine handoffs. Correct use will be encouraged by an existing implementation plan. Adoption of structured handoff elements into routine practice will be limited because our clusters will include only a fraction of our patients. Consequently, average clinicians will participate in only one or two handoffs per week under trial conditions. We nonetheless plan to assess learning by assessing the interaction between time and treatment effect.

Additionally, there are frequently discrepancies between the actual fidelity of the implemented intervention and the documentation of adherence to the intervention.³² While we will be measuring documented adherence as our measure of treatment received, we additionally plan for a subset of handoffs to be observed to better understand the use of the cognitive aid and how similar the usual care and intervention arm handoffs appear in style and content. Observations will further provide some information about handoff performance, but we do not plan to specifically evaluate information transfer which would provide mechanistic data.

An additional limitation will be the availability of 30-day outcomes in our electronic health records which only cover the Memorial Hermann Health System. Outcomes for patients cared for elsewhere will not be captured. But presumably, missing data will be comparably distributed across our treatment groups. Missing data will thus reduce power but not introduce bias.

In summary, this is a trial protocol for a cluster randomised trial of an intraoperative anaesthesia handoff cognitive aid versus usual care handoffs with a planned Bayesian analysis designed to measure the impact of the cognitive aid on postoperative complications in a wide range of adults undergoing non-cardiac surgery.

Trial status

The trial began enrolling on 28 November 2024 with planned enrolment through 16 April 2026 and final data collection 30 days later.

Ethics and dissemination

Our trial was approved by the University of Texas Health Science Center at Houston Committee for the Protection of Human Subjects (HSC-MS-24-0552). Informed consent was waived because the trial does not increase patient risk and would be impractical with individual consent. Our trial was registered on 29 July 2024 at clinicaltrials.gov (NCT06533111).

Dissemination will be planned through publication in the academic medicine literature and at national and international anaesthesiology meetings. Data will be maintained by the research group but will be available on request in de-identified format.