Variability and relative contribution of surgeon and anesthesia specific time components to total procedural time in cardiac surgery

Abstract

Objectives:

Decreasing variability in time intensive tasks during cardiac surgery may reduce total procedural time, lower costs, reduce clinician burnout, and improve patient access. The relative contribution and variability of surgeon and anesthesia control times to total procedural time is unknown.

Methods:

669 patients undergoing coronary artery bypass graft surgery were enrolled. Using linear regression, we estimated adjusted surgeon and anesthesia control times controlling for patient and procedural covariates. The primary end point compared overall surgeon and anesthesia control times. The secondary end point compared the variability in adjusted surgeon and anesthesiologist control times. Sensitivity analyses quantified the relative importance of the specific surgeon and anesthesiologist in the adjusted linear models.

Results:

The median surgeon control time was 4.1 hours (interquartile range: 3.4 to 4.9 hours) compared to a median anesthesia control time of 1.0 hours (interquartile range: 0.8 to 1.2 hours, p < 0.001). Using linear regression, the variability in adjusted surgeon control time amongst surgeons (range: 1.8 hours) was 3.5-fold greater than the variability in adjusted anesthesia control time amongst anesthesiologists (range: 0.5 hours, p < 0.001). The specific surgeon and anesthesiologist accounted for 50% of the explanatory power of the predictive models (p < 0.001).

Conclusions:

Surgeon control time variability is significantly greater than anesthesia control time variability and strongly associated with the surgeon performing the procedure. While these results suggest surgeon control time variability is an attractive operational target, further studies are needed to determine practitioner specific and modifiable attributes to reduce variability and improve efficiency.

Central Message

Surgeon control time has significantly greater variability and contributes more to total procedural time than anesthesia control time in CABG, making it a target for improving operational efficiency.

Introduction

Operating room (OR) time efficiency is a high-priority for optimal patient care, clinician wellness, and fiscal sustainability.^1,2 For example, prolonged OR hours have been associated with increased burnout in cardiac surgical trainees, attendings, anesthesiologists, and operating room nurses.^3–7 Reducing total procedural time (TPT) in cardiac surgery may reduce OR work hours, enable higher OR throughput, improve patient access to care, reduce cost, and decrease burnout.⁸

Total procedural time in surgery is partially controlled by anesthesiologists (anesthesia control time, ACT) and partially controlled by surgeons (surgeon control time, SCT).^9–14 ACT is the period between a patient’s arrival in the operating room and the completion of all anesthesia-related tasks.⁸ SCT is the period between procedure start (often demarcated by “skin incision” time) and procedure end.⁹ The relative contributions of SCT and ACT to TPT in cardiac surgery are unknown. Classification of significant variability in ACT or SCT is critical for identifying modifiable operational tasks that could increase efficiency, particularly if such tasks are time-intensive.¹⁵

Our primary objective was to calculate the relative contributions of ACT and SCT to TPT in a cohort of patients undergoing isolated coronary arterial bypass grafting (CABG). We hypothesized that SCT would have a significantly larger contribution to TPT than ACT. Our secondary objective was to calculate the variability in ACT and SCT amongst anesthesiologists and surgeons, respectively. Our secondary hypothesis was variability amongst surgeons in SCT would be significantly greater than the variability amongst anesthesiologists in ACT.

Methods

The Institutional Review Board (IRB) of Stanford University approved the study protocol and publication of data (Protocol #64908, approval date: 03/14/2022). Patient written consent for the publication of the study data was waived by the IRB per their protocols.

Patients, procedures, surgeons and anesthesiologists

All patients undergoing primary sternotomy for isolated CABG on cardiopulmonary bypass at Stanford University School of Medicine from July 1^st, 2017 to June 30^th, 2022 were included. Patients receiving concomitant cardiac surgical procedures (e.g. valve replacement), robot-assisted surgery, a non-sternotomy surgical approach, a reoperative sternotomy, “off-pump” CABG, or emergency surgery were excluded from analysis. Additionally, CABGs performed by surgeons or anesthesiologists with fewer than five observations in the dataset were excluded.

Patient demographics, comorbidities and surgical procedural details were extracted from the local Society of Thoracic Surgery (STS) Adult Cardiac Surgery Database (ACSD) as important covariates. Because patient characteristics and procedural complexity may not be uniformly distributed across anesthesiologists and surgeons, we extracted patient- and procedure-specific STS variables used in STS risk calculations for isolated CABG, as well as those likely to be associated with procedural length such as patient age, body mass index (BMI), diabetes, left ventricular ejection fraction, prior mediastinal radiation, preoperative creatinine, number of grafts performed, use of arterial grafts, and use of pulmonary artery catheters. A complete list of extracted variables is available in Supplemental Table 1.

The variable “surgeon experience” for each surgery was estimated by calculating the elapsed time (in years) from the date of completion of surgical training and the day of surgery. The variable “anesthesiologist experience” was similarly calculated. Because these variables were not normally distributed, they were log-transformed for use as a covariate in statistical analysis.

Operational metrics, anesthesia and surgical control times

Operating room timestamps were extracted from the institutional electronic medical record. ACT was calculated as the elapsed time from the patient entering the operating room (“in room” time) as recorded by the circulating registered nurse (RN), to the completion of anesthesia tasks (“handoff to surgical team” time), as recorded by the anesthesia team (see Figure 1A). The handoff timestamp is recorded after completion of airway management, establishment of all invasive monitors and catheters, insertion of additional peripheral intravenous catheters, removal of any central line drapes, and the patient deemed ready for surgical preparation. SCT was calculated as the elapsed time from skin incision (“incision” time) to skin closure (“procedure end” time), both recorded by the circulating RN. We did not examine “surgical preparation time” (elapsed time from “handoff to surgical team” to “incision,” during which the patient is positioned and undergoes surgical preparation and drape application) as this is highly standardized work, not attributable to either surgeons or anesthesiologists, and had lower unadjusted variability than either SCT or ACT (data not shown). Similarly, we did not examine the time from “procedure end” to “out of room,” as these activities (e.g. dressing application, drape removal, transfer of patient to ICU bed) were hypothesized to be highly standardized and a mixed care setting. We performed a post-hoc validation of the accuracy and precision of RN-recorded timestamps using objective data from the anesthesia electronic medical record (see Supplemental Methods).

Figure 1:

(A) Schematic of operational variables. Anesthesia Control Time (ACT) is the elapsed time between patient arrival in the operating room (OR) and handoff to the surgical team for patient preparation. Surgeon control time (SCT) is the elapsed time from incision to skin closure. Additional elements of procedural time include preparation and drape time (between ACT and SCT), and dressing and bed transfer time (after SCT); these times were not analyzed due to small overall time and variability. (B) Variability of median surgeon and anesthesia control times amongst surgeons and anesthesiologists in primary, isolated coronary artery bypass graft surgery (CABG), in hours (hrs). The median surgeon control times for each individual surgeon and median anesthesia control times for each individual anesthesiologist were extracted for each physician. Individual physician median control times are pooled and plotted as a box-and-whiskers plot. The variability in median SCT amongst nine individual surgeons was significantly greater than the variability in median ACT among twenty-six individual anesthesiologists (p < 0.001). Bar represents the median of the median control times for that group (surgeons or anesthesiologists), box represents the interquartile range, whiskers represent the range of non-outliers, dots outside the range represent outliers. Additionally, each individually colored dot represents the median control time of an individual surgeon (SCT) or individual anesthesiologist (ACT). (C) Variability of individual anesthesiologist and surgeon control times. Anesthesia and surgeon control times for each CABG performed were calculated for individual anesthesiologists and surgeons, respectively, and plotted as box-and-whiskers plots. Each box-and-whiskers represents an individual anesthesiologist (left) or surgeon (right). Individual anesthesiologists and surgeons are ordered by median anesthesia and surgeon control time, respectively. Red line represents the overall median anesthesia (left) and surgeon (right) control times for all CABGs. Box-and-whiskers bar represents the median control time for that individual, box represents the interquartile range, whiskers represent the range of non-outliers, and dots outside the range represent outliers. Individually colored dots represent control times for individual CABGs, with each dot representing one CABG.

ACT and SCT for each surgery were then matched by the medical record number and the date of surgery to patient-level data extracted from the STS ACSD database. Patients with multiple operations on the same day (e.g. chest re-exploration) were individually reviewed to associate the correct patient-level data with the operational data from the CABG surgery. ACT and SCT for each case were assigned to the cardiac anesthesiologist and surgeon who performed the procedure. For cases with changes in anesthesiologists, the ACT was assigned to the first anesthesiologist of record. There were no missing data in the final cohort for any of these variables.

Statistical analysis

Because patient factors and procedural complexity may influence ACT and SCT, and these factors are unlikely to be uniformly distributed across anesthesiologists and surgeons, we performed an analysis to estimate adjusted surgeon-specific SCT and anesthesiologist-specific ACT, controlling for underlying patient and procedural factors. We constructed linear regression models to estimate adjusted-ACT and adjusted-SCT, controlling for anesthesiologist, surgeon, anesthesiologist and surgeon clinical experience, and the 27 patient and procedural covariates that may be associated with ACT and SCT (covariates listed in Supplemental Table 1). This analysis provided estimates as to how much of observed variance in control times is attributable to surgeons and anesthesiologists, rather than patient or procedural factors beyond their control.

For the primary outcome, we assessed the difference between ACT and SCT using a Wilcoxon rank-sum test. For the secondary outcome, first we calculated the median unadjusted ACT and SCT for individual anesthesiologists and surgeons, respectively. We then compared the variability in unadjusted median ACT amongst individual anesthesiologists with the variability of the median SCT amongst individual surgeons using Levene’s test for Homogeneity of Variance.¹⁶ Second, to compare the variabilities in adjusted ACTs and SCTs amongst anesthesiologists and surgeons, we extracted the surgeon-specific adjusted SCT and anesthesiologist-specific adjusted ACT coefficients for each surgeon and anesthesiologist from the full linear models and compared the variability of the adjusted SCTs and ACTs with Levene’s test for homogeneity. For the primary and secondary outcomes, a p-value of < 0.05 was considered significant.

We performed three sensitivity analyses to assess the relative importance of the anesthesiologist variable, surgeon variable, and each patient and procedural covariate on ACT and SCT. First, to confirm the overall validity of our hypothesized models, we derived new model coefficients using a randomly selected 60% of the original dataset (derivation cohort), and assessed the derived model performance on the remaining 40% of the dataset (validation cohort) using Pearson’s correlation coefficient. Second, we individually removed surgeon, anesthesiologist, and each individual covariate from the model and assessed the change in model performance as measured by the changes in the model Akaike Information Criterion (AIC), comparing the full model to the reduced nested model with an F-test.^17,18 AIC is a well described method for assessing model fit, penalizing models with extraneous variables that do not substantially improve model explanatory power; a lower (more negative) AIC indicates better model fit. Third, we created parsimonious models for ACT and SCT using backwards selection to optimize the parsimonious models’ AIC. We then assessed the relative importance of each variable in the parsimonious model performance using the Lindemann, Merenda, and Gold method.^19,20

All statistical analysis was performed using R version 4.2.2 (The R Foundation, Vienna, Austria), RStudio version 2022.12.0+353 (Posit, Boston, MA) and Python version 3.10.10 (Python Software Foundation, Fredericksburg, VA). For exploratory analysis investigating patient and procedural factors associated with ACT and SCT using reduced nested models, we performed a Bonferroni-Holm correction to maintain a family wise Type I error rate of 5%, thus a p-value of < 0.05 was considered significant after this correction. To assess the relative importance of covariates in the parsimonious linear SCT and ACT models, we used the relaimpo package in R.^19,20

Results

Patients, procedures, surgeons, anesthesiologists, and timestamp validation

We identified 669 primary, isolated CABG procedures performed by 9 cardiac surgeons and 26 cardiac anesthesiologists. Patient and procedural characteristics are summarized in Table 1. The patient median age was 66 years old (interquartile range [IQR] 58 to 73), median BMI was 27.7 (IQR 24.7 to 30.8), median left ventricular ejection fraction of 58% (IQR 48% to 62%), and median number of vessels with coronary artery disease were 3 (IQR 3 to 3). The median number of grafts performed per patient was 3 (IQR 2 to 3), with 99% (662/669) receiving internal mammary artery grafts, and 20% (137/669) receiving radial artery grafts. Pulmonary artery catheter monitoring was used in 67% (447/669) of surgeries. The median surgeon and anesthesiologist experience levels at the time of CABG were 10.6 years (IQR: 6.0 to 12.3 years) and 5.0 years (IQR: 1.7 to 15.6 years), respectively. In a validation of the accuracy and precision of RN-entered “in-room” and “procedure end” timestamps, compared to correlated objective events in the anesthesia EMR, we found the RN-entered timestamps were both highly accurate and precise (see Supplemental Results).

Table 1:

Patient characteristics. The cohort included 669 patients undergoing primary coronary artery bypass graft surgery without additional surgeries.

	Median [Interquartile Range]
Total Procedures	669
Surgeon Control Time (hours)	4.11 [3.41, 4.86]
Anesthesia Control Time (hours)	0.98 [0.80, 1.21]
Continuous Variables
Patient Variables
Age	66.00 [58.00, 73.00]
Height (cm)	172.69 [165.10, 177.80]
Weight (kg)	80.40 [70.40, 94.90]
STSBMI (kg/m^2)	27.68 [24.71, 30.84]
Preoperative Creatinine (mg/dL)	0.95 [0.83, 1.18]
Left Ventricular Ejection Fraction (%)	58.00 [48.00, 62.00]
Procedural Variables
Number of Vein Grafts Performed	2.00 [1.00, 2.00]
Total Number of Grafts Performed	3.00 [2.00, 3.00]
Clinical Variables
Surgeon Experience (years)	10.59 [5.95, 12.96]
Anesthesiologist Experience (years)	5.03 [1.69, 15.55]
Categorical Variables	Number (%)
Patient Variables
Gender (male)	536 (80.1)
Hispanic Ethnicity
No	577 (86.2)
Not Documented	6 (0.9)
Yes	86 (12.9)
Number of vessels with CAD
One	7 (1.0)
Two	98 (14.6)
Three	564 (84.3)
Pre-operative Hemodialysis	36 (5.4)
Diabetes Mellitus	339 (50.7)
Prior Mediastinal Radiation	8 (1.2)
Peripheral Vascular Disease	90 (13.5)
Cerebrovascular Disease	126 (18.8)
Prior Stroke	58 (8.7)
Prior Myocardial Infarction	342 (51.1)
Heart Failure	244 (36.5)
Hypertension	610 (91.2)
Chronic Lung Disease	76 (11.4)
Prior Intravenous Drug Abuse	51 (7.6)
Immunosuppression	25 (3.7)
Cardiac Diagnosis at Admission
Stable Angina	350 (52.3)
Unstable Angina	84 (12.6)
Non-ST Elevation MI (Non-STEMI)	164 (24.5)
ST Elevation MI (STEMI)	13 (1.9)
Other	14 (2.1)
None	44 (6.6)
Procedural Variables
Radial Artery Graft	137 (20.5)
Internal Mammary Artery Graft	662 (99.0)
Pulmonary Artery Catheter	447 (66.8)

CAD—coronary artery disease; cm—centimeters; dL—deciliter; kg—kilograms; mg—milligrams

Variability in unadjusted surgeon and anesthesia control times and contributions to total procedural time

For the primary hypothesis, the overall median SCT was 4.1 hours (IQR 3.4 hours to 4.9 hours), significantly longer than the overall median ACT of 1.0 hours (IQR 0.8 to 1.2 hours, p < 0.001, see Supplemental Figure 1). Comparing the unadjusted median ACTs among the 26 anesthesiologists, the median ACT of the 50^th percentile anesthesiologist was 1.0 hours, while the median ACT for anesthesiologists at the 25^th and 75^th percentile was 0.9 hours and 1.0 hours, respectively (IQR = 0.1 hours). The minimum and maximum median ACT amongst individual anesthesiologists was 0.6 hours and 1.2 hours, respectively (range = 0.6 hours, see Figure 1B).

Among the nine cardiac surgeons, the median SCT of the 50^th percentile surgeon was 4.3 hours, while the median SCT for the surgeons at the 25th and 75^th percentile was 4.1 hours and 4.8 hours respectively (IQR = 0.7 hours). The minimum and maximum median SCT amongst individual surgeons were 3.2 hours and 5.4 hours, respectively (range = 2.2 hours, see Figure 1B). Variability was significantly larger for unadjusted SCT than ACT (Levene’s test for homogeneity p < 0.001, see Figure 1B).

Comparison of variability in adjusted surgeon and anesthesiologist control times

Because patient and procedural factors associated with SCT and ACT may confound unadjusted results, we fit linear regression models to estimate adjusted SCT and ACT, while controlling for surgeon, anesthesiologist, surgeon and anesthesiologist experience, and all patient and procedural covariates listed in Supplemental Table 1. We extracted the regression surgeon-specific SCT and anesthesiologist-specific ACT coefficients from the full SCT and ACT linear models, enabling us to adjust for surgeon and anesthesiologist independent confounders of ACT and SCT. Using the surgeon with the median adjusted SCT as a reference, other surgeons were associated with estimated differences in adjusted SCT ranging from −0.8 hours to +1.0 hours (range of adjusted SCT: 1.8 hours). Using the anesthesiologist with the median ACT as a reference, other anesthesiologists were associated with estimated differences in ACT ranging from −0.3 hours to +0.2 hours (range of adjusted ACT: 0.5 hours). After controlling for possible patient and procedural confounders, we found that the variability in adjusted surgeon SCT remains significantly higher than the variability in adjusted anesthesiologist ACT (Levene’s test for homogeneity p < 0.001).

Association of surgeon with SCT and anesthesiologist with ACT

We performed three sensitivity analyses assessing the relative importance of the surgeon and anesthesiologist variables with SCT and ACT. First, we used a derivation-validation approach to confirm the validity of our hypothesized SCT and ACT models. The Pearson’s correlation coefficient for the derived SCT and ACT models on the validation cohort were 0.56 and 0.34, respectively (p < 0.001 for each), suggesting our hypothesized models were valid for predicting SCT and ACT (see Supplemental Figure 2).

Second, we removed the surgeon and anesthesiologist variables from their respective SCT and ACT models.¹⁸ We compared the reduced nested models removing surgeon and anesthesiologist to the full model using AIC, with lower (more negative) AIC indicating better model fit, and a Bonferroni-Holm corrected F-test. The full SCT model AIC was −102. Removal of the surgeon covariate substantially decreased model explanatory power for SCT with an increased AIC to +49 suggesting a strong association of surgeon with SCT (Bonferroni-Holm corrected F-test p < 0.001). Using this approach, additional variables associated with SCT included radial artery grafting, number of vein anastomosis, and a history of PVD (see Supplemental Table 2). Surgeon experience was not associated with SCT (unadjusted p = 0.28).

For ACT, the AIC for the full model was −1644. Removing anesthesiologist substantially worsened model explanatory power (AIC −1576, Bonferroni-Holm corrected p < 0.001, Supplemental Table 2), suggesting a strong association of anesthesiologist with ACT. In addition to anesthesiologist, the only other factor significantly associated with ACT was use of pulmonary artery catheter (nested model AIC = −1638, Bonferroni-Holm corrected p < 0.001).

Third, we performed a sensitivity analysis creating a parsimonious model of SCT and ACT using AIC-optimized backwards selection and assessing relative variable contribution in model explanatory power. The AIC-optimized backwards selected SCT model identified eight variables, of which six were significantly associated (p < 0.05) with SCT—surgeon, patient age, weight, PVD, radial artery grafting and the number of vein anastomosis (See Supplemental Table 3). The parsimonious model R² was 0.44. We then assessed the relative importance of each covariate in the parsimonious model in explaining SCT. The surgeon accounted for 50% of the model’s explanatory power, radial artery grafting 34%, and number of vein anastomoses 6% (see Figure 2).

Figure 2:

Relative importance of variables in parsimonious linear models of surgeon and anesthesia control times. (A) Among the eight variables in the parsimonious model of surgeon control time (SCT), the specific surgeon was the most important factor, accounting for 50% of the SCT model’s explanatory power. (B) Similarly, among the nine variables in the parsimonious anesthesia control time (ACT) model, the specific anesthesiologist was the most important variable, accounting for 51% of the ACT model’s explanatory power. Relative variable importance was calculated using the Lindemann, Merenda, and Gold method.

For ACT, the AIC-optimized backwards selection model identified nine variables, of which seven were significantly associated (p < 0.05) with ACT—anesthesiologist, PA catheter insertion, heart failure, immune suppression, patient gender, preoperative creatinine, and surgeon experience (See Supplemental Table 3). The parsimonious model R² was 0.29. Anesthesiologist accounted for 51% of the model’s explanatory power and PA catheter insertion 18% (see Figure 2).

Discussion

In order to identify systems-level variation and targets for interventions to lower total procedural time, we conducted a study of the variability and relative contributions of ACT and SCT to total procedural time in patients undergoing CABG. We found that SCT contributes significantly more to TPT than ACT (4.1 hours vs 1.0 hours, p < 0.001). After controlling for underlying patient and procedural covariates, we found the variability in adjusted SCT amongst nine cardiac surgeons (adjusted range: 1.8 hours) to be significantly greater than the variability in adjusted ACT amongst 26 cardiac anesthesiologists (adjusted range: 0.5 hours, p < 0.001). We identified the variable with the greatest association with SCT was the surgeon performing the CABG, and the greatest association with ACT was the anesthesiologist providing anesthesia. Other factors most associated with SCT and ACT were additional time-consuming procedures, such as radial artery or additional vein grafts (for surgeons) or pulmonary artery catheter insertion (for anesthesiologists). Most other patient and procedure-specific factors were not significantly associated with SCT or ACT. Our findings suggest that both absolute and relative variability in SCT may be a particularly impactful operational variable for effectively lowering TPT in cardiac surgery.

Reducing TPT may improve operational efficiency, patient access to care, resource utilization, and clinician burnout. Identifying systems-level variability enables targeted interventions to lower TPT, particularly for high-variability, time-intensive tasks.^1,15 Our results extend prior operational studies using a variety of factors to estimate TPT in non-cardiac surgery. In one study using machine learning to estimate the contribution of surgeon, procedural, and patient factors in TPT, surgeon features accounted for 43% of model gain in predictive power, while patient factors only contributed 15% of model gain.²¹ In a second study examining a broad variety of surgeries, TPT was significantly associated with the surgeon performing the procedure for 75% of procedures studied, while patient specific variables were only associated with TPT in 19% of studied procedures.¹⁵ To the best of our knowledge, our study is the first to assess the relative contribution of surgeons, anesthesiologists, and patient and procedural factors to TPT specifically in cardiac surgery. Our results extend these prior studies, indicating that the largest contributor to TPT in cardiac surgery is SCT, and the largest factor in variability in SCT is the surgeon performing the operation.

Over our cohort of 26 anesthesiologists, variability in ACT was small. Given an IQR of 0.1 hours between median ACTs from the 25%ile and 75%ile anesthesiologists, reductions in ACT related to anesthesiologist practice variability as reported here would be modestly impactful from an operational efficiency perspective. The most important variable associated with ACT was anesthesiologist, followed by the placement of a pulmonary artery catheter. The latter is plausible, as pulmonary artery catheter placement requires extra time. Large scale changes to perioperative workflows such as dedicated induction rooms and early recovery areas have been shown to reduce ACT by 20–30 minutes, however the magnitude of this reduction is of modest impact in the studied setting given the three-fold larger contribution of SCT and its variation to TPT.^9,22 Additionally, these logistical and infrastructure changes are likely associated with increased financial costs and personnel requirements, further taxing healthcare systems and clinicians.⁹

Specific interventions reducing SCT variability in cardiac surgery are currently not well described and require future study. In one study, a bundled approach standardizing OR instruments, cardiopulmonary bypass line preparation workflows, and initiating cardiopulmonary bypass while performing vein harvesting were associated with reduced SCT in CABG.²³ A preliminary study in cardiac surgery identified specific surgeon-assistant dyads to be associated with reduced SCT, when compared to the same surgeon working with other assistants.²⁴ Identifying characteristics of high performing surgical teams and standardizing team members may decrease SCT. Three recent studies suggest increased intraprocedural turnover in circulating RN and scrub technologist personnel are associated increased operative times.^25–27 In cardiac surgery, reducing intraprocedural staff turnover may reduce the risk of sharps errors, decreasing SCT.²⁷ In orthopedic surgeries, each circulating RN turnover was associated with a 15 to 20 minute increase in procedure time.^25,26 Additionally, working with a surgeon-preferred anesthesiologist also was associated with reduced SCT in orthopedics.²⁶ Finally, in patients undergoing radial artery grafting, parallel processing by performing radial artery harvest during central line placement has also been associated with reduced SCT.²⁸ Given the increasing indications for radial artery use in CABG and strong association of radial artery grafting with SCT in our study, this particular intervention may be highly impactful.

Our study has several strengths. First, the novel operational characterization of ACT and SCT and their variability as components of TPT in cardiac surgery provide clear metrics for future studies targeting systems-level interventions to improve operational efficiency. Second, we have reduced the complexity of accounting for confounders by focusing on a single and common procedure, CABG. To further reduce the risk of reporting confounded results, we have fit linear models to assess the relative contributions of surgeons, anesthesiologists, patient and procedural factors to SCT and ACT. Our sensitivity analyses estimate that the surgeon and anesthesiologist variables comprise 50% of the explanatory power of the parsimonious SCT and ACT models, indicating these findings are unlikely to be related to uncontrolled confounding.²⁹ Third, as a high-volume cardiac surgery center, existing standard intraprocedural workflows should reduce operative variability compared to lower-volume centers.

Study limitations

Our study has limitations. First, our study is a single-center study assessing variability in a single procedure, CABG. Our data may not be generalizable to other centers or other cardiac surgical procedures. While limited, this report may still spur future efforts given the absence of current literature characterizing components of procedural time in cardiac surgery. Second, we cannot exclude the possibility that unmeasured confounding may affect our results. To mitigate this, we extracted standard STS ACSD metrics collected and used them to estimate underlying surgical risk; integrating these standardized metrics should reduce bias.

Third, even after controlling for patient- and procedural-level confounding, we have not directly observed intraoperative workflows, processes, and technical factors that may be associated with SCT. While surgical techniques and team dynamics may vary across surgeons and teams, direct observations will likely be required to identify modifiable factors that can reduce SCT variability. Fourth, our model parameters only explain 44% and 31% of observed SCT and ACT variability, respectively. Accordingly, additional unmeasured factors contribute to ACT and SCT variability. If identified, these covariates may also be important points for operational interventions. Fifth, there are no unified definitions of OR “efficiency,” and the association of SCT and ACT variability with patient outcomes is uncertain. Overly short control times may also be suboptimal if critical procedural components are inadequately addressed. “Efficiency” goals therefore likely include a combination of clinical and procedural time-related endpoints and may vary across institutions; integrating our results with metrics such as the Efficiency-Quality index may serve as a starting point to create more standardized measures.^30,31 Variability in SCT may indicate both operational and quality improvement opportunities; larger datasets will assess the association of SCT with post-operative outcomes to construct and improve these combined metrics.³¹

Conclusions

In a single-center retrospective cohort study analyzing ACT and SCT in patients undergoing isolated CABG via primary sternotomy, we found that SCT contributed approximately four-fold more time to TPT than ACT (See Figure 3). We further found approximately 3.5-fold more variability in SCT compared to ACT when controlling for a comprehensive set of surgeon- and anesthesiologist-independent covariates. Our findings suggest that SCT may be a particularly high yield target to improve operational efficiency. Additionally, prolonged surgical duration has been associated with post-operative neurologic, pulmonary, and infectious complications.^32–37 Further research assessing the association of specific procedural time components and post-operative outcomes may enable focused improvement efforts targeted at specific procedural steps.

Figure 3:

Variability and relative contributions of surgeon and anesthesiologist specific time components to total procedural time. After adjustment for patient and procedural confounders, adjusted surgeon control time contributed significantly more time and variability to total procedure time than anesthesia control time.

Central Picture: Median surgeon control times have higher variability than median anesthesia control times

Perspective Statement.

After accounting for patient and procedural confounding, surgical control time variability in coronary artery bypass grafting remained the largest source of variation in total procedural time. Reducing surgeon control time variability by identifying modifiable surgeon-related factors associated with these fluctuations may reduce overall procedural time and enhance resource utilization.

Glossary of Abbreviations

ACSD

Adult Cardiac Surgery Database

ACT

anesthesia control time

AIC

Akaike Information Criterion

BMI

body mass index

CABG

coronary artery bypass grafting

IQR

interquartile range

IRB

Institutional Review Board

OR

Operating Room

RN

registered nurse

SCT

surgeon control time

STS

Society of Thoracic Surgeons

TPT

total procedural time