Transesophageal Echocardiography, Acute Kidney Injury and Length of Hospitalization Among Adults Undergoing Coronary Artery Bypass Graft Surgery

Abstract

Objective:

To test the association between transesophageal echocardiography (TEE) and incidence of acute kidney injury and length of hospitalization among US adults undergoing isolated coronary artery bypass graft (CABG) surgery.

Design:

This was an observational, retrospective cohort analysis.

Setting:

This study used a multicenter claims dataset from a commercially-insured population undergoing CABG surgery in the US between 2004–2016.

Participants:

Adults aged 18 years or older with continuous insurance enrollment and an absence of renal-related diagnoses prior to the index CABG surgery.

Interventions:

Receipt of TEE within one calendar day of the index CABG surgery date.

Measurements and Main Results:

Of 51,487 CABG surgeries, 5,361 (10.4%; [95% CI: 10.1–10.7%]) developed acute kidney injury and the mean length of hospitalization was 8.8 days (95% CI: 8.7–8.8). The TEE group demonstrated a greater absolute risk difference (RD) for acute kidney injury by multiple linear regression, overall, (RD=+1.0; [95% CI: 0.4–1.5%]; p<0.001) and among a low risk subgroup (RD=+1.0; [95% CI: 0.4–1.6; p=0.002), but not by instrumental variable analysis (RD=+0.9 [95% CI: −1.1–2.9%]; p=0.362). The TEE group demonstrated a longer length of hospitalization by multiple linear regression, overall (+2.0%; [95% CI: 1.1–2.9%]; p<0.001), among a low risk subgroup (+2.2%; [95% CI: 1.2–3.2%]; p<0.001) and by instrumental variable analysis (+10.3%; [95% CI: 7.0–13.7%]; p<0.001).

Conclusions:

TEE monitoring in CABG surgery was not associated with a lower incidence of acute kidney injury or decreased length of hospitalization. These findings highlight the importance of additional work to study the clinical effectiveness of TEE in CABG surgery.

Introduction

Coronary artery bypass graft (CABG) surgery is the most widely performed adult cardiac surgery in the US, with over 160,000 cases performed annually.¹ Renal injury is one of the most common postoperative complications following CABG surgery and is independently associated with both in-hospital and long-term mortality.^2–4 Factors associated with kidney injury during or after CABG surgery include hypotension, low cardiac output, intravascular volume depletion and embolization.^2–6

Transesophageal echocardiography (TEE) is frequently used in CABG surgery for hemodynamic monitoring and management of complications related to cardiopulmonary bypass.^7–15 While complications directly related to TEE are rare,¹⁶ unnecessary procedures performed for incidental findings diagnosed by TEE during CABG surgery has been associated with worse clinical outcomes¹⁷ and guidelines for TEE monitoring in CABG surgery remain indeterminate.^{10, 18} Nevertheless, because of its ability to accurately differentiate between cardiac failure^9–13 and hypovolemia,^{14, 15} routine perioperative monitoring with TEE could potentially improve postoperative renal outcomes if it led to better hemodynamic management during CABG surgery.

There are no randomized or non-randomized comparative effectiveness studies looking at the relationship between TEE monitoring and clinical outcomes in CABG surgery. For that reason, we chose to undertake a retrospective cohort study using US private insurance claims data to determine if TEE monitoring was associated with a lower incidence of acute kidney injury or decreased length of hospitalization following CABG surgery. We anticipated that sicker patients would be more likely to get a TEE around CABG surgery. To reduce the influence of this potential selection bias, we incorporated the statistical technique of instrumental variable analysis.^{19, 20} We hypothesized that monitoring with TEE would be associated with a lower incidence of new-onset acute kidney injury and shorter hospital length of stay among patients undergoing CABG surgery.

Materials and Methods

Data Source

This retrospective, cohort analysis used Optum’s de-identified Clinformatics® Data Mart Database (CDM); a multi-institutional, administrative claims dataset of a commercially-insured population undergoing isolated CABG surgery in the US between 2004 and 2016. Optum’s CDM is a proprietary research database containing de-identified claims from over 60 million individuals. The available data consists of patient sociodemographic characteristics, inpatient medical diagnoses and surgical procedure hospital claims, individual physician billing claims, pharmacy prescriptions and laboratory test results.²¹ The database is geographically diverse, representing all 50 US states. Because these data are de-identified, this study was exempted from review by the Institutional Review Board of the University of Pennsylvania, Philadelphia.

Population

We included all hospital discharges among adults aged 18 or older with a claim for a CABG surgery between January 1, 2004 and December 31, 2016. Surgical discharge claims were identified using International Classification of Disease-9-Clinical Modification (ICD-9-cm)²² procedure codes (36.10–7; 36.19) or International Classification of Disease-10 Procedure Coding System (ICD-10-pcs)²³ procedure codes (Appendix Table 1). Due to limitations of the available data, age was approximated using year of birth. To derive age in years, we supposed each patient was born on January 1^st of the documented year of birth and subtracted the date of admission from January 1^st of that patient’s year of birth.

Patients were excluded if they met any of the following criteria: (1) an ICD-9-cm or ICD-10-pcs code for any cardiac valve repair or replacement (Appendix Table 2); (2) age <18 years of age; (3) missing data on age, sex or state of residence; (4) the territory of Puerto Rico; (5) an ICD-9-cm or ICD-10-cm diagnosis code for any renal or kidney-related condition within the six months prior to the date of index admission for CABG surgery (Appendix Table 3); (6) less than six months of continuous enrollment and (7) states performing fewer than ten CABG surgeries annually (Figure1).

Figure 1: Overview of Study Design Showing Three Analyses.

Our study included three analyses to compare outcomes among patients undergoing isolated CABG surgery with vs without TEE: (1) multiple linear regression analysis among the overall cohort; (2) an instrumental variable regression model and (3) multiple linear regression analysis among a cohort at low risk for development of acute kidney injury and low risk for prolonged length of hospitalization.

Exposure Variable

The primary exposure for this analysis was receipt of a TEE within one calendar day of the index surgery date. We assessed receipt of an intraoperative TEE using Current Procedural Terminology (CPT) codes for TEE (93312–5; 93317–8; 93320–1; 93325) in provider claims associated with that hospitalization matched date of TEE to the date of the cardiothoracic surgery CPT codes (Appendix Table 4). CPT codes are designed to reimburse specialty-specific procedures and have been validated^{24, 25} to identify echocardiography in both surgical²⁶ and non-surgical populations.²⁷

Outcome Variables

The primary outcome of interest, acute kidney injury, was defined using previously validated ICD-9-cm^{28, 29} and ICD-10-cm³⁰ codes for either renal injury (ICD-9: 584; 584.5–9; ICD-10: N17.0-.2; N17.8–9; N99.0) or renal failure requiring dialysis (ICD-9: V45.1; V56.0–1; 3995; ICD-10: T82.4; Y60.2; Y61.2; Y84.1; Z49.0–2; Z99.2; NB50.W0–1ZZ; B50.WYZZ B51.W0ZA B51.W0ZZ B51.W1ZA B51.W1ZZ B51.WYZA B51.WYZZ B51.WZZA B51.WZZZ) within the index hospitalization for isolated CABG surgery.^28–30 Patients were excluded from analysis if an ICD-9-cm or ICD-10-cm diagnosis code for any renal or kidney-related condition was present within the six months prior to the date of index admission for CABG surgery (Appendix Table 3 and Figure 1).

ICD-9/10 codes for acute renal failure are highly specific,^28–30 and ICD-9/10 codes for hemodialysis are both highly sensitive and highly specific,^28–30 Which is why we chose to combine both sets of ICD-9/10 codes to create a composite variable for renal injury. The secondary outcome of interest was hospital length of stay which was defined as the difference between the date of discharge and date of admission.

Covariates

We collected data on admission year and patient age, sex, race, and ethnicity from the hospital discharge claim. Race was categorized as black, white, other, or unknown and ethnicity was classified as either Hispanic or non-Hispanic. Additional data collected from the hospital discharge claim included 14 comorbidities (arrhythmia, congestive heart failure, anemia, coagulopathy, chronic pulmonary disease, diabetes mellitus, electrolyte abnormalities, hypertension, liver disease, neurologic disease, obesity, pulmonary hypertension, peripheral vascular disease, valvular disease), which were determined by secondary ICD-9/10-cm diagnosis codes based on established algorithms.^31–34 Data on the surgical procedure was obtained via ICD-9-p/10-pcs procedure codes in the discharge claim and included the number of bypass grafts performed and whether an internal mammary artery bypass was performed. Data on the US state in which the procedure was performed was obtained from physician provider claims associated with each hospital discharge.

Instrumental Variable

In a non-randomized comparative effectiveness study, conventional regression methods may not account for all potential sources of confounding if patients in different treatment groups differ in ways that may not be completely captured by the study database. In the present study, we anticipated that patients who received TEE during CABG surgery could potentially be sicker than those who did not in ways that could only partially be accounted for in the above regression model. As such, we carried out an instrumental variable analysis.

Use of TEE for CABG varies across US states (Figure 2). Because practice patterns related to TEE use within a given state are largely stable over time, (Appendix Figure 1) a patient’s likelihood of receiving a TEE for CABG surgery can be predicted based on historical TEE use in the state where they are seeking care. After controlling for observed patient factors (e.g., via regression) undergoing CABG in a state where TEE use is common is an instrumental variable if it affects outcomes only by promoting receipt of TEE (the so-called exclusion restriction) and if the state in which treatment is received is otherwise unrelated to unmeasured risk factors conditionally given measured risk factors.¹⁹ Our instrumental variable was the fraction of all CABG patients who received TEE in a given state in the year prior to each index procedure. Based on published rules of thumb, we considered our instrument to be predictive of the study exposure if the F-statistic in this regression was greater than 10.³⁵ To assess whether the distribution of measured potential confounders was similar across values of the instrument, we compared all variables between patients above and below the median value of the instrument, using standardized differences.³⁶ We considered difference of 0.20 (20%) or lower to indicate a small difference between groups in a given covariate.

Figure 2: Geographic Distribution of the Percentage Transesophageal Echocardiography Use in Isolated CABG Surgery Across the United States Between 2004 and 2016.

Brightest orange corresponds to US states with the lowest percent transesophageal echocardiography use and darkest grey corresponds to the highest use. Here, North Dakota had the lowest, (11%; [95% CI: 8–14%]) and Hawaii had the highest, (91%; [95% CI: 85–95%]) of transesophageal echocardiography use in isolated coronary artery bypass graft surgeries performed in the US between 2004 and 2016.

In contrast to standard regression, in which the coefficient for TEE represents the adjusted exposure for the average patient, the coefficient in the instrumental variable analyisis represents the marginal patient.³⁷ In this case, a patient whose indication for TEE is borderline or indeterminate. Therefore, in the instrumental variable analysis, receipt of TEE in the marginal patient is a function of the practice pattern of TEE use within that state from the previous year rather than intrinsic patient characteristics.

Statistics

Chi² and t tests were used to evaluate associations between patient characteristics between the two groups (TEE vs no TEE). Linear regression was used to estimate the association between receipt of TEE and the outcome measures of: (1) new-onset, in-hospital acute kidney injury and (2) hospital length of stay. In accordance with previous work,³⁸ linear probability models (ordinary least squares regression with a binary independent variable) represent an acceptable alternative to logistic regression when the incidence of the independent variable is at least 10%³⁹ and the sample size is sufficiently large that (by the central limit theorem)⁴⁰ the error term is normally distributed.³⁹ Here, we chose to use a linear probability model because: (a) we observed the incidence of acute kidney injury to be 10.4%; (b) our sample size of over 50,000 adequately allowed for normal distribution of the error term and (c) the model allows for the directly-interpretable parameter of absolute risk differences associated with TEE use which is relevant to clinical decision making.

The instrumental variable analysis used two-stage least squares regression.³⁵ All regressions adjusted for the covariates defined above. Because length of stay data was right-skewed (mean greater than the median) these data were log-transformed (using the natural logarithm) prior to the linear regression analysis.⁴¹ Post-regression, because only the dependent variable was log-transformed, the resultant coefficient was exponentiated (antilog of the natural log) and subtracted from one to provide a percentage difference in length of hospitalization for each covariate in the model.⁴¹

Subgroup Analysis

Finally, we carried out a multiple linear regression analysis on a sub-group with less comorbid disease. It is widely reported in the literature that those with congestive heart failure demonstrate worse clinical outcomes and have longer length of hospitalizations compared to those without heart failure.^{42, 43} Moreover, patients with congestive heart failure, chronic pulmonary disease or diabetes mellitus are at increased risk for development of acute kidney injury following CABG surgery.^44–46 We projected that TEE would more likely be performed in patients with these comorbid conditions leading to confounding by indication among patients with congestive heart failure, chronic pulmonary disease or diabetes mellitus. To limit this selection bias we undertook a subgroup analysis, eliminating patients with an ICD-9/10 claim for congestive heart failure, chronic pulmonary disease and diabetes mellitus (Appendix Table 5 and Figure 1).

All statistical tests were two-sided and a p-value of <0.05 was set for statistical significance. Analyses used SAS 9.4 (SAS Institute, Cary, NC), and STATA 15.0 (StataCorp, College Station, TX). Figures were created using Tableau Desktop 9.2 (Tableau, Seattle, WA), Google Draw Version 1.1 (Google, Mountain View, CA) and Sketch Version 47.1 (Sketch, Hague, the Netherlands).

Results

Unadjusted Analysis

Following exclusions (Figure 1), our study cohort included 51,487 patients undergoing isolated CABG surgery. 47.2% (95% confidence interval (CI): 46.8–47.6%) had a TEE claim 47.5% (95% confidence interval (CI): 47.1–48.0%) had a TEE and 52.8% (95% CI: 52.4–53.2%) did not have a TEE claim. TEE in CABG surgery varied across states ranging from 11–92% (Figure 2).

Compared to CABG surgeries performed without a TEE, CABG surgeries performed with a TEE were older patients who more often had congestive heart failure, previous coagulopathic disorders, pulmonary hypertension and valvular heart disease prior to admission (Table 1). Overall, among the entire cohort, 5,361 (10.4%; [95% CI: 10.1–10.7%]) patients developed new-onset, in-hospital, acute kidney injury; and had a mean, overall, length of hospitalization of 8.8 days (95% CI: 8.7–8.8). In-hospital acute kidney injury among patients undergoing CABG with TEE was 11.6% (95% CI: 11.2–12.0) vs 9.3% (95% CI: 9.0–9.7% p<0.001) without a TEE. CABG with TEE was associated with a longer, unadjusted, length of stay (9.0-days; 95% CI 8.9–9.1) compared to no TEE (8.6-days; 95% CI 8.5–8.7); p<0.001).

Table 1:

Baseline Characteristics Among Patients Undergoing Coronary Artery Bypass Graft Surgery With Transesophageal Echocardiography

Covariate		Total N = 51,487	TEE (N = 24,296)	No TEE (N = 27,191)	p Value
Age (mean; 95% CI)		64.99 (64.90–65.07)	65.22 (65.09–65.35)	64.78 (64.66–64.90)	<0.001*
Sex (N; %)	Female	11,663	5,409 (22.3)	6,254 (23.0)	0.046†
	Male	39,824	18,887 (77.7)	20,937 (77.0)
Race (N; %)	White	38,270	17,889 (73.6)	20,381 (75.0)	<0.001†
	Black	4,733	2,131 (8.8)	2,602 (9.6)
	Other	3,948	2,047 (8.4)	1,901 (7.0)
	Unknown	4,536	2,229 (9.2)	2,307 (8.5)
Ethnicity (N; %)	Hispanic	2,984	1,534 (6.3)	1,450 (5.3)	<0.001†
	Non-Hispanic	48,503	22,762 (93.7)	25,741 (94.7)
Medical Covariates
Arrhythmia (N; %)		5,888	2,823 (11.6)	3,065 (11.3)	0.217†
Congestive heart failure (N; %)		3,224	1,709 (7.0)	1,515 (5.6)	<0.001†
Preexisting Coagulopathy (N; %)		2,112	1,134(4.7)	978 (3.6)	<0.001†
Chronic pulmonary disease (N; %)		2,560	1,149 (4.7)	1,411 (5.2)	0.017†
Anemia (N; %)		40	17 (<1)	23 (<1)	0.552†
Diabetes mellitus (N; %)		5,239	2,387 (9.8)	2,852 (10.5)	0.013†
Electrolyte disturbance (N; %)		2,815	1,348 (5.5)	1,467 (5.4)	0.446†
Hypertension (N; %)		12,001	5,319 (21.9)	6,682 (24.6)	<0.001†
Liver disease (N; %)		254	121 (<1)	133 (<1)	0.886†
Neurologic disease (N; %)		400	188 (<1)	212 (<1)	0.940†
Obesity (N; %)		1,493	701 (2.9)	792 (2.9)	0.852†
Pulmonary hypertension (N; %)		250	139 (<1)	111 (<1)	0.008†
Peripheral vascular disease (N; %)		1,048	499 (2.1)	549 (2.0)	0.780†
Cardiac valve disease (N; %)		50,527	555 (2.3)	405 (1.5)	<0.001†
Surgical Covariates
One bypass graft (N; %)		8,896	4,293 (17.7)	4,603 (16.9)	0.026†
Two bypass grafts (N; %)		2,808	1,593 (6.6)	1,215 (4.5)	<0.001†
Three bypass grafts (N; %)		1,810	1,054(4.3)	756 (2.8)	<0.001†
Four or more bypass grafts (N; %)		6,690	2,984 (12.3)	3,706 (13.6)	<0.001†
CABG involving internal mammary artery (either right or left) (N; %)		46,864	22,178 (91.3)	24,686 (90.8)	0.050†

Abbreviations: CABG, coronary artery bypass graft; CI, confidence interval; TEE, transesophageal echocardiography.

^*Indicates analyzed using student’s t test.

^†Indicates analyzed using Chi-square.

Adjusted Regression Analysis: Overall

Following adjustment using multiple linear regression, among the overall cohort of 51,487 patients undergoing CABG surgery, the TEE group demonstrated a greater absolute risk difference (RD) in the incidence of acute kidney injury (RD=+1.0 percentage point [95% CI: 0.4–1.5]; p<0.001). The TEE group also demonstrated a longer length of stay (+2.0%; [95% CI: 1.1–2.9%]; p<0.001) (Table 2; detailed results from this regression appear in Appendix Tables 6 and 7).

Table 2:

Study Outcomes for the Overall Regression, Instrumental Variable Regression and Subgroup Regression Analyses

New-onset, Acute Kidney Injury
	Absolute Risk Difference (percentage point difference)	95% CI	P-value
Multiple linear regression: entire cohort (N=51,487)
TEE	+1.0	0.4–1.5	<0.001
Instrumental variable regression (N=48,652)
TEE	+0.9	−1.1–2.9	0.362
Multiple linear regression: low-risk subgroup (N=41,793)
TEE	+1.0	0.4–1.6	0.002
Length of Hospitalization
	Absolute percent change	95% CI	P-value
Multiple linear regression: entire cohort (N=51,487)
TEE	+2.0%	1.1–2.9	<0.001
Instrumental variable regression (N=48,652)
TEE	+10.3%	7.0–13.7	<0.001
Multiple linear regression: low-risk subgroup (N=41,793)
TEE	+2.2%	1.2–3.2	<0.001

Multiple linear regression: low-risk subgroup (N=41,793)

TEE: Transesophageal echocardiography

Instrumental Variable Analysis

The instrumental variable was highly correlated with the exposure of TEE (Appendix Table 8 and Appendix Figure 1), had an F-statistic of 3,897; p<0.001 and demonstrated good covariate balance among those above and below the median value of the instrument (Appendix Table 9).

In the instrumental variable analysis, CABG surgeries performed with a TEE did not demonstrate a greater RD in the incidence of acute kidney injury; (RD=+0.9 percentage point [95% CI: −1.2–2.9%]; p=0.362. The instrumental variable analysis did reveal a greater absolute percent increase in the length of hospitalization among those undergoing CABG with TEE (+10.3%; [95% CI: 7.0–13.7%]; p<0.001). (Table 2; detailed results from this regression appear in Appendix Tables 10 and 11).

Adjusted Regression Analysis: Low Risk Cohort

The results among the 41,793 patients undergoing CABG surgery at low risk for both prolonged length of hospitalization and low risk for development of acute kidney injury were consistent with the overall cohort. After adjustment, the TEE group demonstrated a greater absolute RD in the incidence of acute kidney injury (RD=+1.0 percentage point; [95% CI: 0.4–1.6%]; p=0.002) and longer hospital length of stay (+2.2%; [95% CI: 1.2–3.2%]; p<0.001) (Table 2 detailed results from this regression appear in Appendix Table 12 and 13).

Discussion

Among 51,487 patients undergoing isolated CABG surgery in the US between 2004 and 2016, we did not observe improved outcomes among CABG surgeries performed with vs without TEE. CABG surgeries performed with TEE demonstrated a modestly greater adjusted risk difference for development of new-onset, acute kidney injury among both the overall cohort and the low-risk subgroup analysis. These results were attenuated in the instrumental variable analysis which accounted for both observed and unobserved differences between patients undergoing isolated CABG with vs without TEE; suggesting that the instrument did eliminate residual, unobserved confounding. The TEE group also demonstrated statistically significantly greater hospital length of stay among both the overall cohort and the low-risk subgroup analysis.

At present, it is controversial whether intraoperative TEE improves outcomes in isolated CABG surgery.^{10, 12, 17, 18, 47} Previous studies of TEE use in CABG surgery consist in the form of descriptive observational^{11, 12, 48, 49} and case series^50–52 investigations However, these studies lacked a comparator group and did not study the impact of TEE use to clinical outcomes in CABG surgery.^{4,11,12,48,50–52} Moreover, while several of these prior observational studies have shown that intraoperative TEE changed the surgical plan in isolated CABG surgery due to diagnosis of previously unrecognized valve disease or identification of a patent foramen ovale,^{11, 12} alteration of the surgical plan does not necessarily translate into clinical outcomes benefit. A 2009 study by Krasuski and colleagues used propensity score matching to compare patients who did vs did not undergo a surgical closure of an incidental, patent foramen ovale (identified by intraoperative TEE) while undergoing CABG or valve surgery. These investigators observed a greater adjusted odds of postoperative stroke¹⁷ among the group that underwent closure of the incidental patent foramen ovale with no long-term survival benefit.¹⁷ Similar to this study, following adjustment for both observed and unobserved covariates, our findings do not suggest a clinical outcomes benefit to the use of intraoperative TEE in the setting of isolated CABG surgery.

Because TEE in isolated CABG surgery is used primarily as a hemodynamic monitor, our work parallels closely to the studies related to use of the pulmonary arterial catheter. Once ubiquitous, the use of the pulmonary arterial catheter decreased following several randomized controlled trials that all demonstrated a lack of clinical benefit;^53–57 which resulted in a substantial decrease in the use of pulmonary arterial catheters in cardiac surgery⁵¹ and near abandonment in non-cardiac, low-risk patients.^{58, 59} Like pulmonary arterial catheter studies, our findings suggest a lack of clinical benefit to the use of invasive hemodynamic monitoring with intraoperative TEE. A plausible explanation for the lack of clinical benefit could be that TEE may be utilized more frequently in patients demonstrating higher burden of comorbidities, inherently at higher risk for perioperative complications and longer length of stay due to complexity of clinical management. However, we are reassured in the validity of our results given the consistency across all supplementary analyses including a group at lower risk for development of postoperative, acute kidney injury.

Our study represents an important first step in the critical appraisal of TEE use as a hemodynamic monitor in the setting of isolated CABG surgery. In the absence of randomized and non-randomized evidence, we chose to undertake an observational, retrospective study to investigate whether TEE use in CABG surgery was associated with a clinical outcomes benefit. In contrast to prior observational studies,^{11, 12} our comparative-effectiveness study aimed to investigate the impact of TEE by undertaking three statistical analyses including an instrumental variable analysis designed to mitigate unobserved confounding. These statistical models not only adjusted for observed differences among patients but also potentially addressed confounding due to unobserved differences. By all analyses, TEE use in CABG surgery was associated with acute kidney injury and longer length of hospitalization.

This study should be interpreted with awareness of several limitations. First, the observational, non-randomized, design of our study using claims data precludes a causal link between TEE and acute kidney injury and length of hospitalization. While multiple linear regression adjusts for differences in observed covariates, it is unable to fully adjust for unobserved differences (such as ejection fraction) which could lead to residual unobserved confounding by indication; particularly if the instrument were not perfect. Since our instrument incorporated state and year, we cannot fully rule out that differences existed across patients either within a state or in a given year that may have influenced outcomes (i.e. we cannot absolutely confirm the exclusion restriction). Second, administrative claims data have a high specificity but low sensitivity for identification of acute kidney injury.^{28, 29} This, combined with the lack of laboratory values (no creatinine value) could have led to underestimation of the true, incidence of renal injury in this cohort. However, given the mean rate of TEE approached 50%, we believe this limitation would be equally distributed between the two groups. Third, because of the multifactorial etiology of kidney injury following cardiac surgery, the use of TEE as a hemodynamic monitoring device to optimize resuscitation, inotropic support and fluid management could be interpreted as indirect. Fourth, although Optum’s CDM represents a large, commercially-insured patient population, these findings are not completely generalizable; either internationally or among the Medicare population.

Conclusions

Our findings may have health policy implications for clinical practice. The evidence to support the routine use of TEE in CABG surgery is equivocal.¹⁸ This study found that among adult patients undergoing isolated CABG surgery in the US, TEE was not associated with decreased incidence of new-onset, in-hospital, acute kidney injury or decreased hospital length of stay. These findings highlight the importance of additional work to study the clinical effectiveness of TEE in isolated CABG surgery.