We performed a retrospective cohort analysis of patients aged ≥18 years who underwent an outpatient preoperative evaluation between July 1, 2021, and June 30, 2023, followed by an elective surgical procedure. Exclusion criteria were urgent and emergent surgical procedures, duplicate visits, and incomplete data. Hartford Healthcare is a 7-hospital integrated health care system in Connecticut. The preoperative evaluation centers are staffed by advanced practice providers, in collaboration with internal medicine hospitalist physicians. The data of interest were collected as part of routine clinical care. The study followed the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) reporting guideline [20].

This study was approved by the Institutional Review Board of Hartford Healthcare (HHC-2023-0113; approved on May 18, 2023), which waived the requirement for written informed consent. The data were deidentified before study analysis was performed. No compensation was provided to study participants.

In busy clinical practice, consistently completing a multistep algorithm can be challenging. To address this issue, we developed an EMR form (Epic) to assist clinicians in performing preoperative assessments (Figure 2). This takes less than 1 minute to complete; displays suggestions when preoperative cardiac testing may be unnecessary (an example is shown in Multimedia Appendix 1); and tracks the completed steps of the algorithm and the point at which it ends, referred to as the algorithm end point (AEP). The electronic form was implemented on July 1, 2022, as a standard part of outpatient preoperative evaluations performed at Hartford Healthcare preoperative evaluation centers. The form has 3 components: basic clinical information (completed for all patients), risk assessment (these steps use the preoperative risk tool results to guide the clinician through the steps of the algorithm), and cardiology consultation data (if performed). The AEPs are shown in Figure 1.

The primary exposure was the completion of the EMR form. The 2 study cohorts were dichotomized based on the date of preoperative evaluation: from July 1, 2021, to June 30, 2022 (pre–EMR form cohort), and from July 1, 2022, to June 30, 2023 (post–EMR form cohort). The following variables were collected: demographic data (age, sex, race, ethnicity, date of the preoperative center visit, and date and type of surgery), comorbidities (atrial fibrillation, CAD, congestive heart failure, cerebrovascular accident, transient ischemic attack, chronic kidney disease, or diabetes mellitus), perioperative risk scores (functional capacity, see Multimedia Appendix 2; American Society of Anesthesiologists physical status; RCRI score; and Gupta MICA score), and surgical risk level (categorized as low, moderate, or high risk, with standard definitions used at our institution; see Multimedia Appendix 3).

The primary preoperative resource utilization outcomes were the completion of preoperative cardiology consultations and cardiac diagnostic testing (echocardiography, stress tests, or cardiac catheterization). These must have occurred within 60 days before the surgery to be considered preoperative (The 60-day timeframe was selected to account for instances where surgery is rescheduled due to delays in obtaining testing, although preoperative evaluations typically occur 30 days before surgery). The primary 30-day postoperative outcome collected were as follows: MACE (defined as a composite measure of acute MI, cardiac revascularization, acute congestive heart failure [CHF], or all-cause mortality), acute MI (as defined by the Standardized Endpoints in Perioperative Medicine initiative; see Multimedia Appendix 4) [26], cardiac revascularization (percutaneous coronary intervention or coronary artery bypass graft surgery), acute CHF (defined as clinical or radiographic evidence of volume overload treated with diuretics), and all-cause mortality. Secondary outcomes were intensive care unit (ICU) utilization, all-cause emergency department visits, and all-cause readmissions within 30 days after surgery. Mortality data were obtained from the Connecticut Department of Public Health [27]. All deaths in Connecticut are reported to the Department of Public Health; hence, we consider this a reliable measure. All other data and outcomes were extracted from EMR reporting.

The appropriateness of cardiology consultations was evaluated in the post–EMR form cohort. Possible cardiology consultation indications were defined as the presence of an active cardiac condition (AEP 1); known cardiac disease with unclear status (AEP 2); concern for myocardial ischemia (AEP 3); new ECG changes or abnormal ECG with no prior workup (AEP 4); and elevated MACE risk with poor functional capacity, for which further testing may change management (AEP 9). All other AEPs were considered “no clear indications” (Figure 1).

The study population was depicted with frequencies and percentages for binary or categorical information and the median and IQR for the numerical data. Group comparisons for binary or categorical information were performed using the chi-square or Fisher exact test if the sample size was small for binary variables. If P<.05 was observed for the first test for categorical variables (>2 classes), a post hoc test was carried out with Bonferroni adjustment. The independent-samples Mann-Whitney U test was used for age comparison between the groups.

Propensity score matching was performed to identify comparable subpopulations. The predictive probability of assigning patients to the pre– versus post–EMR form cohort was generated using the demographics and patient characteristics listed in the Methods section. Propensity score–matched, pre– and post–EMR form cohorts were identified using a 1:1 case-control match on propensity score [28], and baseline characteristics were evaluated to determine whether the 2 subpopulations were comparable. A sensitivity analysis was conducted using the E-value approach to assess the magnitude of the unmeasured confounding bias [29,30]. The lowest E-value is 1, suggesting that no unmeasured confounding exists to explain the current association between the predictor and outcome. A higher E-value indicates a stronger unmeasured confounder association that may explain the current effect [29]. A subanalysis was performed in the post–EMR form cohort to evaluate the association of appropriate cardiology consultation indications versus not with completed cardiology consultations and 30-day MACE. Hypothesis testing was performed with a 2-sided α of .05, and all analyses were performed using IBM SPSS Statistics for Windows (version 29).

In this cohort study of patients presenting for outpatient preoperative evaluations before surgery, completion of a structured, EMR-based preoperative cardiac algorithm was associated with a decreased frequency of preoperative cardiology consultations and echocardiograms without an increase in postoperative MACE and other adverse outcomes.

Our study was observational; however, several factors support the validity of our results. We studied a considerable surgical population over 2 years and used propensity score matching to balance several potential confounders of perioperative risk between cohorts, including age, sex, race, comorbidities, perioperative risk tool results, and inherent surgery-specific risks. Both cohorts had a substantial burden of comorbidities (~36%), and a high proportion of patients underwent moderate- or high-risk surgical procedures (~76%). The postoperative outcomes were similar between the pre– and post–EMR form cohorts, with a cumulative low risk of postoperative MACE of 0.4% (pre–EMR form: 45/11,645, 0.4% vs post–EMR form: 46/11,645, 0.4%; P=.92), suggesting that our initiative decreased unnecessary consultations and testing while maintaining an excellent quality of care. Consistent with our results, several other studies also show that inappropriate cardiology consultations and stress tests do not lower the risk of postoperative MACE [5-8,10-12].

A subanalysis of “appropriate” versus “no clear indications” for cardiology consultation in the postintervention cohort showed that many consultations were still requested without a clear indication, highlighting an opportunity to improve the process. Interestingly, the MACE rates did not differ regardless of whether a cardiology consultation was completed, even when there was an appropriate reason for the consultation (Table 4). Similar findings have been reported in the context of preoperative cardiology consultations in patients hospitalized for hip fracture surgery [8]. We suggest that preoperative cardiology consultations should be requested only if required for clinical management and not just because a surgical procedure is planned.

Our study provides a template to guide clinicians in adhering to preoperative algorithms to reduce low-value care. Since the EMR form data can be used to determine the algorithm steps completed, future research could use a similar process to evaluate the ACC/AHA algorithm [4], which has not been prospectively validated despite its wide use.

Our study had several limitations. Due to the retrospective design, the possibility of selection bias and residual confounding remains despite balancing the measured baseline characteristics using propensity scoring. However, we also calculated the E-value, which suggests a low likelihood of unmeasured confounders. The high baseline rate of preoperative cardiology consultations in our study population (23%) may not reflect clinical practice elsewhere. However, our literature review shows a significant variation with rates of 8.7% in low-risk gastrointestinal endoscopic procedures [7], 51.8% in a population undergoing low-risk bariatric surgery [15], and from 6.9% to 87.5% in a study of patients undergoing vascular surgery across 29 hospitals [31]. Our study observed a lower rate of complications compared to NSQIP data [32]; however, NSQIP uses random sampling [33] as compared to all consecutive patients in our study, including approximately 25% of patients undergoing low-risk surgical procedures. Lastly, our data are from a single health care system and thus may not be generalizable to other care settings.

The use of a novel electronic form for the preoperative cardiac risk algorithm is associated with decreased cardiology consultations and testing without an increase in postoperative cardiac complications.