Multicomponent Deimplementation Strategy to Reduce Low-Value Preoperative Testing

This quality improvement study investigates if a multicomponent deimplementation strategy directed at clinicians is associated with a reduction in low-value testing before general surgery operations.

Key Points

Question

Is a multicomponent deimplementation strategy directed at clinicians associated with a reduction in low-value testing before general surgery operations?

Findings

In this quality improvement study including 1143 adult patients, deployment of a multicomponent deimplementation strategy was associated with a reduction in the use of low-value tests before 3 general surgery operations. The resulting changes in testing practices were independent of temporal trends within or outside the study hospital.

Meaning

A multicomponent deimplementation strategy informed by preidentified determinants of practice and including the introduction of evidence-based guidelines and decisional-support materials, stakeholder engagement, and interactive presentations with consensus building was associated with a decrease in low-value preoperative testing.

Abstract

Importance

Routine preoperative blood tests and electrocardiograms before low-risk surgery do not prevent adverse events or change management but waste resources and can cause patient harm. Given this, multispecialty organizations recommend against routine testing before low-risk surgery.

Objective

To determine whether a multicomponent deimplementation strategy (the intervention) would reduce low-value preoperative testing before low-risk general surgery operations.

Design, Setting, and Participants

This study had a pre-post quality improvement interventional design using interrupted time series and difference-in-difference analytic approaches. The setting was a single academic, quaternary referral hospital with 2 freestanding ambulatory surgery centers and a central preoperative clinic. Included in the study were adult patients undergoing nonurgent outpatient inguinal hernia repairs, lumpectomy, or laparoscopic cholecystectomy between June 2022 and August 2023. Eligible clinicians included those treating at least 1 patient during both the preintervention and postintervention periods.

Interventions

All clinicians were exposed to the multicomponent deimplementation intervention, and their testing practices were compared before and after the intervention. The strategy components were evidenced-based decisional support, multidisciplinary stakeholder engagement, educational sessions, and consensus building with surgeons and physician assistants staffing a preoperative clinic.

Main Outcomes and Measures

The primary end point of the trial was the rate of unnecessary preoperative tests across each trial period.

Results

A total of 1143 patients (mean [SD] age, 58.7 [15.5] years; 643 female [56.3%]) underwent 261 operations (23%) in the preintervention period, 510 (45%) in the intervention period, and 372 (33%) in the postintervention period. Unnecessary testing rates decreased over each period (intervention testing rate, −16%; 95% CI, −4% to −27%; P = .01; postintervention testing rate, −27%; 95% CI, −17% to −38%; P = .003) and within each test category. The decrease in overall testing was not observed at other hospitals in the state on adjusted difference-in-difference analysis.

Conclusions and Relevance

In this quality improvement study, a multicomponent deimplementation strategy was associated with a reduction in unnecessary preoperative testing before low-risk general surgery operations. The resulting changes in testing practice patterns were not associated with temporal trends within or outside the study hospital. Results suggest that this intervention was effective, applicable to common general surgery operations, and adaptable for expansion into appropriate clinical settings.

Introduction

Low-value care is defined as tests or treatments that provide little or no benefit to patients and are associated with harm or costs.^1,2 A common source of low-value care within surgery is routine preoperative testing, particularly before low-risk ambulatory surgery.³ Numerous high-quality studies have demonstrated that routine preoperative tests before low-risk operations do not prevent adverse events or case cancellation.^4,5,6,7,8,9 These unnecessary tests cost health systems and patients time and resources and often lead to care cascades that can cause harm.^10,11,12 As a result, multiple surgical, anesthesia, and medicine specialty organizations, in partnership with the Choosing Wisely initiative, recommend against the use of preoperative testing in healthy patients before low-risk operations.¹³

Despite these recommendations, routine preoperative testing is common, with varying testing rates across surgeons and hospitals.^14,15 Behavioral determinants of low-value preoperative testing have been previously described.^16,17 These reasons for unnecessary testing, such as a lack of knowledge regarding the evidence base and misplaced intentions to prevent harm, provide an opportunity to use deimplementation strategies that aim to decrease unnecessary testing.² Although the determinants of the behavior have been established, interventions targeting these determinants of preoperative testing have not previously been implemented and rigorously evaluated across diverse procedure types.¹⁸ These types of efforts have primarily been published outside the peer-reviewed literature, restricted to narrow procedures (eg, cataracts) or populations (eg, veterans), or lack analytical frameworks designed to describe impact within established care contexts.^19,20,21 In particular, the impact on the appropriateness of testing, based on patient comorbidities and functional status absent from claims data, has not been assessed, to our knowledge. An intervention to reduce low-value preoperative testing has not been pursued in a large cohort of patients undergoing common surgical procedures under general anesthesia.

This pre-post quality improvement study tested a multicomponent deimplementation intervention at an academic, quaternary referral center to decrease unnecessary preoperative tests for patients undergoing 1 of 3 low-risk general surgery operations. The intervention components were grounded in prior ethnographic research of multiple stakeholder groups, addressing local determinants of practice and intrainstitutional processes.²² We hypothesized that the intervention would be associated with a decrease overall testing rates and unnecessary testing independent of any temporal trends within or outside the study hospital.

Methods

Study Design

This single institution, pre-post quality improvement interventional study evaluated preoperative testing practices for patients undergoing 1 of 3 low-risk general surgery operations before and after the rollout of a multicomponent deimplementation strategy. We used interrupted time series and difference-in-difference analytic approaches to account for temporal trends within or outside the study hospital. The study was approved by the University of Michigan institutional review board, and the requirement for written informed consent was waived owing to the use of deidentified patient data. Study procedures were reported according to the Standards for Quality Improvement Reporting Excellence (SQUIRE) reporting guidelines.²³

Participants and Setting

The study was undertaken in a large, quaternary referral hospital system, including 2 freestanding ambulatory surgery centers and a preoperative clinic. Before study activities, the intervention hospital had been identified as a high utilizer of preoperative testing via statewide claims data with a baseline testing rate in 2021 of 66%.¹⁴ Adult patients who underwent outpatient, nonurgent inguinal hernia repairs, lumpectomy, or laparoscopic cholecystectomy from June 2022 to August 2023 were included. Participant race and ethnicity information was identified in the claims data as Asian, Black, White, and other (which included American Indian or Alaska Native and Native Hawaiian or Other Pacific Islander) or undisclosed. Race and ethnicity were included to assess and control for associations between these patient characteristics and outcomes. Patients seen through the emergency department or inpatient services were excluded. Patients of surgeons who operated only before or only after the intervention were also excluded. The intervention was not directed at patients. Fifteen surgeons in 2 subspecialties (surgical oncology and minimally invasive surgery) and all physician assistants (PAs) staffing 1 of 3 preoperative clinics were exposed to the intervention. The preoperative clinic provides an assessment separate from the surgeon’s preoperative visit with their patient. PAs, supervised by a surgeon who directs the preoperative clinic, document a preoperative history and physical, assess patients’ preoperative risk, order additional testing and consultations as necessary, follow up the results of their workup, and perform preoperative teaching (eg, expected postoperative recovery, drain teaching, etc). On preliminary medical record review, surgical trainees (residents, fellows) did not account for any preoperative laboratory testing in this elective outpatient setting. This was consistent throughout the analysis; therefore, trainees were not involved in the study as participants. The full study period was 15 months. The 4-month preintervention period was from June to September 2022. The intervention period, wherein the deimplementation strategies were deployed, began with stakeholder engagement in October 2022, and presentations to clinicians occurred from January to March 2023.²⁴ Once these activities were concluded, data were collected during the postintervention period, lasting 5 months from April to August 2023.

Deimplementation Strategy Components (Exposure)

The components of this deimplementation strategy (the intervention) were chosen based on determinants of overuse discovered from ethnographic observations and clinician interviews at the study site.²² The intervention consisted of (1) the introduction of evidenced-based testing guidelines and decision support materials presented on a dedicated study website, (2) stakeholder engagement with multidisciplinary perioperative clinicians, including anesthesiology and surgical leaders, and (3) group presentations and consensus-building with surgeons and PAs staffing the preoperative clinic. The decision support materials were based on the Choosing Wisely guidelines, multiple national organizations, and other toolkits designed to reduce low-value preoperative testing (eAppendix in Supplement 1).^13,25,26 These were iteratively tailored to the local context based on feedback from surgeons and anesthesiologists during stakeholder engagement (Figure 1). The guidelines were compiled by a team of clinicians, presented to stakeholders, iteratively revised based on that feedback, and formatted for point-of-care use. They were then posted on the dedicated website and provided to participants via a conveniently accessible link. The recommendations and website were finalized before conducting presentations. Educational sessions and consensus building with surgeons occurred during monthly divisional (surgical oncology and minimally invasive surgery) meetings. These were two 30-minute presentations that consisted of sharing the guidelines, the data motivating this study’s activities, and providing time to answer questions. There was a follow-up meeting with both groups of surgeons after study activities were concluded to share results and obtain reactions and feedback. Engagement with preoperative clinic PAs consisted of 3 discrete virtual group meetings lasting 20 to 30 minutes, describing the intervention’s goals, collected outcomes, baseline testing rates, and discussing clinicians’ questions and concerns. Follow-up meetings with these groups were conducted after the study timeline to exchange feedback and maintain stakeholder involvement.

Figure 1. Intervention Timeline.

Intervention components include educational materials on a dedicated website, stakeholder engagement and input, and presentations of educational materials, trial purpose, and dialogue with health care professionals. PA indicates physician assistant.

Data Sources and Outcomes

Data were abstracted via medical record review using a standardized data collection instrument. Data were abstracted by study team members (R.J.K., C.R., C.P., A.V., E.K., D.N., H.P.). Senior medical students trained junior students using a standardized data collection instrument and consistent instructions across the study’s timeline. Data elements included the patient’s age, comorbidities, American Society of Anesthesiologists (ASA) class, procedure, attending surgeon, whether a test was ordered, and, if so, which clinician ordered a given test. Preoperative testing included complete blood cell counts (CBCs), basic metabolic panels (BMPs), comprehensive metabolic panels (CMPs), and electrocardiograms. The primary end point of the intervention was the rate of unnecessary preoperative tests in eligible patients, ie, the proportion of patients receiving any unnecessary test divided by the total number of patients in each implementation period. Testing rates were analyzed as a proportion of operations performed and appropriateness within a given type of test. Patients’ age, comorbidities, and ASA class were used to assess the appropriateness of their preoperative tests. For example, any test administered to patients with ASA class I or II was categorized as unnecessary, while testing for patients with ASA class III or higher was determined to be appropriate based on their comorbidities (eg, a patient with diabetes may have an appropriate preoperative BMP) (eAppendix in Supplement 1). Specialist status was also incorporated into the appropriateness assessment (ie, a hematologist’s preoperative CBC or hepatologist’s preoperative CMP was always classified as appropriate). The assessment of appropriateness was deliberately constructed to be conservative in assigning tests as appropriate by accounting for multiple comorbidities, specialty designations, and clinical scenarios.

Secondary outcomes were assessed using 2 additional data sources. Emergency department visits within 7 days postoperatively and readmissions within 30 days postoperatively were collected via intrainstitutional medical record abstraction databases and outside hospital emergency department databases within the same state. Additionally, administrative claims data available via the Michigan Value Collaborative were used to compare overall testing rates at the intervention hospital with those at other hospitals before and after the intervention. This database captures payment and utilization data for an episode of care from paid, adjudicated claims.^27,28 These data were compared with those obtained from medical record review and found to give comparable testing estimates during similar periods. The 3 target operations and preoperative tests were identified using current procedural terminology codes.

Statistical Analysis

We compared the characteristics of patients across the study periods using analysis of variance for multicategory comparisons of continuous data and χ² tests for categorical variables. A generalized multilevel logistic regression model was used to conduct an interrupted time series analysis, comparing rates of unnecessary testing before the intervention with postintervention rates, adjusted for age, sex, operation, and binary assessment of ASA class (I or II compared with III or IV).²⁹ Time was included as a continuous variable, and we included Gaussian random effects for the attending surgeons. This approach was chosen to account for existing trends in testing and covariates influencing testing appropriateness. The mean marginal effect of each intervention stage was reported.

A separate hospital-level difference-in-difference analysis compared over-time trends in overall testing as a proportion of the 3 target operations at the intervention hospital with other hospitals in the statewide collaborative, to evaluate whether observed changes could be explained by secular trends in testing, as opposed to the intervention. The sample available in this database did not include surgeon designations, as compared with the primary medical record review analysis; therefore, a general linear model with a log link function and hospital-level clustered SEs was used instead of a multilevel model. The covariates in this model included sex, Charlson-Dayo Score, operation type, and time by month. A 2-sided P value <.05 was considered statistically significant. All statistical analyses were conducted using Stata, version 18 (StataCorp).³⁰

Results

Patient Cohort

The final cohort included 1143 patients (mean [SD] age, 58.7 [15.5] years; 643 female [56.3%]; 500 male [43.7%]) across the 3 study phases, with 418 (37%) undergoing inguinal hernia repairs, 252 (22%) laparoscopic cholecystectomies, and 473 (41%) lumpectomies. There were 261 operations (23%) in the preintervention period, 510 (45%) in the intervention period, and 372 (33%) in the postimplementation period (Figure 1). There were no significant differences between patients across the 3 study periods by age, race, ASA class, or comorbidities (Table 1). Participants identified with the following races and ethnicities: 46 Asian (4.0%), 101 Black (8.8%), 939 White (82.2%), and 57 other or undisclosed (5.0%). There were more patients undergoing laparoscopic cholecystectomies in the preintervention period, contributing to the relatively higher proportion of female patients in that period as well (female vs male, 172 of 261 [66%] vs 89 of 261 [34%]; P = .001 for both). Nearly all patients (1112 of 1143 [97%]) were evaluated in the preoperative clinic.

Table 1. Baseline Characteristics of Patient Groups by Implementation Stage.

Patient characteristics	Preintervention	Intervention	Postintervention	P value
No./total No. (%)	261/1143 (23)	510/1143 (45)	372/1143 (32)	NA
Operation, No. (%)
Lumpectomy	105 (40)	206 (40)	162 (44)	<.001
Inguinal hernia repair	73 (28)	221 (43)	124 (33)
Laparoscopic cholecystectomy	83 (32)	83 (16)	86 (23)
Age, mean (SD), y	57.4 (15.9)	59.5 (15.5)	58.6 (15.3)	.19
Sex, No. (%)
Female	172 (66)	281 (55)	190 (51)	<.001
Male	89 (34)	229 (45)	182 (49)
Race, No. (%)
Asian	12 (5)	20 (4)	14 (4)	.19
Black	19 (7)	34 (7)	48 (13)
White	217 (84)	429 (85)	293 (79)
Other/undiscloseda	13 (5)	27 (5)	17 (5)
ASA class, No. (%)
I	11 (4)	19 (4)	13 (4)	.18
II	174 (67)	319 (63)	210 (57)
III	76 (29)	171 (34)	148 (40)
IV	0	1 (0)	1 (0)
Comorbidities, No. (%)
Anemia/thrombocytopenia	37 (14)	64 (13)	53 (14)	.73
Cardiovascular disease	39 (15)	83 (16)	68 (18.3)	.52
Anticoagulant use	16 (6)	33 (7)	28 (8)	.75
Kidney disease	13 (5)	32 (6)	29 (8)	.36
Diabetes/endocrinopathy	92 (35)	177 (35)	158 (43)	.13
Liver disease	20 (8)	21 (4)	28 (8)	.05
Vascular disease risk factors	109 (42)	183 (36)	145 (39)	.23

Abbreviations: ASA, American Society of Anesthesiologists; NA, not applicable.

^a Other race includes American Indian or Alaska Native and Native Hawaiian or Other Pacific Islander. Race and ethnicity data were identified in the claims data.

Appropriateness of Testing (Primary Outcome) and Overall Testing Rates

In the preintervention period, 37% of patients (98 of 261) received an unnecessary test. This proportion decreased to 25% (126 of 510) in the intervention period and 14% (52 of 372) in the postimplementation period. This trend was present within each test category (Figure 2). Unnecessary testing decreased within each category of operation in each period, from 54%, 46%, and 21% in laparoscopic cholecystectomy, 41%, 31%, and 15% in inguinal hernia repairs, and 20%, 10%, and 9% in lumpectomies, respectively. The rates of necessary testing were consistent throughout the study period: 15% (38 of 261), 12% (59 of 510), and 13% (50 of 372) in the 3 intervention periods, respectively. We also observed that variation in monthly testing rates was significantly less in the postintervention period, with the SD around the mean monthly rate of unnecessary testing in the intervention period being nearly 10 times higher than the SD in the postintervention period (0.07 vs 0.008).

Figure 2. Overall Testing Decreased and Appropriateness Increased Across Type of Test.

For all comparison groups, there was significant change at level P ≤.01 in both overall testing and appropriateness.

Overall testing rates also decreased over each period, from 134 of 261 (51%) preintervention to 185 of 510 (36%) during the intervention and 102 of 372 (27%) postintervention (P < .001). This change was relatively consistent across each type of test (Figure 2). Scaling to the number of operations in each period, there were 0.65 tests per patient during the preintervention period, 0.42 tests per patient during intervention, and 0.22 tests per patient postimplementation.

Eight of 11 attending surgeons (73%) showed improved rates of unnecessary testing. The 3 surgeons who did not change their testing patterns had lower than average rates before the intervention; 2 continued to have unnecessary tests ordered for approximately 20% of their patients postimplementation. The PAs in the preoperative clinic were responsible for 55% (54 of 98) of the unnecessary tests in the preintervention period, and attending surgeons ordered 32% (31 of 98). These relative proportions remained stable throughout the intervention period, but within the PA group, the appropriateness of the tests improved from 21% to 35% to 61%, while the attending surgeons increased from 14% of tests being appropriate up to 15%, and finally to 39% in the postintervention period.

In the adjusted mixed-effect generalized model, the association between the study period and receipt of any unnecessary test demonstrated a significant decrease in the likelihood of receiving an unnecessary test as the intervention progressed from one period to another (odds ratio [OR], 0.40; 95% CI, 0.20-0.81; P = .01 and OR, 0.13; 95% CI, 0.04-0.42; P = .003, respectively), with representative unadjusted trend in Figure 3. The average marginal effect of the intervention period decreased the likelihood of receiving an unnecessary test by 16% (95% CI, 4%-27%), and the postintervention period decreased this likelihood by 27% (95% CI, 17%-38%), compared with the preintervention period.

Figure 3. Monthly Rate of the Proportion of Patients Receiving Unnecessary Testing Across Study Periods (Unadjusted With Linear Lines of Best Fit).

Apr indicates April; Aug, August; Dec, December; Feb, February; Jan, January; Jul, July; Jun, June; Mar, March; Nov, November; Oct, October; Sep, September.

Comparison With Other Hospitals

The mean rate of overall preoperative testing, as assessed by claims data at nonintervention institutions, remained stable across time periods (37%, 37%, 38%, respectively). In contrast, the mean preoperative testing rates at the intervention institution decreased across the study periods (38%, 36%, 33%, respectively). When comparing the probability of preoperative testing, the intervention institution significantly reduced the probability of preoperative testing postintervention period compared with nonintervention institutions in the same state when accounting for existing trends in testing practices (OR [SE], 0.792 [−4.29]; P < .001) (Table 2).

Table 2. Difference-in-Difference Model Comparing the Likelihood of Overall Testing in the Intervention Hospital With Other Hospitals in Statewide Collaborative.

Model variable	OR of receiving any test (SE)	P value
Female	0.998 (−0.06)	.95
Charlson-Dayo score	1.103 (11.74)a	<.001
Operation (inguinal hernia reference)
Lumpectomy	1.074 (0.85)	.39
Laparascopic cholecystectomy	1.058 (1.09)	.28
Time, mo	1.002 (0.27)	.78
Time period
Intervention	0.940 (−1.34)	.18
Postintervention	0.952 (−0.67)	.50
Institution-intervention interaction
Intervention hospital, intervention	0.867 (−2.24)b	.03
Intervention hospital, postintervention	0.792 (−4.29)a	<.001
Institution	1.062 (0.73)	.46

Abbreviation: OR, odds ratio.

^a P < .001.

^b P < .05.

Adverse Events

There was no difference in 7-day emergency department visits between the preintervention period (14 [5%]), the intervention period (10 [2%]), and the postintervention period (35 [2%]). Similarly, 30-day readmission rates remained steady across study periods (2% [6] in preintervention; 1% [6] during the intervention; and 1% [3] postintervention; P = .33).

Discussion

This quality improvement interventional study has 2 main findings. First, this multicomponent deimplementation intervention was associated with a reduction in unnecessary preoperative testing before common, low-risk general surgery operations. This reduction in testing was independent of temporal testing trends within or outside the study hospital, further affirming the outcome of the strategy. Second, the intervention was associated with a decrease in the number of patients receiving any test and the variation in the testing rates over time.

The association of the multicomponent deimplementation strategy with a reduction in low-value preoperative testing before low-risk general surgery operations demonstrates how established principles can be extended into new clinical settings.³¹ Prior investigations have effectively reduced testing in select populations and operations, primarily within Veterans Health Administration hospitals and in older adults undergoing cataract surgery.⁵ These results support the ability to deimplement low-value testing in 3 of the top 10 most common general surgery procedures. Reducing low-value testing in these high-volume cases has the potential for significant impact.³² Further, analyzing a data source that captured testing appropriateness allowed for a more meaningful assessment of the impact of testing deimplementation. Although there was a shift to slightly more patients having higher ASA class over the study timeline, this was addressed by how testing appropriateness was determined and by its inclusion in the mixed-effect model. The associated decrease in overall testing and increase in testing appropriateness was significant, as low-value testing interventions are pursued across specialties and measured more deliberately. Importantly, there was not a reduction in appropriate testing. The trends in overall testing rates determined from claims data mirrored the trends we observed in more detailed data abstraction by medical record review, suggesting that administrative data could be used to monitor the effects of future deimplementation interventions.

This study offers an example of using best deimplementation practices in health care settings to effect change. Settings with higher rates of low-value care provide the best opportunities to move the needle on guideline uptake.^28,33,34 These baseline data and contextual planning in the preintervention stage are crucial to disrupting existing practices.³¹ A foundation of this intervention design was mapping determinants of behavior to evidenced-based behavioral intervention strategies before deploying an intervention.^14,22 The social network interactions between surgeons, PAs, and anesthetists were key to designing strategy components addressing facilitators of persistent high testing; leveraging this type of knowledge should be the standard approach.^35,36 Additionally, basing strategy selection in data-driven guidelines not only improves patients’ care but also lends creditability to the study activities themselves.³⁷ The higher variation in monthly testing rates ultimately becoming more uniform in the postintervention period potentially reflects acceptance and new habits. Interventions taking a similar evidence-based and contextualized approach have also succeeded in increasing preoperative optimization of patients with hernia, decreasing postoperative opioid prescriptions and renewing surgical checklist utilization.^38,39,40 Finally, the intervention itself did not require extensive financial or clinician time investments, making the effect sizes found even more encouraging, and laying the groundwork for expansion.

The success of this intervention establishes a foundation for expanding the strategy components to other practices across diverse types of health care settings.⁴¹ Future work will test which aspects of low-value testing deimplementation strategies can be translated without adaptation and which need to be context specific. Additional expansion opportunities include low-risk surgery outside of general surgery within the same institution (ie, knee arthroscopy, tubal ligation). The maintenance of these improved testing rates will also be investigated at the intervention institution. Although we plan to test further the intervention’s effect in a cluster randomized design across 16 Michigan hospitals, these analytic techniques provide preliminary evidence that the observed outcome was associated with the intervention itself.

Limitations

There are a few limitations to this study. First, it describes a single academic practice setting involving primarily subspecialized surgeons who each perform a narrow range of procedures. The high participation rates in the presentations and engagement with the intervention activities may be challenging to recreate in other settings and may have influenced deimplementation success to some extent. The dedicated preoperative surgical clinic is an example of a distinct practice pattern, but not necessarily a unique one. The intervention could be easily adapted to target other settings or physicians, such as primary care physicians who routinely preform preoperative evaluations. Additionally, this study was not designed to be powered to detect differences in the very rare adverse events after low-risk ambulatory operations. We included these data to ensure that the rates of these events did not increase after the intervention, but we cannot draw broader conclusions based on this analysis. Specifically, we did not measure intraoperative events or case cancellations, but previously appropriately powered studies have demonstrated that routine preoperative testing does not prevent case cancelations or adverse events. We also relied on robust safety data previously published in adequately powered trials.^{4,5,6,21,42,43} Finally, the intervention outcome was compared to a historical rather than a concurrent cohort. We did not think a clinician-level randomized design would be possible without significant spillover to control clinicians. We used quasi-experimental causal inference analytics to minimize the risk that the observed outcome could be due to temporal trends rather than the intervention.

Conclusions

Results of this study show that a quality improvement intervention using a multicomponent deimplementation strategy was associated with a decrease in low-value preoperative testing, a major source of health care waste. These efforts were associated with a decrease in overall testing rates and an increase in the appropriateness of the remaining tests among general surgery patients undergoing common operations. The improvement was accomplished by ensuring that efforts were evidence based, tailored to the local context, and supported by stakeholders. This intervention advances efforts to decrease low-value preoperative testing by demonstrating effectiveness and safety in a large population of general surgery patients, and these efforts should continue to expand into appropriate clinical settings.

Supplement 1.

eAppendix. Decision Support Materials

Supplement 2.

Data Sharing Statement.