Electronic Health Record–Based Nudge Intervention and Axillary Surgery in Older Women With Breast Cancer

This nonrandomized controlled trial examines rates of sentinel lymph node biopsy among women 70 years and older with breast cancer before and after deployment of an electronic health record nudge targeting surgeons.

Key Points

Question

Can an electronic health record (EHR)–based nudge intervention decrease rates of axillary surgery in women 70 years and older with early-stage, clinically node-negative, hormone receptor–positive, and HER2-negative breast cancer?

Findings

In this nonrandomized controlled trial including 387 patients, 7 surgeons, and 8 clinical settings, a 12-month initiative that deployed a nudge intervention into the EHR schedule significantly decreased the use of low-value sentinel lymph node biopsy by nearly 50% relative to the 12 months before the nudge deployment.

Meaning

A nudge that targets surgeons before counseling patients on surgical management can be used to decrease use of a low-value surgery and reduce overtreatment.

Abstract

Importance

Choosing Wisely recommendations advocate against routine use of axillary staging in older women with early-stage, clinically node-negative (cN0), hormone receptor–positive (HR+), and HER2-negative breast cancer. However, rates of sentinel lymph node biopsy (SLNB) in this population remain persistently high.

Objective

To evaluate whether an electronic health record (EHR)–based nudge intervention targeting surgeons in their first outpatient visit with patients meeting Choosing Wisely criteria decreases rates of SLNB.

Design, Setting, and Participants

This nonrandomized controlled trial was a hybrid type 1 effectiveness-implementation study with subsequent postintervention semistructured interviews and lasted from October 2021 to October 2023. Data came from EHRs at 8 outpatient clinics within an integrated health care system; participants included 7 breast surgical oncologists. Data were collected for female patients meeting Choosing Wisely criteria for omission of SLNB (aged ≥70 years with cT1 and cT2, cN0, HR+/HER2− breast cancer). The study included a 12-month preintervention control period; baseline surveys assessing perceived acceptability, appropriateness, and feasibility of the designed intervention; and a 12-month intervention period.

Intervention

A column nudge was embedded into the surgeon’s schedule in the EHR identifying patients meeting Choosing Wisely criteria for potential SLNB omission.

Main Outcomes and Measures

The primary outcome was rate of SLNB following nudge deployment into the EHR.

Results

Similar baseline demographic and tumor characteristics were observed before (control period, n = 194) and after (intervention period, n = 193) nudge deployment. Patients in both the control and intervention period had a median (IQR) age of 75 (72-79) years. Compared with the control period, unadjusted rates of SLNB decreased by 23.1 percentage points (46.9% SLNB rate prenudge to 23.8% after; 95% CI, −32.9 to −13.8) in the intervention period. An interrupted time series model showed a reduction in the rate of SLNB following nudge deployment (adjusted odds ratio, 0.26; 95% CI, 0.07 to 0.90; P = .03). The participating surgeons scored the intervention highly on acceptability, appropriateness, and feasibility. Dominant themes from semistructured interviews indicated that the intervention helped remind the surgeons of potential Choosing Wisely applicability without the need for additional clicks or actions on the day of the patient visit, which facilitated use.

Conclusions and Relevance

This study showed that a nudge intervention in the EHR significantly decreased low-value axillary surgery in older women with early-stage, cN0, HR+/HER2− breast cancer. This user-friendly and easily implementable EHR-based intervention could be a beneficial approach for decreasing low-value care in other practice settings or patient populations.

Trial Registration

ClinicalTrials.gov Identifier: NCT06006910

Introduction

The Choosing Wisely campaign began in 2012 with the stated goal of sparking conversations between clinicians and patients about unnecessary or low-value care, defined as the use of health services in which the potential harms outweigh the benefits.¹ As a part of this campaign, the Society of Surgical Oncology has recommended against the routine use of sentinel lymph node biopsy (SLNB) in women 70 years and older with early-stage, clinically node-negative (cN0), hormone receptor–positive (HR+) breast cancer. These recommendations were developed because axillary staging did not impact survival and rates of SLN positivity were low because of the tumor’s biological phenotype.^2,3,4 Even in older patients with tumors that exhibit concerning clinicopathologic features, limited nodal involvement does not often alter receipt of chemotherapy independently from genomic testing.⁵ Despite these recommendations, most women still receive axillary surgery.^6,7

Early efforts to deimplement low-value SLNB use involved surgeon-directed educational initiatives⁸ and a nomogram predictor of nodal involvement based on clinical variables.⁹ While both of these approaches modestly decreased SLNB use, rates remained high at 46% and 62%, respectively, at the conclusion of the initiatives. Barriers to widespread deimplementation of SLNB are multifactorial and include the perceived effect of SLNB on adjuvant therapy decision-making, specific patient and tumor factors that increase the chances of having a positive SLN, and concern regarding adjuvant endocrine therapy adherence (ie, surgeons are more likely to perform SLNB in patients who may have trouble adhering to endocrine therapy).^{10,11,12,13,14}

One approach to reduce low-value care, in general, is the application of electronic health record (EHR) “nudges,” which represent a form of clinical decision support that may alter clinicians’ medical decision-making without restricting choice. When properly designed, these tools can be straightforward, user-friendly, and capable of delivering timely reminders at the point of care.^15,16 Nudges, such as those that guide choice through defaults, enable choice, or frame information,¹⁷ have shown promise in reducing low-value care in a variety of settings, such as deprescribing, but have rarely been applied to reduce low-value surgeries.^18,19

In this prospective, nonrandomized controlled trial, we sought to determine if a nudge intervention embedded in the EHR patient schedule could decrease low-value SLNB use. Following the intervention period, we conducted semistructured interviews to identify barriers and facilitators surrounding use of the nudge.

Methods

Study Design

We conducted a hybrid type 1 effectiveness-implementation study employing a prospective, nonrandomized controlled trial (NCT06006910) designed to assess the effectiveness, feasibility, acceptability, and appropriateness of an EHR-based nudge to reduce low-value SLNB (Figure 1). As such, in a hybrid type 1 effectiveness-implementation study, the primary aim was to assess the effectiveness of the intervention with a secondary aim to understand the implementation of and interaction with the nudge intervention itself. Different types of effectiveness-implementation designs are reviewed in Curran et al.²⁰ The study was conducted at 8 surgical oncology clinics at a single integrated health care system. The trial ran from October 2021 through October 2023 (pre-intervention control period: October 2021 through October 2022; intervention period with nudge deployment: October 2022 through October 2023). This trial followed the Transparent Reporting of Evaluations With Nonrandomized Designs (TREND) reporting guideline. The protocol was approved by the UPMC quality improvement review committee, which exempted informed consent of participants involved in the study (trial protocol in Supplement 1).

Figure 1. Schematic of the Study Design and CONSORT Diagram.

This 2-year study consisted of a 12-month control period (without the nudge deployed in the electronic health record [EHR]) followed by a 12-month intervention period (with the nudge deployed in the EHR). In between those periods, the study investigator gave a brief introductory session and the surgeons completed a preintervention survey.

Study Participants and Setting

Eligible surgeons were faculty members of the Division of Breast Surgical Oncology (exclusively practicing breast surgery) who saw new patients with breast cancer at 8 subspecialty clinics across the region. Eligible patients were those who had a new consultative encounter with the breast surgical oncologist and who met the following Choosing Wisely criteria: aged 70 years and older and a new diagnosis of HR+ and human epidermal growth factor receptor 2–negative (HER2−), cN0, early-stage (clinical T stage I and II) breast cancer. As a part of routine care, all patients underwent a physical examination, mammogram, and breast and axillary ultrasonography to confirm the disease was clinically and radiographically node negative. Patients were excluded from analysis if the tumor was deemed an ipsilateral recurrence. Consecutive patients meeting these criteria in the control and intervention periods were included in the study.

At our institution and for patients included in this study, it was not common for patients to see both surgical and medical oncology on the same day. For these patients, a standard workflow includes surgical consultation in the outpatient setting, surgery, and then appointments with medical and radiation oncology unless not indicated.

Intervention

The nudge intervention was a new column in the Epic (Epic Systems) EHR schedule view that flagged eligible patients (based on the patient’s age, diagnosis code of estrogen receptor–positive (ER+) breast cancer, and a new breast cancer visit) at their first outpatient surgical consultation with a caution sign/red clipboard icon. When the surgeons hovered over the icon, a text box displayed that reminded the surgeons to consider omission of SLNB after detailed review of core biopsy pathology and ultrasonographic imaging (see eFigure 1 in Supplement 2 for a representative image of the nudge in the EHR). Following the 12-month preintervention control period, the research team delivered a brief session to the surgeons introducing the column nudge rationale and design (which lasted less than 30 minutes in totality).

Outcomes

The primary outcome was the change in the rate of SLNB in the intervention period among all eligible patient encounters. The investigators, starting at the time of study commencement in the control period, reviewed the EHR weekly to record the recommendation for breast and axillary surgery and verified the procedure performed at the time of surgery.

Additional clinical data were collected on usage of downstream health services in the control and intervention periods, such as rates of pathological node positivity, receipt of adjuvant radiation and systemic therapies, results of genomic testing, and proportion of patients who experienced symptoms warranting lymphedema consultation within the first 3 months after surgery. Accordingly, we also recorded reasoning, as documented in the EHR, for why surgeons performed SLNB.

Secondary outcomes included acceptability, appropriateness, and feasibility of the intervention from the group of breast surgeons. To assess these outcomes, we used a validated survey instrument²¹ along with a series of study-specific supplemental questions (eAppendix 1 in Supplement 2). Each outcome measure, which include the acceptability of intervention measure (AIM), intervention appropriateness measure (IAM), and the feasibility of intervention measure (FIM), consisted of 4 items measured on a 5-point Likert scale for a total of 12 questions. Sample survey questions included, “The nudge seems to meet my approval” (acceptability) or “The nudge seems implementable” (feasibility). These surveys were administered prior to the intervention start to gauge the surgeons’ perception of these items before interaction with the intervention and to make changes as needed.

To assess barriers to and facilitators of implementation, we conducted semistructured interviews with each participating surgeon after the intervention period. The interview guide was informed by a conceptual framework based on literature review, prior work by the authors, and version 2 of the Consolidated Framework for Implementation Research (CFIR2) (eFigure 2 in Supplement 2).²² The interview questions were derived from relevant domains or constructs from CFIR2 (eAppendix 2 in Supplement 2). Each interview was conducted by a member of the research team via video conference, lasted approximately 30 minutes, and was recorded with participant consent.

Covariables

At the time of medical record review, we recorded patient age, patient risk analysis index (RAI; frailty index for estimating postoperative morbidity and mortality), tumor ER and progesterone receptor (PR) histochemical score (H-score, assessed from the core biopsy specimen), tumor Ki-67 (assessed from the core biopsy specimen), the mean Magee equation score (an estimate of Oncotype DX Recurrence Score [Exact Sciences] based on pathologic and clinical variables), tumor grade, tumor clinical T stage, tumor histology, type of breast surgery, and outpatient clinic location. To assess for patients experiencing symptoms warranting a lymphedema evaluation, we used a previously validated natural language understanding approach that uses context- and language-based processing of free text embedded within the EHR.²³

Statistical Analysis

Characteristics of the patients and tumors as well as outpatient practice locations (academic vs community setting) were examined using proportions for categorial variables and median values with IQRs for continuous variables. All variables were compared from patient encounters in control and intervention periods using Fisher exact and Mann-Whitney testing.

The change in the rate of SLNB was analyzed using an unadjusted logistic regression model, an adjusted logistic regression model with the intervention as a binary variable, and an adjusted interrupted time series model with time as a continuous variable. In both adjusted models, the final set of covariates were selected based on bayesian information criteria. In the adjusted logistic regression model, we considered the intervention as a binary exposure and adjusted for monthly patient and tumor characteristics and type of breast surgery. We then fitted an adjusted interrupted time series model, which included the continuous time (per month basis) effect and an interaction term between the intervention and time to adjust for the intrinsic trend in the rate of SLNB over the study period.

In each of the 3 models, rate difference and odds ratio (OR) were both estimated along with the 95% CI. The Wald CI was reported for OR, while the CI for risk difference was estimated numerically from 1000 bootstrapped samples by sampling from the observed data with replacement. All analysis was conducted using R version 4.3.2 (R Foundation). A P value less than .05 was considered statistically significant.

The preintervention survey data were analyzed descriptively. A median score of 16 or higher for each measure (ie, AIM, IAM, and FIM), aggregated across all surgeons, indicated that the intervention component was acceptable, appropriate, or feasible, respectively. The study-specific questions included in the survey were scored on a sliding scale from 0 to 100 (with 100 indicating that a factor highly influences decision-making on whether to perform SLNB or not).

Qualitative Analysis

Interviews were transcribed verbatim and deidentified before analysis. Members of the investigative team used a combination of deductive and inductive strategies to develop the codebook and code interviews and identify relevant themes, informed by the CFIR2 framework. Two investigators independently coded all transcripts and reviewed coding to resolve any discrepancies.

Results

Surgeon and Patient Baseline Characteristics

eTable 1 in Supplement 2 describes the demographic characteristics of the 7 breast surgeons participating in the intervention period and the semistructured interviews. Similar numbers of patients meeting inclusion criteria for the study were seen by these surgeons in the control and intervention periods. Patients in both the control and intervention period had a median (IQR) age of 75 (72-79) years. Tumor characteristics were also similar with no significant differences observed (eTable 2 in Supplement 2).

Rates of SLNB Use

SLNB was performed in 91 of 194 eligible patients (46.9%) in the control period and in 46 of 193 eligible patients (23.8%) in the intervention period (Figure 2), demonstrating a 49.3% decrease in use of SLNB. There was no difference in SLNB utilization in the control or intervention period when stratifying by race and ethnicity. In the adjusted interrupted time series, after adjusting for secular trends in SLNB use, the intervention was associated with a significant decrease in use of SLNB (adjusted odds ratio [aOR], 0.26; 95% CI, 0.07-0.90; P = .03) (Table). The intervention resulted in an immediate level change in the SLNB rate after deployment of the intervention but did not show a significant change in the slope in the intervention period compared with the control period. These results indicate that the intervention was associated with the decrease in SLNB rates after adjusting for the secular rate of SLNB. Additional follow-up after the conclusion of the intervention showed that rates of SLNB continued to decrease (6-month mean of 15.6%), indicating durability in the effect of the intervention (eFigure 3 in Supplement 2).

Figure 2. Rates of Sentinel Lymph Node Biopsy (SLNB) Use Before and After Nudge Intervention Deployment in the Electronic Health Record (EHR).

A, Rate of SLNB in the control period and intervention period per month over the 24-month study period with the horizontal dashed line indicating the mean rate of SLNB use over the 12-month period; the bar graph shows the mean (SD) SLNB rates in the respective period with a 49.3% decrease in SLNB usage after deployment of the intervention. B, Breakdown of the rate of SLNB per month across the study period plotted with the number of unique patients seen in a given month. Dotted line between months 12 and 13 in both panels denotes the time at which the nudge intervention was deployed into the EHR.

Table. Unadjusted and Adjusted Models for Change in the Rate of SLNB Between the Control and Intervention Periods.

Model	SLNB rate		Difference for intervention relative to control (95% CI)	Adjusted OR (95% CI)	P value
	Control period	Intervention period
Unadjusted model	46.9	23.8	−23.1 (−32.9 to −13.8)	0.35 (0.23 to 0.55)	<.001
Adjusted logistic regression model with binary intervention timea	52.9	16.8	−36.1 (−70.7 to −23.8)	0.14 (0.07 to 0.28)	<.001
Adjusted interrupted time series modelb	52.6	15.5	−37.1 (−72.5 to −26.0)	0.26 (0.07 to 0.90)	.03

Abbreviations: ER estrogen receptor; H-score, histochemical score; OR, odds ratio; PR, progesterone receptor; SLNB, sentinel lymph node biopsy.

^a Adjusted for age, clinical Magee equations, type of breast surgery, practice location, ER H-score, and PR H-score selected using bayesian information criterion.

^b Adjusted for age, tumor Ki-67, clinical Magee equations, type of breast surgery, practice location, ER H-score, and PR H-score selected using bayesian information criterion.

Investigation into the documented and inferred reasons for why surgeons still performed SLNB revealed the surgeons tended to cite factors related to tumor biology (such as higher Ki-67, lower ER/PR expression, and tumor grade) more frequently in the intervention period, which was further corroborated by multivariable analysis (eTables 3 and 4 in Supplement 2).

Effect of SLNB Omission on Downstream Care

Rates of pathological node positivity were higher in intervention period (8.8% in the control period, 15.2% in the intervention period), and all patients with node positivity were staged with pN1 disease (no patient had more than 2 positive nodes and thus no patients were staged with pN2-3 disease). The recommended adjuvant systemic therapy, including recommendations for chemotherapy and/or abemaciclib, and radiotherapy were similar in the 2 groups (eTable 5 in Supplement 2). Recommendations for adjuvant chemotherapy were, in all cases, driven by genomic testing independent of nodal status. With the significant decreases in rates of SLNB use, we also observed a decrease in the number of patients who experienced symptoms worthy of a referral for lymphedema evaluation (6.2% of patients in the control period; 3.6% of patients in the intervention period).

Survey Results

The intervention met criteria for acceptability, appropriateness, and feasibility among the participating surgeons with aggregate scores exceeding 16 of 20 (Figure 3). Scores for feasibility of the intervention were highest among the 3 components.

Figure 3. Preintervention Survey Results.

Results from the preintervention survey, which included (A) a validated survey mechanism about the acceptability of intervention measure (AIM), intervention appropriateness measure (IAM), and feasibility of intervention measure (FIM), which uses an aggregated score within each measure (out of 20, with scores >16 indicating that the designed intervention was acceptable, appropriate, and feasible). Plotted bar indicates median score; error bars, 95% CI; circles, individual surgeon scores. B, Study-specific survey questions asking the surgeons which factors most influenced their decision-making about performing SLNB; higher scores on the 100-point sliding scale indicated the given item highly influenced their practice. Solid line indicates median score; dashed line, IQR; circles, individual surgeon scores.

Surgeon Interaction With the Intervention

In the semistructured interviews, we identified 4 themes related to surgeon engagement and interaction with the intervention: (1) intervention as a “reminder” for patient applicability of omission of SLNB; (2) logistics of intervention implementation and engagement; (3) patient perceptions and influence; and (4) impact of axillary staging decision-making on interactions with medical and radiation oncology colleagues (Box).

Box. Key Themes, Corresponding CFIR Domains, and Representative Quotes From Semistructured Interviews of Breast Surgeons.

Each quote represents a response from an individual surgeon.

Themes From Deductive Coding

Intervention as a “Reminder” for Patient Applicability to Omission of SLNB (CFIR2 Domain: Innovation)

“I see a huge value in the concept of the nudge offering one more layer of support that what decisions they are making are supported by people who think about this specific problem.”

“It was kind of a trigger to me to look at the patient age first and foremost as a reminder. Then, I would have it primed when I went to look deeper at the rest of the chart.”

“It made me more thoughtful because just this little quick reminder on the EHR prompted me to think about it. So, it was a good reminder, and it did exactly what its job was, which was to nudge me to think about biology rather than just reflexively doing SLNB because that is ‘complete care.’ I had to keep in mind that complete care doesn’t always equate with better care.”

Logistics of Intervention Implementation and Engagement (CFIR2 Domain: Innovation)

“I liked that I didn’t have to look for it. There are other nonrelated electronic medical record things that need to be done, with additional buttons to click to close out a patient’s chart, that you have to remember to do. Whereas for this, it was right there, and I didn’t have to click any additional buttons. I don’t have to do anything more than just see it. It didn’t slow anything down, and it didn’t require anything other than the normal routine. It was just like an eyeball reminder that you should consider this.”

“I can’t say that the [EHR intervention] added a meaningful amount of time to my workflow such that it was irritating. In fact, it was sort of welcome in the sense that it gave me a sense that I’m not crazy, and I’m not trying to do anything that’s outside of standard of care, because this is, in fact, the patient who would theoretically qualify for omission.”

Patient Perceptions and Framing (CFIR2 Domain: Inner Setting)

“I have had some patients when we talk about this have a negative reaction to it, depending on how you present it. Sometimes it can come off as we don’t think you’re going to live much longer, so we’re not going to do the surgery that younger patients would get. But I try to explain it more that we don’t want to give you the risks of additional anesthesia and the risk of lymphedema with axillary surgery if it’s not going to change any of your treatment recommendations or long-term outcomes.”

“I would say that the majority of patients are very happy to have less surgery done. There are some patients who do have the concern that less is being done to them because of their age, and they don’t want to have less done to them because of their age. But oftentimes, when I counsel them about the risk of lymphedema and paresthesia, I think they become more comfortable.”

“It’s a dichotomous opinion. There are either patients who are in love with the omission concept and then there are some patients, especially the ones who ‘feel’ younger, mentally and emotionally, that tend to give me the impression that they feel like we are falling short of standards of care because we are trying to do or omit something that otherwise they felt all their other friends and people they’ve spoken to about breast cancer treatment are getting.”

Impact of Axillary Staging Decision-Making on Interactions With Medical and Radiation Oncology Colleagues (CFIR2 Domain: Outer Setting)

“We don’t get to make this decision in a vacuum, so there have been instances in which patients who haven’t had axillary staging have gone on to seek opinions from our colleagues in medical and radiation oncology and were flat out told that they were treated incorrectly by the surgeon, and that they should have been staged. That’s no good for us, and it’s no good for patients to hear that, so continuing to educate and talk to other specialties is important.”

“I am more likely to email the medical oncology and radiation oncology partners that I work with for tumors that are biologically concerning if I want to get more backup and input on the decision and how it may affect their treatment planning. But this is difficult to do for all patients given that we are all busy.”

“If I’m going to be sending somebody back to the community as it has been in the past, sometimes I will have patients return to me, especially in the last 3 to 4 years, that will say ‘I am very upset with you because I am now committed to radiation therapy because my radiation oncologist said you didn’t operate on my lymph nodes.’ So now they feel compelled to do the radiation, whereas if I just got that lymph node surgery, they may have felt differently.”

Themes From Inductive Coding

Shift to Looking More Critically at Tumor Biology and Aging Phenotype to Drive Decision-Making

“Really, we know that the biology of breast cancer is what drives the prognosis and the outcome, rather than the axillary staging. Now we know it really is the biology of the cancer.”

“I think that a subset of patients sometimes will question when we say we can omit the axillary staging surgery, especially if you have a really healthy 70-year-old who could live 30 more years. So, I think you have to take into account the health of the patient and their functionality.”

“I don’t use it literally [the age 70 cutoff included in the Choosing Wisely criteria]. I’m not looking for a patient who’s chronologically more than 70. I’m looking for a patient who is biologically more than 70.”

“The intervention pushed me to thinking of the biology of the tumor and about the multidisciplinary care, and that’s what sometimes gives me pause with omission, in addition to the level of debilitation or not.”

“So, if I saw the icon, I would be reminded to look deeper at the estrogen receptor level, not just if it was positive or negative. I would then holistically look more into other components of the tumor, including the grade and Ki-67.”

“Choosing Wisely didn’t necessarily subclassify or specify further criteria for omission, as in it was just an umbrella term. But not all ER+/HER2− breast cancer is the same, right? It comes down to the biological phenotype.”

Standardizing/Formalizing the Approach to This Omission of Axillary Surgery in This Population

“There are a lot of things that we could do better in breast surgery. There’s a there’s a lot we can learn from the way nudge has formalized my approach to axillary staging in this specific subset of patients that we can apply to other areas of breast cancer surgery that would benefit people.”

“I think the intervention has made me make the same decision consistently in this population of patients.”

“I think it [the intervention] is a good thing, because it has, to a great degree, helped further standardize the patterns of care that we are delivering and reducing exposure to inappropriate care.”

Respondents indicated that while they always look at the patient’s age when reviewing medical records and planning for surgery, the intervention helped serve as a reminder that SLNB may not be warranted for a given patient (“It served as a reminder to me that every time I’m seeing patients in the office, and I think that’s really helpful just to push it into the front of my mind”). Further, respondents described engagement with the intervention as a visual cue and welcomed the fact that there were no additional clicks or justifications they had to write, which facilitated continued use given that it added little or no extra time to their medical record review (“I think a benefit of the nudge is that it was present, but it didn’t slow anything down and didn’t require anything other than the normal routine”).

Using inductive coding, we identified 2 additional themes: (1) intervention initiating a shift to thinking more critically about the tumor biology and patient’s biological age and (2) intervention standardizing individual surgeon approaches to omission of SLNB. Respondents remarked on the concept of biological age in concert with functional status, level of debility (if any), and comorbidities that may move them away from using age 70 years as a strict cutoff. As the intervention pushed surgeons to consider tumor biology more over the study period, many commented that the intervention helped standardize their approach to SLNB omission in this cohort of patients.

Discussion

In this nonrandomized controlled trial, an EHR-based nudge intervention significantly decreased the rate of SLNB use in older women who met Choosing Wisely recommendations for omission of SLNB. The decrease in SLNB rates after deployment of the intervention did not result in changes to the recommended adjuvant therapy but did decrease the number of patients who experienced symptoms worthy of further evaluation for lymphedema. The designed intervention met criteria for acceptability, appropriateness, and feasibility. Thematic analysis from semistructured interviews indicated that the surgeons felt the intervention’s most helpful aspects included serving as a point-of-care reminder without having to complete additional tasks within the EHR, as well as prompting a critical evaluation of the tumor biology and patient’s biological age.

This study provides a pragmatic approach to facilitate the deimplementation of a low-value surgery, in line with recently suggested strategies.²⁴ Prior deimplementation efforts have focused attention on specific levels: patient-, surgeon-, and system-level interventions. We opted to focus our efforts on surgeons themselves because evidence shows that surgeons, in this setting, serve as the therapeutic keystone by analyzing critical patient and tumor factors, facilitating multidisciplinary conversations with medical and radiation oncologists, and framing conversations on utility of SLNB with patients.^10,13,25 While educational initiatives^8,26 and nomogram-based risk calculators for predicting nodal positivity^9,27,28 may help surgeons, our approach demonstrated a larger and more durable reduction in SLNB use because it was not a 1-time effort but a continuous, day-of-clinic-visit reminder present over 12 months.

The effectiveness of nudges typically depends on the effort required to engage with the nudge and the degree to which the nudge interrupts standard clinical workflow.¹⁹ The designed nudge intervention in this study did not require high cognitive burden on the part of the surgeon, did not disrupt existing workflows, and did not require additional clicks or justifications,¹⁸ all of which were remarked on during the semistructured interviews with the surgeons. With the effect of this approach, it may be reasonable to extend this intervention to facilitate axillary surgery de-escalation in other patient populations or clinical practice settings. This is particularly important given the recent reporting of the SOUND trial, which provided evidence that patients with small tumors and a cN0 axilla can forego axillary surgery irrespective of age.²⁹

Our semistructured interviews indicated that a beneficial aspect of the intervention was the push for surgeons to standardize their approach when evaluating patient factors (such as biological age and functional status) with tumor factors (such as size, grade, and Ki-67). Instead of “routine use,” the intervention prompted deeper consideration of the impact on downstream care. Another key theme was the need for continued multidisciplinary conversations about how omission of SLNB affects decisions concerning radiation and systemic therapy. In the absence of pN2 disease, as in this study, chemotherapy decision-making is almost entirely driven by genomic testing and not by nodal staging.^30,31,32,33 With the recent results of the MONARCHE study³⁴ and NATALEE study for use of adjuvant CDK4/6 inhibitors, future multidisciplinary consideration is warranted to further adjudicate systemic therapy decisions should omission of SLNB occur.

Limitations and Strengths

This study has important limitations. First, the sample size of the surgeons in the intervention group is limited; thus, the nudge intervention should be evaluated in a larger context in future studies. Given the nonrandomized nature of the study, there exist alternative explanations for the decrease in SLNB rates outside and in addition to the intervention itself. This intervention comes at a time of de-escalation in breast surgery across several indications, including for SLNB in older patients. Specifically in this patient population, evidence continues to accumulate for the low-value nature of axillary staging^5,35,36,37 even years after the Society of Surgical Oncology’s adoption of the Choosing Wisely recommendations, building momentum for continued de-escalation. Additionally, there could be an element of conformity or social pressure as this intervention was deployed within a single division. While we do not believe this had an appreciable effect on its own, we cannot rule out that rates of SLNB changed from the brief introductory session alone. Despite these limitations, there are several strengths, including the duration of the study; the innovation of the intervention, which has not previously been tested in this context; and the information gathered in the semistructured interviews to further inform implementation of the intervention.

Conclusion

This study showed that the nudge we deployed in the EHR decreased low-value SLNB use in older women without affecting the receipt of adjuvant therapies. Over the course of the intervention period, surgeons reported that the nudge helped remind them of the potential for SLNB omission, standardize their practice, and look more toward the tumor’s biological phenotype to guide surgical decision-making. Based on these findings, an EHR nudge is an evidence-based solution to deimplement SLNB and may be extended to other practice settings (such as in community clinics for general surgeons who practice breast surgery) or to other patient populations (such as all postmenopausal patients with small ER+ tumors according to the SOUND trial).

Supplement 1.

Trial protocol

Supplement 2.

eFigure 1. Image of Epic EHR showing column nudge deployment and design

eAppendix 1. Copy of the pre-study survey mechanism along with study-specific questions

eFigure 2. Conceptual framework used to guide semi-structured interview questions

eAppendix 2. Copy of the interview guide & script for post-intervention semi-structured interviews

eTable 1. Demographic characteristics of the participating surgeons

eTable 2. Baseline patient and tumor characteristics based included in the control and intervention periods

eFigure 3. Follow up of SLNB rates after the conclusion of the intervention period

eTable 3. Multivariable analysis of factors associated with SLNB use before and after the nudge intervention

eTable 4. Documented and inferred reasons from the EHR notes why surgeons still performed SLNB before and after the nudge

eTable 5. Recommended adjuvant therapy in the control and intervention periods

eReferences

Supplement 3.

Data sharing statement