Factors associated with provision of non-beneficial surgery: A national survey of surgeons

Abstract

Objective:

We tested the association of systems factors with the surgeon’s likelihood of offering surgical intervention for older adults with life-limiting acute surgical conditions.

Summary Background Data:

Use of surgical treatments in the last year of life is frequent. Improved risk prediction and clinician communication are solutions proposed to improve serious illness care, yet systems factors may also drive receipt of non-beneficial treatment.

Methods:

We mailed a national survey to 5200 surgeons randomly selected from the American College of Surgeons database comprised of a clinical vignette describing a seriously ill older adult with an acute surgical condition, which utilized a 2×2 factorial design to assess patient and systems factors on receipt of surgical treatment to surgeons.

Results:

Two thousand, one hundred sixty-one surgeons responded for a weighted response rate of 53%. For an 87-year-old patient with fulminant colitis and advanced dementia or stage IV lung cancer, 40% of surgeons were inclined to offer an operation to remove the patient’s colon while 60% were inclined to offer comfort-focused care only. Surgeons were more likely to offer surgery when an operating room was readily available (OR 4.05, p<.001) and the family requests “do everything” (OR 2.18, p<.001).

Conclusion:

Factors outside the surgeon’s control contribute to non-beneficial surgery, consistent with our model of clinical momentum. Further characterization of the systems in which these decisions occur might expose novel strategies to improve serious illness care for older patients and their families.

Mini-Abstract:

In this national survey of surgeons, surgeons report factors outside their control, such as availability of an operating room and a family’s request to “do everything,” influence their decision to offer non-beneficial surgery. Such findings are consistent with our model of clinical momentum. Efforts to improve overtreatment should include understanding of and intervention on systems factors, alongside communication interventions.

Introduction

Approximately one third of Medicare beneficiaries will have a surgical procedure in the year before death.¹ Overtreatment at the end of life is a problem in part because it exposes patients to burdensome treatments, i.e. pain from surgery, when most Americans would prefer a comfort-focused strategy.^2,3 This unwanted care is inconsistent with patient’s wishes and has significant financial burdens and healthcare costs.^4,5 Resources may be better spent on high quality end-of-life care, especially for patients for whom surgery has limited benefit.^6,7 Provision of non-beneficial surgery is often attributed to ineffective communication.⁸ In theory, clarifying options and outcomes while incorporating patients’ overall health goals would reduce non-beneficial surgery. Multiple attempts to improve communication for people with life-limiting illnesses have been evaluated.^9–11 Despite this, few studies have shown that improving communication will yield major reductions in overtreatment or improve end-of-life care.^12,13

In focus groups with surgeons about major surgical decisions, respondents invoked systems factors – e.g., expectations from consulting physicians and family members, inadequate time for decision making – that generated pressure to intervene when they believed surgery was not in the patient’s best interest. They noted an acute surgical diagnosis starts a trajectory of care towards surgical intervention that is difficult to stop.¹⁴ Our research team developed a conceptual model to describe these forces that promote overtreatment at the end of life.¹⁵ We call this phenomenon clinical momentum – an accumulation of events and responses in a system primed for intervention (Figure 1).^14,15 The model draws from work in medical decision-making and behavioral psychology.¹⁵

Figure 1:

Conceptual framework of clinical momentum snowballing over time in a surgical setting.¹⁵ Recognition-primed decision making allows an abnormal symptom to prompt reflexive management with the appropriate treatment. As additional abnormalities are identified, they are treated with subsequent interventions, creating new diagnoses and leading to more interventions, whereby a cascade ensues. When a surgical problem is identified, it prompts surgical consultation. Clinicians use the “fix-it” model to describe a problem that can be fixed to restore normalcy without considering the patient’s overall health and trajectory; patients and families start to view surgery as a fix to all problems. Once enough problems and subsequent interventions arise, sunk cost effects make it difficult consider an alternate pathway, and surgeons feel compelled to operate even when they think surgery is not in the patient’s best interest. As momentum accumulates over time (x-axis), the probability of a desirable outcome decreases (y-axis).

Momentum starts when a problem is identified. Confronted with abnormal labs or vital signs, the clinician recognizes patterns and reflexively applies the relevant treatment.¹⁶ In decision psychology, this is known as “recognition-primed decision-making.” This valuable skill for emergent, time-sensitive settings, leaves little room to incorporate patient preferences.¹⁶ Once an abnormality is identified, clinicians intervene to restore normalcy; previously described as the “fix-it” model.¹⁷ Clinicians also use this model to discuss treatment with patients, describing it as a correction for an isolated abnormality, without contextualizing the disease or intervention within the patient’s overall health trajectory.¹⁸ Each intervention or abnormality can produce “cascade effects,” allowing subsequent interventions to ensue without stopping to consider alternative pathways.¹⁹ When enough time, energy, and resources are invested, “sunk costs” explain a commitment to continuing.²⁰ These components are guided by clinical practice norms, i.e., what is “usually done,” and frequently reinforced by hospital policies and procedures.²¹

We theorize systems factors contribute to non-beneficial surgical intervention in seriously ill older adults. Using a large national survey, we sought to test the effect of and identify associations between these factors and surgeons’ willingness to offer surgery.

Methods

Survey Design

We conducted focus groups and interviews with surgeons about high-stakes surgical decisions.^8,14 Participants identified patient and system-level factors contributing to non-beneficial surgery near the end of life including concerns about prognostication and end-of-life communication, managing patient and family emotions, availability of palliative care, and a clinical momentum that promotes surgical intervention despite limited treatment benefit. We generated a survey with fixed-choice questions to assess the prevalence and generalizability of these factors identified in our qualitative studies.

The survey includes a clinical vignette wherein consultation is requested for a hypothetical 87-year-old patient in the intensive care unit with fulminant Clostridium difficile colitis (Appendix 1).⁸ We aimed to test the effect of two factors on surgeons’ decision to offer surgery: patient comorbidity and systems factors aligned with our model of clinical momentum. For patient comorbidity, we tested stage IV lung cancer against advanced Alzheimer’s disease. Both conditions are considered terminal illnesses; however, they may have different impacts on the likelihood of surgeons offering comfort-focused care. For systems factors, we tested the effect of clinical momentum by including or omitting a statement that the referring intensivist believes surgery is indicated. We used a 2 by 2 factorial design to generate four versions of the vignette and randomized and distributed these evenly among the sample (Table 1).

Table 1.

2×2 factorial design to test patient factors and systemic factors embedded in the vignette on surgeon’s inclination to offer operative treatment to a patient with an acute surgical condition

		System Factor: Intensivist Endorses Surgery
Patient Factor: Comorbidity		No Intensivist Endorsement	Intensivist Endorsement
	Alzheimer’s disease	Alzheimer’s disease + No Endorsement	Alzheimer’s disease + Endorsement
	Stage IV lung cancer	Lung cancer + No endorsement	Lung cancer + Endorsement

We evaluated the face validity of all questions in an iterative fashion using cognitive interviewing with 10 attending surgeons from surgical subspecialties including oncology, trauma surgery, minimally invasive surgery, and endocrine surgery. We queried respondents regarding their interpretation of each question and solicited feedback on the organization of the survey. We modified question content after each interview and re-tested it to ensure interpretation of all questions was aligned with the original intent.

The final survey included 10 primary domains with 45 sub-questions about the surgeon’s treatment plan and the importance of various patient, surgeon, and system-level factors on the decision to offer surgery (Table 2). This includes factors embedded in the vignette and direct questions about the likelihood of specific factors influencing their response in the clinical vignette. We could not directly test the model of clinical momentum in a survey because it represents latent properties of clinical care, which are difficult to appreciate by the people in the system. Instead, we used systems factors identified by surgeons in focus groups that fit the model well and incorporates several components of the model. We used a broad definition of systems factors to encompass factors outside the surgeon’s control ranging from a family’s request to do everything to intensivist referrals (Table 2). We used a 4-point Likert scale for each response frame. Demographic information included years in practice, geographic region, and volume of older patients seen in the surgeon’s practice.

Table 2.

System-level factors tested in the survey and how it relates to the clinical momentum model

Systems-level factors tested in the survey	Relationship to clinical momentum model
Availability of the on-call team and OR (Q3a)	Cascade effects
Time required to discuss non-operative treatments (Q3b)	Recognition-primed decision making, cascade effects
Pressure from your department/hospital to increase operative volume (Q3c)	Patterns of usual care, “fix-it” model
Availability of palliative care resources (Q3d)	Patterns of usual care, cascade effects
Family’s request to “do everything” (Q4g)	“Fix-it” model, cascade effects, sunk costs effects
Intensivist endorses surgery (clinical vignette)	“Fix-it” model

Data Collection

We distributed two rounds of surveys to randomly selected Fellows of the American College of Surgeons (ACS). The first round was sent in February 2017 to a random sample of 2,800 surgeons and mistakenly had a substantial age bias. A second round of surveys was sent in February of 2018 to a random sample of 2,400 surgeons. The second round included two additional questions: “How frequently are you asked to perform non-beneficial surgery on seriously ill, older patients with acute surgical conditions?” and, “How many non-beneficial surgeries of this type do you perform per year?” Otherwise, the survey questions in each round were identical.

Each mailing contained a stamped return envelope and a $2.50 LED-lighted pen for incentive. We sent additional mailings via US mail and electronic mail to non-responders, with a $5 gift card. We then used internet search to identify publicly available contact information for remaining non-responders. Survey data were collected and managed using REDCap, a web-based data capture tool.²²

Analysis

We analyzed the combined results of surveys from 2017 and 2018. Given pandemic related delays, the analysis occurred between June 2021 and June 2022. Instead of discarding first round results, we weighted answers from all respondents to match the age and demographics of surgeons from the ACS member database. All analyses used the weighted sample.

We used the American Association for Public Opinion Research standard definitions to calculate the weighted adjusted response rate (ARR). The ARR = I/ [I+e(T-R)-NE], where R is eligible respondents, e is the estimated proportion of eligible non-respondents, T is the total number of surveys distributed, and NE is the number of ineligible respondents. We estimated e=(s+n)/(s+n+i), where s is the number of eligible respondents who completed the survey, n is the number of eligible respondents who did not complete the survey, and I is the number of ineligible respondents.²³ The ARR was weighted using the number of surveys distributed in each round.

Responses to survey questions containing a 4-point Likert scale were dichotomized by combining “Very likely” with “Somewhat likely” and “Somewhat unlikely” with “not at all likely.” Respondents who indicated they were “somewhat likely or very likely” to offer surgery and “somewhat unlikely or not at all likely” to offer comfort care based on their clinical vignette were designated as surgeons who would offer surgical treatment. Respondents who indicated they were “somewhat likely or very likely” to offer comfort care and “somewhat unlikely or not at all likely” to offer surgery were designated as surgeons who would offer comfort care. All respondents were included in the descriptive analysis, which included surgeon characteristics and their perspectives on decision-making and preoperative communication (Tables 3 and 4). Our primary outcome was surgeons’ decision to offer surgery over comfort care. Surgeons who indicated “somewhat likely or very likely” or “somewhat unlikely or not at all likely” to both questions were not included in the bivariate and multivariable analyses of this outcome given ambiguity about whether they would offer surgery or comfort care (Table 5 and 6).

Table 3:

Respondent characteristics (n=2044)^a

Characteristic	n (%)
Male	1571 (77)
Years in practice
≤10	743 (36)
11–20	584 (29)
21–30	490 (24)
>30	218 (11)
Practice setting
Private practice	769 (38)
Academic practice	734 (36)
Private practice with academic affiliation	315 (15)
Other	217 (11)
Geographic location
Northeast	440 (22)
Midwest	447 (22)
South	717 (35)
West	425 (21)
Territories	7 (<1)
No. of patients age ≥65 cared for each month
<6	90 (4)
6–10	220 (11)
11–19	442 (22)
>20	1283 (63)

^aThe unweighted number of respondents is 2044; the weighted number of respondents is 2036.

Table 4:

Surgeons’ perspectives on decision making and preoperative communication for a hypothetical 87-year-old patient with either end-stage dementia or stage IV lung cancer with Clostridium difficile associated toxic megacolon (n=1945)^a

For this patient, how confident are you in your ability to:	Somewhat confident or very confident No. (%)b	A little confident or not at all confident No. (%)
Assess surgical risk, including estimated survival	1893 (95)	101 (5)
Describe poor prognosis, including death	1986 (99)	9 (1)
Direct the conversation towards the treatment decision you feel is in the patient’s best interest	1926 (97)	69 (3)
How much do you agree with the following statements?	Somewhat or very much No. (%)	A little or not at all No. (%)
As a surgeon, my job is to offer surgery and let the patient or family choose	727 (36)	1268 (64)
As a surgeon, I am ultimately in control of preventing non-beneficial surgery	1761 (88)	234 (12)
Over-treatment is a serious problem in this country for patients near the end of life	1885 (95)	110 (5)
If this were your patient, how likely would you be to do each of the following:	Somewhat likely or very likely No. (%)	Somewhat unlikely or not at all likely No. (%)
Offer this patient an operation to remove her colon	754 (38)	1241 (62)
Offer a choice between surgery and comfort care	1366 (68)	629 (32)
Offer comfort care only	1031 (52)	963 (48)
Request a palliative care consult	1640 (82)	355 (18)
Request a second opinion with another surgeon	381 (19)	1614 (81)

^aThe unweighted response to survey questions in this table is 1945 respondents; the weighted response is 1995 respondents.

^bPercentages are calculated by row throughout the table.

Table 5:

Effect of the clinical vignette on surgeons’ decision to offer surgery^a

Characteristic	Somewhat likely or very likely to offer surgery n = 589 No. (%)b	Somewhat likely or very likely to offer comfort care n = 870 No. (%)	P-value
Clinical vignette
Dementia only	128 (32)	273 (68)
Dementia, intensivist endorses surgery	153 (43)	207 (57)	.07
Lung cancer only	146 (43)	197 (57)
Lung cancer, intensivist endorses surgery	163 (46)	193 (54)
Patient comorbidity
Dementia	281 (37)	480 (63)	.07
Lung cancer	309 (44)	390 (56)
Intensivist presence
Intensivist	316 (44)	400 (56)	.07
No Intensivist	273 (37)	470 (63)

^aThe unweighted number of respondents who answered the question is 1500. The weighted number of respondents is 1460.

^bPercentages are calculated by row.

Table 6.

Bivariate and multivariable analyses of survey factors associated with the decision to offer surgical intervention for older adults with life-limiting acute surgical conditions (n=1410)^b

Patient or Systemic Factor	Would offer surgery n= 560 (%)d	Would offer comfort care n= 812 (%)d	P-value	Adjusted odds ratioc (95% CI)	P-value
Availability of OR time and personnel
Important	215 (38)	120 (15)	<.001a	4.05 (2.48–6.61)	<.001a
Time required to discuss non-operative treatments
Important	195 (35)	220 (27)	.05a	0.82 (051–1.31)	.42
Family member requests “do everything”
Important	422 (75)	469 (58)	<.001a	2.18 (1.42–3.35)	<.001a
Desire to preserve referral relationships
Important	131 (23)	178 (22)	.69	0.82 (0.15–1.34)	.45
Surgeons control receipt of non-beneficial surgery
Surgeons are in control	472 (84)	741 (91)	.01a	0.56 (0.32–0.97)	.04a
Surgeon specialty
General	459 (82)	577 (71)		ref
Oncology	61 (11)	107 (13)	.001a	0.67 (0.37–1.23)	.20
Vascular/Cardiac/Transplant	40 (7)	120 (15)		0.47 (0.26–0.84)	.01a
Other	1 (<1)	8 (<1)		0.04 (0.00–0.68)	.03a
Surgeon Years of Experience
≤10	216 (39)	248 (31)		ref
11–20	170 (30)	239 (29)	.04a	0.72 (0.42–1.29)	.26
21–30	131 (23)	216 (27)		0.63 (0.38–1.04)	.08
>30	43 (8)	109 (13)		0.37 (0.22–0.63)	<.001a

^aRepresents statistically significant p-values <.05

^bn=1410 reflects the unweighted number of respondents who answered all the questions in this table. The weighted total can be added by row.

^cMultivariable logistic regression was used to model the odds that a surgeon indicated they would offer surgery only and not comfort care by each survey factor, while adjusting for surgeon practice location and gender

^dPercentages are calculated by column

We computed descriptive statistics by grouping complete responses within the domains of the survey to maximize sample size. For example, surgeons who answered all questions related to the clinical vignette were included in analysis examining factors embedded in the clinical vignette with the decision to offer surgery. We used chi-square tests for all bivariate analyses. We used multivariable logistic regression models to examine the association of patient, surgeon, and system factors with the odds of offering surgery while adjusting for surgeon practice, location, and gender. Stata Version 14 (StataCorp, College Station, TX) was used for all analyses and p-values <.05 were considered statistically significant.

Results

An unweighted total of 2,161 surgeons responded to the survey with a weight adjusted response rate of 53%. The majority, 77%, were male. They were evenly dispersed throughout the country with varying years in practice (Table 3). Surgeons’ practice setting included private practice (38%), academic practice (36%), and a combination of both (15%). More than half of survey respondents reported caring for 20 or more patients who are 65 years or older each month.

Nearly all surgeons felt confident in their abilities to assess surgical risk (95%), describe poor prognosis (99%), and steer the conversation towards the treatment decision they felt was in the patient’s best interest (97%) (Table 4). Although 95% agreed that over-treatment is a serious problem in this country for patients near end of life, 88% felt they were in control of preventing non-beneficial surgery. Thirty-six percent agreed or strongly agreed with the statement their “job is to offer surgery and let the patient and/or family choose.”

We evaluated the weighted response of 1,995 surgeons responding to the clinical vignette. Among 3 distinct question prompts, 38% of respondents were somewhat likely or very likely to offer this patient an operation while 68% were likely to offer a choice between surgery and comfort care, and just over half (52%) were likely to offer comfort care only. When we excluded 544 responses due to ambiguity, specifically, they would simultaneously offer surgery and comfort care only, 40% of surgeons would offer surgery while 60% of surgeons would offer comfort-care only. Eighty-two percent of surgeons were likely to request a palliative care consult and only 19% were likely to request a second opinion with another surgeon (Table 4). In bivariate analysis, we saw no statistically significant effect of factors randomly embedded in the vignette on the surgeon’s propensity to offer surgery. Surgeons who were told an intensivist believed the patient’s problem was surgical were not more likely to offer surgery (44% v. 56%, p=.07) (Table 5). Of those who indicated they would offer surgery, 48% had a vignette with a dementia patient and 52% had a vignette with a lung cancer patient (p=.07). When controlling for other factors, surgeons were more likely to offer surgery to patients with stage IV lung cancer compared to patients with dementia (Odds Ratio (OR)=1.54, 95% CI (1.06–2.21)). Intensivist endorsement of a surgical problem did not increase the likelihood of offering surgery (OR=1.31, CI (0.09–1.90)).

In bivariate analysis of surgeon-reported factors, more surgeons who would offer surgery considered family members’ requests to “do everything” an important consideration in their decision making compared to surgeons who indicated they would offer comfort care (75% v. 58%, p<.001). We found a statistically significant association between a surgeons’ decision to offer surgery and the availability of operating room time and personnel surgery compared to those who would offer comfort care (38% v. 15%, p<.001) (Table 6).

In multivariable analysis, we found a persistent association between surgeons offering surgery and their report that a family member’s request to “do everything” was important (OR=2.18, 95% CI (1.42–3.35)). Additionally, a significant association existed between availability of an operating room and a surgeons’ decision to offer surgery (OR=4.05, 95% CI (2.48–6.61)). If surgeons felt in control over non-beneficial surgery, they were less likely to offer surgery (OR=0.56, 95% CI (0.32–0.95)). Surgeons with >30 years of experience were less likely to offer surgery than those with ≤10 years of experience (OR=0.37, 95% CI (0.22–0.63)).

Discussion

When posed with a question about offering urgent surgery for a terminally ill 87-year-old woman, 40% of surgeons would offer surgery. Although the vast majority of surgeons reported overtreatment is a serious problem for patients near the end of life and favored engaging palliative care, only 60% endorsed offering comfort care. Despite consensus that surgeons are in control of preventing non-beneficial surgery, it is curious that many offered surgery for this seriously ill patient with high mortality and grave comorbidities. Respondents offering surgery were more likely to endorse factors influencing this decision such as the ready availability of an operating room and a family member’s request to “do everything.” These findings highlight a disconnect between the surgeon’s assessment of the patient’s best interest and willingness to offer intervention that conflicts with this assessment. It is possible some believed the chance of days to weeks of life prolongation was beneficial. Nonetheless our previous work⁸ and enthusiasm for comfort care reported herein suggest this explanation is unlikely. Instead, factors perceived as outside their control appear to prompt surgeons to offer surgery they believe is non-beneficial, suggesting the systems in which these choices are offered needs further examination.

Surgeons, working in a system primed for intervention, might find it difficult to disrupt momentum built over the course of a patient’s hospitalization. Our vignette describes a familiar situation characterized by clinical momentum: after a series of events starting with a urinary tract infection, the surgeon is consulted for a surgical emergency to fix-it, as if the patient has one singular problem and surgery will restore her to health.^17,24 Because clinicians commonly use a fix-it narrative to describe treatment, families see each new problem as a discrete physiologic aberration.²⁵ Oftentimes, someone other than the surgeon will call the OR with good intentions to expedite care. Surgeons have previously described this as a factor that pushes them to operate against their judgment, reflecting the fast pace of clinical care whereby it is easier to continue the path towards intervention than to enter a challenging conversation to disrupt the building momentum.¹⁴ Moreover, deliberation about surgery typically focuses on risks and operative endpoints, e.g. 30-day survival, not a broader discussion of the patient’s overall health.^17,18,26 While the fix-it model is useful for healthy patients with appendicitis, it is misleading for patients with life-limiting illness.^15,17,18 Toxic megacolon is a culmination of events related to the overall health of the patient. While colectomy might prolong her life for a few days, surgery will not change her prognosis; a fact lost when colectomy is considered in isolation.

Offering surgery as a choice suggests the surgeon believes surgery is a reasonable, but risky, option, which is categorically different than non-beneficial.^27–29 Although surgeons likely offer choices to support autonomy, it deprives patients and families of the surgeon’s expertise, better expressed as “I’m worried surgery is not a good idea.”^29–32 Given the fast pace of acute care and the number of clinicians involved, surgeons will need stronger communication skills to disrupt this momentum (Appendix 2). Surgeons can start by intentionally communicating a clear change.^33,34 This strategy, called reframing, signals an acute shift and is designed to reorient patients and family, using phrases like, “things are different now.”^8,15,34,35 Avoiding words like fix-it seems obvious, yet surgeons use other words with similar implications – remove, repair, and treat.³⁶ Surgeons would do better to address the goals of intervention and whether surgery can meet those goals. For example, “Surgery might extend her life for days to weeks.”¹⁸

For patients and families, a request “to do everything” may not result in care that is desirable. Because this request is rarely explored, surgeons will interpret “do everything” as a request for surgery and continued life-sustaining treatment, without considering that “everything” ranges from “everything that might provide maximum relief of suffering, even if it might unintentionally shorten life” to “everything that has any possible potential to prolong life even a small amount, regardless of the patient’s suffering.”³⁷ Families often request “do everything” because they fear their loved one will be neglected or anything other than maximal treatment is not “good” care.^28,37 Asking for everything reflects a system that venerates intervention. Although patients and families generally prefer comfort over prolonged life-sustaining treatment at the end of life,^3,4,38,39 when clinicians ask, “Do you want us to do everything?” they say yes to signal they have not abandoned their loved one. Instead, surgeons might explore the request to operate, “What are you hoping surgery might do for her?”

For policy makers, the pace of acute care characterized by clinical momentum, provided by multiple teams, each charged with managing a different problem, can generate care detached from the patient’s overall circumstances. Hospital protocols are designed to prioritize efficiency,^16,31 e.g., a ready OR, and fee-for-service payments incentivize operating over meaningful discussion.¹³ Moreover, the pressure to discharge patients exposes older adults with life-limiting illness to the revolving door between acute and post-acute care, despite poor rehabilitation potential.^40,41 For patients who survive in the short term, extending biological life for days or weeks, a dependence on institutions could separate them from things that made life worth living.^37,40 Better end-of-life care will require systems-level interventions.^5,16 Dis-incentivizing procedural care, decision support to flag patients with short-term mortality, and initiatives like the Geriatric Surgery Verification program with a clear requirement to document the “goals” of surgery, e.g., “prolong life”, rather than a reductionist “remove the colon,” would all support improved care for older adults with life-limiting illnesses.⁴²

Our survey has several limitations. Even with a strong conceptual model and robust preliminary data, it is difficult to test underlying features of clinical care with a survey. We embedded factors from our model in the vignette, i.e., an intensivist endorsing surgery that were subtle to mirror the latent phenomenon of clinical momentum. Yet this may have been too refined for respondents to perceive. Moreover, surveys inherently struggle with external validity between respondent report and what occurs in clinical practice. Future studies could utilize direct observation with systems engineering to characterize system factors in real time. Finally, survey data were collected pre-pandemic. Given limited enthusiasm to redirect resources for patients at the end-of-life during times of extreme scarcity,^43,44 we suspect clinical practice has not changed.

Conclusion

Although surgeons view overtreatment at the end-of-life as a serious problem, more than one-third of respondents in our national survey would offer surgery to a seriously ill patient with an acute surgical problem. Factors outside the surgeon’s control, such as an available OR and a request to “do everything,” were associated with offering surgery, suggesting the need to evaluate the system in which these events occur.