What if We Don’t Operate? Outcomes of Nonoperatively Managed Emergency General Surgery Patients

Abstract

Background:

Although two-thirds of patients with emergency general surgery (EGS) conditions are managed nonoperatively, their long-term outcomes are not well described. We describe outcomes of nonoperative management (NOM) in a cohort of older EGS patients and estimate the projected risk of operative management (OM) using the NSQIP Surgical Risk Calculator (SRC).

Materials and Methods:

We studied single-center inpatients aged 65 and older with an EGS consult who not undergo an operation (1/2019–12/2020). For each patient, we recorded the surgeon’s recommendation as either an operation was “ Not Needed” (medical management preferred) or “Not Recommended” (risk outweighed benefits). Our main outcome of interest was mortality at 30 days and 1 year. Our secondary outcome of interest was SRC-projected 30-day postoperative mortality risk (median % [IQR]), calculated using hypothetical low-risk and high-risk operations.

Results:

We included 204 patients (60% female, median age 75), for whom an operation was “Not Needed” in 81% and “Not Recommended” in 19%. In this cohort, 11% died at 30 days and 23% died at 1 year. Mortality was higher for the “Not Recommended” cohort (37% vs 5% at 30 days and 53% vs 16% at 1 year, p<0.05). The SRC-projected 30-day postoperative mortality risk was 3.7% [1.3–8.7] for low-risk and 5.8% [2–11.8] for high-risk operations.

Conclusions:

NOM in older EGS patients is associated with very high risk of short and long-term mortality, particularly if a surgeon advised that risks of surgery outweighed benefits. The SRC may underestimate risk in the highest-risk patients.

Introduction

Patients over 65 years old represent a disproportionately large share of the emergency general surgery (EGS) disease burden in the United States. Nearly one million older patients are admitted annually with EGS diseases, and approximately 70% are managed nonoperatively.(1) Older EGS patients have unique needs that must be addressed as part of their care, including mobility, functional dependence, cognitive function, decision-making capacity, complex multimorbidity, and frailty. These issues may complicate their course of treatment, and are associated with higher risk when considering operative management.(2–5) Older patients may also present in more complicated stages of disease, and as a consequence may be more likely to fail non-operative management, suggesting that aggressive intervention could be warranted despite the higher risk.(6–9) Clinical decision-making in these cases is complex, and the accurate prediction of older patients’ risk of post-operative adverse outcomes plays a large role in weighing operative vs. non-operative management. Unfortunately, high-quality prognostication methods for nonoperatively managed patients is limited.

The most common prognostication method used for surgical patients is the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) Surgical Risk Calculator (SRC). The SRC uses pre-operative patient factors to estimate the risk of mortality, readmission, reoperation, and functional decline.(10) Several studies have supported its accuracy in EGS patients, although some reports suggest the prediction accuracy is lower in older patients.(11–15) Additional modifications have been made to the SRC to address specific needs and concerns in the older patient.(16, 17) Unfortunately, the SRC has a few major limitations in this patient population: outcomes are limited to 30 days, non-operatively managed patients are not included in the calculation of SRC algorithms, and the anticipated operative procedure must be known for calculation, which is not always the case in EGS diseases. For the most complex patients, estimation of outcomes for both operative and nonoperative strategies is imperative for shared decision-making.(18, 19) Unfortunately, there are no reliable methods to prognosticate outcomes in patients who are treated nonoperatively.

The primary aim of our study was to explore health profiles and outcomes for older EGS patients who were treated with non-operative management (NOM) and compare their outcomes to our best projection of operative management (OM) outcomes using the NSQIP SRC. We hypothesized that the observed mortality rates would be higher than those projected by the SRC, possibly identifying limitation for using the SRC to prognosticate for some of our sickest patients. Our secondary aim was to explore rates of palliative care consultation rates in our population and hypothesized that palliative care would be underutilized.

Methods

Patient Inclusion

We conducted a retrospective chart review of inpatients aged 65 and older at a tertiary academic medical center who were seen by or admitted to the EGS service between January 2019 and December 2020. Our intention was to capture all patients for whom an urgent surgical evaluation was requested but no operation was performed. Inpatient encounters were identified by the presence of any note written or cosigned by the EGS service attending during the inpatient admission. In our hospital, the same attending surgeon pool covers EGS, trauma, and surgical/trauma intensive care units; any patient who had a concurrent trauma admission or who was administered critical care services but did not need an evaluation for a possible operation was excluded. We also excluded patients whose surgical care and decision-making was provided by a service other than EGS, including patients who were managed by surgical oncology, trauma surgery, colorectal surgery. We also excluded patient encounters for those who had already received an operative intervention and had a consultation for follow-up care. For patients with more than one inpatient encounter in our inclusion period, we selected their first encounter.

Variable Inclusion: EGS Diagnoses and Variable Collection

EGS consult questions were grouped into categories shown in the Supplementary Table. Additional patient factors were collected from the chart for use in the SRC. These variables included patient comorbidities, baseline functional and cognitive status, age, biological sex, ASA class, body mass index, the presence of sepsis, acute kidney injury, ventilator dependence, disseminated cancer, diabetes, hypertension requiring medication, congestive heart failure, severe COPD, and renal failure requiring dialysis. We also recorded whether the patient acted as their own decision-maker, or whether a proxy healthcare agent was involved. We identified whether a palliative care consult was obtained during the acute hospitalization.

We then reviewed the surgeon’s notes regarding their recommendation for non-operative treatment and categorized their determination as “OM Not Needed” or “OM Not Recommended.” Patients were categorized as “OM Not Needed” if it was documented that there was no clinical indication for an operation. The “OM Not Needed” group generally included patients whose presenting symptoms resolved without operative intervention, whose acute disease was suspected to be secondary to a medical cause, and patients who were treated preferentially with alternative non-operative strategies (such as percutaneous drains or antibiotics). Patients categorized as “OM Not Recommended” required documentation by the EGS attending that an operation was not being offered due to severe comorbidities, severe acute medical disease that increased risk of perioperative complications to an unacceptably high level, or because any proposed procedure was deemed to be futile. In rare cases, the surgeon documented that even with no “acute need for an operation,” the patient’s clinical risk was so high that no operation would be offered even if the patient clinically deteriorated. These patients were classified as “OM Not Recommended.”

ACS NSQIP Surgical Risk Calculator Prognostication

To determine prognostic projections from the ACS NSQIP SRC, a surgical procedure must be entered. However, in EGS, the final operative procedure is nearly always unknown when decision-making about whether surgery should be offered occurs. Even in “simple” diseases such as acute cholecystitis, there is potential for “possible” additional procedures within a surgery, which is manifested in the consent process as “laparoscopic cholecystectomy, possible open, possible cholangiogram,” and in the worst cases could include bile duct explorations or even a hepaticojejunostomy with a Roux-en-Y reconstruction. Smaller operations tend to carry lower risk while larger and longer operations tend to be higher risk. To account for this possible range of risk, we included a range of possible procedures for our projected postoperative mortality projection. To address this uncertainty, all patients in the cohort were assigned a hypothetical “low risk” and “high risk” Current Procedural Terminology (CPT) procedure code for use in the SRC. The “low risk” procedures represented less invasive procedures to treat the diagnosis in question, while the “high risk” procedures were generally more invasive. For example, in acute cholecystitis, the “low risk” option was a laparoscopic cholecystectomy, while the “high risk” option was an open cholecystectomy. For some diagnoses for which there was only one possible operative procedure (for example: an abscess, which would be treated by incision and drainage), the procedure was classified as “low-risk”. Examples of low-risk and high-risk operations that were utilized are presented in Table 1.

Table 1:

Examples of “Low-Risk” and “High-Risk” Procedures Used in Surgical Risk Calculation

Consult Question Type	“Low-Risk” Procedure	“High-Risk” Procedure
Acute Appendicitis	Laparoscopic Appendectomy	Open Appendectomy
Acute Cholecystitis	Laparoscopic Cholecystectomy	Open Cholecystectomy
Perforated Ulcer	Gastrorrhaphy, suture of perforated duodenal or gastric ulcer, wound or injury	Partial, Distal Gastrectomy with Vagotomy
Small Bowel Obstruction	Lysis of Adhesions	Small Bowel Resection

All factors obtained from chart review for each patient were then entered in the SRC along with the assigned “low risk” and “high risk” procedure code for that patient’s diagnosis. The SRC calculated risk of readmission, 30-day mortality, discharge to SNF, functional decline, new mobility aid use, and length of stay; these were recorded for both the “low risk” and “high risk” procedures for each patient.

Outcomes

The primary outcomes of our study were observed mortality at 30 days and at 1 year. Secondary outcomes of interest included length of stay (LOS), presence of a palliative care consultation, readmission, discharge to SNF, and whether the patient later underwent an operation to treat the disease for which the initial consult was requested. Data were analyzed in a priori determined groups including our main comparison (“OM Not Needed” vs. “OM Not Recommended”), as well as the subgroup of patients for whom palliative care was utilized.

Analysis

We first performed a descriptive comparison between patients with “OM Not Needed” and “OM Not Recommended” to assess differences in patient factors and outcomes. We included demographic factors such as age and sex, along with other factors such as number of comorbidities and functional status. Data are reported as n (%) for categorical variables and median (interquartile range) for continuous variables. We used a chi-square test to compare categorical variables. If groups were small (<5), then a Fisher’s exact test was used. A Wilcoxon rank-sum (Mann-Whitney two-sample statistic) was used to compare continuous variables. All calculations were performed using Microsoft Excel (Redmond, WA) or Stata IC 16.0 (StataCorp, College Station, TX).

We then described, for the entire population, along with the OM Not Needed, OM Not Recommended, and palliative care subgroups, the projected risk of 30-day perioperative mortality of “low risk” and “high risk” operative management using predicted outcomes from the SRC, alongside the observed outcomes of our nonoperatively managed patient cohort. No statistical comparisons were performed for this descriptive evaluation.

This study was approved by our hospital’s Institutional Review Board which included a waiver of consent.

Results

We identified 2242 patient encounters with a note written by the EGS service; 573 encounters occurred in patients who were aged 65 or older. Patients who had an operation during their admission, were discharged from the emergency department, transferred to another facility from the emergency department, or left against medical advice were then excluded, leaving 286 encounters. After application of all exclusion criteria, including duplicate admissions and patients treated by other surgical services, a pool of 204 patients who were managed without an operation were included in our final cohort.

Our patient cohort is described in Table 2. The median age was 75 years (IQR 68–83) and 123 patients (60.3%) were female. The median number of comorbidities was 7 (IQR 5–11) and 88 patients (43.1%) were functionally dependent prior to admission. Most patients served as their own medical decision maker (84%). The reason for NOM was recorded as “OM Not Needed” in 166 patients (81.4%) and “OM Not Recommended” in 38 patients (18.6%). The patient or their health care agent declined operative management in 11 cases (5.4%), 6 of whom were in the OM Not Recommended group. The most common reasons for EGS consult were acute cholecystitis (33 (16.2%)) and small bowel obstruction (57 (27.9%)); the distribution of patients within each consult category is presented in a Supplementary Table. The median length of stay was 5 days (IQR 3–9), and 32 patients (15.7%) received a palliative care consultation during the admission. At the time of discharge, 137 patients (67.2%) were discharged home, 47 (23.0%) were discharged to SNF, and 5 (2.5%) were discharged to hospice. The overall 30-day mortality was 10.8% and 1-year mortality was 23.0%. Thirty-three patients (16.2%) were readmitted, with a median of 46 days (IQR 17–94) to readmission, and 16 (7.8%) had a later operation to treat the disease for which EGS was originally consulted, with a median of 70 days (IQR 29–109) to the eventual operation.

Table 2:

Patient Demographics and Outcomes

Study Population	Total N = 204	Operative Management Not Needed N=166	Operative Management Not Recommended N=38	Palliative Care Group** N = 32	p-value*
Patient Characteristics
Age, years, median (IQR)	75 (68–83)	74 (68–83)	78 (69–83)	76 (70–85)	0.46
Female, n (%)	123 (60.3)	103 (62.1)	20 (52.6)	19 (59)	0.29
Functional Status Prior to Admission, n (%)					0.07
Independent	116 (56.9)	100 (60.2)	16 (42.1)	9 (28.1)
Partially Dependent	61 (29.9)	44 (26.5)	17 (44.7)	13 (40.6)
Totally Dependent	27 (13.2)	22 (13.3)	5 (13.2)	10 (31.3)
Number of Comorbidities, median (IQR)	7 (5–11)	7.5 (5–11)	6 (5–10)	7 (6–11)	0.16
Patient was Own Decisionmaker, n (%)	172 (84.3)	143 (86.1)	29 (76.3)	17 (53.1)	0.13
Outcomes
Discharge Disposition, n (%)					<0.001 +
Died in the Hospital	15 (7.4)	6 (3.6)	9 (23.7)	14 (43.8)
Discharged to Home	137 (67.2)	123 (74.1)	14 (36.8)	5 (15.6)
Discharged to SNF	47 (23.0)	36 (21.7)	11 (29.0)	9 (28.1)
Discharged to Hospice	5 (2.5)	1 (0.6)	4 (10.5)	4 (12.5)
Palliative Care Consult, n (%)	32 (15.7)	17 (10.2)	15 (39.5)	32 (100)	<0.001
Length of Stay, days, median (IQR)	5 (3–9)	5 (3–8)	6 (3–12)	9 (4–20)	0.32
30-day Mortality, n (%)	22 (10.8)	8 (4.8)	14 (36.8)	16 (50)	<0.001
1-year Mortality, n (%)	47 (23.0)	27 (16.3)	20 (52.6)	23 (71.9)	<0.001
Later Readmitted, n (%)	33 (16.2)	28 (16.9)	5 (13.2)	0 (0.0)	0.58
Delayed Operative Management, n (%)	16 (7.8)	12 (7.2)	4 (10.5)	0 (0.0)	0.50

^*P-value is calculated between Operative Management Not Needed and Not Recommended groups. Chi-square test used to compare categorical variables; Wilcoxon rank sum test used to compare continuous variables, unless otherwise noted.

⁺Fisher’s exact test used

^**Patients who received palliative care consultations came from both subgroups of Operative Management Not Needed and Not Recommended.

Patients in the “OM Not Needed” and “OM Not Recommended” groups were of similar age and had similar comorbidity counts (Table 2). The most common comorbid conditions in the “OM Not Needed” group included hypertension (84.9%), hyperlipidemia (56.0%), cancer (34.3%), and chronic obstructive pulmonary disease (COPD, 30.1%); the most common conditions in the “OM Not Recommended” group included hypertension (68.4%), coronary artery disease (39.5%), COPD (39.5%), and arrhythmia (39.5%). Fewer patients in the “OM Not Recommended” group were functionally independent compared with “OM Not Needed” patients, but this did not reach statistical significance (42% vs. 60% independent, p=0.07). Fewer patients in the “OM Not Recommended” group also served as their own decisionmakers, but this was also not statistically significant (76% vs. 86%, p=0.13). Patients for whom OM was not recommended were more likely to receive a palliative care consultation (39.5% vs. 10.2%, p<0.001) and be discharged to hospice care (10.5% vs. 0.6%, p<0.001). These patients also had significantly higher mortality at 30 days and at one year (both p<0.001).

Projected perioperative mortality using the SRC was generally lower than the observed mortality of our NOM cohort (Table 3). The “OM Not Needed” subgroup had higher 30-day mortality than SRC predicted operative outcomes (10.8% NOM vs. 2.8% (IQR 1.2–7.2) for low-risk procedures and 4.2% (IQR 1.7–9.1) for high-risk procedures); the observed readmission rate was 16.2%, which was within the projected perioperative readmission range for high-risk procedures (13.9% (IQR 10.5–17.9) for high-risk and 13.0% (IQR 9.4–15.2) for low-risk procedures). Mortality at 30 days with NOM was 36.8% for the “OM Not Recommended” group, which was very high compared with projected perioperative mortality rates of 8.8% (IQR 2.4–14.9) for a low-risk procedure and 13.1% (IQR 6.1%–25.6%) for a high-risk procedure. Readmissions for the NOM cohort were lower than the SRC predicted, however, this may be due, at least in part, to the higher mortality rate in the NOM patient group.

Table 3:

Outcomes, Operative Management Not Needed and Not Recommended

	Operative Management Not Needed N=166			Operative Management Not Recommended N=38
Outcome	Projected, “Low Risk” Operation N = 166	Projected, “High Risk” Operation N = 166	Observed, with NOM* N = 166	Projected, “Low Risk” Operation N = 38	Projected, “High Risk” Operation N = 38	Observed, With NOM* N = 38
Readmission, %, median (IQR)	13.0 (9.4–15.2)	13.9 (10.5–17.9)	16.2	13.7 (10.6–17.6)	18.7 (15.9–20.5)	13.2
30-day Mortality, %, median (IQR)	2.8 (1.2–7.1)	4.2 (1.7–9.1)	10.8	8.8 (2.4–14.9)	13.1 (6.1–25.6)	36.8
Discharge to SNF, %, median (IQR)	23.8 (13.1–45.6)	30.5 (15.4–48.0)	23.0	27.5 (9.9–53.0)	47.1 (32.0–73.1)	28.9
Length of Stay, days, median (IQR)	6 (4–8)	7 (6–9)	5 (3–9)	5 (2–10)	10 (7–15)	6 (3–11)
Functional Decline, %, median (IQR)	27.6 (7.6–46.6)	35.6 (7.7–54.4)	--	13.2 (2.3–44.4)	23.0 (4.5–61.6)	--
New Mobility Aid Use, %, median (IQR)	16.9 (9.5–26.6)	26 (20.2–40.1)	--	16.1 (7.9–26.2)	39.7 (25.2–55.0)	--

^*Outcomes on readmission, 30-day mortality, and discharge to SNF in the Observed with NOM group are displayed as prevalence observed in the study population

Thirty-two patients (15.7%) received an inpatient palliative care consult (Table 4). Of these patients, 50% died within 30 days and 72% died within one year. Comparisons to projected outcomes from the SRC in this subgroup are presented in Table 4.

Table 4:

Patients with Palliative Care Consultation, NSQIP SRC and NOM Outcomes

	Palliative Care Group N=32
Outcome	Projected, Low Risk Operation N = 32	Projected, High Risk Operation N = 32	Observed, with NOM* N = 32
Readmission, %, median (IQR)	17.3 (14.5–19.1)	18.3 (15.6–20.4)	0.0%
30-day Mortality, %, median (IQR)	10.3 (6.4–18.2)	15.6 (7.0–30.7)	50.0%
1-year Mortality, %, median (IQR)	--	--	71.8%
Discharge to SNF, %, median (IQR)	38.5 (23.1–59.9)	55.8 (33.9–74.5)	28.1%
Length of Stay, days, median (IQR)	9 (5–10)	10 (8–15)	9 (4–20)
Functional Decline, %, median (IQR)	23.6 (7.0–66.9)	12.4 (3.9–69.8)	--
New Mobility Aid Use, %, median (IQR)	17.9 (10.8–34.2)	35.3 (24.7–56.0)	--

NSQIP SRC = National Surgical Quality Improvement Program Surgical Risk Calculatior, NOM = Nonoperative Management

^*Outcomes on readmission, 30-day mortality, 1-year mortality, and discharge to SNF in the Observed with NOM group are displayed as prevalence observed in the study population

Discussion

Approximately 70% of EGS patients are treated with NOM, but the long-term outcomes of these patients is not well described.(1) In our study of nonoperatively managed patients evaluated for EGS conditions, the overall prognosis was extremely poor, as nearly one in four died by one year. For patients who were declined an operation because of high risk or because the operation would be futile, half had died by one year. In our cohort, 32 (15.7%) EGS patients had a palliative care consult during their admission; of these patients, nearly three in four died within one year, suggesting that these services are likely reserved for patients for whom death is likely imminent and treatment options were perceived to be exhausted.

Communication about the risks and expected benefits of all treatment options, including nonoperative management, is essential for surgical decision making.(20) Unfortunately, outcomes of nonoperative management are poorly described in the literature. Our best operative prognostication tool, the NSQIP SRC, derives algorithms from the operatively managed population. Our prior work demonstrated that NOM tends to be utilized in patients with higher levels of baseline frailty.(21) Because of this, we hypothesized and demonstrated that the observed outcomes of the NOM population are poorer than projected by the SRC. Another possible explanation for the poor outcomes in the NOM cohort is that NOM strategies are less likely to be curative and subsequently result in higher risk of long-term mortality. We posit that the surgeon’s intuition remains crucial in these difficult cases. The value of the provider’s intuition is demonstrated across multiple medical specialties involving the “surprise question”, which asks whether the medical team member would be surprised if a particular patient died within one year.(21, 26) This question has not been well studied among surgeons, however Lilley et. al suggest that it may serve as a potentially helpful screening tool for surgeons to examine the broader clinical picture for older patients with an increased risk of 1-year mortality.(21, 26)

The surgeon is ultimately responsible for guiding patients and their families through these difficult choices. From the patient perspective, EGS operations are viewed as a choice between life and death, with no choice but to proceed with surgery.(22) However, the decision to have an operation is not as simple as choosing between life and death. Operatively managed high-risk EGS patients are not ensured a smooth recovery, and they may be faced with accelerated mortality, persistent long-term functional decline, and loss of independence.(23, 24) Unfortunately, studies that attempt to directly compare operative to nonoperative management of EGS conditions are scarce. Nonoperatively managed EGS patients have more baseline frailty than patients selected for operative management, and any observational study will need to overcome these biases.(21) One recent study used Medicare data to match operatively and nonoperatively managed patients via an instrumental variable matching strategy to minimize bias.(25) In this study, the long-term benefit of operative management varied by specific disease type, but for some EGS diseases, particularly colorectal and upper gastrointestinal diseases, nonoperative management had superior outcomes to operative management.

For some patients, the most appropriate medical offering may be palliative care. Palliative care, or specialized medical care which is focused on symptom management for people living with serious illness, can help address each patient’s unique goals of care regarding pain and other needs. Palliative care can be provided as “primary” via the patient’s primary treatment team, or through consultative expertise from specialists. Although palliative care consultation was more common in our patient population (15.7%) than in other studies (0.3–5%),(18, 19, 26) our population also had a notably high risk of mortality and palliative care still may have been underutilized. In older EGS patients, and early involvement of palliative care can play an important role in reconciling a patient’s needs with end-of-life goals.(27, 28) Barriers to provision of palliative care in surgical patients are known to be present at the medical team, patient, and family levels, although it is also known that education on palliative principles at all levels can increase utilization and efficacy of interventions.(18, 19, 29).(29, 30) Communication is at the core of providing high-quality palliative care, especially for surgical patients, however indications and measures of quality for palliative care in surgical patients are still under investigation and have yet to be implemented in practice.(31–33)

There are several limitations to our study. This is a descriptive study comparing observed outcomes to ranges of predicted outcomes by the NSQIP SRC and included a small patient population. While we have provided descriptions of the outcomes and contrasted these outcomes with theoretical outcomes of surgical management, this is not a direct comparison of outcomes and statistical comparisons were not performed. Cause of death was not studied in patients who died, because many of them died outside of the hospital, limiting conclusions that can be drawn about whether the decision to offer surgery played a key role in the patient’s eventual mortality. Despite the lack of these data, it cannot be denied that our patient population was extremely high risk for mortality, regardless of the immediate cause of death. Another limitation was that the surgeon’s reasons for pursuing NOM were obtained retrospectively and may have resulted in inaccurate subgroup assignment. Some patients did not have a diagnosis which would be treatable with an operation (for example, consultation for necrotizing soft tissue infection determined to be cellulitis), and it was difficult to apply the SRC for these diagnoses. Our patient population had higher operative management rates than in studies performed from claims data, suggesting that not all patients admitted with EGS conditions had an EGS consultation. Our patient population was also quite heterogeneous, which limits our ability to examine specific EGS disease subgroups, explore patterns of comorbidities, or perform more adjusted statistical analyses. Despite these limitations, we believe the description of patients who received nonoperative management is extremely important, and the comparison to hypothetical operative outcomes underscores an important topic for future study.

Conclusions

To date, surgical literature and risk calculation focus primarily on patients who undergo operative management, but outcomes for nonoperatively managed patients are not well described. Our study suggests that older patients who require an EGS consult, even if they do not have a surgical indication, are a patient population with high risk of mortality within 1 year. Palliative care principles should be utilized early and systematically to clarify and document goals of care. Future studies should aim to improve prognostication after nonoperative management.