Delay in Emergency Hernia Surgery is Associated with Worse Outcomes

Abstract

Background

Patients requiring emergent surgery for hernia vary widely in presentation and management. The purpose of this study was to determine if the variation in timing of urgent surgery impacts surgical outcomes.

Methods

The national NSQIP database for years 2011-2016 was queried for emergent surgeries for abdominal hernia resulting in obstruction or gangrene by primary post-op diagnosis. Diaphragmatic hernias were excluded. Patients were grouped by surgical timing from admission to day of surgery: same day, next day, and longer delay. Multinomial propensity score weighting was used to address potential differences in underlying covariates’ clustering across the timing groups followed by multivariable logistic regression of morbidity and mortality.

Results

Weighted analysis yielded an effective sample size of 76,364. Hernia types included inguinal (20.9%); femoral (6.7%); umbilical (20.2%); ventral (41.0%); and other (10.4%). Delayed surgery was associated with increased rates of major complications (26.4% versus 20.9%, p<0.001), longer operative times (+12.5 minutes, p<0.001), longer postoperative lengths of stay (+1.6 days, p<0.001), increased re-operations (5.9% versus 4.7%, p=0.019), increased readmissions (7.0% versus 5.7%, p=0.004), and increased 30-day mortality (2.4% versus 1.7%, p=0.002).

When controlling for other factors, next-day surgery (OR = 1.23, 95% CI: 1.05-1.45, p=0.009) and surgery delayed more than one day (OR = 1.40, 95% CI: 1.13-1.73, p<0.002) were associated with an increased odds of a major complication. Mortality and readmission by timing of surgery were not independently significant.

Conclusions

Delay in surgery for emergent hernias increased the odds of major morbidity but not mortality. Patients presenting with hernia and an indication for urgent surgical intervention may benefit from an operation as soon as feasible rather than warrant waiting for further physiologic optimization, medical clearance, or specialized surgical personnel.

INTRODUCTION

Over 40% of abdominal wall hernia repairs performed annually in the United States are for urgent or emergent indications.¹ For hernia-related, high-grade bowel obstructions, immediate intervention is strongly supported by current consensus guidelines and systematic reviews.^2,3 Despite evidence supporting immediate surgery for strangulated abdominal wall hernias, reported surgical practices highlight the heterogeneity of surgical timing with hernia-related obstructions often managed nonoperatively for days prior to surgery.^4-7 While some operative delays may be due to capacity and staffing,⁸ it has also been suggested that short-term delays may be useful for physiologic optimization or “medical clearance.”^9-11

Thus, the optimal timing of surgery in acutely ill patients is more controversial in practice than conventional wisdom dictates. For example, prior studies have demonstrated that the distinction between operable fulminant colitis and colitis requiring medical management alone is often difficult for surgeons with many patients transitioning into an operative paradigm days into an admission. These studies have shown that delays by even two days after admission are associated with increased postoperative complications, hospital costs, and lengths of stay.^9,12-17 For strangulated hernias, consensus guidelines currently recommend immediate surgery.² However, clearly identifying which hernias are strangulated on presentation is difficult, and current clinical practice often prefers short-term observation to identify the trajectory of one’s hernia obstruction. In addition to this unpredictable clinical course, clinical factors such as immediate preoperative acuity and medical comorbidities also influence postoperative outcomes in patients undergoing non-elective surgery. It is tempting to acutely “optimize” patients prior to surgery, but the impact of short delays for physiologic optimization on emergency surgery are poorly understood and not currently reported in the literature.

We investigated whether delays in hernia surgery when presenting acutely obstructed were warranted. The purpose of our study was to compare outcomes of three temporal approaches to emergency hernia surgery – emergency versus next-day versus delayed – while controlling for clinically important preoperative risk factors and markers of patient acuity. Based on prior findings, we hypothesized that we would observe an independent association favoring immediate surgery due to decreased postoperative complications and mortality rates.

METHODS

Data and Case Identification

We identified all adult patients undergoing emergent surgery for an abdominal hernia in the American College of Surgeon’s National Surgical Quality Improvement Program (NSQIP) Participant Use Data File from January 1, 2011 to December 31, 2016. We determined cases to be “emergent” based on the 2011 revised definition of emergency case described in the NSQIP user guide which provided for abstractor discretion to assign it whenever case details, surgeon notes, and anesthesiologist notes corroborated that a case was both unplanned and related to acute decompensation (see Appendix A). Those requiring surgical intervention for an abdominal hernia were determined based on abstractors’ reporting of the operative indication by International Classification of Diseases, 9^th Revision (ICD-9) and 10^th Revision (ICD-10) diagnosis codes (ICD-9: 550.X-552.X; ICD-10: K40.X-46.X). To reduce heterogeneity involved with different surgical approaches and unique presentations, we then excluded diaphragmatic hernias (ICD-9: 551.3, 552.3, 555.3; ICD-10: K44.X). We further excluded any cases that were reported as transfers into the reporting institution as the amount of time as an inpatient at another hospital was unknown. Our initial analysis of the NSQIP file highlighted logical inconsistencies (<1% of total cases) suggestive of coding errors or unusual care paradigms. To reduce the effects of likely erroneous data entry, we excluded any case of emergent surgery where the final diagnosis was reported as a non-obstructing, non-gangrenous abdominal wall hernia (less than 10% of selected sample).

Defining the Primary Exposure Variable

We identified patients as having “immediate” surgery if the NSQIP procedure day was the same as the day of admission, a “next-day” surgical intervention if the procedure was one day later, and a “delayed” intervention if greater than one day after admission. We based this definition on prior studies and our own exploratory analysis demonstrating a similar trend of increasing complication after 2 to 3 days of admission in NSQIP data.^9,12-16 We defined these terms based on the perspective of the consulting surgeon (i.e., time of admission) rather than the patient’s clinical time course (i.e., onset of symptoms), which is not readily obtainable from NSQIP data.

NSQIP Clinically-Abstracted Variables

We collected patient sex, age, and clinically abstracted, standardized comorbidities that have been previously described and can be found in the information accompanying the annual publication of the NSQIP Participant Use File.^18,19 We elected to use clinical assessments of acuity (e.g., preoperative sepsis) over the available laboratory markers in NSQIP due to the former being more indicative of clinically significant abnormalities.¹⁶ We also obtained in-hospital mortality, post-operative complications, and post-operative length of stay (LOS) from the NSQIP dataset using similarly standardized definitions.^18,19 Our primary outcome variable was a major complication occurrence. We defined a “major complication” similar to the American College of Surgeons’ Surgical Risk Calculator (http://www.riskcalculator.facs.org) use of a “serious complication” aggregating all of the standard definitions of individual NSQIP complications excluding superficial surgical site infection and failure to wean from the ventilator.

Statistical analysis

We designed an a priori multinomial propensity score weight-adjusted analysis of primary data that we then performed in R 3.4.1 (R Project for Statistical Computing, Vienna, Austria) to address potential differences in underlying covariates’ clustering across operative timing categories (Hospital Day 0, Hospital Day 1, or > Hospital Day 1). We used a previously reported multinomial propensity score weighting-without-matching methodology²⁰ via the mnps function of the twang package (RAND Corporation, Santa Monica, California) loaded into R Studio 1.01.153 (R Studio, Boston, MA) to assign each individual patient a propensity score relative to their individual representativeness of his or her respective operative approach. We estimated propensity scores via a generalized boosted model – a 2,500-iteration tree-based regression model – with a balance rule based on minimizing the absolute standardized mean difference of all iterative pairwise comparisons. Propensity scores were trimmed to remove outlier cases with inverse probability scores greater than then 99^th percentile as has been previously proposed.²¹ Diagnostic tests were evaluated for post-propensity balance (see Appendix B). Further control by clustering by hospital or provider were not available in NSQIP data.

For our primary endpoints, we compared the univariate association of preoperative risk factors and postoperative outcomes across operative approach using a weight-adjusted modified Wald test or Pearson’s Chi-squared test where appropriate. Propensity scores were then used as inverse probability weights in a doubly robust multivariable logistic regression of major complications, readmission, and mortality on preoperative risk factors by surgical timing. We identified potential covariates by univariate statistical significance and biological plausibility for confounding (e.g., type of hernia, surgical approach, preoperative physiologic compensation, and high-risk comorbidities). We then constructed regression models using backwards stepwise logistic regression on potential covariates by optimization of the Akaike information criterion. Covariates were tested for multicollinearity (variance inflation factor > 10 threshold). Variables that a priori were identified to have a high biological likelihood for confounding (e.g., sex, age, procedure type) were included regardless of p value. Secondary analysis included multivariable linear regression of operative timing on preoperative risk factors by surgical timing. A post hoc analysis was performed for the findings reported below to ensure that inclusion or exclusion of lower-risk hernia types (e.g., inguinal) had no overall effect on statistical conclusions. All regression analyses were performed using Stata/MP 14.2 (StataCorp, College Station, TX).

This study design was reviewed and approved by the Johns Hopkins Medicine Institutional Review Board and in accordance with prior guidelines for analysis using the NSQIP datasets.^22,23

RESULTS

We identified 28,673 emergency hernia surgeries in the NSQIP participant use file that yielded an effective weighted sample size of 76,364 after multinomial propensity score weighting. Hernia types included inguinal (20.9%); femoral (6.7%); umbilical (20.2%); ventral (41.0%); and other (Spigelian, obturator, etc.) (10.4%). Each operative timing category was approximately 1/3 of the weighted dataset (immediate surgery, 37.1%; next-day surgery, 36.4%; delayed surgery, 26.5%). Less than 2% of cases were performed laparoscopically. Patient covariates that remained significant after weighting included age greater than 65 years old, sex, ASA class greater than 2, diabetes, hypertension requiring medication, and COPD (Table 1).

TABLE 1.

Demographic and baseline characteristics for propensity score-weighted patients undergoing emergent hernia surgery at NSQIP institutions by timing of surgery, 2011-2016.

Characteristic (%)	Total N=76,364	Immediate Surgery N=28,331 (37.1)	Next-Day Surgery N=27,812 (36.4)	Surgery > 1 Day Later N=20,198 (26.5)	p
Age > 65 years	44.6	43.2	43.4	48.2	<0.001
Female Sex	53.0	52.1	52.4	55.2	0.04
Body Mass Index (Mean)	32.0	32.0	32.0	32.0	1.0
Functional Status, Independent	93.6	93.8	93.8	93.0	0.4
ASA Class > 2	63.6	61.1	61.1	70.4	<0.001
Comorbidities
Preoperative Sepsis	21.6	21.3	21.5	22.3	0.5
Bleeding Disorder	10.1	9.6	9.8	11.2	0.08
Diabetes	17.7	16.9	17.1	19.7	0.02
Chronic Steroid Use	4.0	3.9	3.9	4.4	0.4
History of Ascites	4.5	4.2	4.3	5.0	0.2
Hypertension	53.3	51.6	51.6	58.1	<0.001
Congestive Heart Failure	2.0	1.9	2.0	2.1	0.6
History of stroke or TIA	2.2	2.4	2.1	2.1	0.8
Dyspnea	7.4	7.0	7.1	8.3	0.2
COPD	8.3	7.8	7.7	9.6	0.008
Smoker	20.2	20.8	20.5	18.8	0.1
Hernia Type
Inguinal	20.9	21.8	21.8	18.3	0.004
Femoral	6.7	7.3	6.8	5.7	0.06
Umbilical	20.2	21.5	21.3	16.9	<0.001
Ventral	41.0	39.0	39.6	45.7	<0.001
Other	10.4	9.6	9.8	12.2
Laparoscopic Approach	1.9	1.8	1.6	2.2	0.2

Covariates less than 2% of the population not shown here: history of weight loss, disseminated cancer, preoperative wound infection, pregnancy, history of angina, history of myocardial infarction, history of peripheral vascular disease, preoperative pneumonia, chronic ventilator dependence, chronic renal failure, on hemodialysis, requiring preoperative transfusions, history of esophageal varices, and preoperative coma.

Surgery delayed more than one day was associated with an increased rate of major complication (26.4% versus 20.9% overall, p < 0.001), a longer operative time (100.8 minutes versus 88.3 minutes overall, p < 0.001), a longer average postoperative length of stay (7.0 days versus 5.4 days overall, p < 0.001), an increased risk of second operation (5.9% versus 4.7% overall, p = 0.019), an increased risk of readmission (7.0% versus 5.7% overall, p = 0.003), and an increased risk of 30-day mortality (2.4% versus 1.7% overall, p = 0.002) (Table 2). Each specific complication reported in NSQIP occurred more frequently in the delayed surgery group (Table 2).

TABLE 2.

30-day postoperative outcomes for propensity score-weighted patients undergoing emergent hernia surgery at NSQIP institutions by timing of surgery, 2011-2016.

Characteristic (%)	Total N=76,364	Immediate Surgery N=28,331 (37.1)	Next-Day Surgery N=27,812 (36.4)	Surgery > 1 Day Later N=20,198 (26.5)	p
Major Complication	20.9	18.1	19.6	26.4	<0.001
Wound Infection	8.0	6.9	7.4	10.5	<0.001
Postoperative Sepsis	6.4	5.7	5.9	8.3	<0.001
Pneumonia	3.7	3.5	3.5	4.4	0.06
Reintubation	2.8	2.3	2.5	3.7	0.01
Pulmonary Embolism	0.7	0.6	0.7	1.0	0.06
Deep Vein Thrombosis	1.2	0.9	1.2	1.6	0.05
Failed Ventilator Wean	4.0	3.4	3.8	5.2	<0.001
Renal Insufficiency	0.8	0.6	0.7	1.1	0.05
Renal Failure	0.9	0.9	0.8	1.1	0.5
Urinary Tract Infection	2.2	1.9	2.1	2.7	0.06
Stroke	0.3	0.2	0.3	0.4	0.6
Cardiac Arrest	0.8	0.6	0.6	1.2	<0.001
Myocardial Infarction	0.8	0.7	0.7	1.0	0.3
Postoperative Bleeding	3.3	2.8	2.9	4.3	0.003
Operative Time (Mean, Minutes)	88.3	82.2	85.4	100.8	<0.001
Postoperative Length of Stay (Mean, Days)	5.4	4.8	4.9	7.0	<0.001
Operative Takeback	4.7	4.2	4.3	5.9	0.019
Readmission	5.7	5.1	5.4	7.0	0.003
30-day Mortality	1.7	1.4	1.5	2.4	<0.001

When we performed doubly robust multivariable logistic regression of a major complication on timing of surgery and covariates of interest (Table 3), we identified an independent 23% increase in the odds of a complication with next-day surgery (OR = 1.23, 95% CI: 1.05-1.45, p = 0.009) and a 40% increase in the odds of a complication with surgery delayed even longer (OR = 1.40, 95% CI: 1.13-1.73, p = 0.002). We further found that when controlling for other factors, operative time was on average 4 minutes longer for next-day surgery (p = 0.01) and 16 minutes longer for surgery delayed more than one day (p < 0.001) (full models not shown). The odds of mortality and readmission by timing of surgery were not independently significant.

TABLE 3.

Doubly robust multivariable logistic regression of a major complication following emergency hernia surgery (propensity weighted sample; n = 8,888 individual additionally weighted observations)

Variable	Unadj. OR (95% CI)	p	Adj. OR (95% CI)	p
Surgical Timing
Same-day	Ref		Ref
Next-day	1.11 (1.03-1.19)	0.005	1.23 (1.05-1.45)	0.009
> 1 day later	1.63 (1.45-1.84)	<0.001	1.40 (1.13-1.73)	0.002
Hernia Type
Inguinal	0.69 (0.62-0.77)	<0.001	0.43 (0.29-0.64)	<0.001
Femoral	1.30 (1.10-1.53)	0.002	0.54 (0.35-0.83)	0.005
Umbilical	0.67 (0.59-0.75)	<0.001	0.38 (0.25-0.57)	<0.001
Ventral	1.24 (1.14-1.35)	<0.001	0.68 (0.46-0.99)	0.05
Other	1.33 (1.18-1.50)	<0.001	0.70 (0.46-1.07)	0.1
Age	1.03 (1.02-1.03)	<0.001	1.02 (1.01-1.02)	<0.001
Female Sex	1.10 (1.00-1.20)	0.03	0.85 (0.74-0.97)	0.02
Body Mass Index	1.00 (1.00-1.01)	0.1	1.01 (1.01-1.02)	<0.001
ASA Class	2.59 (2.43-2.76)	<0.001	1.71 (1.53-1.90)	<0.001
Functional Status	1.94 (1.79-2.11)	<0.001	1.32 (1.16-1.52)	<0.001
Preoperative Transfusion	5.73 (4.03-8.13)	<0.001	5.98 (2.29-15.61)	<0.001
Chronic Steroid Use	2.35 (1.96-2.83)	<0.001	1.39 (1.03-1.86)	0.03
Preoperative Wound Infection	2.82 (2.27-3.51)	<0.001	1.73 (1.21-2.47)	0.003
Preoperative Sepsis	3.10 (2.82-3.40)	<0.001	1.78 (1.55-2.04)	<0.001
Congestive Heart Failure	3.28 (2.70-3.99)	<0.001	1.34 (0.91-1.97)	0.1
Peripheral Vascular Disease	2.92 (1.76-4.86)	<0.001	1.52 (0.96-2.41)	0.07
Dyspnea	1.79 (1.62-1.97)	<0.001	1.24 (1.08-1.43)	0.003
COPD	2.39 (2.10-2.72)	<0.001	1.30 (1.05-1.61)	0.02
Smoker	0.97 (0.87-1.09)	0.641	1.19 (1.00-1.41)	0.05
Chronic Renal Failure	4.57 (3.54-5.89)	<0.001	2.32 (1.53-3.50)	<0.001
Bleeding Disorder	2.34 (2.07-2.64)	<0.001	1.19 (0.99-1.43)	0.07
Ascites	2.71 (2.28-3.23)	<0.001	1.87 (1.43-2.46)	<0.001
Altered Mental Status	6.13 (3.99-9.40)	<0.001	2.26 (1.40-3.65)	0.001

Abbreviations: ASA - American Society of Anesthesiologist classification, COPD – chronic obstructive pulmonary disease

DISCUSSION

This study identified an independent association between timing of urgent hernia surgery and postoperative outcomes in a large, nationally representative database. The odds of a postoperative complication were 23% greater when surgery occurred the next day after admission and 40% greater when surgery occurred more than one day after admission. These results included controlling for both comorbidities and immediate preoperative clinical markers of acuity. We also found longer operating times in the delayed surgery group and consistent increased rates of complications by individual sub-categories of NSQIP postoperative complications. There was no significant difference in mortality or readmission by timing of urgent hernia surgery. As these findings were based off of retrospectively collected data, these temporal associations highlight the need for further prospective work to determine if postoperative outcomes for emergency hernia surgery may be substantially improved if preoperative delays were decreased.

In this study, we examined whether there was an association between timing of urgent hernia surgery and postoperative outcomes using nationally abstracted NSQIP data. The natural assumption is that those operated on immediately should be the most acutely ill patients and therefore have the worst outcomes while those operated on after a day or more were less in need of emergency surgery should do better as their disease process was caught at an earlier point. These findings challenge this assumption by identifying the opposite association in a large dataset that is generalizable to the United States surgical population. Those patients that proceeded to an operation the most expediently did better than those patients that were observed or optimized in the hospital for one or more days. Moreover, even though those with greater comorbidity burden (e.g., COPD, anticoagulation) appear to be more present in the delayed group, the opportunity for further comorbidity optimization was not associated with improved outcomes.

We would suggest that there are a handful of reasonable explanations for why patients requiring urgent surgery may not be operated on immediately: 1) preoperatively resuscitating or decompressing patients in a critical care setting; 2) obtaining medical “clearance” even in an emergency setting; 3) observing borderline cases to assess for disease evolution; and 4) lacking the immediate personnel or facility-based capacity to proceed with emergent surgery.^8-11 While this study is not able to stratify patients by reason for operative delay, it did find a cumulative effect of worse surgical outcomes with longer delays. These findings are consistent with smaller series and systematic reviews of single institution studies that demonstrated even short-term delays prior to hernia surgery lead to worse outcomes for patients.^3-7 These findings also support the high degree of urgency recommended in international consensus guidelines for non-elective hernia surgery.² Hence, without further data, we interpret these findings to argue for immediate surgical intervention in the case of a hernia with acute obstructive symptoms unless there is a clear contraindication. Specifically, we would suggest that operative delays due to lack of immediate surgical capacity are inappropriate and may be placing patients at unnecessary increased surgical risk. Further research is required to more effectively delineate if there are particular subgroups of patients that may benefit from operative delay versus the overall statistical finding of harm in this patient population.

This study also has limitations that should be recognized. First, the data for this study came from prospectively collected NSQIP quality improvement reporting rather than a formal, prospective clinical investigation. NSQIP reporting is susceptible to institutional-specific practices while not having an identifier for cluster analysis,^16,24 and our findings need to be reproduced in future trials. However, our findings being consistent with previous studies cited above is encouraging that the effect of any reporting differences does not affect conclusions. A related limitation of NSQIP reporting is that hospital days are recorded as calendar days rather than absolute time from admission which could inappropriately classify patients presenting near midnight as next-day operations if crossing over into a new calendar day. This limitation is mitigated by the reassuring trend in findings of the step-wise harm of delayed surgery and our proposed biological model. Finally, NSQIP reporting does not capture the entirety of the heterogeneous hernia population (e.g., hernia size, visceral organ involvement). These possibilities for misclassification bias should occur in a random fashion and tend to bias findings toward a null hypothesis, which if occurring in these findings appears to be minimal that it does not affecting our positive results.

Second, the propensity scoring used in this study allows for a more methodologically accurate comparison of operative timing groups but also leads to pan-population enrichment that limits the ability to generalize absolute results from this study or compare to other studies. For example, complication rates reported are not directly comparable outside of the propensity-weighted sample used in this study.

Finally, the most important limitation of this study is the focused context in which it may be used with operative decision-making. While current guidelines strongly recommend surgery for hernias with bowel ischemia, there is much more heterogeneous management of patients presenting without evidence of ischemia and obstructive symptoms alone.² This study only included patients with obstructive hernia symptoms that ultimately required surgery, selecting against those patients successfully managed with a non-operative strategy (with or without subsequent elective repair). As such, our findings are specific to patients for which surgery is inevitable. While we were not able to compare these finding to patients who presented with obstructive hernia symptoms but were ultimately discharged without having surgery performed, we chose our selection criteria to ensure the validity of our findings. Specifically, we selected patients with problems severe enough to inevitably require surgery and have shown that observation is associated with incrementally increased risk in these patients. This form of limited recommendation for immediate surgery in patients who will indeed require surgery is increasingly common in the acute care surgery literature.^9,16,17

CONCLUSIONS

A delay in surgery for obstructed or strangulated hernia substantially and progressively increases the odds of major morbidity but not mortality. Observation does not appear to be benign in patient’s ultimately requiring surgery for obstructed or strangulated hernias and adds incremental risk with the length of observation. Patients presenting with hernia and indications for urgent surgical intervention likely need to be taken to the operating room as soon as is physiologically feasible. Delays for resuscitation and optimization beyond physiologic tolerance of the procedure may increase the risk of major morbidity.