Reducing Under-Diagnosis of Hirschsprung-Associated Enterocolitis: A Novel Scoring System

Abstract

Background:

Hirschsprung-Associated Enterocolitis (HAEC) is a life-threatening and difficult to diagnose complication of Hirschsprung Disease (HSCR). The goal of this study was to evaluate existing HAEC scoring systems and develop a new scoring system.

Study Design:

Retrospective, multi-institutional data collection was performed. For each patient, all encounters were analyzed. Data included demographics, symptomatology, laboratory and radiographic findings, and treatments received. A “true” diagnosis of HAEC was defined as receipt of treatment with rectal irrigations, antibiotics and bowel rest. The Pastor and Frykman scoring systems were evaluated for sensitivity/specificity and univariate and multivariate logistic regression performed to create a new scoring system.

Results:

Four centers worldwide provided data on 200 patients with 1450 encounters and 369 HAEC episodes. Fifty-seven percent of patients experienced one or more episodes of HAEC. Long segment colonic disease was associated with a higher risk of HAEC on univariate analysis (OR 1.92, 95% CI 1.43–2.57). Six variables were significantly associated with HAEC on multivariate analysis. Using published diagnostic cutoffs, sensitivity/specificity for existing systems were found to be 38.2%/96% for Pastor’s and 56.4%/86.9% for Frykman’s score. A new scoring system with a sensitivity/specificity of 67.8%/87.9% was created by stepwise multivariate analysis. The new score outperformed the existing scores by decreasing under-diagnosis in this patient cohort.

Conclusion:

Existing scoring systems perform poorly in identifying episodes of HAEC, resulting in significant under-diagnosis. The proposed scoring system may be better at identifying those under-diagnosed in the clinical setting. Head-to-head comparison of HAEC scoring systems using prospective data collection may be beneficial to achieve standardization in the field.

Introduction

Hirschsprung-Associated Enterocolitis (HAEC) can be life-threatening, especially if the diagnosis is missed or delayed. Unfortunately, symptoms are often non-specific, making the diagnosis more difficult. The clinical symptoms most commonly seen in HAEC were first described by Harald Hirschsprung at the International Congress for Children’s Disease in Berlin¹ and include distention, explosive diarrhea, vomiting, fevers, and lethargy. In the modern age, Ehlalaby² reiterated these diagnostic criteria and several attempts have been made at creating a clinical diagnostic score to aid in the diagnosis of HAEC since then. The Pastor score, published in 2008, was created using the Delphi expert consensus method and includes 16 components: explosive diarrhea, foul-smelling diarrhea, bloody diarrhea, a history of enterocolitis, explosive stool on digital rectal exam, distention on exam, decreased peripheral perfusion, lethargy, fever, air-fluid levels on KUB, dilated loops of bowel, sawtooth appearance, rectosigmoid cutoff sign, pneumatosis, leukocytosis and a left shift.³

Frykman et.al. refined the Pastor score in 2018, suggesting lowering the threshold diagnostic cutoff to four points rather than ten.⁴ Frykman also suggested a new score limited to four criteria: explosive stool, decreased peripheral perfusion, lethargy and dilated loops of bowel.⁴ The American Pediatric Surgical Association Hirschsprung Disease Interest Group has also published guidelines for HAEC diagnosis and treatment, though this system is not a scoring system.⁵ Unfortunately, the existing scoring systems have not been widely adopted in clinical practice. The Pastor score has numerous elements, was designed for research purposes, and is difficult to use at bedside.^5–7 A study performed by Dore et.al. in Spain demonstrated that the score cutoffs for both scores were too restrictive and therefore posed the risk of underdiagnosing HAEC.⁸ Another study from Indonesia by Gunadi et.al. found similar results, with the Pastor score being too restrictive and insensitive.⁷

This study sought to evaluate the diagnostic accuracy of the Pastor and Frykman scores and to create a streamlined diagnostic tool with improved clinical utility. It was hypothesized that the previous two scoring systems would be poor predictors of the diagnosis of HAEC.

Methods

Study Design

This was a multi-institutional, international, retrospective analysis of patients with HAEC. Institutional Review Board approval was obtained at all sites prior to the study and in accord with 45 CFR 46.116(d), requirement for informed consent was waived. Participating sites included Le Bonheur Children’s Hospital (Memphis, TN), Yale New-Haven’s Children’s Hospital (New Haven, CT), The Hospital for Sick Children (Toronto, Canada) and The Royal Children’s Hospital (Melbourne, Australia). Data use agreements between each site and the coordinating center (Le Bonheur Children’s Hospital) were executed prior to any data sharing. Medical records were reviewed for patients aged 18 years old and younger with a diagnosis of Hirschsprung Disease using ICD9 (751.3) and 10 codes (Q43.1) between January 2009 and June 2016.

Data Collection

All data were collected using REDCap (Research Electronic Data Capture) electronic data capture tools hosted by the University of Tennessee Health Science Center.^9,10 REDCap is a secure software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for data integration and interoperability with external sources. Demographic information for each patient was collected, including age at diagnosis, race, gender, disease length, details pertaining to original surgery and pathology at the time of diagnosis. Information was then collected at each hospital or clinic visit for Hirschsprung-related encounters, including symptomatology, laboratory values, radiology results, and treatments provided, based on documentation from the first 24 hours of admission. All 16 criteria for the Pastor and Frykman score were collected [Table 1].^3,4 Data elements not clearly documented in the patient chart were left blank. For each encounter, it was noted whether the provider documented a diagnosis of HAEC.

Table 1:

Clinical Prediction Scores according to Pastor and Frykman. A score ≥ 10 on the Pastor score was the original cutoff, but Frykman later suggested a modified cutoff of ≥4. A score ≥4 is also the cutoff for Frykman’s score.

Clinical Finding	Points
Pastor Score
History of HAEC	1
Diarrhea with Explosive Stool	2
Diarrhea with foul-smelling stool	2
Diarrhea with bloody stool	1
Abdominal distention	2
Fever	1
Lethargy	1
Explosive discharge of gas and stool on DRE	2
Decreased peripheral perfusion	1
Leukocytosis	1
Left shift	1
Dilated loops of bowel	1
Multiple air fluid levels	1
Sawtooth appearance with irregular mucosal lining	1
Rectosigmoid cutoff sign (absence of distal air)	1
Pneumatosis	1
Frykman Score
Diarrhea with explosive stool	5
Lethargy	5
Decreased peripheral perfusion	5
Dilated loops of bowel on KUB	4

Data Analysis

For the purpose of this study, a “true” diagnosis of HAEC was defined as patients that received the gold standard of treatment at the time of admission or initial encounter, consisting of antibiotics (either oral or intravenous), bowel rest and rectal irrigations.¹¹ To account for a spectrum of disease, those placed on a clear liquid diet in addition to antibiotics and rectal irrigations were included as having HAEC. This was compared with to the number of HAEC diagnoses made by the provider. The Pastor and Frykman scores were retrospectively calculated for each episode, and sensitivity and specificity of each score were calculated. A score cutoff ≥4 was used for the Pastor score based on a modified diagnostic threshold published by Frkyman.⁴ A score cutoff ≥4 was also used for the Frykman score as previously published.⁴

Univariate and multivariate logistical regression were then performed to identify significant predictors of an episode of HAEC. Factors with missing data were included in the analysis by the use of the missing-indicator approach in which a dummy variable is utilized to indicate that the data field is missing. Stepwise multivariate regression in SAS (SAS Institute, Cary, North Carolina) was employed to create a novel scoring system. Six variables were identified to include in the new score. For ease of use, the variable coefficients from the logistic regression equation were rounded to the nearest whole number. The score was then calculated by adding each variable multiplied by its coefficient. To determine the cutoff, sensitivity, specificity and area under the receiver operating curve were calculated and the score that maximized area under the curve was chosen. All statistical analysis was performed in R.^12,13 A p-value of <0.05 was considered significant.

Results

Demographics of study population

A total of 200 patients, with 1450 patient encounters, were identified. Each patient was seen an average of 7.25 times during the study period. Basic demographics are listed in Table 2. As expected, male patients were predominant (82%). Disease length was most commonly in the rectum only (27%) or in the sigmoid colon (35%). Over half underwent a Soave procedure (58%) as the surgical procedure of choice.

Table 2:

Patient Demographics.

Demographic	Total (N=200)
Gender, n(%)
Male	163 (82%)
Female	37 (18%)
Race, n(%)
White	77 (38%)
Black	55 (28%)
Hispanic	10 (5%)
Asian	8 (4%)
Multiracial	7 (3.5%)
Other	27 (14%)
Unknown	16 (8%)
Age Range at Diagnosis, n(%)
Less than 60 days	138 (69%)
Less than 1 year	29 (14.5%)
Greater than 1 year	14 (7%)
Unknown	19 (9.5%)
Comorbidities, n(%)
Cardiac	23 (11.5%)
Pulmonary	2 (1%)
Neurologic	10 (5%)
Trisomy 21	18 (9%)
Disease Length, n(%)
Rectum Only	53 (26.5%)
Sigmoid Colon	71 (35%)
Descending/Splenic Flexure	11 (5.5%)
Transverse Colon	3 (1.5%)
Hepatic Flexure/Descending	9 (4.5%)
Cecum	9 (4.5%)
Small Bowel	26 (13%)
Unknown	18 (9%)
Number of Episodes of HAEC, n(%)
0	86 (42%)
1–4	93 (47%)
5–9	16 (8%)
>10	5 (2.5%)

Incidence of HAEC

There were 369 cases (25% of encounters) of HAEC in 114/200 (57%) patients based on the true diagnostic criteria established a priori for this study. The majority of encounters for HAEC were inpatient encounters (67%). No statistically significant differences in sex, race or ethnicity were seen between patients with HAEC and those that did not present with an episode of HAEC. Of the true HAEC diagnoses (those that were treated for HAEC), providers only diagnosed 105/369 (28%) with HAEC; 247/369 (67%) were treated for HAEC but not given a discharge diagnosis of HAEC by the provider, suggesting under-diagnosis. Additionally, twenty patients were diagnosed with HAEC but not treated according to established guidelines.

Univariate Analysis

Univariate analysis was performed to identify factors associated with an episode of HAEC. Table 3 shows the frequency that factors were identified in patients, overall missingness and the results of univariate analysis. A history of HAEC, while present in 45% of patients, was not significantly associated with HAEC (OR 0.99, 95% CI 0.77–1.26, p=0.1). Explosive diarrhea, hypotension, guarding, peritonitis, and radiology findings of a sawtooth pattern, rectosigmoid cutoff sign, pneumatosis and pneumoperitoneum were all infrequently seen (< 1% of cases) and not significantly associated with a diagnosis of HAEC. Long-segment colonic disease (defined as disease proximal to the sigmoid-descending colon junction up to the cecum, including total colonic disease¹⁴) was significantly associated with an increased risk of HAEC (OR 1.92, 95% CI 1.43–2.57, p<0.0001), while small intestinal disease (disease involving the small bowel proximal to the terminal ileum¹⁴) was associated with a decreased risk of HAEC (OR 0.28, 95% CI 0.24–0.5, p<0.0001).

Table 3:

Univariate Analysis of factors associated with HAEC episodes

Factor	Number of patients, n(%)	Missingness, n(%)	OR (95% CI)	p-value
Fever	149 (10.3%)	440 (30.3%)	4.89 (3.4– 7.09)	<0.0001*
Hx HAEC	648 (44.7%)	194 (13.4%)	0.986 (0.768– 1.26)	0.91
Diarrhea (any)	246 (17.0%)	440 (30.3%)	2.17 (1.6– 2.93)	<0.0001*
Watery diarrhea	127 (8.7%)		3.27 (2.26– 4.74)	<0.0001*
Explosive diarrhea	5 (0.3%)		1.96 (0.257– 11.9)	0.46
Foul smelling diarrhea	19 (1.3%)		6.54 (2.56– 18.7)	0.00016*
Bloody Diarrhea	8 (0.6%)		8.92 (2.04– 61)	0.0075*
Change Activity/Lethargy	171 (11.8%)	704 (48.6%)	6.74 (4.66– 9.81)	<0.0001*
Anorexia	235 (16.2%)	519 (35.8%)	6.99 (5.05– 9.74)	<0.0001*
Vomiting	324 (22.3%)	416 (28.7%)	4.28 (3.23– 5.7)	<0.0001*
Obstipation	70 (4.8%)	627 (43.4%)	10.5 (6.02– 19)	<0.0001*
Decreased Perfusion	13 (0.9%)	583 (40.2%)	5.03 (1.62– 18.7)	0.0076*
Distention	273 (18.8%)	243 (16.8%)	6.75 (5.04– 9.07)	<0.0001*
Fever on PE	51 (3.5%)	379 (26.1%)	10.4 (5.36– 22.3)	<0.0001*
Hypotension	5 (0.3%)	459 (31.7%)	3.33 (0.55– 25.4)	0.19
Tachycardia	188 (13.0%)	283 (19.2%)	2.71 (1.96– 3.73)	<0.0001*
Explosive gas/stool on DRE	30 (2%)	1062 (73.2%)	11.9 (4.8– 36)	<0.0001*
Tenderness on PE	117 (8.1%)	271 (18.7%)	6.56 (4.38– 9.97)	<0.0001*
Guarding	7 (0.5%)	779 (53.7%)	2.84 (0.621– 14.5)	0.17
Peritonitis	1 (0.07%)	842 (58.1%)	--	0.98
Air fluid levels AXR	119 (8.2%)	1222 (84.2%)	1.34 (0.775– 2.31)	0.30
Dilated loops AXR	256 (17.7%)	1091 (75.2%)	2.92 (1.83– 4.72)	<0.0001*
Sawtooth pattern AXR	4 (0.3%)	1384 (95.5%)	0.509 (0.057– 10.9)	0.58
Rectosigmoid cutoff AXR	12 (0.8%)	1383 (95.4%)	0.367 (0.08– 1.98)	0.21
Pneumatosis AXR	1 (0.07%)	1285 (88.6%)	--	0.99
Pneumoperitoneum AXR	2 (0.14%)	1095 (75.5%)	0.73 (0.029– 18.6)	0.83
Leukocytosis	154 (10.6%)	1071 (73.9%)	1.74 (1.12– 2.74)	0.014*
Left Shift	81 (5.6%)	1153 (79.5%)	1.74 (1.0– 3.1)	0.053
Length Type:
Short Segment	199 (13.7%)	221 (14.6%)	Reference
Long Segment	103 (7.1%)		1.92 (1.43– 2.57)	<0.0001*
Total Intestinal HSCR	11 (0.8%)		0.276 (0.138– 0.5)	<0.0001*

Percentages represent who often finding was present out of the 1,450 patient encounters. Abbreviations: OR: Odds ratio, CI: Confidence Interval, HAEC: Hirschsprung associated enterocolitis, PE: physical exam, DRE: digital rectal exam, AXR: Abdominal x-ray, HSCR=Hirschsprung Disease

^*p<0.05 was considered significant

Multivariate Analysis and New Score Creation

Step-wise multivariate regression revealed six variables that remained significantly associated with a diagnosis of HAEC: fever (OR 7.4, 95% CI 4.1–13.4, p<0.0001), bloody diarrhea (OR 13.2, 95% CI 2.4–106, p=0.006), obstipation (OR 3.4, 95% CI 1.7–7.4, p=0.001), distention (OR 3.2, 95% CI 1.7–3.3, p=0.0003), dilated loops of bowel on abdominal x-ray (OR 2.2, 95% CI 1.4–3.3, p=0.0005), and leukocytosis (OR 10.3, CI 6.5–16.7, p<0.0001) (Table 4). These variables were then used to create a new score, where the coefficient was rounded to the nearest whole number. The new proposed score is:

$$\begin{gathered} \left(2\text{ × fever}\right)+\left(2\text{ × bloody diarrhea}\right)+\left(2\text{ x leukocytosis}\right)+\text{ obstipation }+\text{ distention }+\text{ dilated loops of bowel on abdominal x-ray} \\ (1\text{ if a sign is present},0\text{ if absent}). \end{gathered}$$

Table 4:

Multivariate Analysis of factors associated with HAEC episodes

Factor	Coeff	Assigned point	OR (95% CI)	p-value
Fever	2.0033	2	7.41 (4.14– 13.4)	<0.0001*
Bloody Diarrhea	2.5823‡	2	13.2 (2.38– 106)	0.0056*
Obstipation	1.2354	1	3.44 (1.66– 7.38)	0.0012*
Distention	1.1660	1	3.21 (1.73– 6.10)	0.00027*
Dilated loops on AXR	0.7687	1	2.16 (1.4– 3.32)	0.00049*
Leukocytosis	2.3337	2	10.3 (6.47– 16.7)	<0.0001*

Abbreviations: Coeff: coefficient/parameter estimate from model, OR: odds ratio, CI: confidence interval, AXR: abdominal x-ray

^*p<0.05 was consider significant

^‡For ease of use, variables were rounded to the nearest whole number, except for bloody diarrhea, which was rounded down so that all variables would be weighted similarly.

Score Performance and Internal Validation

Diagnostic cutoff scores were then determined by finding the sensitivity, specificity, and area under the receiver operating curve (AUC) (Table 5). Of those with HAEC, 81/369 (22%) had a score of zero. The most common score of 2 was seen in 108/369 (29.3%) HAEC encounters. While a score of 1 maximized sensitivity and AUC, only 10.3% of patients with HAEC had a score of 1. A score of 2 maximized specificity compared to a score of 1, with an equivalent AUC. Of the encounters with HAEC, 250 had a score of 2 or more, for a sensitivity of 67.8%. A score of 2 or more was found to be highly predictive of an episode of HAEC (OR 10.8, 95% CI 7.6–15.4, p<0.0001) and was chosen as the diagnostic cutoff.

Table 5:

Performance of the new score in predicting the presence or absence of HAEC episodes at each threshold value

Cut-off value†	Number of patients, n(%)	Incidence of HAEC	OR (95% CI)	Sensitivity (95% CI)	Specificity (95% CI)	Area under the ROC curve (AUC) (95% CI)
0	81 (21.95%)	8.7%	Reference	---	----	---
1	38 (10.3%)	26%	3.7 (2.4–5.7)	78.05% (73–82%)	77.98% (75–80%)	0.7802 (0.7557–0.8047)
2	108 (29.3%)	50.9%	10.8 (7.6–15.4)	67.8% (63–72%)	87.9% (86–90%)	0.7782 (0.7524–0.8039)
3	52 (14.1%)	78.8%	38.7 (31.1–75.3)	38.48% (33–44%)	97.5% (96–98%)	0.6799 (0.6546–0.7052)
4	57 (15.5%)	85.1%	59.3 (30.4–127.5)	24.39% (20–29%)	98.8% (98–99%)	0.6159 (0.5938–0.6381)
5	28 (7.6%)	90.3%	97.1 (33.5–412.4)	8.94% (06–12%)	99.72% (99–100%)	0.5433 (0.5287–0.558)
6	4 (1.1%)	100%	---	1.36% (0–3%)	100% (100–100%)	0.5068 (0.5009–0.5127)
7	1 (0.3%)	100%	---	----	----	----

^†Cut-off value represents greater than or equal to that score

Abbreviations: OR: odds ratio, CI: confidence interval, ROC: Receiver Operating Curve

Score = (2 × fever) + (2 × bloody diarrhea) + (2 × leukocytosis) + obstipation + distention + dilated loops of bowel on abdominal x-ray (1 if a sign is present, 0 if absent).

Comparison to Previous Scoring Systems

The new score was compared with the previous scoring systems. Of the 369 episodes of HAEC, 172 (46%) met the diagnostic cutoff using the new scoring system but did not meet the cutoffs for the Frykman or Pastor scores. Only 54 patients (14.6%) with HAEC met the cutoff for all three scores. Receiver Operating Curves (ROC) were plotted for each score [Figure 1]. The new score has the highest AUC of 0.7782. Sensitivities, specificities, positive predictive values, and negative predictive values were also calculated for each score [Table 6]. Sensitivity was highest for the new score when compared with the Frykman and Pastor scores, as was the negative predictive value. The Pastor score maximized specificity and positive predictive value.

Figure 1: Receiver Operating Curves for all scoring systems.

ROC curve for Pastor’s score applied to this patient population resulted in an area under the curve (AUC) of 0.5367 when using the cutoff of ≥4 set by Frykman and 0.5009 when using the original cut-off of ≥10 set by Pastor. ROC curve for Frykman’s score yielded an AUC of 0.5235. The new score performed significantly better, with an AUC of 0.7782.

Table 6:

Comparison of the different scores in terms of sensitivity, specificity, positive predictive value, negative predictive value and area under the curve.

Statistic	Pastor (≥4)	Pastor (≥10)	Frykman (≥4)	New Score (≥2)
Sensitivity (95% CI)	38.2% (33–43%)	0.5% (0–0.2%)	56.4% (51–61%)	67.8% (63–72%)
Specificity (95% CI)	96.0% (95–97%)	100% (100–100%)	86.9% (85–89%)	87.9% (86–90%)
PPV (95% CI)	76.6% (70–83%)	100% (16–100%)	59.4% (54–65%)	65.6% (61–70%)
NPV (95% CI)	82.0% (80–84%)	74.7% 72–77%)	85.4% (83–87%)	88.9% (87–91%)
AUC	0.5367	0.5009	0.5235	0.7782

Abbreviations: PPV: positive predictive value, NPV: negative predictive value, AUC: area under the ROC curve, CI: confidence interval, ROC: Receiver Operating Curve

Discussion

Two-hundred patients with 1450 encounters were retrospectively evaluated for episodes of HAEC. Patients with long-segment colonic disease were found to be at a higher risk for episodes of HAEC, while small intestinal disease was associated with a decreased risk of HAEC. Multivariate analysis showed that fever, bloody diarrhea, obstipation, distention, dilated loops of bowel on abdominal x-ray and leukocytosis were all independently associated with a diagnosis of HAEC. A novel clinical score was created using those six factors and a cutoff of ≥2 was established.

This is the largest study to date looking at the HAEC diagnostic scoring systems, with 200 patients and 369 episodes of HAEC across 3 different countries. This study confirmed previous reports that found the Pastor score has poor sensitivity and was too restrictive.^7,8 The clinical importance of improved diagnosis was recently demonstrated in a study by Wall et.al. critically examining the implementation of an HAEC triage and treatment protocol.¹¹ The use of their HAEC protocol resulted in fewer hospital admissions and improved successful outpatient management of HAEC, though they found no difference in length of stay or morbidity.¹¹

The new scoring system offers several advantages over the Pastor and Frykman scores. The Pastor score is long and difficult to use clinically and was developed to be primarily a research tool rather than a clinical tool. Additionally, the Pastor score was developed based on expert opinion, not clinical data. Frykman’s score has only four criteria, which improves ease of use, but two of the four criteria were rarely observed in the current cohort. Diarrhea with explosive stool was only documented in five patients (0.3%), of which only two had HAEC (40%); decreased peripheral perfusion was seen in thirteen (0.9%), of which nine had HAEC (69%). Therefore, to meet diagnostic requirements for HAEC, the majority of patients with a score ≥4 in the Frykman scoring system must have lethargy or dilated loops of bowel, both of which are non-specific findings. This new score includes signs that are frequently seen in HAEC, such as fevers, distention, obstipation, and leukocytosis. While bloody diarrhea was rare and seen in only eight patients (0.6%), it was highly specific and of those with bloody diarrhea, six (75%) had HAEC. This new score also maximizes sensitivity at 67.8% compared to 38% for Pastor and 56% for Frykman’s scores. Of those with HAEC, 176 (46%) did not meet the diagnostic threshold for either the Frykman or Pastor scores but did meet the cutoff for the new, proposed score. Forty-six percent of patients would therefore have been undiagnosed with HAEC using the previous scoring systems.

The observation that long-segment colonic disease was associated with an increased risk of HAEC is not a new finding. Elhalaby also previously showed that longer segment disease was associated with an increased risk of HAEC, which has been confirmed in other studies as well as the present analysis.² Reding et.al. also found an increased incidence of HAEC in long-segment disease.¹⁵ While the pathophysiology behind this finding is not entirely clear, it has recently been noted that enteric nervous system defects in Hirschsprung Disease extend beyond the area of aganglionosis, including decreased neuron fiber density, smaller ganglion size, and altered neurotransmitter expression.^16–18 These factors may contribute bowel dysfunction and dysmotility, which contribute to HAEC pathogenesis. Furthermore, long-segment disease is often diagnosed later in life, with more frequent pre-operative episodes of HAEC. Our analysis did not differentiate between pre-operative and post-operative episodes of HAEC.

The most significant limitation to this study was its inherent lack of external validation. A score will perform better on the data used to create it. External validation will be required in the future to further substantiate this model. Additionally, this study was retrospective in nature, and limited by the information available in the medical records. As a multi-institutional study, there were likely practice differences unaccounted for, and differences in radiologic interpretations and treatment algorithms between sites. Because there is no gold standard for the diagnosis of HAEC, this study was also limited to its own definition of HAEC, which was defined as the clinical decision to treat with rectal irrigations, bowel rest and antibiotics. This may have led to an overestimation of the incidence of HAEC. Indeed, we observed that 67% of the patients treated for HAEC did not have HAEC as a discharge diagnosis, suggesting that there is either significant under-diagnosis and/or over-treatment of HAEC. Furthermore, this study did not differentiate between pre-operative and post-operative HAEC episodes. Many of these limitations could be overcome with prospective data collection focused on this question.

Conclusion

Improved diagnosis of HAEC has the potential to reduce hospital admissions, limit morbidity and improve outcomes. This novel scoring system for HAEC has several advantages over previous scoring systems, but most importantly, may help reduce the rate of under-diagnosis of HAEC thereby allowing for earlier diagnosis and treatment of patients. Prospective studies will be required to externally validate the new score prior to widespread clinical implementation.