Development of a Gestational Age–Specific Case Definition for Neonatal Necrotizing Enterocolitis | Gastroenterology | JAMA Pediatrics | JAMA Network
[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.  Derivation of Study Data Set
Derivation of Study Data Set

The final data set included 3866 infants with 3866 abdominal radiography (AXR) images, with 888 infants with necrotizing enterocolitis (NEC) (204 confirmed by laparotomy) and 2978 infants without NEC.

Figure 2.  Clinical and Abdominal Radiography Findings Among Infants With Necrotizing Enterocolitis (NEC) and Without NEC by GA Group
Clinical and Abdominal Radiography Findings Among Infants With Necrotizing Enterocolitis (NEC) and Without NEC by GA Group

The clinical findings of blood in stool, abdominal discoloration, abdominal tenderness, and the grouped variable of increased and/or bilious aspirates and abdominal distension were included in the final model, as were the radiologic findings of pneumatosis and the grouped variable of pneumoperitoneum, fixed loop, and portal venous gas. Gestational age (GA) group 1 included infants with less than 26 weeks’ GA; group 2, infants with 26 to less than 30 weeks’ GA; group 3, infants with 30 to less than 37 weeks’ GA; and group 4, 37 or more weeks’ GA.

Figure 3.  Ordinal Necrotizing Enterocolitis (NEC) Score, Gestational Age (GA)–Specific Case Definition, and Corresponding Probabilities for NEC by GA Group
Ordinal Necrotizing Enterocolitis (NEC) Score, Gestational Age (GA)–Specific Case Definition, and Corresponding Probabilities for NEC by GA Group
Table 1.  Diagnostic Characteristics for Clinical and Radiological Signs of NEC
Diagnostic Characteristics for Clinical and Radiological Signs of NEC
Table 2.  NEC Score and Corresponding PPV and AUC by GA Group
NEC Score and Corresponding PPV and AUC by GA Group
1.
Vermont Oxford Network. Vermont Oxford Network database manual of operations: part 2: data definitions and data forms for infants born in 2013. https://public.vtoxford.org/wp-content/uploads/2014/03/Manual-of-Operations-Part-2-17_1.pdf. Accessed June 19, 2016.
2.
Kliegman  RM, Walsh  MC.  Neonatal necrotizing enterocolitis: pathogenesis, classification, and spectrum of illness.  Curr Probl Pediatr. 1987;17(4):213-288.PubMedGoogle Scholar
3.
Bell  MJ, Ternberg  JL, Feigin  RD,  et al.  Neonatal necrotizing enterocolitis: therapeutic decisions based upon clinical staging.  Ann Surg. 1978;187(1):1-7.PubMedGoogle ScholarCrossref
4.
Leistner  R, Piening  B, Gastmeier  P, Geffers  C, Schwab  F.  Nosocomial infections in very low birthweight infants in Germany: current data from the National Surveillance System NEO-KISS.  Klin Padiatr. 2013;225(2):75-80.PubMedGoogle ScholarCrossref
5.
Guthrie  SO, Gordon  PV, Thomas  V, Thorp  JA, Peabody  J, Clark  RH.  Necrotizing enterocolitis among neonates in the United States.  J Perinatol. 2003;23(4):278-285.PubMedGoogle ScholarCrossref
6.
Cust  AE, Darlow  BA, Donoghue  DA; Australian and New Zealand Neonatal Network (ANZNN).  Outcomes for high risk New Zealand newborn infants in 1998-1999: a population based, national study.  Arch Dis Child Fetal Neonatal Ed. 2003;88(1):F15-F22.PubMedGoogle ScholarCrossref
7.
Luig  M, Lui  K; NSW & ACT NICUS Group.  Epidemiology of necrotizing enterocolitis: part II: risks and susceptibility of premature infants during the surfactant era: a regional study.  J Paediatr Child Health. 2005;41(4):174-179.PubMedGoogle ScholarCrossref
8.
Kastenberg  ZJ, Lee  HC, Profit  J, Gould  JB, Sylvester  KG.  Effect of deregionalized care on mortality in very low-birth-weight infants with necrotizing enterocolitis.  JAMA Pediatr. 2015;169(1):26-32.PubMedGoogle ScholarCrossref
9.
Sharma  R, Hudak  ML, Tepas  JJ  III,  et al.  Impact of gestational age on the clinical presentation and surgical outcome of necrotizing enterocolitis.  J Perinatol. 2006;26(6):342-347.PubMedGoogle ScholarCrossref
10.
Equator Network. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. http://www.equator-network.org/reporting-guidelines/stard/. Accessed June 19, 2016.
11.
Spencer  A, Modi  N.  National neonatal data to support specialist care and improve infant outcomes.  Arch Dis Child Fetal Neonatal Ed. 2013;98(2):F175-F180.PubMedGoogle ScholarCrossref
12.
Shaughnessy  J, Garfield  J, Greer  B. Data handling. In: Bishop  A, Clements  K, Keitel  C, Kilpatrick  J, Laborde  C, eds.  International Handbook of Mathematics Education. New York, NY: Springer-Verlag; 1997:205-237.
13.
George  NI, Lu  T-P, Chang  C-W.  Cost-sensitive performance metric for comparing multiple ordinal classifiers.  Art Intell Res. 2016;5(1):135-143. doi:10.5430/air.v5n1p135Google Scholar
Original Investigation
March 2017

Development of a Gestational Age–Specific Case Definition for Neonatal Necrotizing Enterocolitis

Cheryl Battersby, MRCPCH1; Nick Longford, PhD1; Kate Costeloe, MD2; et al Neena Modi, MD1; for the UK Neonatal Collaborative Necrotising Enterocolitis Study Group
Author Affiliations
  • 1Neonatal Data Analysis Unit, Imperial College London, London, England
  • 2Barts and the London School of Medicine and Dentistry, London, England
JAMA Pediatr. 2017;171(3):256-263. doi:10.1001/jamapediatrics.2016.3633
Key Points

Question  Is it possible to develop a case definition for necrotizing enterocolitis (NEC) that discriminates between infants with and without the disease?

Findings  In this population study using data from 3866 infants, we developed a simple NEC score associated with the probability of NEC and identified gestational age–specific cutoff points. Less mature infants are less likely to present with pneumatosis or blood or mucus in their stool and more likely to present with gasless abdominal radiography findings.

Meaning  Consistent application of a gestational age–specific case definition offers an opportunity to strengthen global efforts to reduce the burden of neonatal NEC.

Abstract

Importance  Necrotizing enterocolitis (NEC) is a major cause of neonatal morbidity and mortality. Preventive and therapeutic research, surveillance, and quality improvement initiatives are hindered by variations in case definitions.

Objective  To develop a gestational age (GA)–specific case definition for NEC.

Design, Setting, and Participants  We conducted a prospective 34-month population study using clinician-recorded findings from the UK National Neonatal Research Database between December 2011 and September 2014 across all 163 neonatal units in England. We split study data into model development and validation data sets and categorized GA into groups (group 1, less than 26 weeks’ GA; group 2, 26 to less than 30 weeks’ GA; group 3, 30 to less than 37 weeks’ GA; group 4, 37 or more weeks’ GA). We entered GA, birth weight z score, and clinical and abdominal radiography findings as candidate variables in a logistic regression model, performed model fitting 1000 times, averaged the predictions, and used estimates from the fitted model to develop an ordinal NEC score and cut points to develop a dichotomous case definition based on the highest area under the receiver operating characteristic curves [AUCs] and positive predictive values [PPVs].

Exposures  Abdominal radiography performed to investigate clinical concerns.

Main Outcomes and Measures  Ordinal NEC likelihood score, dichotomous case definition, and GA-specific probability plots.

Results  Of the 3866 infants, the mean (SD) birth weight was 2049.1 (1941.7) g and mean (SD) GA was 32 (5) weeks; 2032 of 3663 (55.5%) were male. The total included 2978 infants (77.0%) without NEC and 888 (23.0%) with NEC. Infants with NEC in group 1 were less likely to present with pneumatosis (31.1% vs 47.2%; P = .01), blood in stool (11.8% vs 29.6%; P < .001), or mucus in stool (2.1% vs 5.6%; P = .048) but more likely to present with gasless abdominal radiography findings (6.3% vs 0.9%; P = .009) compared with infants with NEC in group 3. In the ordinal NEC score analysis, we allocated 3 points to pneumatosis, 2 points to blood in stool, and 1 point each to abdominal tenderness and abdominal discoloration; 1 point was assigned if 1 or more of pneumoperitoneum, fixed loop, and portal venous gas were present, and 1 point was assigned if both increased and/or bilious aspirates and abdominal distension were present. The cutoff scores for the dichotomous GA-specific case definition were 2 or greater for infants in groups 1 and 2, 3 or greater for infants in group 3, and 4 or greater for infants in group 4. The ordinal NEC score and dichotomous case definition discriminated well between infants with (AUC, 87%) and without (AUC, 80%) NEC. The case definition has a sensitivity of 63.9% (95% CI, 60.6-67.0), a specificity of 96.8% (95% CI, 96.1-97.4), an AUC of 80.0% (95% CI, 79-82), and a PPV of 85.5% (95% CI, 82.6-88.1). Applying the cut points to the 431 infants who underwent a laparotomy yielded a sensitivity of 76.5% (95% CI, 70.0-82.1), a specificity of 74.4% (95% CI, 68.3-80.0), an AUC of 75.0% (95% CI, 71.0- 80.0), and a PPV of 72.9% (95% CI, 66.4-78.7).

Conclusions and Relevance  The risk of NEC and clinical presentation are associated with GA. Adoption of a consistent GA-specific case definition would strengthen global efforts to reduce the population burden of this devastating neonatal disease.

Introduction

Necrotizing enterocolitis (NEC) is a serious gastrointestinal inflammatory disease predominantly but not exclusively affecting the preterm infant. Presenting signs are often nonspecific, and diagnosis of lesser degrees of disease severity can be difficult. The lack of a consistent case definition adds to the difficulties of determining true disease burden, synthesizing the results of clinical trials, progressing preventive and therapeutic research, and evaluating the effectiveness of quality improvement interventions. Commonly used definitions include that of the Vermont Oxford Network1 and Bell staging criteria2,3; of note is that the latter was developed as a criterion for staging after the diagnosis was made, not as a case definition. Other definitions include those from the US Centers for Disease Control and Prevention4 and varying groupings of clinical and radiological findings used by individual study authors.5-8 None of these definitions are evidence-based or validated or incorporate gestational age (GA), although this influences the risk of NEC.9 Our aim was to develop a GA-specific case definition for NEC to facilitate research, surveillance, and quality improvement.

Methods
Study Design, Data Source, and Regulatory Approvals

This study has been reported according to Standards for Reporting Diagnostic Accuracy Studies guidelines.10 Daily clinical information on infants admitted to neonatal units in England is recorded in a point-of-care, clinician-entered electronic patient record. A defined data extract, the Neonatal Data Set (National Health Service Information Standard ISB1595) is transmitted quarterly to the Neonatal Data Analysis Unit at the Imperial College London and Chelsea and Westminster Hospital National Health Service Foundation Trust in London, England, where patient episodes across different hospitals are merged and data are cleaned and entered into the National Neonatal Research Database (NNRD).11 More than 400 data items are held on the NNRD and comprise static basic demographic details (eg, month and year of birth, birth weight, and gestational age) applicable to all infants, episodic data (eg, blood culture, clinical outcomes, and diagnoses), and daily data (eg, respiratory support, feeding, surgical procedures, and drugs received). Each data item is clearly defined in an accompanying meta-data set and mapped to existing national standards as well as International Classification of Diseases codes. Diagnoses include fixed choice and free-text items with the pseudo-anonymized National Health Service number as the principal identifier.

Neonatal units contributing to the NNRD are known as the UK Neonatal Collaborative. The NNRD is approved by the National Research Ethics Service (10/H0803/151), Confidentiality Advisory Group of the Health Research Authority (8-05(f)/2010) and the Caldicott Guardians and Lead Clinicians of contributing hospitals. This investigation is a component of the UK Neonatal Collaborative-NEC study (UK Clinical Research Network Portfolio ID 11853; National Research Ethics Service ref 11/LO/1430). Anonymous data held in the NNRD were captured during the course of clinical care for patient management and other purposes. The research team had no contact with patients and their families, and informed consent was not required (REC approval no LO/1430). We invited clinical leads from all neonatal units in England to participate. Clinical staff prospectively recorded a predefined list of clinical and radiography findings in the electronic patient record of any infant who had abdominal radiography (AXR) to investigate gastrointestinal concerns.

Data Extraction

From the NNRD, we extracted data on all infants who had clinical and AXR findings recorded between November 2011 and September 2014. These findings comprised demographic data (eg, GA and birth weight); diagnoses; procedures; clinical opinion on the certainty of NEC diagnosis; whether NEC was confirmed by visual inspection of bowel, histology and/or autopsy; and daily NEC treatment (medical or surgical). Diagnoses of NEC were made by local clinical and surgical teams across 163 neonatal units and 21 surgical centers. Spontaneous intestinal perforation was considered a distinct entity and was not included in the NEC analyses.

We identified infants who received a laparotomy and in whom the diagnosis of NEC was either confirmed or refuted and infants without laparotomy for whom the final diagnosis was considered unequivocally yes or no by the attending clinician; data were excluded if the diagnosis of NEC was uncertain. For infants who received a laparotomy, we included information from the first AXR image indicating that a laparotomy had been performed. For infants who did not undergo a laparotomy but had multiple AXR images, we used computer-generated code to select 1 image at random for each infant. We performed internal cross-validation for the outcomes of NEC or no NEC by comparing these data with other data held on the infant (eg, diagnoses, daily NEC diagnosis, and procedures) and excluded infants for whom data were inconsistent. For the infants retained in the analysis, we noted the following variables: GA, birth weight, clinical findings (increased and/or bilious aspirates, blood in stool, mucus in stool, abdominal discoloration, abdominal distension, abdominal mass, and abdominal tenderness) and radiological findings (gasless, portal venous gas, fixed loop, pneumatosis, and pneumoperitoneum). We categorized GA into groups (group 1, less than 26 weeks’ GA; group 2, 26 to less than 30 weeks’ GA; group 3, 30 to less than 37 weeks’ GA; group 4, 37 or more weeks’ GA). We calculated birth weight z scores and categorized these into tertiles. We collated variables with very low (defined as less than 10%) or high (defined as greater than 30%) prevalence among infants with and without NEC into a single composite variable.

Statistical Methods

We investigated whether presentation of NEC varied with GA by comparing clinical and radiological findings among infants with and without NEC using the χ2 and Fisher exact tests, as appropriate. For each clinical and radiological sign, we determined the odds ratio for NEC, sensitivity, specificity, positive predictive value (PPV), and area under the receiver operating characteristic curve (AUC).

We split the data set at random using a computer-generated code into model development and validation data sets of equal size. On the model development data set, we used stepwise logistic regression to determine the model that best predicted the probability of NEC. We assessed the quality of this prediction model using the validation data set. We replicated this procedure 1000 times. All available covariates (clinical and radiological findings, birth weight, and GA) were entered into the model, and the covariate with the lowest absolute t ratio was removed at each step. Model fitting and validation were performed 1000 times and the resulting predictions averaged. We calculated the absolute deviation,12 defined as the difference between the probability of NEC based on the model (in the range from 0 to 1 for every infant) and the diagnosis of NEC (1 for NEC and 0 for no NEC). Infants with an absolute deviation greater than 0.5 were considered to have an incorrect prediction. We selected the model with the lowest rate of incorrect predictions.

To determine the points allocated for the clinical and/or radiological signs in an ordinal NEC score, we rounded the coefficients from the model up or down to an integer and assessed the corresponding AUC and PPV. For each GA group, we found the cut point, which minimized misclassification13 by applying 2 conditions: (1) a PPV exceeding 60% with (2) the highest AUC. We conducted a sensitivity analysis by applying the definition only to infants who had their diagnoses confirmed by laparotomy. Finally, we compared the performance of the case definition with that from the Vermont Oxford Network,1 defined as at least 1 clinical sign (bilious aspirate or vomiting, abdominal distension, or blood in the stool) and at least 1 radiological finding (pneumatosis, hepatobiliary gas, or pneumoperitoneum). All analyses were performed in R and Stata version 11.0 (StataCorp). Statistical significance was set at P < .05.

Results

We secured the participation of all 163 neonatal units (43 special care units, 76 high dependency units, and 44 intensive care units) in England across 23 clinical networks. We included 3866 AXR images from 3866 infants, including 2978 (77.0%) without NEC and 888 (23.0%) with NEC, of which 204 were confirmed by laparotomy and the rest by clinician ascertainment (Figure 1). Of the 3866 infants, sex was unknown for 5 and missing for 198; 2032 of 3663 (55.5%) were male, and the mean (SD) birth weight was 2049.1 (1941.7) g. Infants with NEC tended to be preterm (median [interquartile range] GA: infants with NEC, 28 [25-30] weeks; infants without NEC, 35 [29-39] weeks). Among infants with NEC, the most frequent findings were abdominal distension, pneumatosis, increased and/or bilious aspirates, and abdominal tenderness (Table 1). Among infants without NEC, the most frequent clinical findings were abdominal distension and increased and/or bilious aspirates (Table 1). In the whole cohort, a gasless abdomen, portal venous gas, abdominal mass, fixed loop, pneumoperitoneum, and mucus in the stool were rare findings, with prevalences below 10%. Except for gasless abdomen and abdominal mass, all other findings were significantly different between infants with and without NEC (Table 1). No single clinical or radiological finding discriminated between infants with and without NEC, as evidenced by the relatively low AUC. Even pneumatosis, which on its own increased the odds of NEC 76-fold, had an AUC of only 71%. In subsequent analyses, we found that the prediction of NEC was improved using combinations of clinical signs and AXR findings. Comparing infants in group 1 with group 3, infants with NEC born at lower GAs were less likely to present with pneumatosis (31.1% vs 47.2%; P = .01), blood in stool (11.8% vs 29.6%; P < .001), or mucus in stool (2.1% vs 5.6%; P = .048) but more likely to present with gasless AXR findings (6.3% vs 0.9%; P = .009) compared with more mature infants with NEC (Figure 2).

Predictive Model for NEC and Ordinal NEC Score

We grouped 3 radiological findings (pneumoperitoneum, fixed loop, and portal venous gas), each with a very low frequency, to create a new variable, PFP, assigning a value of 1 if 1 or more of these variables were present and a value of 0 if none were present. We grouped 2 clinical findings (increased and/or bilious aspirates and abdominal distension), each with relatively high frequencies, into another variable, AA, assigning a value of 1 if both findings were present and a value of 0 otherwise. The model with the lowest rate of incorrect predictions included 6 variables (blood in stool, abdominal discoloration, abdominal tenderness, pneumatosis, PFP, and AA) and GA group. The coefficients were roughly in the proportion 3:2:1 for pneumatosis (3.855), blood in stool (2.760), and the other signs (1.46 for abdominal discoloration, 1.46 for abdominal tenderness, 1.65 for 1 or more of PFP, and 0.83 for AA). Therefore, to create an ordinal NEC score, we allocated 3 points to pneumatosis and 2 points to blood in stool. One point each was allocated to abdominal tenderness discoloration, abdominal tenderness, PFP, and AA. The sum of points provided the ordinal NEC score. This had an AUC of 0.88 (95% CI, 0.86-0.89), indicating that it discriminated well between infants with and without NEC.

GA-Specific NEC Score Cut Points

The NEC scores were perfectly aligned with the corresponding PPV (ie, the higher the score, the higher the PPV) but not with the AUC (Table 2). Applying the 2 conditions of highest AUC and PPV yielded the same cut point of 2 for GA groups 1 and 2 but different cut points for GA groups 3 and 4. A cut point of 2 provided the highest AUC for GA group 3, but the PPV was only 41.0%; therefore, we selected a cut point of 3 (PPV, 70.7%), with the trade-off of a slightly lower AUC (80% vs 84%). Similarly, for GA group 4, we selected a cut point of 4 even though a cut point of 2 provided the highest AUC because the PPV was only 8.3%, again accepting a slightly lower AUC (75% vs 84%). The final cut points for the NEC score were 2 for GA groups 1 and 2, 3 for GA group 3, and 4 for GA group 4. Applying this to the entire data set yielded a sensitivity of 63.9% (95% CI, 60.6-67.0), a specificity of 96.8% (95% CI, 96.1-97.4), an AUC of 80.0% (95% CI, 79-82), and a PPV of 85.5% (95% CI, 82.6-88.1). Figure 3 provides a schematic summary of the NEC score, its use in a dichotomous case definition with GA-specific cut points, and GA-specific probability plots.

Sensitivity Analysis for the Subset of Infants Who Received a Laparotomy

Applying the cut points to the 431 infants who underwent a laparotomy yielded a sensitivity of 76.5% (95% CI, 70.0-82.1), a specificity of 74.4% (95% CI, 68.3-80.0), an AUC of 75.0% (95% CI, 71.0- 80.0), and a PPV of 72.9% (95% CI, 66.4-78.7).

Comparison With Vermont Oxford Network Definition

The estimated misclassification rates for applying our NEC score and the Vermont Oxford Network definition were 10.5% and 13.7%, respectively. The difference between the estimated error rates was 3.1%, which was statistically significant (P < .001). The Vermont Oxford Network definition applied to the entire data set yielded a sensitivity of 49.7% (95% CI, 46.3-53.0), a specificity of 97.2% (95% CI, 97.6-97.8), an AUC of 73.0% (95% CI, 72.0-75.0), and a PPV of 84.2% (95% CI, 80.7-87.2); the subset of infants who underwent a laparotomy yielded a sensitivity of 68.1% (95% CI, 61.3-74.5), a specificity of 71.8% (95% CI, 65.5-77.6), an AUC of 70.0% (95% CI, 66.0-74.0), and a PPV of 68.5% (95% CI, 61.6-74.8) (eTable in the Supplement).

Discussion

Using a large population sample, we confirmed previous observations that risk and clinical presentation of NEC are associated with GA.9 We showed that a combination of findings rather than a single clinical or AXR finding provided the highest diagnostic accuracy. We developed an ordinal NEC score corresponding to the GA-specific probability of NEC, a dichotomous case definition with GA-specific cutoffs, and showed that these perform favorably compared with the widely used Vermont Oxford case definition. Because our aim was to develop a simple and pragmatic case definition suitable for widespread clinical application, we concluded that it was worth sacrificing some precision to develop a case definition that would be easy to use. Therefore, the coefficients in the model were rounded up or down to an integer to create an ordinal NEC score.

The strength of our study is that it is based on a large population data set using information captured as part of each infant’s clinical care. In contrast to widely used criteria, such as the Vermont Oxford case definition and Bell staging, we incorporated GA into our case definition. Of note is that our conclusions remained robust to the sensitivity analysis performed on the subset of infants who received a laparotomy in whom the diagnosis was secure. We also noted that restricting analysis to infants with NEC confirmed at laparotomy would not represent the population for which the case definition is intended. Until there is a reliable noninvasive diagnostic test or biomarker for the disease in infants who do not receive a laparotomy, this challenge cannot be overcome. This study was based on a pragmatic design using a comprehensive national data set.

The 2 conditions used to select the cut point for the dichotomous case definition were selected a priori. The highest AUC (maximum sensitivity and specificity) was applied to avoid overreporting and underreporting, as false negatives and false positives are viewed as equally important when developing a case definition for which the primary purpose is use in clinical research, surveillance, and quality improvement as opposed to clinical treatment decisions. The AUC is a trade-off between sensitivity and specificity, and therefore, we applied the second condition requiring the PPV to exceed 60%. Positive predictive value is influenced by prevalence, but sensitivity, specificity, and AUC are not. We accept that a more stringent definition with a lower sensitivity and higher specificity may also be appropriate. Our case definition was not developed for use to guide clinical decision making but nonetheless does provide an objective measure that, in part, reflects pooled clinician judgment. Additionally, the GA specific probability plots may provide clinical utility in illustrating the likelihood of a true NEC for any given score.

Limitations

Our study had limitations. We recognize that the use of clinical ascertainment in addition to visual inspection at laparotomy and/or histological evidence to identify unequivocal NEC may be considered a limitation. Furthermore, the decision as to whether an infant had NEC or spontaneous intestinal perforation was determined by the attending medical and surgical teams, although we acknowledge the difficulty of reliably separating spontaneous intestinal perforation from NEC.

Conclusions

Necrotizing enterocolitis is a disease feared by parents and clinicians because it can strike suddenly and with little warning and has a high mortality and morbidity, with preventive and therapeutic options that remain limited. Basic and applied research, complemented by quality improvement initiatives underpinned by rigorous surveillance, are essential to tackle this disease, but meta-analysis and interpretation of studies requires consistency and hence comparability of case definitions. We suggest that the ordinal score and dichotomous case definition we provide offers opportunity to strengthen global efforts to reduce the burden of neonatal NEC.

Back to top
Article Information

Corresponding Author: Neena Modi, MD, Neonatal Data Analysis Unit, Section of Neonatal Medicine, Department of Medicine, Imperial College London, 369 Fulham Rd, Room G.4.2, 4th Floor, Lift Bank D, London SW10 9NH, England (n.modi@imperial.ac.uk).

Accepted for Publication: September 21, 2016.

Correction: This article was corrected on May 1, 2017, to fix errors in the estimates of sensitivity and specificity.

Published Online: January 3, 2017. doi:10.1001/jamapediatrics.2016.3633

Author Contributions: Ms Battersby had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Battersby, Costeloe, Modi.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Battersby, Costeloe, Modi.

Critical revision of the manuscript for important intellectual content: Battersby, Longford, Modi.

Statistical analysis: Battersby, Longford.

Obtained funding: Modi.

Administrative, technical, or material support: Battersby, Modi.

Study supervision: Costeloe, Modi.

Conflict of Interest Disclosures: In the last 5 years, Dr Modi has received grants from the National Institute for Health Research, Medical Research Council, the British Heart Foundation, Action Medical Research, HCA International, Westminster Medical School Research Trust, Bliss, the Department of Health of the United Kingdom, NHS England, and NHS London as well as consultancy fees from Ferring International. No other disclosures were reported.

Funding/Support: This study represents independent research funded by the National Institute for Health Research under a program grant for applied research, medicines for neonates (grant reference No. RP-PG-0707-10010) (Dr Modi).

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed are those of the authors and not the National Health Service, the National Institute for Health Research, or the Department of Health of the United Kingdom.

Group Information: Data from the following UK Neonatal Collaborative Necrotising Enterocolitis Study Group members were included: Cheshire and Merseyside: Arrowe Park Hospital: O. Rackham; Countess of Chester Hospital: S. Brearey; Leighton Hospital: A. Thirumurugan; Liverpool Women’s NHS Foundation Trust: N. Subhedar; Macclesfield District General Hospital: G. Whitehead; Ormskirk District General Hospital: T. McBride; Warrington Hospital: C. Zipitis, D. Webb, and H. Satish; and Whiston Hospital: L. Chilukuri. East of England Perinatal Network: Basildon Hospital: N. Sharief; Bedford Hospital: R. Kadalraja and A. Mittal; Broomfield Hospital: R. N. Mahesh Babu; Colchester General Hospital: S. Dalton; Hinchingbrooke Hospital: H. Dixon; Ipswich Hospital: M. James; James Paget Hospital: V. Jayalal; Lister Hospital: J. Kefas; Luton and Dunstable Hospital: J. Birch; Norfolk and Norwich University Hospital: M. Dyke; Peterborough City Hospital: S. Babiker; Princess Alexandra Hospital: T. Soe; Queen Elizabeth Hospital: S. Rubin; Rosie Maternity Hospital: A. Ogilvy-Stuart; Southend Hospital: A. Khan; Watford General Hospital: S. Narayanan; and West Suffolk Hospital: I. Evans. Greater Manchester: North Manchester General Hospital: N. Panasa; Royal Oldham Hospital: J. Moise and N. Maddock; Royal Albert Edward Infirmary: C. Zipitis; Royal Bolton Hospital: C. Turner; St Mary’s Hospital: N. Edi-Osagie and E. Gasiorowski; Stepping Hill Hospital: C. Heal; Tameside General Hospital: J. Birch and A. Date; and University Hospital of South Manchester: A. Elazabi. Kent and Medway: Darent Valley Hospital: A. Hasib; Maidstone and Tunbridge Wells Hospital: H. Kisat; Medway Maritime Hospital: G. Ramadan; and East Kent Hospitals University NHS Foundation Trust: V. Vasu. Lancashire and Cumbria: Furness General Hospital: A. Olabi; Royal Lancaster Infirmary: J. Fedee; Lancashire Women and Newborn Centre: S. Sivashankar; Royal Preston Hospital: R. Gupta; and Victoria Hospital: C. Rawlingson. Midlands Central: George Eliot Hospital and University Hospital Coventry: P. Satodia; Kettering General Hospital: P. Rao; Northampton General Hospital and Warwick Hospital: F. Thompson and S. Gupta; and Queen’s Hospital Burton on Trent: A. Manzoor. Midlands North Staffordshire, Shropshire, and Black Country: Manor Hospital: A. K. Bhaduri; New Cross Hospital: A. Skinner; Royal Shrewsbury Hospital: S. Deshpande; Russells Hall Hospital: T. Pillay; Staffordshire General Hospital: K. K. Tewary; and University Hospital of North Staffordshire: K. Palmer. Midlands South West: Alexandra Hospital: A. Short; Worcestershire Royal Hospital: A. Gallagher; Birmingham City Hospital: J. Nycyk; Birmingham Heartlands Hospital: R. Mupanemunda; Good Hope Hospital: J. Meran; Birmingham Women’s Hospital: I. Morgan and A. Bedford-Russell; and Hereford County Hospital: H. C. Underhill. North Central London: Barnet Hospital and Chase Farm Hospital: T. Wickham; The Royal Free Hospital: V. van Someren; University College Hospital: S. Watkin; and Whittington Hospital: R. Blumberg. North East London: Homerton Hospital: N. Aladangady; King George Hospital and Queen’s Hospital: B. Sharma; Newham General Hospital and North Middlesex University Hospital: L. Alsford; and The Royal London Hospital and Whipps Cross University Hospital: C. Sullivan. North Trent: Barnsley District General Hospital: S. Hamdan; Bassetlaw District General Hospital: H. Mulenga; Diana Princess of Wales Hospital: P. Adiotomre; Scunthorpe General Hospital: A. Jackson; Doncaster Royal Infirmary: J. S. Ahmed; Chesterfield and North Derbyshire Royal Hospital: A. Foo; Rotherham District General Hospital: C. Harrison; and Sheffield Teaching Hospital: E. Pilling. North West London: Chelsea and Westminster Hospital: S. Uthaya; Ealing Hospital: R. Mathur; Hillingdon Hospital: M. Cruwys; Northwick Park Hospital: C. Philipp and R. Nicholl; and West Middlesex University Hospital: E. Eyre. Northern: Cumberland Infirmary and West Cumberland Hospital; P. Whitehead and M. Ben-Hamida; Darlington Memorial Hospital and University Hospital of North Durham: D. A. Bowes; James Cook University Hospital: N. Sabrine; Queen Elizabeth Hospital: D. Bosman; Royal Victoria Infirmary: N. Embleton; South Tyneside District Hospital: R. Bolton; Sunderland Royal Hospital: M. Abu-Harb; University Hospital of North Tees: C. Harikumar; and Wansbeck General Hospital: J. Olivier. Peninsula: Derriford Hospital: N. Maxwell; North Devon District Hospital: Y. Cherinet; Royal Cornwall Hospital: P. Munyard; Royal Devon and Exeter Hospital: N. Osbourne; and Torbay Hospital: M. Raman. South East London: Guy’s and St Thomas’ Hospital: K. Turnock; King’s College Hospital: A. Hickey; Princess Royal University Hospital and Queen Elizabeth Hospital: O. Banjoko; and University Hospital Lewisham: J. Kuna. South West London: Croydon University Hospital: A. Kumar; Epsom General Hospital: K. Watts; St Helier Hospital: R. Shephard; Kingston Hospital: D. Lindo; and St George’s Hospital: L. De Rooy. South Central (North and South): Basingstoke and North Hampshire Hospital: R. Wigfield; Dorset County Hospital: P. Wylie; Milton Keynes Foundation Trust Hospital: I. Misra; Oxford University Hospitals and Horton Hospital: N. Shettihalli; John Radcliffe Hospital: E. Adams; Poole Hospital NHS Foundation Trust: M. Khashu; Princess Anne Hospital: F. Pearson; Queen Alexandra Hospital: C. Groves; Royal Berkshire Hospital: P. de Halpert; Royal Hampshire County Hospital: D. Schapira; Salisbury District Hospital: N. Brown; St Mary’s Hospital Isle of Wight: C. Burtwell; St Richard’s Hospital: N. Brennan; Stoke Mandeville Hospital: S. Salgia; and Wexham Park Hospital: R. Sanghavi. Surrey and Sussex: Conquest Hospital: G. Whincup; East Surrey Hospital: K. Khader; Frimley Park Hospital: A. Mallik; Princess Royal Hospital and Royal Sussex County Hospital: P. Amess; Royal Surrey County Hospital: M. Hardo; St Peter’s Hospital: P. Reynolds; and Worthing Hospital: E. Vamvakiti. Trent: King’s Mill Hospital: V. Noble; Lincoln County Hospital and Pilgrim Hospital: A. S. Rao; Nottingham City Hospital and Nottingham University Hospital: S. Wardle and J. Dorling; and Royal Derby Hospital: M. Ratnayaka. Western: Gloucestershire Royal Hospital: J. Holman; Great Western Hospital: S. Zengeya; Royal United Hospital: S. Jones; Southmead Hospital: P. Mannix; St Michael’s Hospital: P. Cairns; Taunton and Somerset Hospital: R. J. Mann; and Yeovil District Hospital: M. Eaton. Yorkshire: Airedale General Hospital: M. Babirecki; Bradford Royal Infirmary: S. Oddie; Calderdale Royal Hospital: K. Schwarz; Dewsbury and District Hospital: D. Gibson; Harrogate District Hospital: C. Jampala; Hull Royal Infirmary: K. Green and J. Preece; Leeds Neonatal Service: K. Johnson; Scarborough General Hospital: A. Hawkridge; and York District Hospital: G. Millman.

Additional Contributions: We acknowledge the contribution of clinical staff from neonatal units in England; assistance from analysts Yevgeniy (Eugene) Statnikov, MSc, and Daniel Gray, MA, and manager Richard Colquhoun, BSc (Neonatal Data Analysis Unit, Imperial College London, London, England); and support from members of the boards of the Neonatal Data Analysis Unit and Medicines for Neonates Programme, including Zoe Chivers, MSc (Bliss Charity, London, England), Deborah Ashby, PhD (Imperial College London, London, England), Peter Brocklehurst, FRCOG (University of Birmingham, Birmingham, England), Elizabeth Draper, PhD (University of Leicester, Leicester, England),Michael Goldacre, FRCP (University of Oxford, Oxford, England), Jacquie Kemp, MSc (NHS England, Leeds, England), Azeem Majeed, MD (Imperial College London, London, England), Stavros Petrou, PhD (The University of Warwick, Coventry, England), Andrew Wilkinson, FRCPCH (University of Oxford, Oxford, England), and Alys Young, PhD (The University of Manchester, Manchester, England). None of the contributors were compensated for their work.

References
1.
Vermont Oxford Network. Vermont Oxford Network database manual of operations: part 2: data definitions and data forms for infants born in 2013. https://public.vtoxford.org/wp-content/uploads/2014/03/Manual-of-Operations-Part-2-17_1.pdf. Accessed June 19, 2016.
2.
Kliegman  RM, Walsh  MC.  Neonatal necrotizing enterocolitis: pathogenesis, classification, and spectrum of illness.  Curr Probl Pediatr. 1987;17(4):213-288.PubMedGoogle Scholar
3.
Bell  MJ, Ternberg  JL, Feigin  RD,  et al.  Neonatal necrotizing enterocolitis: therapeutic decisions based upon clinical staging.  Ann Surg. 1978;187(1):1-7.PubMedGoogle ScholarCrossref
4.
Leistner  R, Piening  B, Gastmeier  P, Geffers  C, Schwab  F.  Nosocomial infections in very low birthweight infants in Germany: current data from the National Surveillance System NEO-KISS.  Klin Padiatr. 2013;225(2):75-80.PubMedGoogle ScholarCrossref
5.
Guthrie  SO, Gordon  PV, Thomas  V, Thorp  JA, Peabody  J, Clark  RH.  Necrotizing enterocolitis among neonates in the United States.  J Perinatol. 2003;23(4):278-285.PubMedGoogle ScholarCrossref
6.
Cust  AE, Darlow  BA, Donoghue  DA; Australian and New Zealand Neonatal Network (ANZNN).  Outcomes for high risk New Zealand newborn infants in 1998-1999: a population based, national study.  Arch Dis Child Fetal Neonatal Ed. 2003;88(1):F15-F22.PubMedGoogle ScholarCrossref
7.
Luig  M, Lui  K; NSW & ACT NICUS Group.  Epidemiology of necrotizing enterocolitis: part II: risks and susceptibility of premature infants during the surfactant era: a regional study.  J Paediatr Child Health. 2005;41(4):174-179.PubMedGoogle ScholarCrossref
8.
Kastenberg  ZJ, Lee  HC, Profit  J, Gould  JB, Sylvester  KG.  Effect of deregionalized care on mortality in very low-birth-weight infants with necrotizing enterocolitis.  JAMA Pediatr. 2015;169(1):26-32.PubMedGoogle ScholarCrossref
9.
Sharma  R, Hudak  ML, Tepas  JJ  III,  et al.  Impact of gestational age on the clinical presentation and surgical outcome of necrotizing enterocolitis.  J Perinatol. 2006;26(6):342-347.PubMedGoogle ScholarCrossref
10.
Equator Network. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. http://www.equator-network.org/reporting-guidelines/stard/. Accessed June 19, 2016.
11.
Spencer  A, Modi  N.  National neonatal data to support specialist care and improve infant outcomes.  Arch Dis Child Fetal Neonatal Ed. 2013;98(2):F175-F180.PubMedGoogle ScholarCrossref
12.
Shaughnessy  J, Garfield  J, Greer  B. Data handling. In: Bishop  A, Clements  K, Keitel  C, Kilpatrick  J, Laborde  C, eds.  International Handbook of Mathematics Education. New York, NY: Springer-Verlag; 1997:205-237.
13.
George  NI, Lu  T-P, Chang  C-W.  Cost-sensitive performance metric for comparing multiple ordinal classifiers.  Art Intell Res. 2016;5(1):135-143. doi:10.5430/air.v5n1p135Google Scholar
×