Secondary Logo

Journal Logo

Original Articles: Nutrition

Cost-effectiveness Analysis of Feeding Guidelines for Infants Following Intestinal Surgery

Shores, Darla R.; Mogul, Douglas; Allen, Julia; Delarmente, Benjo A.; Padula, William§

Author Information
Journal of Pediatric Gastroenterology and Nutrition: May 2020 - Volume 70 - Issue 5 - p 657-663
doi: 10.1097/MPG.0000000000002642


What Is Known

  • Infants with intestinal failure are at risk of developing intestinal failure–associated liver disease.
  • Intestinal failure impacts health care costs and quality of life.
  • Improving enteral nutrition can reduce risk.

What Is New

  • Postoperative feeding guideline use in infants with intestinal failure reduced the risk of intestinal failure–associated liver disease, and was cost-saving in both the short- and long term.

Infants at risk of intestinal failure due to surgical gastrointestinal disorders such as necrotizing enterocolitis (NEC), gastroschisis, spontaneous intestinal perforation, intestinal atresia, or volvulus frequently have feeding difficulty due to malabsorption and/or postoperative dysmotility (1,2). Such infants often have prolonged parenteral nutrition (PN) requirements, increasing the risk of sepsis from central line–associated bloodstream infections, intestinal failure–associated liver disease (IFALD), and prolonged hospitalization (3,4). When coupled with limited enteral nutrition (EN), moderate-severe IFALD can progress to chronic liver disease and liver failure that necessitates multivisceral transplantation of the liver and small bowel (5). EN, which improves intestinal barrier function, reduces the risk of IFALD (6).

To improve the time to achieve enteral autonomy and reduce the risk of IFALD, we implemented postoperative feeding guidelines in our neonatal intensive care unit (NICU) that promoted faster advancement of EN compared to standard practice. We previously reported significantly reduced time to reach 50% calories from EN and to achieve enteral autonomy (100% enteral feeding), reduced IFALD incidence, and reduced severity of IFALD in the feeding guideline group compared to similar case mix of historical controls (7). As NEC can occur postoperatively in 10% of subjects, this was also evaluated as a safety measure.

The morbidity and economic impact of complications related to NEC and sepsis in infants can be substantial, and are at least in part due to nutrition-related complications associated with prolonged PN use and hospitalization (8–10). The short-term and long-term economic impact of improving EN and reducing liver disease is not well described (11). The aim of this study was to assess the costs, resource utilization, and resultant short-term and long-term benefits associated with the feeding guideline use by performing a cost-effectiveness analysis (CEA). We hypothesized that investment in provider time and health care resources to maintain adherence to these guidelines was cost-saving.


Study Design

A CEA was performed using a Markov model built in Microsoft Excel (Microsoft, Redmond, WA). The model was designed to reflect the health care sector and societal perspectives to assess the value and efficiency of guideline use versus standard feeding (Fig. 1). These alternatives explored short- and long-term outcomes in pediatrics for up to 160 days. The model was designed to produce an incremental cost-effectiveness ratio (ICER) in terms of US Dollars per total hospital days, referred to as hospital “stay.”

Short-term health state transitions. Infants entered into the model in the immediate postoperative period (point 1). Infants could remain without IFALD and without other complications (necrotizing enterocolitis [NEC] or sepsis), or they could transition to any combination of those health states (points 2–5). Ultimately, infants transitioned to discharge or death as a final disposition (points 6 or 7). IFALD = intestinal failure–associated liver disease.

Study Population

We used historical and prospective cohorts of surgical infants fed via recently developed guidelines (N = 81) from October 2013 to December 2016 that were implemented as standard of care, compared to historical controls (N = 83) from January 2007 to June 2013 (7). Previous results have been published (7). Subjects were younger than 6 months who had surgical intervention at our institution for NEC, spontaneous intestinal perforation, gastroschisis, intestinal atresia, and volvulus in the NICU. There was no statistical difference in demographics and baseline characteristics between groups (Table 1). The mix of primary diagnoses was similar, and the majority of subjects in both groups underwent small bowel resection (60% vs 53%). Length of residual small bowel was similar between groups.

Demographic and clinical characteristics

Feeding Guidelines

The feeding guideline algorithm was previously published (12). The guidelines used higher initial EN volumes (20 mL · kg−1 · day−1) and daily advancement (20 mL · kg−1 · day−1) if tolerated, along with implementation of weekly multidisciplinary rounds to sustain adherence. Before guideline implementation, the median initial feeding volume was 10 mL · kg−1 · day−1, and baseline median time to reach 50% of calories from EN was 11 days (12). Breast milk and donor milk were preferentially used in the guidelines, followed by elemental formula if feeding tolerance occurred. In both groups, only soy lipid emulsion was used and it was restricted once cholestasis occurred (direct bilirubin >2 mg/dL). Calories from PN were reduced as EN was advanced.

Decision Model—Short-term Outcomes

The health outcomes of interest were moderate-severe IFALD (direct bilirubin >5 mg/dL in setting of PN use and absence of other known liver disease), postoperative complications, defined as NEC or sepsis after resuming EN, and death. The health states captured in the model included: No IFALD and No Complications (no postoperative NEC or sepsis); +IFALD, No Complications; +IFALD, 1 Complication (either postoperative NEC or sepsis); +IFALD, 2 Complications (both postoperative NEC and sepsis); No IFALD, Complications (no IFALD but postoperative NEC or sepsis); Death; and Discharge. Infants began cycling in the model in the immediate postoperative period (time zero) in the No IFALD/No Complications state and could then progress to the complication health states 2 to 5. They ultimately transitioned to hospital discharge or death. Each cycle within the model was 1 hospital day, and the total time horizon was 160 days (or cycles), the average NICU stay for a premature surgical infant. The simulation was developed using 100 patients in the guideline group and 100 patients in the standard group.

Model Input Parameters


The costs associated with each health state were tabulated using both cost data obtained from the hospital and marginal costs of treating postoperative complications (NEC and sepsis) available in the literature (Supplemental Digital Content 1, The deterministic base case cost per day of treating patients with no IFALD and no complications was determined using data obtained from Johns Hopkins Hospital (DataMart), and provided the base cost for all other treatment costs. The cost of baseline physician payments for neonatal critical care per day was determined using current procedural terminology physician billing codes 99468, 99469, 99471, and 99472 (13). We assumed that given the distance in time between experimental and control arms data collection, that despite major interruptions such as the Affordable Care Act, there are no changes in utilization or pricing of the treatment that would induce different behaviors than what is exhibited in the model. All costs were measured in 2017 dollars in the model to standardize costs between periods (14).

Treatment for NEC and sepsis was similar and included cessation of EN, systemic antibiotics, and PN (15). The daily marginal costs for treating these complications were estimated using published data (16). NEC resulted in 16 days longer NICU stay compared to infants without NEC and was associated with a marginal cost of $15,440 (16). Sepsis had an increased length of stay of 17 days compared to those without sepsis and was associated with a marginal cost of $10,055. The addition of a second comorbidity in infants already diagnosed with 1 complication resulted in an additional cost of $16,543. The marginal cost of these complications was estimated using the actual direct costs for charges associated with treatment and cost of ancillary staff, but did not include physician services that were billed elsewhere. The average marginal cost per day of treating each complication was added to the baseline daily NICU cost for health states. The cost of the guideline implementation was estimated using a weekly current procedural terminology physician consultant billing code 99291 for critical care, which was then converted into a daily cost of $35 (13). This cost was added to the daily baseline cost of each cycle in the guideline group.

From a societal perspective, indirect costs due to foregone daily earnings by a caregiver were incorporated in the analysis. Each infant was assumed to have 1 caregiver at bedside at any given time throughout the hospitalization. Average weekly earnings in the United States in December 2016 were obtained from the Bureau of Labor Statistics’ Current Employment Statistics (national), and were converted to a daily average for each daily cycle until discharge or death (17).

Measure of Effectiveness

The most appropriate measures for estimating health utilities in children are still being developed, and there is even less data for infants (18). Several studies have used length of stay as a measure of effectiveness when performing economic evaluations of health states or clinical interventions in neonates given the lack of published utility scores (8–10,16). In order to measure the effectiveness of the guideline and standard interventions, the total hospital days were calculated within the model, stratified by number of complications.

Transition Probabilities

Transition probabilities for all transitions from the baseline No IFALD/No Complications state to each health state, including death, were calculated using the probabilities of developing each of the various complications obtained in the initial study (Supplemental Digital Content 1, The overall probabilities of developing complications were transformed into daily probabilities. The model assumed infants would develop complications by postoperative day 40, based on clinical experience (7). The probabilities for transitioning to discharge were calculated using a triangular probability distribution, created using the range of lengths of stay for patients with each type of complication, and centered around the median length of stay.

Cost-effectiveness and Sensitivity Analysis

A base-case ICER was calculated using deterministic parameters for both the health care sector and societal perspective models. The ICER is equal to the difference in costs between the 2 strategies given the number of complications, divided by the differences in effectiveness (length of stay in days).

Univariate and multivariate probabilistic sensitivity analyses were used to address uncertainty in the model. The univariate sensitivity analysis used standard deviations of base case parameters established by ±15% of the inputs. The multivariate probabilistic sensitivity analysis was conducted by drawing random probabilistic values of inputted parameters with transition probability values drawn using a beta distribution and cost values drawn using a gamma distribution using a standard deviation of ±15% of each parameter. A Monte Carlo simulation was performed with 10,000 trials.

Long-term Outcomes Assessment

To assess the long-term cost effectiveness of the feeding guidelines, an additional cost-utility analysis was performed. This 2-step model has previously been used to evaluate short-term and long-term outcomes in infants with NEC and/or sepsis (19,20). The long-term disability associated with postoperative NEC and sepsis was neurodevelopmental impairment (19). The long-term disability associated with IFALD and cirrhosis was liver failure, which can lead to death or liver transplantation (21). Therefore, a second decision-tree was created (Supplemental Digital Content 2, in which infants could develop cirrhosis (Yes/No) and Disability (Yes/No), which could be mild, moderate, or severe. Utility scores, life expectancy, costs, and probabilities associated with long-term complications NEC and sepsis in infants were obtained from the literature, with the assumption that the probabilities of impairment were similar in the United States and United Kingdom (19). Costs of mild, moderate, and severe disability were adjusted from 2006 GBP to 2017 USD (Supplemental Digital Content 1, (13). Probabilities of progression from IFALD to cirrhosis, costs, and utility scores related to cirrhosis during childhood were obtained from the literature (21–23). Costs of treating cirrhosis were adjusted to 2017 USD (Supplemental Digital Content 1, (13). Although not equivalent, adult utility data were used to estimate quality-adjusted life years (QALYs) as these complications were most like to happen later in childhood or adolescence. For the model, it was assumed that annual costs of treating cirrhosis in children were similar to that of adults. The ICER was calculated the differences in annual long-term costs compared to differences in QALYs between the 2 interventions.


Feeding surgical infants via the guidelines reduced costs (Supplemental Digital Content 3, We previously reported that infants in the guideline group had significantly faster advancement of EN, fewer days of PN, reduced IFALD incidence and severity, and fewer episodes of sepsis compared to the standard feeding group (7). The incidence of postoperative NEC was <10% and similar between groups. In the health care sector perspective model, the standard intervention for postoperative feeding for 100 infants incurred a cost of $31,258,902 and resulted in a total of 8296 hospital days. The guideline intervention incurred a cost of $29,295,553 and resulted in a total of 8096 hospital days. The ICER for comparing the cost-effectiveness of these 2 interventions was $−9832 per hospital stay (total hospital days). The guidelines resulted in both a lowered total cost and a lowered number of hospital days per 100 patients and are thus a dominant treatment when compared to standard care. Similarly, in the societal perspective model, the standard intervention incurred a cost of $32,316,869 and resulted in 8296 hospital days, whereas the new guidelines incurred a cost of $30,328,053 and resulted in 8096 hospital days, giving an ICER of $−9959 per hospital stay.

Sensitivity Analyses

In the univariate sensitivity analysis (Fig. 2), the greatest variation in the ICER came from varying the cost per day in the NICU. By varying the cost ±15%, the ICERs ranged from $−6640 to $−13,027 per hospital stay. The results support the guidelines as a dominant intervention. As few infants in the guideline cohort had complications, varying the transition probability of transitioning from No IFALD/No Complications to IFALD and 1 Complication and IFALD and 2 Complications made little change in the ICER.

Tornado diagram of univariate sensitivity analysis. The base case input parameters were adjusted by ±15%. Varying the cost of the baseline NICU charge and treatment of complications had the most impact on the incremental cost-effectiveness ratio. NICU = neonatal intensive care unit; PN = parenteral nutrition.

The multivariate probabilistic sensitivity analysis conducted with 10,000 Monte Carlo simulations found 90.5% of all simulated ICERs were cost-saving (Fig. 3). Similarly, in the societal perspective model, 90.5% of all simulated ICERs were cost-saving.

Cost-effectiveness acceptability curve of feeding guidelines. More than 90% of the incremental cost-effectiveness ratios in the Monte Carlo simulations were cost-saving from both the hospital and societal perspective (hospital perspective displayed).

Long-term Results

When comparing long-term outcomes (ie, cirrhosis, transplantation), the guideline intervention was again dominant. A base-case model using distributions of infants into each outcome group based on the deterministic results from the short-term model found that annual long-term costs and utilities in the standard group were $4030 and 0.90, respectively (Supplemental Digital Content 3, Annual long-term costs and utilities in the guideline group were $2830 and 0.91. The ICER was $−91,756/QALY.


The implementation of the postoperative feeding guidelines for surgical infants not only improved clinical outcomes in the short- and long term, but was also cost-effective. Use of the guidelines reduced the time to achieve enteral feeding goals, thereby reducing the risk of IFALD and sepsis, which ultimately reduced hospital costs and long-term disability from chronic liver disease in the model (7).

Both IFALD and sepsis are associated with longer lengths of hospital stay, higher hospital resource utilization, and worsened long-term clinical outcomes. In our analysis, guideline use resulted in $−9832 per hospital stay in the health care sector perspective model and $−9959 per hospital stay in the societal perspective model. In the multivariate probabilistic sensitivity analysis, 90% of the ICERs were in the dominant quadrant. In the univariate sensitivity analysis, the parameter variability that most affected the ICER was the baseline NICU cost per day. Differences in hospital costs at other hospitals may therefore affect the cost-effectiveness of a feeding intervention; however, even when the baseline NICU cost was increased by 15%, the guideline intervention still dominated the standard of care. Although the reduction in sepsis may in part be attributed to recent improvements in preventative line care bundles, infants with faster achievement of enteral autonomy may have also benefited from fewer line-days due to faster discontinuation of PN and improved intestinal barrier function.

Along with reducing IFALD, feeding guidelines may improve other outcomes. Withholding of enteral feeding is strongly associated with extrauterine growth restriction in preterm infants (24). Feeding guidelines have reduced practice variability, improved delivery of EN, and improved outcomes in multiple high-risk neonatal populations, including very-low-birth-weight infants and infants with congenital heart disease, without increasing the risk of NEC (25–28). In our group, implementation of multidisciplinary rounds with a subspecialist feeding champion improved adherence to the guidelines, without a substantial increase in hospital charges for consultation.

Reducing the risk of IFALD can be expected to decrease the incidence of chronic liver disease including cirrhosis and liver cancer. These conditions are well-understood to be associated with substantial impairments in quality of life, neurocognition, and growth as well as impairments in additional organs including the kidney, heart, and lungs (29). Ultimately, these sequelae of end-stage liver disease may necessitate a combined liver-intestinal transplant which, at a minimum, commits the individual to a lifetime of medical therapy with systemic toxicities, but has relatively poor long-term survival (ie, 50% 5-year survival) compared to transplants that involve only the liver (30).

As alternative lipids that include fish oil become more readily available in the United States, their use may also affect the development and resolution of severe cholestasis and further decrease the risk of chronic liver disease. In our study, only soy lipid was used, but lipid minimization strategies were employed if cholestasis occurred. There has yet to be conclusive evidence that the use of alternative lipids to prevent cholestasis in neonates is successful in the short-term (31). Lipid strategies are, however, likely more impactful in those infants in requiring long-term PN.

In addition to the direct cost savings to hospitals and the health care sector in the short- and long term, there are reimbursement-related implications of reducing incidence of hospital-acquired conditions to be considered. Hospital payments from both public and private payers are increasingly being linked to quality measures, further incentivizing practices that improve quality of care (32).


The probabilities used to construct the patient simulation model were calculated using a single cohort study. Certain cost data were unique to our institution, which may affect generalizability. The societal perspective model only incorporated indirect costs due to foregone earnings and did not include other nondirect medical costs, such as transportation. Costs due to foregone daily earnings reflected national averages and did not reflect variation in patients’ caregiver characteristics. To account for limitations in the data, univariate and multivariate sensitivity analyses were conducted. The use of hospital stay as the measure of effectiveness for may also be viewed as a limitation; however, there is lack of published data on utility scores for infants in neonatal intensive care (18). Because of the lack of published, validated health utility measures for health states of infants, a CEA and not cost-utility analysis was undertaken for the short-term. For our long-term analysis, the adult utility scores used to estimate QALY in pediatric conditions have not been validated, but superior methods were not available.


Feeding guidelines for surgical infants with gastrointestinal disorders can be a cost-effective intervention that can both reduce NICU complications and decrease costs associated with life-threatening comorbidities. In our models, use of the feeding guidelines resulted in both cost savings and reduction in hospital stay in the short term, and cost savings and increase in QALYs in the long term. These results support our hypothesis that a systematic approach to feeding in surgical infants is cost-effective, but further data from a larger cohort are needed. Future investigations should also evaluate readmission rates and include a more rigorous analysis from the societal perspective.


1. Gonzalez-Hernandez J, Prajapati P, Ogola G, et al. Predicting time to full enteral nutrition in children after significant bowel resection. J Pediatr Surg 2017; 52:764–767.
2. Duggan CP, Jaksic T. Pediatric intestinal failure. N Engl J Med 2017; 377:666–675.
3. Lee WS, Sokol RJ. Intestinal microbiota, lipids, and the pathogenesis of intestinal failure-associated liver disease. J Pediatr 2015; 167:519–526.
4. Javid PJ, Malone FR, Dick AA, et al. A contemporary analysis of parenteral nutrition-associated liver disease in surgical infants. J Pediatr Surg 2011; 46:1913–1917.
5. Kaufman SS, Pehlivanova M, Fennelly EM, et al. Predicting liver failure in parenteral nutrition-dependent short bowel syndrome of infancy. J Pediatr 2010; 156:580.e1–585.e1.
6. Goulet O, Olieman J, Ksiazyk J, et al. Neonatal short bowel syndrome as a model of intestinal failure: physiological background for enteral feeding. Clin Nutr 2013; 32:162–171.
7. Shores DR, Alaish SM, Aucott SW, et al. Postoperative enteral nutrition guidelines reduce the risk of intestinal failure-associated liver disease in surgical infants. J Pediatr 2018; 195:140.e1–147.e1.
8. Bisquera JA, Cooper TR, Berseth CL. Impact of necrotizing enterocolitis on length of stay and hospital charges in very low birth weight infants. Pediatrics 2002; 109:423–428.
9. Ganapathy V, Hay JW, Kim JH. Costs of necrotizing enterocolitis and cost-effectiveness of exclusively human milk-based products in feeding extremely premature infants. Breastfeed Med 2012; 7:29–37.
10. Goudie A, Dynan L, Brady PW, et al. Attributable cost and length of stay for central line-associated bloodstream infections. Pediatrics 2014; 133:e1525–e1532.
11. Lauriti G, Zani A, Aufieri R, et al. Incidence, prevention, and treatment of parenteral nutrition-associated cholestasis and intestinal failure-associated liver disease in infants and children: a systematic review. JPEN J Parenter Enteral Nutr 2014; 38:70–85.
12. Shores DR, Bullard JE, Aucott SW, et al. Implementation of feeding guidelines in infants at risk of intestinal failure. J Perinatol 2015; 35:941–948.
13. CPT/Relative Value Search. American Medical Association. 2017. Accessed April 10, 2017.
14. CPI Inflation Calculator. Bureau of Labor Statistics. Accessed May 1, 2017.
15. Lee JS, Polin RA. Treatment and prevention of necrotizing enterocolitis. Semin Neonatol 2003; 8:449–459.
16. Johnson TJ, Patel AL, Jegier BJ, et al. Cost of morbidities in very low birth weight infants. J Pediatr 2013; 162:243–249.
17. Current Employment Statistics - CES (National). n.d. Bureau of Labor Statistics. 2016. Accessed August 2, 2018.
18. Thorrington D, Eames K. Measuring health utilities in children and adolescents: a systematic review of the literature. PLoS One 2015; 10:e0135672.
19. Rice SJ, Craig D, McCormick F, et al. Economic evaluation of enhanced staff contact for the promotion of breastfeeding for low birth weight infants. Int J Technol Assess Health Care 2010; 26:133–140.
20. Renfrew MJ, Craig D, Dyson L, et al. Breastfeeding promotion for infants in neonatal units: a systematic review and economic analysis. Health Technol Assess 2009; 13:1–146. iii–iv.
21. Nandivada P, Chang MI, Potemkin AK, et al. The natural history of cirrhosis from parenteral nutrition-associated liver disease after resolution of cholestasis with parenteral fish oil therapy. Ann Surg 2015; 261:172–179.
22. Economic cost of advanced liver disease. Science Daily. 2011. Accessed April 12, 2017.
23. McLernon DJ, Dillon J, Donnan PT. Health-state utilities in liver disease: a systematic review. Med Decis Making 2008; 28:582–592.
24. McKenzie BL, Edmonds L, Thomson R, et al. Nutrition practices and predictors of postnatal growth in preterm infants during hospitalization: a longitudinal study. J Pediatr Gastroenterol Nutr 2018; 66:312–317.
25. Morgan J, Young L, McGuire W. Slow advancement of enteral feed volumes to prevent necrotising enterocolitis in very low birth weight infants. Cochrane Database Syst Rev 2015; CD001241.
26. Thoene MK, Lyden E, Anderson-Berry A. Improving nutrition outcomes for infants < 1500 grams with a progressive, evidenced-based enteral feeding protocol. Nutr Clin Pract 2018; 33:647–655.
27. Barr PA, Mally PV, Caprio MC. Standardized nutrition protocol for very low-birth-weight infants resulted in less use of parenteral nutrition and associated complications, better growth, and lower rates of necrotizing enterocolitis. JPEN J Parenter Enteral Nutr 2019; 43:540–549.
28. Kalra R, Vohra R, Negi M, et al. Feasibility of initiating early enteral nutrition after congenital heart surgery in neonates and infants. Clin Nutr ESPEN 2018; 25:100–102.
29. Chapin CA, Bass LM. Cirrhosis and portal hypertension in the pediatric population. Clin Liver Dis 2018; 22:735–752.
30. Hawksworth JS, Desai CS, Khan KM, et al. Visceral transplantation in patients with intestinal-failure associated liver disease: evolving indications, graft selection, and outcomes. Am J Transplant 2018; 18:1312–1320.
31. Hojsak I, Colomb V, Braegger C, et al. ESPGHAN Committee on Nutrition Position Paper. Intravenous lipid emulsions and risk of hepatotoxicity in infants and children: a systematic review and meta-analysis. J Pediatr Gastroenterol Nutr 2016; 62:776–792.
32. Sorensen A, Jarrett N, Tant E, et al. HAC-POA policy effects on hospitals, other payers, and patients. Medicare Medicaid Res Rev 2014; 4: pii: mmrr2014-004-03-a07.

intestinal failure; intestinal failure–associated liver disease; short bowel syndrome

Supplemental Digital Content

Copyright © 2020 by European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition