Characteristics and Outcomes of Children Discharged With Nasoenteral Feeding Tubes

We conducted a retrospective cohort study using the MarketScan Medicaid Database (IBM Watson Health, Cambridge, MA). This database contains deidentified inpatient, outpatient, home health, and prescription drug claims for Medicaid enrollees in 10 to 12 geographically dispersed states (inclusion varies by year), with population demographics comparable to national Medicaid statistics. Individuals can be tracked over time and across encounters. The MarketScan Database contains no identifiable geographic or personal health information. The study was deemed exempt from institutional review board review per policies at the lead author’s institution.

We included children younger than 18 years old discharged from the hospital from an inpatient hospitalization between January 1, 2016, and December 1, 2019, with a durable medical equipment (DME) claim for feeding tube supplies within 30 days of discharge. Specific billing codes and their application to our inclusion-exclusion criteria are shown in Supplemental Table 4. In brief, home enteral feeding tube use was identified in DME claims using Healthcare Common Procedure Coding System codes for feeding tube supplies or formula. To focus specifically on children with new NET use after hospitalization, we excluded children requiring home enteral tube use in the 90 days before admission and those with a surgical feeding tube (eg, GT) at the time of discharge. Existing surgical tubes were identified from DME claims using Healthcare Common Procedure Coding System codes for a GT or jejunostomy tube at any time from 90 days before admission through 30 days postdischarge. Additionally, children with a diagnosis or procedure code indicating the presence of a surgical tube in the 90 days before admission through the date of discharge were excluded using International Statistical Classification of Diseases, Tenth Revision (ICD-10), ICD-10 Procedure Coding System (ICD-10-PCS), and Current Procedural Terminology codes.

Children were required to be continuously enrolled in Medicaid for at least 30 days after discharge from the index hospitalization to allow measurement of both home NET use and postdischarge complications. Children without continuous enrollment during the 90-day lookback period were not excluded as children without insurance are commonly screened for Medicaid eligibility when admitted to the hospital and may be enrolled under presumptive eligibility during hospitalization.

We excluded index birth hospitalizations because NET use after discharge from the NICU is well described elsewhere.^3,15,16  Furthermore, these infants represent a distinct population given that they typically experience much longer hospital LOS while receiving care within specialized units and teams. For children experiencing more than 1 discharge in the 30 days preceding the initial qualifying DME claim, the most proximal discharge to the claim was used to determine the index hospitalization. Because a child may require intermittent home NET use, multiple index hospitalizations for a single child were included as long as each index hospitalization met the requirements for new NET use (ie, no home NET use in the 90 days before admission).

Patient demographics included age, sex, race and ethnicity, medical complexity, and payment arrangement (fee for service or capitation). Race and ethnicity represent social constructs and were included for descriptive purposes. Their inclusion does not imply biologic causal associations but reflects the social dimensions of medicine and existing disparities in health care use resulting from racism.^17,18  Medical complexity was defined as the presence of a complex chronic condition (CCC) during the index hospitalization or on any claim 90 days before admission using ICD-10 codes as developed by Feudtner et al.¹⁹  The CCC category for technology dependence does not include codes corresponding to NET use.

Index hospitalizations were characterized by reason for hospitalization and LOS. Primary diagnoses assigned to the hospitalization were categorized using the Pediatric Clinical Classification System (PECCS), which organizes ICD-10 diagnosis codes into mutually exclusive, clinically meaningful conditions per methodology described by Gill et al.²⁰ 

Our primary postdischarge outcome was a composite of all-cause 30-day revisits, including ED-only return (ie, home disposition) or readmission (direct admissions and those from the ED). Revisit types were also examined separately, and revisits coded as tube-related complications were characterized as either tube displacement/dysfunction or pneumonia/pneumonitis based on the primary diagnosis associated with the revisit (Supplemental Table 5). We evaluated these complications specifically as they are among the most likely or serious complications of NET use to result in a revisit.²¹  Revisit primary diagnoses were also characterized using PECCS. Finally, we examined progression from NET to GT use within 12 months of the index hospitalization, with conversion to GT identified using Current Procedural Terminology and ICD-10-PCS codes.

Descriptive statistics were used to summarize patient and index hospitalization characteristics. Differences between hospitalizations with and without a 30-day revisit were compared using χ² and Wilcoxon rank-sum tests. The 10 most common diagnoses (classified using PECCS) were reported for index hospitalizations, 30-day ED-only returns, and 30-day readmissions. Multivariable logistic regression was used to identify factors associated with 30-day revisits (all-cause and tube displacement/dysfunction; pneumonia/pneumonitis revisits not modeled owing to low event rate). Models were developed using backward selection from an initial list of possible covariates including age, sex, race/ethnicity, number of CCCs, CCC body system, PECCS code, and LOS. Variables were retained in the final model if they were statistically significant at the <.05 level and are reported as odds ratios. Time to GT conversion was examined using Kaplan-Meier methods controlling for continuous Medicaid enrollment up to 12 months after discharge. Multivariable Cox regression models were used to assess factors associated with GT conversion and were developed similarly to the logistic regression models. Post hoc sensitivity analysis excluded children with malignancy given their unique use patterns and access to extensive outpatient support. Finally, we performed subgroup analyses to assess for differences among children with malignancy, without a CCC, and <1 year of age. All analyses were conducted using SAS version 9.4 (SAS Institute, Inc, Cary, NC), and P < .05 was considered statistically significant.

Between 2016 and 2019, there were 1815 index hospitalizations, representing 1791 unique patients, resulting in a discharge with new home NET use (Table 1). This represented 0.3% of all nonbirth hospitalizations resulting in a discharge to home (Fig 1). More than one-third of hospitalizations were for patients younger than 1 year of age, and most were younger than 5 years old. A CCC was present in more than 80% of hospitalizations, and one-half were for children with 2 or more CCCs. The most common CCC category present was cardiovascular, followed by technology, gastrointestinal, and neuromuscular.

Median LOS was 6 days (interquartile range 3–11) before discharge with home NET use (Table 1). The most common primary diagnosis associated with index hospitalization was failure to thrive (11%), followed by malnutrition (6.8%) and acute bronchiolitis (5.9%) (Table 2).

Nearly one-half (890/1815, 49%) of all discharges with a new NET requirement had an ED-only return (480/1815, 26.4%) or hospital readmission (560/1815, 30.9%) within 30 days (Table 3). One hundred and fifty of the 890 discharges with a 30-day revisit experienced both an ED-only return and readmission as separate encounters. Only 195 (10.7%) experienced a 30-day revisit for tube displacement/dysfunction, and only 3 (0.2%) required readmission for this reason. Revisits with a primary diagnosis of pneumonia/pneumonitis were less frequent (54/1815, 3%), but were more likely to be readmitted (45/1815, 2.5%). The most common reason for an ED-only return was rehabilitation care related to prosthetic devices (20.6%) (Table 2). Nine of the 99 encounters assigned to this PECCS category were for an ICD-10 code associated with tube displacement/dysfunction (Z46.89 Encounter for fitting and adjustment of other specified devices). The most common reason for readmission was chemotherapy (7.1%).

Presence of a CCC, race/ethnicity, and Medicaid payment arrangement were not associated with 30-day revisit risk in adjusted analysis. Male sex (odds ratio, 1.21; 1.01–1.47), age (5–12 years vs. <1 year, 0.58, 0.42–0.79), malignancy (3.95, 2.79–5.6), and shorter LOS (0.95 per additional day, 0.94–0.97) were associated with increased revisit risk (Supplemental Table 6). Younger age, absence of a CCC, presence of a metabolic CCC, and shorter LOS were significantly associated with an increased risk of a return visit for tube displacement/dysfunction, whereas having a renal CCC was associated with a lower risk.

Sensitivity analysis excluding malignancy did not meaningfully affect revisit rates or diagnoses apart from the expected omission of chemotherapy as a return diagnosis (Supplemental Tables 7, 8, and 9). In subgroup analyses (Supplemental Tables 10, 11, and 12), children with malignancy did experience higher all-cause revisit rates, but fewer revisits for tube displacement/dysfunction. Results were generally similar to the overall cohort for children without a CCC and those <1 year, with the exception that children without a CCC were nearly twice as likely to return to the ED for tube displacement/dysfunction.

A minority of patients discharged with new NETs progressed to GT placement (16.9% of those enrolled at 6 months and 22.3% at a year) (Fig 2). Factors associated with GT progression at 1 year included presence of 2 or more CCCs (hazard ratio, 1.57; 1.04–2.36) and presence of a neuromuscular CCC (1.58; 1.19–2.08). Factors associated with a decreased likelihood of progression to GT included those aged 13–17 years (0.49 vs. <1 year; 0.27–0.89), gastrointestinal CCC (0.64; 0.46–0.89) and longer index hospitalization LOS (0.97 per day; 0.95–0.99) (Supplemental Table 13). Sensitivity and subgroup analyses yielded comparable findings, with fewer than one quarter of children with or without malignancy, lacking a CCC, and <1 year of age progressing to GT within 12 months (Supplemental Tables 7–12).

Within a multistate cohort of children receiving Medicaid, new home NET use after hospitalization was infrequent overall, and most common among younger children and those with a CCC. Nearly half of all discharges with home NET use were associated with an ED return or readmission within 30 days, though revisits coded as tube-related complications were less common. Less than one quarter of patients with new NET feeding progressed to GT placement within a year of discharge.

Home NET use was associated with a variety of conditions. Common diagnoses included failure to thrive, malnutrition, and dysphagia. Although indicative of feeding difficulties, these diagnoses are nonspecific and commonly associated with an underlying chronic condition (eg, malnutrition secondary to inflammatory bowel disease), which is consistent with the high prevalence of CCCs in our study population. Interestingly, acute respiratory conditions (ie, bronchiolitis, respiratory failure, and pneumonia) were among the most frequent index hospitalization diagnoses. Inpatient NET use for hydration has been well-described for bronchiolitis,²²  although resuming oral hydration is often considered a necessary criterion for discharge. This suggests that some clinicians use home NET to facilitate discharge for children who may not be able to maintain adequate oral intake after their respiratory status has improved. That these represent acute rather than chronic conditions is notable and highlights how medical treatments traditionally reserved for inpatient settings are increasingly used outside of the hospital. For example, some institutions have demonstrated success in discharging children with bronchiolitis on oxygen to be weaned as outpatients.²³  Such changes in practice are associated with reduced costs and the potential for more efficient and family-centered care but should be balanced against the high revisit risk observed in our study.

Previous studies have reported that the need for home health care and technology assistance (including enteral feeding tube use) are associated with increased odds of 30-day unplanned readmission.²⁴  Consistent with this, we observed all-cause revisit rates 3 to 10 times greater than those reported for other common conditions.^25–28  Still, less than 10% of all readmissions and less than half of all ED-only returns were assigned a primary diagnosis code specifically indicating a tube-related complication. We may have underestimated complication rates if revisits were assigned a primary ICD-10 diagnosis code corresponding to the resultant symptom (eg, nausea, feeding difficulties) rather than the tube itself. However, these rates are similar to those reported among patients discharged with a NET from the NICU,^3,15,16  where this practice has steadily gained acceptance as a strategy to reduce LOS in this population.²⁹  Our results suggest that home NETs are being used across a broader range of patients, but further studies are needed to better characterize its safety.

Surprisingly, number of CCCs was not associated with all-cause revisits, although malignancy was, likely reflecting the many potential indications for ED return or readmission in patients with cancer such as febrile neutropenia and planned readmissions for chemotherapy. In contrast, revisits identified as resulting from tube displacement/dysfunction were less frequent among children with a CCC and those with malignancy. The reasons for this are unknown, but it is possible that chronic medical complexity is associated with unmeasured home resources and capacity (eg, home nursing, medically-experienced caregivers), or in the case of malignancy specifically, these complications could be addressed in oncology clinic rather than the ED given the enhanced treatment capabilities typically available in these settings. Taken together, our findings highlight the multifaceted nature of medical complexity, both in terms of how it is defined and its impact on health care utilization patterns.

The majority of patients did not progress to GT within a year. This suggests that for most patients, home NET feeding is not being used as a “bridge” to planned gastrostomy. Recent survey data show a similar pattern, with clinicians citing the need for short-term enteral support or need related to a self-limiting illness as the most common indications for home NET use.⁶  Furthermore, a majority indicated they rely on patient factors, including lack of progress on oral feeding, presence of a neurodisability, and NET complications, rather than a specific duration of need to determine the necessity of transition to a GT.⁶  Within our cohort, children with higher levels of medical complexity were more likely to progress to GT, as were those with neuromuscular CCC, and younger patients, likely reflecting a perceived long-term need for enteral support.

Our study should be interpreted considering its limitations. Most importantly, identifying new home NET use from claims data presented a number of challenges. Many relevant billing codes are concurrently complex, ambiguous, and inconsistently used. For example, although procedural charges for surgical placement of GT are expected to be reliable, the ICD-10 diagnostic code “gastrostomy status” is less likely to be billed during an admission for an unrelated problem. Furthermore, nasogastric tube insertion is not a separately billable procedure. Patients discharged with a nasogastric tube placed while being an inpatient would have neither an inpatient nor associated DME charge for the tube itself. For this reason, we considered DME charges for feeding supplies (eg, feeding pump) as a surrogate for NET use in patients without a surgical tube. Inconsistency in the timing of DME charges following discharge was an unanticipated challenge and may have contributed to some degree of misclassification of both our exposure and outcomes. For example, the finding that 1.9% of return ED visits were for a gastrostomy complication may indicate coding inaccuracies or shortcomings in our NET identification strategy that inappropriately included children discharged with a GT. Alternatively, it may reflect patients who both progressed to GT and subsequently returned to the ED with a tube complication within 30 days of index hospitalization. Discerning between these potential explanations was not possible given limitations in the data but appeared to affect a relatively small number of cases. We were also unable to distinguish between different NET tube locations (ie, nasogastric/-duodenal/-jejunal). Other limitations include the inability to identify duration of need or nonnutritional use of NETs (eg, medication administration), the lack of geographic identifiers in our data source that precluded us from assessing for regional differences in clinical practice, and that Medicaid programs, eligibility, and billing practices differ substantially between states.

Home NET use among children after hospital discharge is uncommon but associated with a wide variety of acute and chronic conditions. Although overall ED return and readmission rates are high for patients discharged with a new home NET, revisits coded as tube-related complications are substantially lower, and most patients do not progress to gastrostomy. These findings suggest that home NET use may be a reasonable option for select patients who require ongoing enteral nutritional support and are otherwise ready for discharge, although further work is needed to clarify the optimal patient populations and conditions for its use.