Gastrostomy Tube Placement and Resource Use in Neonatal Hospitalizations With Down Syndrome Free

This was a retrospective, serial cross-sectional analysis of the National Inpatient Sample (NIS) database from 2006 to 2017 (the latest available data year at the time of analysis). The NIS is part of the Healthcare Cost and Utilization Project (HCUP), which is sponsored and regulated by the Agency for Healthcare Research and Quality (www.hcup-us.ahrq.gov/nisoverview.jsp). The NIS is the largest publicly available all-payer database of hospital discharges from community, nonrehabilitation hospitals in the United States. The NIS is released every year by HCUP, and it includes a 20% stratified sample of all discharges (about 7 million) from US community hospitals. The number of states and hospitals contributing data to NIS increased from 38 states (1045 hospitals) in 2006 to 48 states (4584 hospitals) in 2017. In 2012, the NIS design changed from including all discharges within a 20% stratified random sample of hospitals to a 20% self-weighted, stratified, systematic random sample of discharges from all hospitals. These discharges captured in the NIS contain 35 million hospitalizations annually and represent ∼97% of all US hospitalizations when weighted. We used the “TRENDWT” variable provided by HCUP for the years before 2012 to make estimates comparable to the new NIS design, which began in 2012.⁶  Each hospitalization in this database is deidentified and maintained as a unique entry that has 1 primary diagnosis and <30 secondary diagnoses, along with up to 25 procedure codes using International Classification of Diseases, Ninth Revision (ICD-9) and 10th Revision (ICD-10). The NIS database has been extensively used by researchers and policymakers to make national estimates of health care utilization and outcomes in numerous conditions.^7–11  To confirm our result, we also analyzed the Kids’ Inpatient Database (KID) from the year 2006 to 2016. Like NIS, the KID is the largest publicly available, all-payer, pediatric inpatient care database in the United States. The KID is released every 3 years and yields nearly 7 million hospitalizations each year of patients aged <21 years. Since the NIS data are released every year, it provides more time points to better assess the trend as compared with the KID (released every 3 years). In the KID data, age is a part of sampling strategy where normal newborns are sampled at 10% rate and the complicated newborns are sampled at 80%. Therefore, we chose the NIS data for the primary analysis and the KID data for confirmation. The study involved publicly available deidentified data and was thus exempt from the institutional review board review.

Neonates (aged ≤28 days) were identified using the Neomat variable, which identifies discharges with neonatal and/or maternal diagnoses and procedures. All neonatal hospitalizations with GA ≥35 weeks at birth who were assigned a diagnostic code for DS were included in the study. Newborn hospitalizations with DS were identified using ICD-9 and ICD-10 codes. Neonates with DS who underwent GT placement were identified using ICD-9 and ICD-10 procedure codes as described in Supplemental Table 5. To avoid duplication of data, we identified and excluded neonates who were transferred out to short-term hospitals, skilled nursing facilities, intermediate care facilities, or another type of facility, using the “DISPUNIFORM” variable in keeping with previous studies that used HCUP databases.^9,11,12  Details of the population derivation are shown in Supplemental Fig 2.

Patient-level characteristics such as sex, race, median household income as per zip code, primary payer (Medicare/Medicaid, private insurance, self-pay, or other), and hospital-level characteristics such as hospital location (rural or urban), and teaching status (rural, urban nonteaching and urban teaching), hospital bed size (small, medium, and large), and hospital region (Northeast, Midwest, South, and West) were abstracted. The distribution of the various states in each Census region is available at https://www.hcup-s.ahrq.gov/db/nation/nis/NIS_Introduction_2017.jsp#table2app1. Race was categorized as White, non-Hispanic Black, Hispanic, and others. Comorbidities or complications were identified with ICD-9 and ICD-10 diagnostic and procedure codes as shown in Supplemental Table 5. Resource utilization was assessed by using the length of stay (LOS) and inflation-adjusted hospital costs for each neonatal hospitalization with DS. To calculate the estimated cost of hospitalization, the NIS data were merged with cost-to-charge ratios available from the HCUP. We estimated the cost of each inpatient stay by multiplying the total hospital charge with the cost-to-charge ratio provided by HCUP.¹³  Adjusted cost for each year was calculated in terms of the 2017 cost after adjusting for inflation according to the latest consumer price index data released by the US government.¹⁴  This enabled us to standardize the costs over the study period.

The primary outcome was the frequency of GT placement in neonatal hospitalizations born at ≥35 weeks’ GA with DS and its changes over time from 2006 to 2017. The secondary outcomes were overall resource utilization measured in terms of LOS and inflation-adjusted hospital cost and their trends over time during the study period. For trend analysis, the 12-year study period was divided into 3-year epochs or time periods. The administrative incidence of neonatal hospitalizations with DS was calculated by dividing the number of newborn hospitalizations with DS for each year with the corresponding total number of live births and expressed as per 10 000 live births. The live birth data were obtained from the Centers for Disease Control.¹⁵  The frequency of GT placement was expressed as a proportion of neonatal hospitalizations with DS patients who received GT for each year of the study.

Medians with interquartile range (IQR) were reported for continuous variables and proportions with SEM were reported for categorical variables. In bivariate analysis, the χ² test was used to compare proportions and the Mann-Whitney U test to compare continuous variables between neonatal hospitalizations with DS with and without GT. To assess the changes over time, unadjusted trend analysis was conducted using the Cochran-Armitage trend test for categorical variables and the Jonckheere-Terpstra trend test for continuous variables. Trends were further analyzed with multivariable logistic regression with calendar year included as a covariate to determine the effect of increasing calendar year on the frequency of GT placement and patient and hospital-level factors associated with the outcome of GT placement. Furthermore, multiple linear regression was used to determine the impact of GT placement on LOS and inflation-adjusted hospital costs in neonatal hospitalizations with DS. Because of the possibility of the data on LOS and inflation-adjusted cost being skewed, the multiple linear regression analysis was repeated with log-transformed data.¹⁶  Covariates that were significantly associated with GT placement in bivariate analysis and those that were deemed relevant were included in the multivariable regression analysis. These covariates were sex, GA, race, type of insurance or primary payer, household income, calendar year, hospital bed-size, hospital location, teaching status, severe CHD, need for invasive mechanical ventilation continuously for >96 hours, gastroesophageal reflux disease, surgery for duodenal obstruction, sepsis, surgery for Hirschsprung disease, hypothyroidism, laryngomalacia, tracheobronchomalacia, and pulmonary hypertension. Severe congenital heart disease was defined as any CHD that required cardiac surgery.¹⁷  These covariates were identified using ICD-9 and ICD-10 diagnostic and procedure codes. Statistical analyses were performed using SPSS V26.0 (IBM Corp, Armonk, NY). A P value of <.05 was considered significant for all analyses.

Among 41 900 237 total live births ≥35 weeks’ GA, 51 473 had an ICD-9 or 10 code for DS, yielding an overall prevalence of 12.3 per 10 000 live births ≥35 weeks’ GA from 2006 to 2017. Out of the 51 473 assigned a diagnostic code for DS, 1913 (3.7%) had a procedure code for GT placement (Supplemental Fig 2). The demographic and hospital characteristics and their changes over time for all neonatal hospitalizations ≥35 weeks’ GA with DS in the United States from 2006 to 2017 are described in Table 1. Briefly, 48.4% were females, 49.4% were of White race, 85.2% had GA of ≥37 weeks, 52.3% used Medicaid or other nonprivate health insurance, and the majority of the neonates with DS were in large bed size (63.2%), teaching hospitals (70.9%).

GT placement in newborns hospitalized with DS more than tripled from 1.7% in 2006 to 5.6% in 2017 (P <.001), whereas the prevalence of DS increased from 10.3 to 12.9 per 10 000 live births (P <.001) as shown in Fig 1. Consistent with the KID data, GT placement in newborns with DS increased significantly from 2.5% in 2006 to 5.4 in 2016 (P <.001), whereas the prevalence of DS increased significantly from 11.4 in 2006 to 13.2 per 10 000 live births in 2016 (P <.001) (Supplemental Fig 3). In multivariable logistic regression with GT placement as the outcome, each 1-year increase in the calendar year was associated with increased odds of GT placement (odds ratio, 1.10, 95% confidence interval [CI], 1.06–1.15) as shown in Table 2.

Bivariate analysis of the hospital course, comorbidities, LOS, and inflation-adjusted hospital cost between DS hospitalizations with and without GT placement is shown in Supplemental Table 6. When compared with neonatal hospitalizations without GT, those with GT were more likely to be 35 to 36 weeks’ GA (24.8% vs 14.4%, P <.001), have sepsis (21.6% vs 7.7%, P <.001), duodenal obstruction (16.8% vs 1.8%, P <.001), severe CHD (27.5% vs 11.1%, P <.001), invasive mechanical ventilation lasting >96 hours (36.2% vs 5.9%, P <.001), and pulmonary hypertension (27.6% vs 13.5%, P < .001). Among those who survived to discharge, those with GT had longer LOS (41 vs 5 days, P <.001) and higher inflation-adjusted hospital cost ($95 529 vs $7556, P <.001) when compared with those without GT. In multivariable logistic regression to determine the predictors of GT placement in neonatal hospitalization with DS (Table 2), GA 35 to 36 weeks and multiple comorbidities such as sepsis, surgery for duodenal obstruction, gastroesophageal reflux, severe CHD, and invasive mechanical ventilation and hypothyroidism were associated with increased odds of GT.

In unadjusted trend analysis (Table 3) among neonatal hospitalizations with DS, mortality before discharge decreased from 7.8% in 2006 to 2008 to 1.3% in 2015 to 2017 (P = .009). Median LOS increased from 35 days (IQR, 25–60 days; geometric mean 36 ± 2 days) to 46 days (IQR, 29–60 days; geometric mean 46 ± 2 days; P = .02) and the median inflation-adjusted cost of hospitalization increased from $74 214 (IQR, 50 693–148 772; geometric mean 79 781 ± 2) in 2006 to 2008 to $111 360 (IQR, 66 947–203 803; geometric mean 114 789 ± 2) in 2015 to 2017 (P = .025). In multiple linear regression analysis that adjusted for the previously described covariates, GT placement was associated with increased LOS (β = 24.7 days, 95% CI: 23.2–26.2; P <.001) (Table 4). Similarly, GT placement was associated with increased inflation-adjusted hospital cost (β = $60 431, 95% CI: 55 742–65 138; P < .001) (Supplemental Table 7). In multiple linear regression with log-transformed data, LOS and inflation-adjusted hospital cost were 45.9% (Supplemental Table 9) and 41.7% (Supplemental Table 8) higher in DS hospitalizations with GT when compared with those without.

Using the largest health care database in the United States, we found the overall prevalence of GT placement in newborn hospitalizations with DS was low, but the frequency of GT placement outpaced the increase in DS hospitalizations and more than tripled from 2008 to 2017. Further, those who received GT appeared sicker and had a more complicated hospital course, had longer LOS, and had higher inflation-adjusted hospital costs when compared with those who did not receive GT. There was a significant increase in both LOS and inflation-adjusted hospital costs during the study period. This is the largest analysis on GT in newborn hospitalizations with DS and it extends the depth and breadth of our knowledge on the hospital course and resource use by newborns with DS who receive GT before discharge.

The overall GT placement frequency of 3.7% comports with the rate of 5.2% of DS newborns without major anomalies discharged from the hospital on NG or GT feeds during the initial birth hospitalizations reported by McAndrews et al using data from 277 NICUs in 34 states in the United States from 2005 to 2012.⁴  However, it is significantly lower than the rate of 13.7% in the first year of life of DS babies reported by Poskanzer et al.⁵  The study by Poskanzer et al is a single-center study over the first year of life, whereas that of the current study is a nationally representative, population-based study that assessed GT placement before discharge after admission during the neonatal period. The current study demonstrated that the frequency of GT placement significantly increased from 1.7% to 5.6%. Significant increases in GT placement have been documented in pediatric patients in general,¹⁸  in children with chronic childhood conditions,¹⁹  preterm infants,²⁰  very low birth weight infants,²¹  and preterm infants with bronchopulmonary dysplasia.²²  The reasons for this increase are not clear but it has been suggested the introduction of the less invasive laparoscopic GT placement has made GT more attractive to surgeons and parents.²³  Second, the increase in GT may be related to the publication of a position paper in 2006 by the North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition that recommended NG and nasojejunal feeding tubes should not be used long term because of multiple complications associated with their use.²⁴  Furthermore, mortality significantly decreased during the study period, suggesting that more critically ill newborns with DS who would have otherwise died in the previous years are surviving to discharge. Thus, the increase in GT placement is likely a reflection of the increasing survival of newborns with DS and the intensity of care these babies receive during the neonatal period. Although the technology of GT is life-saving and ensures adequate nutritional management and ease of medication administration, it adds up to the already medically complex care of children with DS. Most of these infants are cared for by their families at home, and our findings highlight the need for coordinated multidisciplinary outpatient care and targeted interventions to improve caregiver comfort and mastery of skills necessary to care for these GTs to improve outcomes for these infants.

We identified prematurity and multiple comorbidities that were significantly associated with GT in newborns with DS. Although these comorbidities increase the odds of GT placement, the placement of GT reflects a more complicated hospital course and may be a marker for the severity of illness during the neonatal hospital stay. The first step to targeting high-risk patients for any intervention is identifying them and the current study provides data on such predictors that can be used for risk stratification and early identification of those newborns with DS who might need GT. Further, these predictors could be incorporated into risk prediction models and scoring systems to identify the need for GT in future studies.

The current study provides, for the first time, invaluable information on resource utilization by newborns with DS who receive GT. Neonatal hospitalizations with DS who received GT had longer LOS and higher inflation-adjusted hospital costs when compared with those without GT. Fox et al also found that GT placement in pediatric hospitalizations was associated with increased LOS.¹⁸  Those with GT were more likely to have multiple comorbidities and complications during their hospital stay. Thus, GT placement in DS may be a marker for the severity of illness and/or the intensity of care received during hospitalization during the neonatal period. In adjusted analysis, DS hospitalizations with GT had a significant increase in both LOS and inflation-adjusted hospital cost. There is a direct relationship between LOS and hospital cost, but the results of the current study should be interpreted within the context of increased survival of newborns with DS. Mortality declined from 8.1% to 1.3%, and this suggests that sicker and critically ill newborns survived because of advances in perinatal, neonatal, anesthetic, and surgical care. This would naturally increase the LOS and invariably increase hospital costs. The trends in resource use constitute a useful instrument for observing possible changes in the process of hospital care for newborns with DS and can influence changes in policy that lead to increased hospital reimbursement for the care of these infants.

This study has several strengths. First, it is the largest and the first population-based study using a nationally representative database to evaluate the trends and predictors of GT placement in newborn hospitalizations with DS. These findings will be valuable to clinicians, as well as the families, for medical decision-making, the early identification of those who need GT, and possibly to shorten the duration of hospitalization. Further, we evaluated trends in resource use, and this could inform changes in policy that lead to adequate reimbursement of hospitals. The data on resource utilization also provides baseline benchmarks for future quality improvement initiatives to decrease LOS.

We recognize the limitations of this retrospective study based on a national database. The information available from NIS is dependent on the quality of documentation and coding based on ICD-9 and 10 codes. However, DS and a procedure such as GT placement are major diagnoses and unlikely to be missed or misrepresented. Indeed, the ICD-9 code for DS has a positive predictive value of 93%.²⁵  Because of the retrospective nature of the study, we were unable to differentiate the effects of GT placement from the effects of the other comorbidities to which it was highly associated with LOS and hospital costs. Although both NIS and KID showed significant increase in GT placement in DS, the numbers are different. Differences in sampling methods between NIS and KID may explain the magnitude of this difference. Newborns with feeding difficulties are sometimes discharged from the hospital on NG tube feedings, but this group was not included in the current study because of the lack of ICD codes for identifying such infants. Thus, the true proportion of newborns with DS with feeding difficulties not feeding fully by mouth is higher than the proportion with GT reported in the current study. The data on resource use provided in this study are limited to neonatal admissions and do not include resource use occasioned by readmission after the neonatal period. We did not evaluate the impact of GT placement in the neonatal on future health care resource use because a GT placed in the neonatal period can theoretically decrease the need to return to the hospital postdischarge because of more stable feeding and medication administration. Further, the NIS database is a compendium of discharges, and each patient could be represented in the database more than once because of readmission. We believe this was marginal and unlikely to affect the results of the study because the inclusion criteria were limited to admission at age ≤28 days.

In summary, there was a significant increase in the frequency of placement of GT in neonatal hospitalizations with DS that outpaced the increase in the administrative prevalence of DS during the study period. Furthermore, there was a significant increase in resource use by DS hospitalization of patients who received GT. Future studies should probe the reasons for the significant increase in GT placements, to devise strategies to identify those babies in need of GT early and, therefore, potentially decrease resource utilization. Additionally, studies that may help to determine the component of resource utilization related to the placement of GT and how GT feeding compares with NG feeding in terms of LOS and hospital costs are warranted.

The authors thank the HCUP of the Agency for Research and Healthcare Quality for the use of the data from the NIS.