Etiologies of Poor Weight Gain and Ultimate Diagnosis in Children Admitted for Growth Faltering

This is a retrospective study of children 2 weeks to 2 years of age admitted to a tertiary-care children’s hospital with a diagnosis of GF. Encounters from January 1, 2013, to December 31, 2017, were identified from the electronic medical record by International Classification of Diseases, Ninth and 10th Editions, codes: failure to thrive in the newborn (779.34/P92.6), failure to thrive in a child >28 days (783.41/R62.51), and abnormal weight loss (783.21/R63.4). Any diagnosis entered during evaluation was used to identify all patients with a primary complaint of GF. Our institution does not have specific admission criteria for GF. Most often, patients are referred to the emergency department (ED) after having failed outpatient management or are referred to a gastroenterology (GI) clinic for subspecialty consultation and directly admitted from the clinic. Rarely, patients presented to the ED for evaluation of other concerns and were found to have GF and were admitted for evaluation, or they were admitted after referral by child protective services because of concern for neglect.

Only index (first) admissions for GF were included. Only patients admitted to pediatric hospital medicine or GI were included.

Because the goal of this study was to evaluate the etiology of isolated GF, patients with known significant medical conditions were excluded (Supplemental Table 5), as were infants born at <35 weeks’ gestation. Significant medical conditions were defined using criteria described by Feudtner et al to describe complex chronic conditions.⁷  Milk protein intolerance (MPI), reflux, and unspecified developmental delay were not criteria for exclusion. Encounters were also excluded if documentation indicated an absence of GF or if there was no follow-up in the 2 years after admission. Follow-up was defined as any visits to outpatient clinics, EDs, or urgent cares, and any subsequent hospital admissions. We had access to records from all pediatric subspecialists, all pediatric ED visits, all admissions to our children’s hospital, and to most other regional hospitals that admit children. We were also able to see outpatient follow-up visits within the 2 largest outpatient pediatric practice groups in our area through our electronic medical record (Cerner) and via Epic Care Everywhere.

Data were extracted from the chart by 1 of 5 reviewers and training was done on a sample data set before starting chart review. Fleiss κ calculated for interrater reliability on final diagnosis in a random sample of 20 charts was 0.818 among all 5 reviewers. Diagnoses of organic disease, mechanical feeding difficulty, and neglect were reviewed by at least 2 reviewers. Documentation was reviewed to identify components of the history and physical exam and to determine diagnosis. Data up to 2 years from admission were reviewed to determine final diagnosis, but components of the history and physical were only evaluated from the index hospitalization. Data were collected, deidentified, and stored using Research Electronic Data Capture.

Diagnoses were grouped into 5 categories, with subcategories, defined below. These categories were developed iteratively during initial analysis of 100 charts and agreed upon by all reviewers.

• Insufficient Intake: includes breastfeeding failure, inadequate formula feeding, improper formula mixing, inappropriate diet, and neglect. These patients were able to gain weight once offered adequate calories, without significant feeding interventions, and continued to gain weight appropriately after discharge. No further diagnoses were made in follow-up. Neglect was defined as documented concern for neglect in primary service documentation with associated child advocacy and/or social work consult that also documented concern for neglect.

• Organic disease

  • ∘ Genetic disorder: any chromosomal anomaly or single gene defect contributing to poor growth. Variants of unknown significance or disorders with no effect on growth or development are not included.

  • ∘ Malabsorption: includes MPI, protein losing enteropathy, celiac disease, chronic diarrhea, and allergic colitis. These patients frequently required formula changes and even total parenteral nutrition during acute hospitalization.

  • ∘ Reflux: patients were unable to take adequate calories because of severe fussiness or emesis and required medication and/or thickened feeds to gain weight. Other etiologies, including MPI, were considered and ruled out.

  • ∘ Other: diseases not included in the above categories but noted to be causative of poor weight gain.

• Mechanical feeding difficulty: these patients were unable to safely take adequate calories without feeding interventions including thickening of feeds, chin and cheek support, feeding tube placement, or ear, nose, and throat procedures. Patients were able to gain weight after these interventions and were not subsequently found to have other diagnoses. Specific diagnoses include oropharyngeal dysphagia, aspiration, laryngomalacia, and oral aversion.

• Mixed etiology: documentation of >1 class of etiology leading to poor weight gain.

• Unknown: No definitive diagnosis was given in documentation. Often, these patients continued to follow up with subspecialty providers, especially genetics, but no definitive diagnosis was made within the 2-year study window.

Statistical analysis was performed in IBM SPSS Statistics. Descriptive statistics were summarized as frequencies for categorical data or as mean (SD) or median (interquartile range) for normally distributed or non-normally distributed continuous data, respectively. Analysis of variance was used to examine associations between demographic or clinical factors and diagnosis. If analysis of variance was statistically significant, post-hoc pairwise comparisons were performed using Pearson’s χ² or Fisher’s exact test for categorical variables and Mann-Whitney U test for continuous variables. Variables that were associated with specific diagnoses with P < .05 were then included in multivariable modeling using logistic regression. When individual variables were strongly correlated with each other, these were combined into a single variable for multivariate analysis, wherein the presence of either individual variable was scored as presence of the combined variable. Past medical history of eczema and rash on physical exam were combined into “eczema/rash,” history of NICU admission and low birth weight (LBW) were combined into “neonatal history,” hypotonia and dysmorphic features were combined as “hypotonia/dysmorphism,” coughing with feeds and prolonged feeds were combined as “dysphagia”, and fussiness with feeds and feeding refusal were combined as “aversion”. Adequacy of daily weight gain was evaluated as a subgroup analysis because this could only be assessed for patients with length of stay (LOS) >24 hours and age <12 months.

A total of 1387 unique encounters were identified, of which 562 met inclusion/exclusion criteria (Fig 1). A total of 65 patients were lost to follow-up and could not be analyzed at 2 years postadmission, leaving 497 cases for complete analysis. See Supplemental Table 6 for demographic and clinical characteristics of included patients.

We identified 5 classes of diagnoses: insufficient intake (traditionally nonorganic), organic disease, mechanical feeding difficulty (MFD), mixed diagnoses, and those whose diagnosis remained unknown at 2 years (Table 1).

The majority of patients (59%) had GF because of insufficient intake. Twenty percent of patients had an organic disease including malabsorption (9%), genetic disorders (7%), reflux (2.6%), or other organic etiologies (1.8%) (Supplemental Table 7). Ten percent of patients had MFD, 2% of patients had mixed etiologies, and 8% of patients had diagnoses that remained unknown.

Among patients lost to follow-up, 75% were diagnosed with insufficient intake at the time of discharge (Supplemental Table 6). Seventeen percent had organic disease, most often MPI or reflux, and 5% had MFD.

Among patients with a single definitive diagnosis (excluding mixed and unknown etiologies), 86.6% were diagnosed during initial admission (Table 2). Insufficient intake and MFD were most often diagnosed during admission (93.5% and 90%). Of note, a diagnosis of neglect (a subtype of insufficient intake) was made less frequently (66% of the time) and was often diagnosed only after returning to care with continued GF or other signs of maltreatment.

Organic diagnoses were less frequently identified during admission. Sixty-five percent of diagnoses were made definitively during admission; 22% were neither diagnosed nor suspected (Table 2). Genetic disorders were most often undiagnosed during initial admission. Thirty-eight percent of these patients were diagnosed by testing sent during admission. An additional 25% of patients were suspected of having an underlying genetic disorder during admission, but diagnostic testing was inconclusive or deferred. Thirty-six percent of patients ultimately diagnosed with genetic disorders were not suspected during admission. At discharge, these patients were most often diagnosed with MFD (58%) or insufficient intake (33%).

Univariate analysis was used to identify demographic features and components of the history and physical exam associated with ultimate diagnosis (Table 3). Patients with insufficient intake were more likely to have public insurance, whereas organic diagnoses and MFD were all associated with private insurance.

Multivariable regression modeling was performed on components of the history and physical associated with ultimate diagnosis (Table 4). Insufficient intake was associated with more severe GF on presentation (lower z score for weight), unremarkable past medical history, and unremarkable physical exam. Malabsorption was positively associated with less severe GF (higher z score for weight), history of eczema or rash on presentation, and symptoms of vomiting or diarrhea. Genetic disease was associated with a history of NICU admission, LBW, dysphagia, and hypotonia and dysmorphic features on exam. MFD was associated with a history of reflux, dysphagia, aversion, and stridor on exam.

Initial hospital outcomes were also associated with ultimate diagnosis (Table 4). Patients with insufficient intake had shorter LOS, whereas patients with genetic disease and MFD had longer LOS, though only the association with insufficient intake remained significant in multivariable analysis. Patients with insufficient intake were also more likely to have adequate weight gain for age during hospitalization (81%), whereas patients with genetic disease were less likely to have adequate weight gain (43%). Association remained significant in subgroup regression modeling.

This is the largest study to date of patients admitted for GF, and the first to review data after admission to determine ultimate diagnosis. Previous studies have evaluated the diagnosis at hospital discharge, but many complex medical conditions cannot be diagnosed in a short hospital admission, so determining the long-term outcome of patients provides valuable information. Our results confirm that GF is most often because of insufficient caloric intake, as shown in previous studies, and this diagnosis is accurately made during hospital admission.^2,3,5,8,9  However, we also found that 20% of otherwise-healthy infants and toddlers admitted for GF have an underlying organic diagnosis, consistent with previous studies showing 5% to 50% of patients having organic disease.^2,3,8–12  Interestingly, 7% of our patients were ultimately diagnosed with a genetic disorder, a higher rate than previously reported, though data are limited.^2,8  This higher rate is likely because of the long-term follow-up in this study. The true rate of genetic disorders may be even greater; of the 39 patients with unknown diagnoses at 2 years postadmission, 3 were later diagnosed with genetic disorders and 5 had variants of unknown significance on chromosome microarray or exome sequencing without definitive diagnosis.

Previous studies found that organic diagnosis were almost always suspected on admission.^2,5,11  However, in this study, 22% of organic diseases were not suspected during admission and over one-third of patients with genetic disorders were not suspected of having a genetic disorder at the time of discharge. This suggests that our diagnostic accuracy for organic disease is not as high as previously reported.

However, this finding does not mean that more aggressive testing is indicated in admissions for GF, or that inpatient genetic testing should be the norm. Diagnosis of most of the genetic disorders identified in this study would not have changed acute inpatient management. Furthermore, many diagnoses could only be made with specific testing sent on the basis of symptoms that developed after discharge, and indiscriminate testing would not have identified these disorders.

Rather than indicating a need for more low-value testing, these results reinforce the importance of close outpatient follow-up for GF and the need for better predictive tools to determine which patients are likely to have significant underlying disease.

Some previous studies have attempted to identify patients most likely to have insufficient intake, but these focused on evaluating parenting and attachment, parental attitudes toward feeding, and even typologies of family structure.^3,13–15  Such evaluations are subjective and prone to bias. Many review articles suggest that patients with “asymmetric” GF (decrease in weight without corresponding decrease in length and/or head circumference) are more likely to have insufficient intake, but this has not been validated, and in this study, the relationship between z scores for length or head circumference and z score for weight did not correlate with ultimate diagnosis.^8,13,16  Weight gain during admission has also been used to identify patients with insufficient intake; however, a recent study found that greater than average weight gain during admission was not specific for a diagnosis of inadequate intake.⁶  Here, we identified components of the history and physical exam that were positively associated with ultimate diagnosis.

Patients with genetic disorders were more likely to have a history of NICU stay and/or LBW, dysphagia, and hypotonia or dysmorphic features on initial physical exam. Thirty-one of 36 patients (86%) with a genetic disorder had at least 1 of these features, but of these, only 62% were suspected of having a genetic disorder during their admission, delaying appropriate referrals at discharge. Patients with genetic disorders were also less likely to have adequate weight gain for age during admission. These features should prompt suspicion for genetic disease and consideration of genetics consultation or close follow-up with medical genetics for continued evaluation after discharge.

Similarly, a diagnosis of malabsorption was associated with less severe GF, symptoms of vomiting and/or diarrhea, and a history of eczema or rash on physical exam. Patients with these findings might benefit from a trial of elemental formula and/or outpatient follow-up with GI.

The retrospective nature of this study introduces several limitations, especially reliance on diagnosis codes for patient identification. The codes used were stringently reviewed and multiple codes considered. Also, all diagnosis codes input at any point in hospitalization were included, which should mitigate bias in final coding. However, because of the variable and subjective nature of the diagnosis of GF, it is possible that patients were included in this study who do not meet all definitions of GF. Also, because our institution has no standardized criteria for admission for GF, this study population is heterogeneous and may differ from patient populations at centers with standardized admission parameters.

Sixty-five patients were also lost to follow-up and could not be analyzed at 2 years postadmission. These patients were more likely to have been diagnosed with insufficient intake and they may have been lost to follow-up because they gained adequate weight and did not require reevaluation. This could lead to an undercount of this diagnosis and subsequent overrepresentation of other etiologies in our study. Given that we were able to see outpatient records for the vast majority of patients and all subspecialist records, the likelihood that patients with significant medical diagnoses were missed is low.

Although this study is the largest of its kind, given the rarity of individual organic diagnoses, we were not able to analyze many possible associations between history and physical exam findings and ultimate diagnosis. The heterogeneity of diagnostic groupings also made identifying associated variables difficult. Factors that were associated with genetic disease included signs of poor muscle tone (hypotonia and feeding difficulty) and dysmorphic features, which are classically associated with chromosomal anomalies. It is unlikely that these features are truly predictive of a genetic disorder such as Schwachman-Diamond syndrome, although this diagnosis is included in the same category.

Ultimately, although the broad outlines of this work are consistent with previously published literature, we have shown that organic disease, especially genetic disorders, are not as accurately suspected in admission as previously thought. With this large sample size, evaluated over a long period of time, we have detected more complex, difficult-to-diagnose disorders that were likely missed in previous studies. We have also identified features correlated with ultimate diagnosis that were present at initial admission, even if the diagnosis was not suspected at the time. Thus, the history and physical remain the most powerful diagnostic tools in the evaluation of GF.

This information can be used to better risk-stratify patients admitted for GF and tailor both inpatient evaluation and outpatient referrals. Ultimately, this data, if validated and expanded, could be used to generate diagnostic algorithms or practice guidelines to better standardize inpatient care of GF and reduce unnecessary testing.