Clinical Characteristics of US Adolescents Hospitalized for Eating Disorders 2010–2022

We used the PEDSnet database (www.pedsnet.org) to analyze data from 2010 to 2022 inclusive. PEDSnet is a national, multispecialty database of 12 children’s hospitals in the United States that collaboratively collects data primarily for clinical research. PEDSnet uses a pediatric-specific Common Data Model for the storage of PEDSnet data (PEDSnet Common Data Model–PEDSnet). Data domains include demographic data, diagnoses, procedures, and person-reported data such as surveys.

For overall data, we included all patient hospitalizations, including for patients who were rehospitalized during the study period. Because we were interested in studying adolescents, we included hospitalizations for patients aged 12 to 21 years who were hospitalized at any time from 2010 to 2022 with a primary diagnosis of 1 or more of the following: AN, BN, orthorexia nervosa, self-induced purging, self-induced vomiting to lose weight, binge ED, psychogenic overeating, nocturnal sleep-related ED, ARFID, and ED not further specified. We did not include pica because it may represent an underlying medical problem such as iron deficiency or malnutrition, nor did we include rumination disorder because it is typically classified as a gastrointestinal disorder. A previous study of 2 children’s hospitals involving manual chart review suggested that very few patients with primary EDs were coded as more general diagnoses such as weight loss.¹⁷  We stratified patients by year according to the date of hospitalization. We excluded hospitalizations for patients with the following diagnoses: systemic lupus erythematosus, inflammatory bowel disease/Crohn’s disease/ulcerative colitis, cystic fibrosis, malignancy, and those with tracheostomies because we wanted to be sure that we excluded patients who may have been classified as having disordered eating for a medical reason. We excluded patients who had a previous or known history of EDs who were hospitalized primarily for another reason.

For patient-level data, we collected each patient’s data only once (the first hospitalization), including data related to the patient’s demographic characteristics (age, race, documented gender, and type of insurance [public/private/other]) at the time of their hospitalization. Using discharge diagnosis data, we collected data regarding patients’ clinical characteristics, including type of ED, BMI percentage, presence of cardiac manifestations, and presence of coexisting psychiatric diagnoses such as depression, anxiety, suicidal thoughts, and posttraumatic stress disorder.

For the type of ED, we used related diagnoses for classification using group consensus among authors. For example, in the AN group, we included diagnoses such as AN, AN restricting type, and AN with dangerously low weight. We defined atypical AN as patients with median BMI percentage >85%.

For malnutrition severity, we used the proposed classification of malnutrition severity according to percentage median BMI (mBMI) (no malnutrition: >90% mBMI percentage; mild: 80%–90% mBMI percentage; moderate: 70%–80% mBMI percentage; and severe: <70% mBMI percentage).⁸  We generated descriptive statistics using R and Excel (Microsoft, Redmond, WA).

For the psychiatric conditions, we noted diagnoses such as anxiety, depression, obsessive-compulsive disorder, and others. For simplicity, we grouped the diagnoses of “panic disorder,” “panic disorder with agoraphobia,” and “panic disorder without agoraphobia” together in a group called panic disorder. Similarly, we grouped the diagnoses of “mood disorder” and “episodic mood disorder” together in a group called mood disorder.

For hospital stay characteristics, we collected data regarding length of stay and discharge disposition, using all hospitalizations (13 403) as the denominator. For rehospitalization rates, we used the number of unique patients (8652) as the denominator to count patients who were rehospitalized multiple times only once.

A simple linear regression model was run with the statsmodels package of Python¹⁸  to quantify how patient numbers changed by year in the time span covered by this study.

This study was deemed exempt by the Nemours institutional review board and this approval was accepted by the PEDSnet institutional review board.

This study included 13 403 hospitalizations by 8652 patients. The number of hospitalizations for adolescents for EDs increased steadily every year from 294 in 2010 to a maximum of 2135 in 2021 (Fig 1 and Supplemental Table 2). This represents an increase of more than sevenfold. The only decrease in the number of hospitalizations noted in our study was from 2135 in 2021 to 1783 in 2022. The linear regression model was significant (F [1–11] = 114.2, P < .001) and had an R² of 0.912. Over the timespan covered by this study, predicted patient count was equal to 108.5385 + 131.7692 (year – 2009), meaning that hospitalizations at the sites included in PEDSNet increased by an average of about 131 per year.

Demographic and clinical data are found in Table 1. Mean age was 15.8 years. Most patients were described as female (85.9%), white (71.8%), and not Hispanic/Latino (84.3%). Most patients (60.5%) had private insurance. The median BMI percentage of patients was 90.3%, which falls just above the mild malnutrition (80%–90% mBMI) category.

In terms of ED type, hospitalizations for patients with AN represented 57.5% of all hospitalizations in the study. Hospitalizations for patients with AN represented the largest group by ED type during every year of the study. Of the patients classified as having AN, 2133 of 5188 (41.1%) were classified as having atypical AN on the basis of mBMI percentage, whereas 1803 of 5188 (34.8%) were classified as having typical AN; the remaining 1252 patients (24.1%) did not have a BMI measurement available. Because patients could be diagnosed with >1 type of ED, the total number of patients listed in Table 1 for this category (9028) exceeded the number of patients in the study (8652). The number of patients with BN did not significantly change during the study period. However, the number of hospitalizations for patients with ARFID increased substantially since being tracked in 2015, even increasing from 2021 to 2022, a period during which the total number of patient hospitalizations for EDs fell (Fig 2).

In terms of severity of malnutrition (Fig 3 and Supp Table 3), just over half of patients had no malnutrition (50.9%), with 25.0% having mild, 18.6% having moderate, and 5.4% having severe malnutrition. Patients with no malnutrition and those with mild or moderate malnutrition represented most of the increase in patients over time, whereas the number of patients with severe malnutrition remained relatively stable over time.

The rate of cardiac manifestations of EDs is also listed in Table 1. The most common cardiac conditions were bradycardia (44.1% of patients), orthostatic hypotension (34.1%), conduction disorder of the heart (8.4%), tachycardia (6.8%), cardiac arrhythmia (4.2%), and abnormal electrocardiogram not otherwise specified (3.5%). These conditions were also not mutually exclusive.

The rate of psychiatric diagnoses in the patients studied was high (Table 1). The most common psychiatric conditions in patients in the study included depression (58.5% of patients), anxiety (57.0%), posttraumatic stress disorder (10.0%), and obsessive-compulsive disorder (8.1%). Many patients were diagnosed with >1 of these conditions. Suicidal thoughts were documented in 21.9% of patients.

Data regarding the hospital stay for these patients are included in Table 1. Median length of stay was 11 days. Most patients (71.9%) were discharged home from the hospital. A significant number (23.6%) of the 8652 patients studied were rehospitalized for reasons related to the ED within 1 year of initial hospitalization.

This study provides significant insight into the demographic and clinical characteristics of adolescents hospitalized in the United States for a primary diagnosis of ED from 2010 to 2022. Although the number and severity of hospitalized patients does not necessarily reflect trends in overall incidence of adolescent EDs, these data can provide important information about the patients who require acute treatment. Other studies have shown increased numbers of hospitalizations among adolescents with EDs during the COVID-19 pandemic,^14,16  though our study is the first to have examined trends as early as 2010. We were not surprised to note the dramatic increase in numbers of these patients during the COVID-19 pandemic, a time of unprecedented socioemotional stress, which can be an exacerbating factor in development or worsening of EDs.¹⁹ 

Importantly, the trend toward increased numbers of adolescents hospitalized for EDs had begun years before the COVID-19 pandemic. Explanations for this finding may be multifactorial, especially because genetic and environmental factors are known to play a role in the development of EDs.²⁰  Increased recognition of EDs, including updated hospitalization criteria, may have played a role. Another contributing factor may be the spread of social media and its focus on body presentation. This idea has received attention in the lay press as social media executives have testified before Congress about the effects of their platforms on adolescents’ behavior.²¹  Certain “pro-ana” (pro-anorexia) and “pro-mia” (pro-bulimia) blogs and Web sites offer vulnerable readers, especially female adolescents, strategies for avoiding treatment, including how to strategically respond to health care providers’ questions about their eating behavior.²² 

The demographic data generated by this study are notable. Most of the patients in this study were white, non-Hispanic females, though other populations, including males, adolescents of color, and transgender teens, are increasingly recognized as being at risk for EDs. Issues related to access to care may be 1 reason why male adolescents and those who are nonwhite were not hospitalized as frequently in our study. In fact, some studies have demonstrated that males and nonwhite patients may be at risk for more severe physiologic sequelae related to EDs.²³  Future studies should further explore differences in presentation among males and females, as well as patients of different racial and ethnic backgrounds.^24,25  We also note the significant rate of rehospitalization within 1 year among patients in our study. Future studies should attempt to identify risk factors for rehospitalization in patients with EDs.

We note that the severity of malnutrition of patients hospitalized for EDs seemed relatively stable or even slightly decreased over time. We surmise that patients may have been referred for hospitalization earlier in their disease course before their nutritional status worsened significantly. Several explanations for this finding are possible, including increased attention to EDs in the lay press or the fact that families were eating together more frequently so that parents and caregivers could observe eating behaviors more directly.¹⁹  The average mBMI percentage of patients included in this study was 90.3%, which likely reflects the diversity of eating disorder diagnoses represented in this sample, including atypical AN (41.1% of the patients with AN of all types were classified as atypical, though the true number was likely higher, because nearly one-quarter had no BMI documented). Although patients with typical AN are inherently underweight, patients with atypical AN and BN cannot be underweight to meet criteria for diagnosis, and patients with ARFID may or may not be underweight. It is important to note that previous studies have demonstrated that severity and rapidity of weight loss appear to be more closely correlated with medical instability than absolute weight at presentation; indeed, the fact that the overall mBMI percentage among patients in this study is slightly above the cutoff for mild malnutrition is significant, underscoring the importance of recognizing the impact of EDs regardless of a patient’s weight status.

One notable finding from our study is the dramatic increase in the number of patients with ARFID since 2015, even though these patients represent only 9.8% of the total number of patients hospitalized for EDs. This may represent a bias in recognition or classification, because ARFID, formerly known as selective ED, was added to the Diagnostic and Statistical Manual only in 2013.²⁶  Patients with ARFID differ from those with other EDs in that patients with ARFID do not have body image disturbance, and therefore fear of weight gain is not a primary driver of their ED behaviors; rather, these patients often exhibit extreme picky eating, lack of interest in food, or severe anxiety related to aversive consequences associated with eating, such as fear of choking or vomiting. Given the increase in recognition of ARFID, this trend deserves monitoring over the next several years.

We were not surprised by the high rate of psychiatric comorbidities, including suicidal ideation, among adolescents hospitalized for EDs. Previous research has found a strong association between anxiety and depression with more severe ED symptoms and poorer prognosis.²⁷  AN has been found to be the type of ED most highly associated with mortality²⁸  and 1 study found that 1 in 5 patients who died prematurely from EDs did so because of suicide.²  Our data likely underestimate the rate of suicidal ideation among adolescents hospitalized for EDs because the patient may not admit to having these thoughts on direct questioning or suicidal thoughts may not be included as a specific diagnosis. These data reinforce the importance of robust screening for suicidal thoughts and behaviors among adolescents hospitalized for EDs.

Our study has several limitations. We investigated trends, not causations, and there may be important exacerbating factors that we could not capture. For example, the impact of social media may have had significant effects, both positive and negative, on the mental health of many adolescents during the study period, especially during the COVID-19 pandemic.²⁹  Another limitation involves our source of data; although it captured data from a large, multistate network that includes many areas of the country, many US adolescents are hospitalized at community or general hospitals, and data from these patients were not captured by this study. PEDSnet may not represent a perfectly generalizable national sample for hospitalized adolescents, although the hospitals in this study group are large, diverse, and represent different areas of the country. We also recognize the limitation of using documented diagnoses for research purposes instead of objective data such as laboratory values. We may have missed patients who presented with weight loss or malnutrition caused by an ED who were coded as weight loss or malnutrition, though on the basis of a previous study,¹⁷  we believe these patients to be very few. In addition, only 6778 of the 8652 patients studied (78.3%) had BMI documented, so we could calculate malnutrition severity only for these patients. Our rates of patients who were rehospitalized likely underestimated the true number, because some patients may have required hospitalization at hospitals not included in PEDSnet or after 2022. Because of local trends in prevalence of COVID-19 during the pandemic, we were not able to formally include the pandemic’s impact in our analysis. Regarding the diagnosis of ARFID, we are unable to estimate the patients who would now be classified as having ARFID but were given other diagnoses before 2015. Finally, we did not capture transgender or nonbinary patients in our study because these data were not available; previous research has suggested that transgender patients are at higher risk for EDs than the general population.³⁰ 

The number of adolescents hospitalized for EDs in the United States has increased gradually since 2010 and dramatically during the COVID-19 pandemic, whereas the severity of malnutrition in these patients has remained stable. Most adolescents hospitalized for EDs continue to be white females, though males and patients of color must be recognized, as well. Patients with ARFID are being increasingly hospitalized. As the COVID-19 pandemic hopefully recedes, trends in adolescents hospitalized for EDs will likely remain dynamic.

This study was performed using the PEDSnet database, and we thank their significant contributions.