Epidemiology of Intensive Nutritional Interventions and Restraints in Hospitalizations for Anorexia Nervosa: Annual National Trends of the United States

ABSTRACT

Objective

Despite the clinical relevance of enteral and parenteral tube feeding and restraint use in anorexia nervosa (AN), national‐level data on their use have remained limited, particularly for total parenteral nutrition (TPN) and percutaneous endoscopic gastrostomy (PEG). This study systematically characterizes the national prevalence and temporal trends in intensive nutritional interventions, defined as nasogastric tube feeding (NGTF), TPN, and PEG, and physical restraint among hospitalizations for AN in the United States (US).

Method

We analyzed 66,785 hospitalizations for AN from the 2017 to 2022 National Inpatient Sample using modified Poisson and logistic regression models to estimate prevalence, temporal trends, and associated factors. Subgroup and sensitivity analyses were conducted by age groups, race or ethnicity, and hospital characteristics.

Results

Intensive nutritional interventions and restraints of any type were used in 6.8% of hospitalizations. The prevalence was 3.8% for NGTF, 1.7% for physical restraint, 1.1% for TPN, and 0.7% for PEG. The overall rate of any intervention increased from 5.8% in 2017 to 8.1% in 2022, with a significant annual increase of 0.43 percentage points (p = 0.007), primarily driven by NGTF and restraint. Adult patients with AN showed the highest prevalence of any intervention at 7.2%, mainly due to TPN and PEG. Higher intervention rates were observed in urban teaching hospitals.

Discussion

This study provides the first national‐level evidence that intensive nutritional interventions and restraints, particularly NGTF and restraint, are increasingly used in US hospitalizations for AN. The lower national estimates compared to single‐center studies likely reflect referral bias, supported by subgroup analysis. Increased use of intensive nutritional interventions and restraints is controversial and underscores the need to consider less intensive or coercive alternatives and clearer clinical ethics guidelines.

Summary

• This nationwide cohort study of 66,785 anorexia nervosa hospitalizations in the United States found that 6.8% involved intensive nutritional interventions and restraints, including nasogastric tube feeding (NGTF), total parenteral nutrition (TPN), percutaneous endoscopic gastrostomy (PEG), and physical restraint.

• The overall use of intensive nutritional interventions and restraints significantly increased from 2017 to 2022, particularly for NGTF and physical restraint, while TPN and PEG remained stable. The median initiation day of intensive nutritional interventions was Day 1 for NGTF, Day 2 for TPN, and Day 3 for PEG.

• The highest prevalence of intensive nutritional interventions and restraints was observed in adult patients with anorexia nervosa, primarily driven by the use of TPN and PEG, contrasting with previous reports focusing on adolescents.

• Psychiatric comorbidity, high physical comorbidity burden, and treatment non‐adherence were significantly associated with increased use of intensive nutritional interventions and restraints, highlighting ethical concerns and the need for less intrusive alternatives.

1. Introduction

Eating disorders account for some of the costliest and longest hospital stays among psychiatric disorders (Owens et al. 2016). Among these, anorexia nervosa (AN) carries the highest mortality, and both intensive nutritional interventions, including nasogastric tube feeding (NGTF), percutaneous endoscopic gastrostomy (PEG) tube feeding, and total parenteral nutrition (TPN), and coercive interventions, such as physical restraint, are used in clinical practice (Attia and Walsh 2025; Tsiandoulas et al. 2023). Although some patients with AN regard such interventions as necessary (Halse et al. 2005; Tan et al. 2010) and accept them voluntarily, NGTF is effective in some populations (Rigaud et al. 2007; Rizzo et al. 2019), some cases may involve elements of coercion or compulsion. Also, apart from restraint or other coercive interventions, NGTF is associated with symptom severity of AN (Clausen 2020; Seed et al. 2016). The use of PEG or TPN is not preferred or recommended by professional guidelines (Attia and Walsh 2025; American Psychiatric Association 2023). Moreover, restraints and TPN carry inherent physical risks such as deep venous thrombosis and catheter‐related infections. PEG generally follows a benign clinical course after tube removal in most cases; however, case reports have described potential complications, including granuloma formation and tube site infection (Alsunaid et al. 2021). In addition, prolonged placement exceeding 6 months may result in gastrocutaneous fistula formation after tube removal (Papavramidis et al. 2011).

Given the recent increase in AN cases (Tsiandoulas et al. 2023), the prevalence of individuals with AN across a wide range of illness severity is expected to rise. In severe and enduring AN, repeated involuntary interventions may cause greater harm than the improvements in quality of life they are intended to achieve (Bianchi et al. 2021; Tsiandoulas et al. 2023). However, the importance of harm reduction extends beyond severe and enduring AN to the broader treatment of AN, where minimizing involuntary interventions and enhancing motivation are essential to reducing potential harms and strengthening the therapeutic alliance (Clausen 2020). Nevertheless, translating these findings into practice remains challenging because key epidemiologic data remain limited. A previous systematic review showed that 13%–44% of inpatients with AN experienced involuntary treatment (Clausen and Jones 2014), but these estimates were based on single‐ or multi‐center inpatient data, such as university hospital psychiatric wards or specialized eating disorder units, and none of the included studies was based on a national‐level cohort. Notably, although the use of restraints is generally higher in adolescent psychiatric populations (Furre et al. 2014), only a small number of studies have examined the prevalence of intensive nutritional interventions and restraints specifically among adolescents with eating disorders (Ayton et al. 2009; Carney et al. 2006). Following the systematic review, only one study, using a national cohort from Denmark, reported that involuntary tube feeding and mechanical restraint were used in 5.0% and 6.3% of patients with AN, respectively (Clausen et al. 2018). In contrast, higher prevalence rates were observed in studies conducted at single centers or regional eating disorder units. A Norwegian single‐center study reported that 31% of adolescent patients experienced physical restraint, of whom 21% were subject to meal‐related physical restraint (Blikshavn et al. 2020), and another from the same institution showed that 18% of patients received involuntary NGTF (Halvorsen et al. 2016). In the United Kingdom, a single‐center study reported that 30% of patients underwent involuntary NGTF (Ayton et al. 2009), although this study was conducted in a highly specialized adolescent eating disorder service and was not intended to estimate national prevalence. Also, an Australian study reported a prevalence of 24% (Carney et al. 2006). A retrospective analysis from four Scottish eating disorder units found NGTF in 12.7% of adolescent and 8.5% of adult hospitalizations (Morris et al. 2015). Due to the limited national cohort studies, the existence of differences in prevalence rate of intensive nutritional interventions and restraints for AN between national cohort and single center cohort is still unknown.

Also, in the United States, available data are restricted to single‐center case series, with involuntary admission rates ranging from 5.2% to 16.6% (Watson et al. 2000; Westmoreland and Mehler 2016), and prior reviews have consistently noted the absence of robust national‐level estimates (Elzakkers et al. 2014). These scarce evidence of national estimates of intensive nutritional interventions and restraints in AN results in even more restricted knowledge of their annual trends. The Danish national cohort did not show a clear increase in involuntary measures over time (Clausen et al. 2018), although it included broader coercive measures such as involuntary admissions and detentions. Furthermore, disease‐specific intensive nutritional interventions such as TPN and PEG have not been reported, to our knowledge, thereby limiting understanding of the full scope of intensive nutritional interventions and restraints for AN (Clausen et al. 2018; Fuller et al. 2023).

To address these gaps, we conducted a nationwide analysis of acute hospitalizations for AN in the US community hospitals, aiming to estimate the prevalence and assess temporal trends of inpatient intensive nutritional interventions, including NGTF, TPN, PEG, and physical restraints.

2. Method

2.1. Details of NIS Cohort

We analyzed data from the 2017–2022 National Inpatient Sample (NIS) collected by the Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality, the largest all‐payer inpatient database in the United States, capturing approximately 35 million hospitalizations annually (Healthcare Cost and Utilization Project 2024a, 2024b, 2024c). Complex survey design was incorporated following methodological recommendations, such as using weights for national estimates; accounting for the design of the NIS when calculating standard errors; and accounting for clustering or nesting of observations (Khera et al. 2017). Importantly, although the NIS captures 97% of hospitalizations in US hospitals, it does not cover residential treatment programs conducted in facilities that are not licensed as hospitals (Kaye and Bulik 2021; Attia et al. 2016).

2.2. Identification of Hospitalizations With Anorexia Nervosa

Hospitalizations with AN and related interventions were identified using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM); International Classification of Diseases, Tenth Revision, Procedure Coding System (ICD‐10‐PCS); Major Diagnostic Categories (MDC); and Diagnosis Related Groups (DRG). For the detailed codes, see Table S1.

Following the previous research using NIS dataset (Kalla et al. 2017; Owens et al. 2016), hospitalizations with AN were identified using the ICD‐10‐CM codes that start with F50.0. Hospitalizations that have this code in any diagnosis data element were regarded as hospitalizations with AN. Since the likelihood of underlying neurodegenerative or geriatric psychiatric conditions increases in older populations, potentially reducing the specificity of AN identification in administrative data (Beydoun et al. 2015; Hou et al. 2019), patients aged ≥ 60 years were excluded. Patients aged < 10 years old were also excluded (Petkova et al. 2019; World Health Organization 2024).

Exclusion criteria included body mass index over 20 in adults, body mass index over the 5th percentile for age in non‐adults, and hospitalizations related to pregnancy, childbirth, and the puerperium.

2.3. Definitions and Identification of Intensive Nutritional Interventions and Restraints in Hospitalizations With Anorexia Nervosa

In this study, we aimed to estimate the frequency of interventions beyond simple meal based nutritional rehabilitation. To this end, intensive nutritional interventions were operationalized as NGTF, PEG, TPN. In addition, the use of physical restraints was examined. This definition excludes interventions such as nutritional advice or observation of eating behavior, which do not involve invasive or coercive measures. These interventions were specified based on previous studies (Attia and Walsh 2025; Blikshavn et al. 2020), and were also confirmed by the counts of frequently provided procedures in the study cohort. These interventions involve the use of physical instruments or surgical procedures, some of which may require application of mental health legislation depending on the country or region (Himmerich et al. 2024). To examine the overall trend, we defined “total” as hospitalizations in which any one of these four interventions was used. In this study, each intervention and its related terms were defined and characterized as follows.

2.3.1. Enteral Nutrition

Delivery of nutrients through the gastrointestinal tract via a feeding tube when oral intake is insufficient or not possible. In this study this primarily refers to NGTF, and also PEG. Enteral nutrition describes the route of nutritional delivery and does not in itself indicate whether treatment is voluntary or involuntary.

2.3.2. Percutaneous Endoscopic Gastrostomy

A surgically placed gastrostomy tube inserted through the abdominal wall to allow longer term enteral feeding. PEG placement is an invasive procedural intervention used to establish durable enteral access.

2.3.3. Parenteral Nutrition

Administration of nutrients intravenously through vascular access, bypassing the gastrointestinal tract. TPN refers to complete nutritional support delivered through central venous access. Parenteral nutrition represents invasive medical nutritional support rather than a measure of coercion.

2.3.4. Physical Restraint

Mechanical methods that restrict a patient's freedom of movement (Luccarelli et al. 2023). In hospital settings restraint may be used for several clinical reasons, including prevention of self harm, management of agitation, or enabling urgent medical treatment. Because restraint restricts liberty it is considered a coercive intervention.

2.3.5. Involuntary Treatment

Medical or psychiatric treatment delivered without the patient's valid consent, usually under legal or best interest frameworks when a patient lacks decision making capacity or refuses treatment. Involuntary treatment refers to the legal status of care rather than the specific procedure performed.

Importantly, the National Inpatient Sample records procedure and diagnostic codes, including restraint status, but does not capture consent, decision making capacity, or legal authorisation. Accordingly, although the dataset can identify enteral feeding procedures such as NGTF, PEG placement, parenteral nutrition including TPN, and restraint, it cannot determine whether these interventions were delivered voluntarily or involuntarily. Furthermore, because NGTF, TPN, and PEG are identified using ICD‐10‐PCS procedure codes, the NIS procedure day variables permit analysis of intervention initiation timing, although procedure end day information is unavailable. By contrast, physical restraint is identified using ICD‐10‐CM diagnosis codes, for which corresponding procedure day information is unavailable. Consequently, initiation timing for restraint could not be analyzed.

2.4. Other Covariates

Following covariates were included: age, race/ethnic, sex, primary payer, income quartile, weekend admission, primary psychiatric diagnosis, psychiatric comorbidity based on the HCUP Clinical Classifications Software Refined (CCSR) (Healthcare Cost and Utilization Project 2024a, 2024b, 2024c), substance use disorders (SUDs) comorbidity based on the HCUP CCSR, weighted summary score of Elixhauser comorbidity index by van Walraven (van Walraven et al. 2009) based on the HCUP Elixhauser Comorbidity Software Refined for ICD‐10‐CM (excluding psychiatric and substance use comorbidity) (Healthcare Cost and Utilization Project 2024a, 2024b, 2024c), adherence to treatment, non‐elective admission, provision of emergency department services, hospital division, hospital bedsize, and teaching status of hospital. Individuals with higher Elixhauser comorbidity index scores were regarded as having a high comorbidity burden.

Whether the principal diagnosis was attributable to a psychiatric condition was determined based on either the MDC code or the DRG code. Given that AN leads to malnutrition and may be classified under MDC 10 (Endocrine, Nutritional, and Metabolic Diseases and Disorders), cases were considered to have a psychiatric principal diagnosis not only if they fell under MDC 19 (Mental Diseases and Disorders) but also if they were assigned to DRG 640 or 641, which correspond to miscellaneous disorders of nutrition, metabolism, fluids, and electrolytes.

All hospitalization costs were adjusted for inflation to 2022 US dollars using the Gross Domestic Product (GDP) Price Index. The deflator values were based on estimates published by the US Bureau of Economic Analysis. Year‐specific GDP deflators from 2016 to 2022 were scaled relative to the 2022 value, and nominal costs were divided by these deflators to obtain inflation‐adjusted costs.

2.5. Trend Analyses

To examine annual trends in the use of each intervention from 2017 to 2022, we fitted modified Poisson regression models accounting for survey design and discharge‐level sampling weights. Models were estimated both with and without covariate adjustment. Calendar year was treated as a continuous variable to estimate linear trends. To aid interpretability, we converted the relative annual risk change into absolute annual percentage point change by multiplying the risk ratio by the baseline probability in 2017. Trends of characteristics were also assessed by modified Poisson regression models without covariate adjustment.

To further assess the magnitude of change in intervention use, we estimated the absolute difference in predicted probabilities between 2017 and 2022 using marginal effects from logistic regression models. For each outcome, logistic regression models with year specified as a categorical variable were constructed, with and without covariate adjustment. Predicted probabilities for 2017 and 2022 were derived, and the difference was calculated. This absolute difference was then divided by five to reflect the average annual percentage point change over the study period.

2.6. Association Between Characteristics of Hospitalizations and Intensive Nutritional Interventions and Restraints

To examine associations between characteristics of hospitalizations and the use of intensive nutritional interventions and restraints, we conducted modified Poisson regression analyses using data pooled across all years. Regression models included all covariates as predictors, and estimates were reported as adjusted risk ratios (aRR). Proportions of each comorbidity were compared between admissions with and without any intensive nutritional interventions and restraints using Rao–Scott chi‐square tests. Holm‐adjusted p values were calculated to account for multiple comparisons. Length of stay was compared between admissions with and without intensive nutritional interventions and restraints using survey‐weighted t‐tests with 95% confidence intervals. To further explore the relationship between age and intensive nutritional interventions and restraints, we modeled nonlinear associations using natural cubic splines with four degrees of freedom. All models were adjusted for characteristics of hospitalizations, and age 18 years was defined as the reference.

2.7. Subgroup Analyses

In order to identify the population most affected by intensive nutritional interventions and restraints, subgroup analyses were conducted using age groups, race, primary psychiatric diagnosis, and hospital location and teaching status. For age groups, patients were categorized into adolescents (< 18 years), young adults (18–27 years), and adults (≥ 28 years). These age groups were used as pragmatic demographic strata and do not represent illness duration or clinical staging. To compare the average annual use of interventions between 2017 and 2022 across subgroups, we conducted chi‐square tests. In addition, subgroup‐specific trends over time were evaluated to assess whether the use of interventions changed differentially across strata.

2.8. Sensitivity Analyses

To assess the robustness of our findings, we conducted a series of sensitivity analyses targeting both diagnostic specificity and the classification of intensive nutritional interventions and restraints. To evaluate the reliability of diagnostic classification, we first restricted the sample to hospitalizations with personality and impulse control disorders, as this is the diagnostic group under which eating disorders are categorized in the DRG system. Second, we also restricted the sample to hospitalizations with documented weight loss based on the Elixhauser comorbidity index, reflecting cases with explicit coding of undernutrition. Third, since differentiating psychotic disorders from AN is important (World Health Organization 2024), hospitalizations with these conditions in the Elixhauser comorbidity index were excluded.

To test the robustness of classification, we performed two additional analyses. First, we excluded hospitalizations involving neurological conditions, particularly those potentially indicative of neurodegenerative disorders. This decision was based on the consideration that physical restraints are often applied in these hospitalizations, and that NGTF or PEG tube feeding may be used in the context of feeding difficulties related to such conditions. We excluded diseases of nervous systems considered suggestive of neurodegenerative diseases, while retaining those related to migraine, sleep disorders, nerve root and peripheral nerve disorders, nonspecific pain syndromes, and non‐intractable epilepsy, which were not deemed relevant in this context. Additionally, we excluded hospitalizations involving major operating room procedures, which were coded beginning in 2019, to ensure that observed trends in the use of physical restraints were not confounded by surgical care.

2.9. Other Statistical Methods and Ethical Consideration

Missing values were handled with multiple imputation. This study was approved by the institutional review board of Tokyo Metropolitan Matsuzawa Hospital (FY2024‐No.28). All statistical analyses were two‐sided, with p < 0.05 considered significant, and performed using R 4.4.1 (R Core Team).

3. Results

A total of 66,785 hospitalizations with AN were included (Figure 1). The median age was 21 years (interquartile range, 16–35), 92% were female, and 80% were White (Table 1). Seventy‐three percent of patients were hospitalized for psychiatric treatment or for the management of malnutrition. Eighty‐six percent of admissions occurred in urban teaching hospitals. The mean length of stay was 11 days. During the study period, the proportion of hospitalizations for AN increased among younger, non‐White individuals without comorbid SUDs. Hospitalizations involving emergency department services also became more frequent (Figure S1 and Table S2).

FIGURE 1.

Study flow. AN, anorexia nervosa; BMI, body mass index. All numbers are shown as the weighted population.

TABLE 1.

Characteristics of study hospitalizations.

Characteristic	Overall a (n = 66,785)	Intensive nutritional interventions and restraints		SMD, 95% CI	p
		Provided (n = 4535)	Not provided (n = 62,250)
Age (Median, IQR)	21 (16, 35)	23 (16, 37)	21 (16, 35)	0.059 (0.029, 0.090)	0.327
Sex				0.027 (0.003, 0.057)	0.441
Female	61,615 (92%)	4215 (93%)	57,400 (92%)
Male	5150 (7.7%)	320 (7.1%)	4830 (7.8%)
Unknown	20 (< 0.1%)	0 (0%)	20 (< 0.1%)
Racial/ethnics groups				0.101 (0.071, 0.132)	0.166
White	50,730 (80%)	3470 (80%)	47,260 (80%)
Black	2190 (3.4%)	185 (4.3%)	2005 (3.4%)
Hispanic	5805 (9.1%)	335 (7.8%)	5470 (9.2%)
Asian or Pacific Islander	2125 (3.3%)	110 (2.5%)	2015 (3.4%)
Native American	240 (0.4%)	30 (0.7%)	210 (0.4%)
Other	2440 (3.8%)	190 (4.4%)	2250 (3.8%)
Unknown	3255 (4.9%)	215 (4.7%)	3040 (4.9%)
Primary payer				0.094 (0.064, 0.125)	0.327
Medicare	7920 (12%)	620 (14%)	7300 (12%)
Medicaid	16,595 (25%)	1100 (24%)	15,495 (25%)
Private insurance	38,515 (58%)	2540 (56%)	35,975 (58%)
Self‐pay	1770 (2.7%)	130 (2.9%)	1640 (2.6%)
No charge	115 (0.2%)	0 (0%)	115 (0.2%)
Other	1755 (2.6%)	145 (3.2%)	1610 (2.6%)
Unknown	115 (0.2%)	0 (0%)	115 (0.2%)
Income quartile				0.109 (0.078, 0.139)	0.033
1	9560 (15%)	570 (13%)	8990 (15%)
2	13,475 (20%)	915 (21%)	12,560 (20%)
3	17,210 (26%)	1020 (23%)	16,190 (26%)
4	25,620 (39%)	1910 (43%)	23,710 (39%)
Unknown	920 (1.4%)	120 (2.6%)	800 (1.3%)
Weekend admission	11,420 (17%)	670 (15%)	10,750 (17%)	0.068 (0.038, 0.098)	0.060
Primary psychiatric diagnosis	48,845 (73%)	3015 (66%)	45,830 (74%)	0.156 (0.126, 0.187)	< 0.001
Psychiatric comorbidity	57,170 (86%)	3965 (87%)	53,205 (85%)	0.057 (0.027, 0.087)	0.135
SUD comorbidity	19,400 (29%)	975 (21%)	18,425 (30%)	0.187 (0.156, 0.217)	< 0.001
Elixhauser comorbidity index (Mean, SD)	5.2 (5.5)	7.3 (6.0)	5.0 (5.4)	0.397 (0.367, 0.427)	< 0.001
Non‐adherence	5140 (7.7%)	485 (11%)	4655 (7.5%)	0.112 (0.082, 0.142)	0.002
Non‐elective admission	59,325 (89%)	4105 (91%)	55,220 (89%)	0.051 (0.021, 0.081)	0.187
Unknown	260 (0.4%)	5 (0.1%)	255 (0.4%)
ED service provided	28,300 (49%)	1975 (52%)	26,325 (49%)	0.068 (0.035, 0.101)	0.108
Unknown	9265 (14%)	765 (17%)	8500 (14%)
Hospital division				0.226 (0.195, 0.256)	< 0.001
New England	5270 (7.9%)	470 (10%)	4800 (7.7%)
Middle Atlantic	12,815 (19%)	905 (20%)	11,910 (19%)
East North Central	8760 (13%)	565 (12%)	8195 (13%)
West North Central	6790 (10%)	270 (6.0%)	6520 (10%)
South Atlantic	9920 (15%)	820 (18%)	9100 (15%)
East South Central	2370 (3.5%)	120 (2.6%)	2250 (3.6%)
West South Central	4040 (6.0%)	200 (4.4%)	3840 (6.2%)
Mountain	6065 (9.1%)	410 (9.0%)	5655 (9.1%)
Pacific	10,755 (16%)	775 (17%)	9980 (16%)
Hospital bedsize				0.164 (0.134, 0.194)	0.001
Small	15,100 (23%)	765 (17%)	14,335 (23%)
Medium	16,935 (25%)	1335 (29%)	15,600 (25%)
Large	34,750 (52%)	2435 (54%)	32,315 (52%)
Teaching status of hospital				0.189 (0.158, 0.219)	< 0.001
Rural	2390 (3.6%)	90 (2.0%)	2300 (3.7%)
Urban nonteaching	7080 (11%)	300 (6.6%)	6780 (11%)
Urban teaching	57,315 (86%)	4145 (91%)	53,170 (85%)

Note: Bold values indicate statistical significance at p.

Abbreviations: CI, confidence interval; ED, emergency department; IQR, interquartile range; SD, standard deviation; SMD, standardized mean difference; SUD, substance use disorder.

^a Data are presented as n (%) unless otherwise stated.

On average, at least one intensive nutritional intervention and restraints were provided in 6.8% (95% CI 6.3–7.3) of hospitalizations from 2017 to 2022 (Figure 2). For each category of intensive nutritional interventions and restraints, physical restraint was provided in 1.7% (1.4–2.0), NGTF in 3.8% (3.3–4.0), TPN in 1.1% (0.9–1.3), and PEG in 0.7% (0.6–0.8). Among those receiving intensive nutritional interventions and restraints, 6.1% (275 out of 4535 hospitalizations) underwent more than one category of intensive nutritional interventions and restraints. The most frequent combination of intensive nutritional interventions and restraints was NGTF with restraint, observed in 6.7% (155/2530) of NGTF cases. The mean length of stay was 10.1 days (95% CI, 9.6–10.5) for admissions without intensive nutritional interventions and restraints and 21.9 days (95% CI, 20.1–23.6) for those with such interventions, showing significantly longer hospitalizations among admissions that received intensive nutritional interventions and restraints (p < 0.001). The initiation day of intensive nutritional interventions was Day 1 (interquartile range (IQR), 0–4) for NGTF, Day 2 (IQR, 0–6) for TPN, and Day 3 (IQR, 1–8) for PEG. Among physical comorbidities, the presence of congestive heart failure, cardiac arrhythmias, valvular disease, fluid and electrolyte disorders, and nutritional anemia not secondary to blood loss was significantly associated with the receipt of intensive nutritional interventions and restraints (Table S3).

FIGURE 2.

Combinations of intensive nutritional interventions and restraints in hospitalizations with anorexia nervosa, 2017–2022. NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition. Only point estimates are shown. All numbers are shown as the weighted population. Percentages indicate proportions relative to the total weighted population (n = 66,785).

Overall, any intensive nutritional interventions and restraints increased from 5.8% in 2017 to 8.1% in 2022, with an annual absolute increase of 0.43% (95% CI, 0.14–0.73; p = 0.007 for trend). TPN and PEG usage remained stable, while hospitalizations involving physical restraint increased from 1.0% to 2.2% and NGTF from 3.3% to 5.1% over the same period, both showing significant trends (Figures 3 and 4). Except for NGTF, these significant trends persisted after covariate adjustment. The adjusted annual change in the concurrent use of NGTF and restraint per hospitalization was 0.03% (95% CI, −0.01 to 0.06; p = 0.167 for trend) (Figure S2). Marginal effects models (Figure S3) and sensitivity analyses (Figure S4) showed similar trends.

FIGURE 3.

Annual trends in the use of intensive nutritional interventions and restraints among hospitalizations with anorexia nervosa. NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition.

FIGURE 4.

Forest plots of annual absolute changes in intensive nutritional interventions and restraints among hospitalizations with anorexia nervosa estimated by trend analyses. NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition.

Significantly high aRR for receiving intensive nutritional interventions and restraints were seen in hospitalizations with high comorbidity index, psychiatric comorbidity, no SUD comorbidity, and non‐adherence to treatment (Table 2). Some hospital characteristics, such as medium to large bed size and urban teaching status, were associated with a higher likelihood of providing intensive nutritional interventions and restraints. For both restraint and NGTF, psychiatric comorbidity and non‐adherence were consistently associated with increased risk, whereas these associations were not observed for TPN and PEG. Age was negatively associated with the risk of restraint and NGTF, but positively associated with the risk of TPN and PEG (Table 2; Figure 5). Emergency department services were provided in 49% of admissions. The provision of emergency department services was not correlated with the overall use of intensive nutritional interventions and restraints, and TPN use was significantly less frequent when emergency department services were provided. For the negative association between SUD and intensive nutritional interventions and restraints, a statistically significant negative association was identified specifically in the adult patients (Table S4). In particular, NGTF, TPN, and PEG, but not physical restraint, were less frequently used among patients with SUD in the adult age group.

TABLE 2.

Association between characteristics of hospitalizations and intensive nutritional interventions and restraints.

Characteristics	Total			Restraint			NGTF			TPN			PEG
	aRR	95% CI	p	aRR	95% CI	p	aRR	95% CI	p	aRR	95% CI	p	aRR	95% CI	p
Age [year]	1.00	(0.99–1.01)	0.838	0.98	(0.96–0.99)	0.009	0.99	(0.97–1.00)	0.013	1.03	(1.02–1.05)	< 0.001	1.04	(1.02–1.06)	< 0.001
Racial/ethnics groups [ref: White]
Black	1.05	(0.69–1.61)	0.821	1.40	(0.73–2.71)	0.312	0.76	(0.39–1.45)	0.402	2.35	(0.95–5.76)	0.063	0.28	(0.04–2.10)	0.216
Hispanic	0.83	(0.62–1.12)	0.228	0.53	(0.27–1.03)	0.062	0.98	(0.68–1.40)	0.904	0.72	(0.30–1.76)	0.469	0.41	(0.09–1.78)	0.233
Asian or Pacific Islander	0.70	(0.44–1.14)	0.152	0.33	(0.08–1.31)	0.116	0.61	(0.32–1.19)	0.150	1.19	(0.36–3.94)	0.780	1.21	(0.30–4.84)	0.784
Native American	1.87	(0.84–4.15)	0.124	Not estimable		NA	2.27	(0.75–6.89)	0.149	3.83	(0.95–15.40)	0.059	Not estimable		NA
Other	1.04	(0.72–1.48)	0.849	0.63	(0.27–1.46)	0.283	1.35	(0.90–2.04)	0.147	0.36	(0.05–2.58)	0.308	0.44	(0.06–3.32)	0.424
Male [ref: Female]	0.91	(0.69–1.21)	0.529	0.83	(0.46–1.52)	0.549	0.80	(0.54–1.19)	0.271	1.14	(0.54–2.39)	0.725	1.49	(0.66–3.41)	0.340
Payer [ref: Medicare]
Medicaid	1.04	(0.78–1.38)	0.804	1.34	(0.75–2.40)	0.316	0.87	(0.57–1.35)	0.544	0.77	(0.41–1.45)	0.427	1.76	(0.84–3.72)	0.137
Private insurance	0.91	(0.70–1.18)	0.481	0.97	(0.56–1.68)	0.926	0.79	(0.53–1.18)	0.257	0.72	(0.42–1.23)	0.235	1.61	(0.78–3.33)	0.199
Self‐pay	1.29	(0.80–2.08)	0.289	2.91	(1.38–6.14)	0.005	1.64	(0.88–3.07)	0.122	Not estimable		NA	0.6	(0.08–4.55)	0.618
No charge	Not estimable		NA	Not estimable		NA	Not estimable		NA	Not estimable		NA	Not estimable		NA
Other	1.33	(0.86–2.06)	0.203	1.98	(0.83–4.70)	0.122	1.10	(0.59–2.06)	0.772	0.5	(0.12–2.04)	0.335	1.69	(0.40–7.22)	0.476
Income quartile [ref: 1]
2	1.14	(0.87–1.49)	0.330	0.84	(0.50–1.40)	0.500	1.24	(0.85–1.81)	0.261	1.31	(0.62–2.80)	0.479	1.24	(0.65–2.35)	0.521
3	0.95	(0.73–1.23)	0.687	0.73	(0.45–1.20)	0.217	0.98	(0.68–1.42)	0.923	1.7	(0.83–3.48)	0.148	0.61	(0.29–1.25)	0.177
4	1.13	(0.89–1.44)	0.299	0.98	(0.64–1.50)	0.936	1.29	(0.91–1.83)	0.154	1.87	(0.97–3.59)	0.061	0.44	(0.21–0.92)	0.029
Weekend admission	0.85	(0.69–1.04)	0.112	0.87	(0.58–1.29)	0.480	0.88	(0.67–1.16)	0.351	0.86	(0.50–1.46)	0.571	0.9	(0.50–1.64)	0.735
Primary psychiatric diagnosis	0.88	(0.73–1.05)	0.161	0.76	(0.54–1.08)	0.122	1.47	(1.11–1.94)	0.007	0.26	(0.15–0.43)	< 0.001	0.67	(0.36–1.24)	0.204
Comorbidity index	1.06	(1.05–1.07)	< 0.001	1.02	(1.00–1.05)	0.110	1.07	(1.06–1.09)	< 0.001	1.07	(1.04–1.10)	< 0.001	1.08	(1.04–1.11)	< 0.001
Psychiatric comorbidity	1.40	(1.11–1.77)	0.005	2.92	(1.54–5.56)	0.001	1.45	(1.06–1.98)	0.020	0.83	(0.52–1.32)	0.423	1	(0.52–1.91)	0.997
SUD comorbidity	0.61	(0.50–0.74)	< 0.001	0.96	(0.68–1.35)	0.800	0.57	(0.43–0.75)	< 0.001	0.45	(0.27–0.74)	0.002	0.38	(0.20–0.70)	0.002
Non‐adherence	1.64	(1.25–2.14)	< 0.001	2.62	(1.68–4.10)	< 0.001	1.61	(1.11–2.32)	0.011	1.32	(0.67–2.58)	0.418	0.64	(0.20–2.07)	0.455
Non‐elective admission	1.00	(0.76–1.30)	0.977	0.97	(0.51–1.83)	0.914	1.11	(0.77–1.59)	0.569	1.14	(0.60–2.18)	0.691	0.42	(0.20–0.88)	0.022
ED service provided	1.01	(0.84–1.21)	0.898	1.30	(0.90–1.89)	0.167	1.02	(0.79–1.32)	0.868	0.57	(0.36–0.91)	0.019	1.62	(0.88–2.97)	0.118
Hospital division [ref: New England]
Middle Atlantic	0.67	(0.49–0.93)	0.015	0.76	(0.38–1.52)	0.443	0.49	(0.33–0.72)	< 0.001	0.62	(0.26–1.47)	0.277	4.23	(0.54–32.98)	0.169
East North Central	0.59	(0.41–0.85)	0.004	0.54	(0.26–1.16)	0.113	0.38	(0.24–0.62)	< 0.001	0.68	(0.27–1.69)	0.404	7.35	(0.94–57.68)	0.058
West North Central	0.41	(0.27–0.64)	< 0.001	0.33	(0.14–0.79)	0.013	0.23	(0.11–0.45)	< 0.001	0.69	(0.25–1.85)	0.458	6.23	(0.75–51.45)	0.09
South Atlantic	0.73	(0.51–1.05)	0.094	0.53	(0.24–1.16)	0.113	0.60	(0.38–0.94)	0.026	0.83	(0.32–2.14)	0.703	5.61	(0.68–46.36)	0.109
East South Central	0.47	(0.29–0.78)	0.004	0.06	(0.01–0.53)	0.010	0.57	(0.31–1.04)	0.067	Not estimable		NA	3.55	(0.35–36.58)	0.286
West South Central	0.47	(0.31–0.73)	< 0.001	0.63	(0.28–1.44)	0.272	0.18	(0.09–0.35)	< 0.001	1.26	(0.48–3.34)	0.638	5.61	(0.65–48.76)	0.118
Mountain	0.66	(0.46–0.95)	0.026	0.72	(0.33–1.58)	0.414	0.48	(0.27–0.84)	0.010	1.31	(0.54–3.18)	0.547	1.73	(0.17–17.23)	0.639
Pacific	0.61	(0.42–0.90)	0.013	0.50	(0.21–1.18)	0.115	0.51	(0.33–0.81)	0.004	0.94	(0.38–2.34)	0.898	1.48	(0.13–16.50)	0.748
Hospital bedsize [ref: Small]
Medium	1.43	(1.10–1.86)	0.007	1.02	(0.61–1.71)	0.946	1.89	(1.31–2.74)	< 0.001	0.89	(0.47–1.66)	0.704	0.89	(0.42–1.88)	0.755
Large	1.34	(1.07–1.68)	0.012	1.51	(0.99–2.31)	0.054	1.38	(0.97–1.94)	0.070	1.34	(0.80–2.25)	0.271	1.04	(0.56–1.93)	0.912
Teaching status of hospital [ref: Rural]
Urban nonteaching	1.12	(0.60–2.10)	0.727	6.35	(0.79–51.13)	0.083	0.67	(0.27–1.67)	0.386	0.66	(0.21–2.11)	0.488	4.66	(0.56–38.65)	0.154
Urban teaching	2.13	(1.25–3.63)	0.005	7.89	(1.08–57.90)	0.042	1.76	(0.86–3.62)	0.123	1.06	(0.39–2.90)	0.904	6.75	(0.89–51.16)	0.065

Note: Boldface indicates statistically significant results.

Abbreviations: aRR, adjusted risk ratio; CI, confidence interval; ED, emergency department; NA, not available; NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; SUD, substance use disorder; TPN, total parenteral nutrition.

FIGURE 5.

Adjusted risk ratios for intensive nutritional interventions and restraints across the age span. Solid lines and shaded areas represent adjusted risk ratios and 95% confidence intervals, respectively, with age 18 years as the reference. NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition.

In the subgroup analyses, the adult age group showed the highest (7.2%; 95% CI, 6.4–8.0) and significantly increasing prevalence of intensive nutritional interventions and restraints, primarily due to elevated use of TPN and PEG (Table 3 and Table 4). On the other hand, intensive nutritional interventions and restraints in the adolescent and young adult age groups mainly consisted of restraint and NGTF. Proportion and trends varied by race. Although not statistically significant, the use of intensive nutritional interventions and restraints among Hispanic patients with AN showed an increasing trend. NGTF was more frequently used among patients with a primary psychiatric diagnosis, although other types of intensive nutritional interventions and restraints were more commonly observed among patients without a primary psychiatric diagnosis. In contrast to patients without a primary psychiatric diagnosis, those with a primary psychiatric diagnosis demonstrated a significant increasing trend in the overall use of interventions. Physical restraint and NGTF were most frequently used in urban teaching hospitals, and these facilities showed a significant increase in the use of any intensive nutritional interventions and restraints.

TABLE 3.

Proportion of intensive nutritional interventions and restraints in each subgroup.

Subgroup	Total			Restraint			NGTF			TPN			PEG
	Estimate	95% CI	p	Estimate	95% CI	p	Estimate	95% CI	p	Estimate	95% CI	p	Estimate	95% CI	p
Age group			0.459			0.1			< 0.001			< 0.001			< 0.001
Adolescents (< 18 years)	6.6	(5.6–7.6)		1.8	(1.4–2.2)		4.9	(4.0–5.8)		0.2	(0.1–0.3)		0.1	(0.0–0.2)
Young adults (18–27 years)	6.5	(5.6–7.4)		2.0	(1.5–2.5)		3.5	(2.8–4.2)		0.7	(0.4–1.0)		0.7	(0.4–1.0)
Adults (≥ 28 years)	7.2	(6.4–8.0)		1.3	(1.0–1.6)		2.9	(2.4–3.4)		2.2	(1.8–2.6)		1.2	(0.9–1.5)
Race			0.12			< 0.001			0.041			0.023			0.375
White	6.8	(6.2–7.4)		1.8	(1.5–2.1)		3.7	(3.2–4.2)		1.1	(0.9–1.3)		0.7	(0.5–0.9)
Black	8.4	(5.4–11.4)		3.0	(1.3–4.7)		3.4	(1.6–5.2)		2.5	(0.7–4.3)		0.2	(0.0–0.6)
Hispanic	5.8	(4.4–7.2)		1.0	(0.4–1.6)		3.8	(2.6–5.0)		0.7	(0.2–1.2)		0.3	(0.0–0.6)
Asian or Pacific Islander	5.2	(3.1–7.3)		1.2	(0.2–2.2)		2.4	(0.9–3.9)		0.9	(0.0–1.8)		0.7	(0.0–1.5)
Native American	12.5	(2.3–22.7)		0.0	(0.0–0.0)		6.3	(0.0–14.7)		4.2	(0.0–9.9)		2.1	(0.0–6.2)
Other	7.8	(5.4–10.2)		1.2	(0.2–2.2)		6.1	(3.9–8.3)		0.2	(0.0–0.6)		0.4	(0.0–1.0)
Primary psychiatric diagnosis			< 0.001			0.035			0.022			< 0.001			< 0.001
Yes	6.2	(5.6–8.8)		1.5	(1.2–1.8)		4.0	(3.5–4.5)		0.5	(0.4–0.6)		0.5	(0.4–0.6)
No	8.5	(7.5–9.9)		2.1	(1.6–2.6)		3.1	(2.4–3.8)		2.6	(2.1–3.1)		1.3	(0.9–1.7)
Hospital location and teaching status			< 0.001			0.04			< 0.001			0.686			0.711
Rural	3.8	(2.1–5.5)		0.2	(0.0–0.6)		1.9	(0.7–3.1)		1.5	(0.4–2.6)		0.4	(0.0–1.0)
Urban nonteaching	4.2	(3.0–5.4)		1.1	(0.4–1.8)		1.6	(0.9–2.3)		1.1	(0.5–1.7)		0.8	(0.3–1.3)
Urban teaching	7.2	(6.6–7.8)		1.8	(1.5–2.1)		4.1	(3.6–4.6)		1.1	(0.9–1.3)		0.7	(0.5–0.9)

Note: Boldface indicates statistically significant results.

Abbreviations: CI, confidence interval; NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition.

TABLE 4.

Results from stratified trend analyses estimating annual changes in percentages of intensive nutritional interventions and restraints for hospitalizations with anorexia nervosa.

Subgroup a	Total			Restraint			NGTF			TPN			PEG
	Annual change	95% CI	p	Annual change	95% CI	p	Annual change	95% CI	p	Annual change	95% CI	p	Annual change	95% CI	p
Age group
Adolescents (< 18 years)	0.48	(−0.01 to 0.97)	0.057	0.22	(0.03 to 0.41)	0.023	0.28	(−0.16 to 0.71)	0.211	−0.08	(−0.27 to 0.12)	0.494	−0.15	(−0.29 to −0.01)	0.139
Young adults	0.10	(−0.43 to 0.63)	0.707	0.13	(−0.03 to 0.28)	0.101	−0.02	(−0.43 to 0.39)	0.927	−0.10	(−0.31 to 0.12)	0.400	−0.15	(−0.41 to 0.12)	0.314
Adults (≥ 28 years)	0.37	(0.02 to 0.72)	0.047	0.13	(−0.02 to 0.29)	0.089	0.23	(−0.08 to 0.53)	0.141	−0.02	(−0.17 to 0.14)	0.847	0.05	(−0.06 to 0.15)	0.407
Race
White	0.27	(−0.01 to 0.54)	0.057	0.16	(0.05 to 0.27)	0.007	0.12	(−0.1 to 0.35)	0.283	−0.06	(−0.17 to 0.05)	0.343	−0.01	(−0.1 to 0.09)	0.911
Black	0.48	(−1.74 to 2.7)	0.665	0.01	(−0.61 to 0.63)	0.985	−0.56	(−2.87 to 1.74)	0.634	0.6	(−1.51 to 2.7)	0.529	Not estimable		NA
Hispanic	0.82	(−0.16 to 1.79)	0.089	0.12	(−0.36 to 0.6)	0.600	0.51	(−0.47 to 1.5)	0.285	−0.13	(−0.34 to 0.08)	0.199	0.02	(−0.19 to 0.23)	0.441
Asian or Pacific Islander	1.02	(−0.82 to 2.86)	0.203	Not estimable		NA	0.79	(−0.91 to 2.49)	0.186	Not estimable		NA	−2.9	(−2.99 to −2.85)	< 0.001
Native American	−0.45	(−0.93 to 0)	0.002	−0.35	(−0.66 to −0.03)	0.702	Not estimable		NA	Not estimable		NA	Not estimable		NA
Other	−0.02	(−1.37 to 1.33)	0.980	0.01	(−0.12 to 0.13)	0.630	0.14	(−0.9 to 1.19)	0.773	Not estimable		NA	−2.9	(−2.89 to −2.84)	< 0.001
Primary psychiatric diagnosis
Yes	0.37	(0.09 to 0.65)	0.015	0.14	(0.01 to 0.27)	0.040	0.25	(−0.01 to 0.5)	0.064	−0.04	(−0.11 to 0.03)	0.325	0.02	(−0.05 to 0.09)	0.613
No	0.31	(−0.21 to 0.84)	0.249	0.18	(0.04 to 0.31)	0.010	0.05	(−0.36 to 0.45)	0.827	−0.06	(−0.32 to 0.2)	0.657	−0.15	(−0.42 to 0.12)	0.308
Hospital location and teaching status
Rural	−0.02	(−0.28 to 0.24)	0.870	Not estimable		NA	−0.17	(−0.54 to 0.21)	0.406	Not estimable		NA	Not estimable		NA
Urban nonteaching	0.33	(−0.36 to 1.02)	0.338	0.14	(−0.26 to 0.53)	0.483	0.04	(−0.33 to 0.41)	0.836	−0.08	(−0.47 to 0.3)	0.685	0.0	(−0.33 to 0.33)	0.989
Urban teaching	0.38	(0.05 to 0.7)	0.030	0.18	(0.07 to 0.29)	0.003	0.22	(−0.08 to 0.51)	0.156	−0.06	(−0.17 to 0.04)	0.255	−0.04	(−0.14 to 0.07)	0.495

Note: Boldface indicates statistically significant results.

Abbreviations: CI, confidence interval; NA, not available; NGTF, nasogastric tube feeding; PEG, percutaneous endoscopic gastrostomy; TPN, total parenteral nutrition.

^a When the incidence of intensive nutritional interventions and restraints in 2017 within a given superscript was zero, trend analyses could not be conducted. Accordingly, these results are reported as not estimable.

4. Discussion

To our knowledge, this is the first study to systematically characterize the national‐level prevalence and trends of NGTF, TPN, PEG, and physical restraint in AN treatment in the United States. In this nationwide study of more than 66,000 hospitalizations for AN, at least one intensive nutritional intervention or restraint was estimated at 6.8%, with a significant increase observed between 2017 and 2022. Importantly, this increase was driven by rising use of NGTF and physical restraint, whereas the use of TPN and PEG remained stable over time. These interventions were most common among hospitalized patients at urban teaching facilities with psychiatric comorbidity, high physical comorbidity burden, and non‐adherence to treatment.

4.1. Prevalence and Trends of Intensive Nutritional Interventions and Restraints

On average, intensive nutritional interventions and restraints were provided in 6.8% of US hospitalizations from 2017 to 2022, a prevalence similar to the Danish national cohort's (Clausen et al. 2018), though slightly higher than the present US cohort. A national audit from England reported that 622 patients underwent NGTF under physical restraint across NHS inpatient settings between June 2020 and May 2021, providing a more representative description of UK practice (Fuller et al. 2023). In contrast, the current US cohort showed an average of 26 hospitalizations per year involving NGTF under physical restraint. Although direct comparisons are challenging due to differences in national population sizes and the lack of information on the inpatient population in the UK report, these disparities may reflect differences in legal requirements, thresholds for coercive care, and institutional norms surrounding the use of intensive nutritional interventions and restraints (Carney et al. 2006; Clausen 2020; Douzenis and Michopoulos 2015; Elzakkers et al. 2014; Westmoreland and Mehler 2016). In particular, in the United States, legal standards for involuntary NGTF differ by state (Ramasamy 2021), with some jurisdictions requiring separate court hearings similar to that required for electroconvulsive therapy, reflecting heightened judicial scrutiny over intrusive interventions (Westmoreland et al. 2017), while others permit such interventions under general mental health legislation. This heterogeneity underscores the importance of contextualizing findings within specific regulatory and cultural environments. However, the prevalence of physical restraint in hospitalizations for AN in the current study (1.7%) exceeded the national average for all hospitalizations in the United States (0.7%) (Luccarelli et al. 2023), indicating that the high rate of coercive interventions among hospitalizations with AN remained evident in this cohort. Importantly, the relatively low overall prevalence suggests that most hospitalizations were managed without intensive interventions, indicating that such treatments were used selectively rather than routinely. At the same time, because the dataset does not include legal admission status or decisional capacity, these findings should be interpreted with caution.

In this cohort, there was no increasing trend in the combined use of NGTF and restraint. One possible explanation is that restraint may be used alone to prevent excessive physical activity following voluntary oral intake, whereas NGTF may be implemented without restraint in cases where self removal of the feeding tube is not observed. These patterns suggest that NGTF and restraint are applied independently in nutritional management according to the individual behavioral characteristics of patients with AN. Furthermore, clinicians' attitudes toward coercion vary between individuals and also across wards (Husum et al. 2023). The use of interventions that inherently involve coercion is likely to be strongly influenced by such differences in clinicians' attitudes. Consequently, their use may not necessarily coincide with NGTF, for which there is some evidence supporting its effectiveness for weight restoration (Rigaud et al. 2007). In addition, the choice of whether to use NGTF or restraint may also be influenced by institution‐specific practices.

Significantly elevated aRR for receiving intensive nutritional interventions and restraints were observed among hospitalizations involving psychiatric comorbidity, higher physical comorbidity burden, non‐adherence to treatment, and absence of SUD comorbidity. These findings, consistent with prior literature (Atti et al. 2021; Carney et al. 2006; Clausen et al. 2018), suggest that intensive nutritional interventions and restraints are more likely to be applied when clinical severity is compounded by treatment resistance or when managing complex psychiatric profiles. Additional subgroup analyses further showed that patients with a primary psychiatric diagnosis were more likely to receive NGTF and exhibited a significant increasing trend in intervention use, suggesting that these interventions may reflect not only medical instability but also behavioral and psychiatric management needs. Of note, the use of intensive nutritional interventions was associated with a higher burden of physical comorbidities, whereas no such association was observed for physical restraint. This finding suggests that intensive nutritional interventions tend to be applied in the context of severe medical instability or specific gastrointestinal complications.

In subgroup analyses, the negative association between SUD comorbidity and the use of intensive nutritional interventions and restraints was most pronounced in the adult age group, where reduced use of NGTF, TPN, and PEG was significantly associated with the presence of SUD. This negative association may reflect clinical prioritization of detoxification during acute substance intoxication over eating disorder treatment (Dooley‐Hash et al. 2019). Treatment for AN is often deferred until substance use has been stabilized (Woodside and Staab 2006). Moreover, in adult AN, SUD itself might be frequently chronic and accompanied by complex medical complications, such as liver cirrhosis and related metabolic disturbances in alcohol dependence, which may further complicate clinical management. As a result, intensive nutritional interventions and restraints for AN may be postponed or not initiated during hospitalizations in which substance use constitutes the primary clinical concern.

Institutional characteristics were also associated with the use of intensive nutritional interventions and restraints, with higher rates observed in medium to large hospitals, particularly urban teaching facilities. However, this association should be interpreted cautiously. Large urban teaching hospitals often function as referral centers and may care for patients with greater medical and psychiatric complexity. Because the dataset lacks key clinical severity indicators such as BMI, medical instability, and treatment history, differences across hospitals may reflect case mix rather than institutional treatment practices. This consideration may also help explain discrepancies between national estimates and rates reported by single center or regional studies, as specialized centers are more likely to treat more severe or refractory cases, potentially inflating the observed rates of intensive nutritional interventions and restraints (Elzakkers et al. 2014). Accordingly, the higher prevalence observed in large urban teaching facilities should be interpreted as consistent with possible referral bias, rather than as evidence of different treatment thresholds.

Interestingly, use of emergency department services was not associated with the overall use of intensive nutritional interventions and restraints and was associated with lower TPN use. This finding may reflect postadmission care pathways and practice patterns. Admissions through the emergency department are more likely to involve urgent and medically unstable presentations, in which care may focus on short term, life saving management of acute complications, such as cardiopulmonary instability or fluid and electrolyte imbalance, rather than on time intensive nutritional interventions (Trent et al. 2013). By contrast, admissions without emergency department use may include more planned admissions, allowing more intensive treatment for AN itself.

Although TPN and PEG utilization remained stable from 2017 to 2022, physical restraint and NGTF increased significantly, raising ethical concerns (Tsiandoulas et al. 2023). These interventions are associated with physical discomfort, emotional distress, and potential iatrogenic harm. Although previous studies from Denmark's national cohort and the single center cohort of the United Kingdom have documented the prevalence of involuntary treatment (Ayton et al. 2009; Clausen et al. 2018), most of these studies did not report the prevalence of TPN and PEG. This may be because the cohort of the United Kingdom examined an adolescent specialist eating disorder setting, whereas Denmark's national cohort used registries specifically focused on coercive practices rather than data covering the full spectrum of acute inpatient treatment for AN. Consequently, these studies may not have captured interventions such as TPN and PEG in acute hospital based care. The present study fills this gap and demonstrates that even though TPN and PEG utilization remained stable over time, NGTF and restraint significantly increased. These trends persisted after covariate adjustment, suggesting a general rise in certain intensive practices, potentially driven by the increasing burden of AN hospitalizations on inpatient care systems. Of note, the marked increase in hospitalizations for eating disorders since the onset of the COVID‐19 pandemic may have contributed to these trends (Tsiandoulas et al. 2023). Both the overall characteristics and subgroup trend analyses indicated an increasing proportion of Hispanic hospitalizations with AN (Hernández et al. 2022) over time, as well as a rising trend in the use of intensive nutritional interventions and restraints among them. Neither trend reached statistical significance, but their consistency may reflect a potential emerging disparity that warrants further attention.

4.2. Prevalence and Trends of Intensive Nutritional Interventions and Restraints in the Adult Patients With AN

Previous studies showed that involuntary NGTF and physical restraint peaked among patients aged 15–18 years old (Clausen et al. 2018). It is known that shorter duration of illness and younger age predict better prognosis in AN patients (Steinhausen 2002). In line with these findings, prior work suggests that earlier age at admission is associated with a higher likelihood of subsequent involuntary treatment, potentially reflecting greater treatment persistence and heightened concern among clinicians and families when patients are younger, as well as increased family‐related conflict during adolescence (Mac Donald et al. 2023). In adolescent patients with AN who have impaired capacity secondary to low body weight, time‐limited coercive feeding to restore capacity and ascertain patient preferences can be ethically justified as a last resort option (Atti et al. 2021; Bunch 2023).

Although prior studies identified younger age as having the highest prevalence of intensive nutritional interventions and restraints (Touyz et al. 2024; Fuller et al. 2023), in the current US cohort, the adult age group showed the highest prevalence of intensive nutritional interventions and restraints. This discrepancy between the adolescent and adult age groups across studies may be attributable to differences in the types of intensive nutritional interventions examined. In the present study, we evaluated NGTF, TPN, and PEG as intensive nutritional interventions. NGTF was significantly more prevalent in the adolescents, whereas the prevalence of TPN and PEG increased significantly from the young adult age group to the adult age group. Previous studies primarily focused on the prevalence of NGTF, while the use of TPN and PEG was not reported. Although nasogastric feeding remains the standard approach in the non‐oral nutritional management of AN, a recent systematic review indicates that other enteral routes, such as PEG or jejunal feeding via nasojejunal or gastrostomy tubes with jejunal extensions, are occasionally used when clinically indicated (Dhopatkar et al. 2025). Of note, psychiatric comorbidity and non‐adherence were strongly associated with the use of NGTF and restraint, but not with TPN or PEG. This divergence may be explained by the fact that NGTF and restraint are more frequently employed as acute, behaviorally responsive measures, whereas TPN and PEG may reflect longer‐term nutritional strategies implemented under different clinical and ethical rationales. PEG and TPN are not standard nutritional interventions for AN and are generally reserved for specific gastrointestinal or medical indications. They should not be interpreted as markers of coercion or of chronic‐stage AN without evidence of the clinical indication. When considering intensive nutritional interventions in AN, it is essential to consider the distribution of use according to both age groups and the specific types of interventions.

4.3. Limitations and Recommendation for Future Research

Several limitations warrant consideration. First, this study lacks outcome and severity data such as survival, body mass index, and time to readmission, which are essential for evaluating the ethical justifiability of intensive nutritional interventions and restraints. In the absence of such outcome data, the observed increase in the use of these interventions should be interpreted with careful ethical and clinical consideration. For physical restraint, factors such as self‐injurious behavior, aggression, and agitated delirium may be more directly associated with restraint use than body mass index, but these variables could not be incorporated into the analysis. In ICD‐11, body mass index‐based subcategories were introduced, although such classifications were not available in ICD‐10. Although ICD‐11 includes fewer severity categories than DSM‐5, the incorporation of body mass index‐based severity coding into future administrative databases will be important. Of note, use of NGTF or restraint requires clear practical safeguards, including multidisciplinary and psychologically informed care frameworks in which such interventions are embedded, rather than delivered in isolation. Moreover, when coercive treatment is considered, structured processes such as ethics consultations, risk review meetings, and advance care planning are crucial to ensure ethically defensible practice. Given that fewer than 10% of patients required such interventions and that their use was more common in urban teaching hospitals, targeted training and support through consultation–liaison teams in these settings may represent a pragmatic strategy to promote good practice and minimize unnecessary use of intensive or coercive measures. Future national cohort studies should incorporate a broader range of outcome data whenever possible, including patient‐centered outcomes such as physical and psychological recovery, alongside contextual clinical and ethical information, to support more balanced and clinically informed ethical discussions and to better assess the proportionality and effectiveness of intensive nutritional interventions and restraints. Second, the data do not distinguish between medical and psychiatric admissions or between voluntary and involuntary procedures. Changes over time in the proportion of psychiatric admissions could therefore have influenced the observed long‐term trends, and this possibility cannot be excluded. In addition, legal admission status may affect the use of these interventions. Its inclusion in future data collection would help clarify the relationship between legal admission status and intensive nutritional interventions and restraints. At the same time, with respect to involuntary procedures, existing literature suggests that even when NGTF is performed without formally documented coercion, patients frequently perceive it as nonconsensual (Neiderman et al. 2001). Moreover, physical restraints are inherently coercive and showed parallel trends with NGTF, suggesting that increased use of NGTF may not reflect a rise in voluntary acceptance. Third, the NIS excludes residential treatment programs conducted in facilities that are not licensed as hospitals, as well as long‐term care facilities. This is an important limitation because most specialty beds for eating disorders in the United States are located in the residential treatment sector, where lengths of stay are typically longer than in acute hospitals. Residential treatment programs have increased in number in recent years, and their quality varies considerably (Attia et al. 2016). However, the NIS does not capture intensive nutritional interventions or restraints used in these settings. Therefore, our findings should not be generalized to the residential treatment sector or to the broader population of eating disorder admissions to 24‐h care in the United States. Future cohort studies should include residential treatment programs and long‐term care facilities to evaluate these interventions across the full continuum of eating disorder care. Fourth, as the NIS is a claims‐based database, procedures such as physical restraints may be underestimated (Luccarelli et al. 2023), and the accuracy of restraint‐related coding remains uncertain. In addition, physical restraint is regulated differently across jurisdictions and may therefore be undercoded. Validation studies are warranted to assess the accuracy of these records. Nevertheless, procedures such as NGTF, TPN, and PEG are closely tied to reimbursement claims and are therefore more likely to be recorded with high sensitivity and specificity, accurately reflecting their actual use. Fifth, the NIS dataset did not allow identification of how many patients had been receiving NGTF or TPN prior to admission. However, patients with preexisting gastrostomy at admission would be coded as Z93.1. None of the PEG cases in our cohort carried this code, indicating that all PEG cases represented new placements. Finally, future research should more explicitly examine international and regional variation in the use of intensive nutritional interventions and restraints, both internationally and at the subnational level. Such variation is likely to reflect differences in legal frameworks, service delivery models, clinical training, and cultural attitudes toward autonomy and coercion (Takimoto 2022). Investigating these contextual factors may help to inform the development of ethical safeguards, evidence‐based service models, and improved training for clinicians working with individuals with severe and enduring AN.

5. Conclusions

This nationally representative study revealed that the use of intensive nutritional interventions and restraints for AN, particularly NGTF and physical restraint, has increased in recent years in the United States. The highest prevalence was observed among hospitalizations with adult age groups, underscoring the need for renewed attention to the ethical implications of such practices. While intensive nutritional interventions and restraints may be clinically warranted in specific contexts, their increasing use highlights a critical need for ongoing surveillance, development of less intrusive alternatives, and stronger evidence linking these interventions to improved outcomes. Ensuring ethically sound care for patients with AN will require sustained policy, clinical, and research efforts.

Author Contributions

Hiroyasu Ino: conceptualization, methodology, formal analysis, writing – original draft. Eisuke Nakazawa: conceptualization, methodology, formal analysis, writing – review and editing, supervision, project administration. Nancy S. Jecker: formal analysis, writing – review and editing. Yoshiyuki Takimoto: formal analysis, writing – review and editing.

Funding

This work was supported by Tokyo Metropolitan Hospital Organization, R071402018; Uehiro Foundation on Ethics and Education.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Lists of the codes for clinical variables.

Figure S1: Annual trends of characteristics of hospitalizations with anorexia nervosa.

Table S2: Results of trend analyses of characteristics.

Table S3: Comparison of physical comorbidities between admissions with and without intensive nutritional interventions and restraints.

Figure S2: Annual trends in the use of nasogastric tube feeding, physical restraint, and their combination among hospitalizations with anorexia nervosa.

Figure S3: Forest plots of annual absolute changes in intensive nutritional interventions and restraints among patients with anorexia nervosa estimated by marginal effects.

Figure S4: Results of sensitivity analyses.

Table S4: Association between substance use disorder comorbidity and intensive nutritional interventions and restraints in each age groups.

Ino, H. , Jecker N. S., Takimoto Y., and Nakazawa E.. 2026. “Epidemiology of Intensive Nutritional Interventions and Restraints in Hospitalizations for Anorexia Nervosa: Annual National Trends of the United States.” International Journal of Eating Disorders 59, no. 10: 2161–2180. 10.1002/eat.70149.

Data Availability Statement

This study used the Healthcare Cost and Utilization Project (HCUP) NIS, available through the HCUP Central Distributor upon completion of the Data Use Agreement (https://www.hcup-us.ahrq.gov).