Inpatient admissions and mortality of anorexia nervosa patients according to their preceding psychiatric and somatic diagnoses

Anorexia nervosa (AN) is associated with increased risk of mortality, but little is known about the risk of inpatient admissions and mortality outcomes in individuals with diagnoses of both AN and other psychiatric and somatic conditions. We aimed to investigate the inpatient admissions and mortality among people with AN and other diagnosed conditions using Danish national registers.

inpatient admission and mortality.Our findings highlight the need for comprehensive, multidisciplinary care of patients with AN considering the spectrum of other diagnosed conditions to improve health outcomes.

K E Y W O R D S
admissions, anorexia nervosa, comorbidity, mortality

| INTRODUCTION
Anorexia nervosa (AN) is a serious eating disorder characterized by extremely low body weight and an intense fear of gaining weight.Lifetime risk of AN is estimated as 0.3% for males and 4% for females. 1 In Denmark, incidence rates doubled between 1995 and 2010. 2 Patients with AN often face severe health consequences as a direct result of extreme weight loss and malnutrition. 3ealthcare utilization among individuals with AN is elevated compared to individuals without an eating disorder (ED), and the need for rehospitalization is common. 4revious studies have highlighted the high degree of comorbidity associated with EDs for both psychiatric [5][6][7][8] and somatic conditions 9,10 ; however, most have considered EDs in general, rather than AN specifically, which warrants specific investigation.Readmission rates among adolescent patients with AN are substantial, with around 45% requiring at least one readmission. 11Other disorders in AN patients may contribute to these healthcare needs.However, psychiatric comorbidity was not found to differ among patients with AN needing versus not needing rapid re-hospitalization after their initial diagnosis, 4 potentially suggesting a role for somatic conditions or the number of other diagnoses as opposed to specific conditions.
Additionally, AN is associated with an increased risk of mortality, 5,12,13 and has the highest relative mortality of all EDs. 14,15A Danish study reported that males and females with AN die 9.6 and 10.5 years younger, respectively, than those of the same sex and age in the general Danish population. 12Other disorders play a role in this increased risk. 16Swedish studies reported that mortality risk in individuals with AN is particularly elevated among patients with other psychiatric and somatic conditions, [17][18][19] with those with both types of conditions at greatest risk. 17ur overarching goal was to describe the other conditions with which individuals with AN are commonly diagnosed.Additionally, we aimed to provide a comprehensive investigation of the risk of inpatient hospital admissions (due to (i) AN, (ii) any psychiatric cause, and (iii) any somatic cause) and mortality among individuals with AN, according to other the number and combinations of other psychiatric disorders and somatic diagnoses, using data from the Danish national registers.We examined the outcomes in relative and absolute terms.

| Study design and population
We conducted a population-based cohort study using the Danish national registers.The study population comprised individuals with a diagnosis of AN who were born

Significant outcomes
• Our study comprehensively describes other diagnosed conditions in individuals with anorexia nervosa in Denmark and elucidates their association with higher rates of inpatient admission and mortality.• Our results emphasize the healthcare needs of individuals with anorexia nervosa considering the spectrum of other diagnosed conditions.

Limitations
• It is probable that psychiatric and somatic conditions may be under-detected in the Danish national registers, as we have no data on people who do not seek medical advice for conditions, or who are treated solely in primary care.• The Danish registers do not include data on remission or recovery.Here, we consider other diagnosed conditions at any point throughout life, which could lead to the inclusion of disorders that are no longer "active" at the time of anorexia nervosa diagnosis.• Due to complexity in diagnoses of eating disorders, some of the results may be observed due to diagnostic instability/complexity.
between January 1, 1977 and December 31, 2010.It is possible to link Danish registers using the unique civil registration number assigned to individuals or upon immigration.AN cases were identified using two sources: (i) the Danish National Patient Register (DNPR) 20 and (ii) the Danish Psychiatric Central Research Register (PCRR). 21 summary of the registers is provided in the Supplementary Material.Briefly, these registers contain information on inpatient contacts only from their inception until 1994; from 1995 onwards, they also contain data on outpatient and emergency room visits.AN cases were identified using Eighth Revision of the International Classification of Diseases (ICD-8) 22 code 306.05 and ICD-10 codes 23 F50.0 and F50.1, with date of diagnosis considered to be the first admission date (inpatient, outpatient or emergency visit) for AN registered when an individual was at least 6 years old-the youngest age an AN diagnosis was considered valid in this study.
Access to individual-level register data is governed by Danish authorities.These include the Danish Data Protection Agency (AU DT ID: 2015-57-0002; AU LBNR 565), the Danish Health Data Authority (SDS FSE ID 98), and Statistics Denmark (project 703934).Each scientific project must be approved before initiation, and approval is granted to a specific Danish research institution.According to Danish law, review by an ethics board and patient consent are not required for purely register-based studies.All data were deidentified and not recognizable at an individual level.

| Other diagnoses received by individuals with AN
We ascertained other selected conditions diagnosed (primary or secondary diagnosis) in each individual prior to their AN diagnosis.We operationalized prior conditions according to: (i) the number of these conditions each individual had (0, 1, 2, 3, 4, 5, ≥6) and (ii) common combinations of these conditions (observed in at least 100 members of the cohort).
][26][27] Using the DNPR and PCRR, we also identified first diagnosis of the 10 ICD-10 subgroups of Mental and Behavioral Disorders (Chapter F codes): organic disorders, substance-use disorders, schizophrenia, mood disorders, anxiety disorders, other eating disorders (OED), personality disorders, intellectual disabilities, developmental disorders, and behavioral disorders.See Supplementary Table 2 for the corresponding ICD-10 diagnostic codes, equivalent ICD-8 codes, and the earliest age of diagnosis. 28

| Inpatient admission and mortality
Inpatient admission date was defined as the first inpatient or emergency admission after the initial AN diagnosis date, extracted from the DNPR.We considered three types of admissions: admission due to (i) AN, (ii) any psychiatric disorder, and (iii) any somatic condition.Allcause mortality and date of death were extracted from the Danish Civil Registration System (CRS). 29

| Other covariates
Age, sex, and urbanicity were obtained from the CRS.We grouped the cohort into five age groups based on age at first AN diagnosis: <12 years old, 12-19 years old, 20-29 years old, 30-39 years old, and ≥40 years old.

| Statistical analysis
Follow-up started on the first date an individual received an AN diagnosis.Follow-up ended at the first of the following: emigration date, date of death, or end-of-followup on December 31, 2018.For admissions, it also ended at hospital admission date.

| Relative risks
We calculated the risk of AN admission, any psychiatric admission, any somatic admission, and mortality by number of prior diagnosed conditions and common combinations of other disorders, compared to individuals with AN only.Cox proportional regression models produced hazard ratios (HRs) and 95% confidence intervals (95% CIs), adjusted for sex, and birth year, as well as age and urbanicity at AN diagnosis.
Due to small numbers of cases, it was not possible to consider combinations of diagnosed conditions and mortality.

| Cumulative incidence
Using competing risks survival analyses, we calculated cumulative incidence of inpatient admission for the three types of admissions (considering death as competing event), and all-cause mortality among AN patients over the follow-up period.

| Sensitivity analysis
We carried out four sensitivity analyses.First, we restricted the population to those living in Denmark for 2 years prior to their initial AN diagnosis to ensure that previous diagnoses would be captured for at least 2 years before follow-up.Second, we restricted to those born in 1989 onwards, meaning all AN diagnoses and readmissions were made after introduction ICD-10 codes and of outpatient and emergency appointments in the registers.Third, for the outcomes of AN admission and mortality, we expanded the definition of other diagnoses to include those diagnosed both before and after AN diagnosis (not for other psychiatric and somatic admissions, as diagnoses made after the AN diagnosis would contribute to both the exposure and outcome).Fourth, for the outcome of AN admission, we carried out analyses stratified by type of initial contact (inpatient vs. outpatient).

| RESULTS
The study population comprised 11,489 individuals diagnosed with AN (Table 1).Most cases were female (93.4%, n = 10,732), and the mean age at AN diagnosis was 18.3 years (SD = 5.1).Prior to AN diagnosis, the most common somatic conditions were respiratory and musculoskeletal conditions (33.5% and 24.0%, respectively), whereas the most common psychiatric disorders were OED and anxiety disorders (34.5% and 32.7%, respectively).At the time of AN diagnosis, 70% had received a diagnosis of at least one other disorder.The most common combinations of other previously diagnosed disorders at the time of AN diagnosis were respiratory disorders (7.3%), OED (4.5%), and anxiety disorders (3.1%).Supplementary Table 3 30 3.1 | Relative risks

| Number of other diagnoses
The risk of inpatient admission due to AN increased with number of prior diagnoses.The HR of inpatient AN admission was 1.04 (95% CI 0.95-1.14)for AN cases with one previously diagnosed condition compared to those with AN only (Table 2).The risk for people with ≥6 previously diagnosed conditions was doubled (HR 2.07, 95% CI 1.71-2.50).A similar pattern was seen for the number of previous diagnoses and psychiatric or somatic admissions (Table 2).For the risk of mortality, no prior diagnoses and one prior diagnosis were combined due to small numbers of cases.Cases with two prior diagnoses had a 56% increased risk of mortality (1.56) compared to those with no or one prior diagnoses (Table 2); however, the 95% CI included 1 (0.89-2.74).For those with ≥6 prior diagnoses for other conditions, the HR was 7.09 (95% CI 3.07-16.37).

| Combinations of other diagnoses
The risk of inpatient admission due to AN was higher for individuals with 10 out of the 13 most common combinations of conditions diagnosed prior to AN, compared to individuals with AN only (Figure 1).However, elevations in point estimates were modest and ranged from just 1% (1.01, 95% CI 0.77-1.33)for congenital conditions to 27% (1.27, 95% CI 0.90-1.77)for the combination of OED and respiratory conditions, and all 95% CIs included 1.
The risk of inpatient admission due to any psychiatric disorder was higher for individuals with 12 out of the 13 most common diagnostic combinations compared to individuals with AN only (with 95% CIs above 1 for four combinations; Supplementary Figure 1).The greatest increases in risk were seen for comorbid anxiety (1.41, 95% CI 1.19-1.68)and mood disorders (1.38, 95% CI 1.13-1.68).The HR for endocrine conditions suggested a reduced risk (0.85) but the 95% CI included 1 (0.59-1.20).
The risk of admission due to a somatic disorder was elevated for 12 of the 13 most common combinations of other diagnoses (with 95% CIs above 1 for five combinations; Supplementary Figure 2).The greatest increase in risk was seen for respiratory and musculoskeletal disorders (1.54, 95% CI 1.25-1.68).

| Cumulative incidence
Twenty years after the initial AN diagnosis, 23% of individuals had had an admission for AN, 12% for any psychiatric disorder and 55% for any somatic disorder; whereas 0.1% had died (Figure 2).

| Number of comorbid conditions
In the same period, 29%-32% of individuals with 5 prior conditions or fewer had been admitted due to AN, whereas among individuals with ≥6 prior conditions, this proportion was 37% (Supplementary Figure 3).The cumulative incidence for all psychiatric and somatic admissions and for mortality are shown in Supplementary Figures 4, 5, and 6 respectively.Within 20 years of the initial AN diagnosis, 43% of individuals with no prior diagnoses had been admitted due to any psychiatric disorder; however, among those with ≥6 prior diagnoses, 74% had been admitted.In contrast, 90% of individuals with no prior diagnoses had been admitted due to any somatic disorder; and, among those with ≥6 prior diagnoses, 91% had been admitted.For those with no prior diagnoses, 1% died.While the proportion who died in this period was between 1% and 3% among those with 1-4 prior diagnoses, among those with 5 and ≥6 prior diagnoses, 15% and 22% died, respectively.

| Combinations of other diagnoses
At least 24% of individuals within each combination of prior disorders were admitted due to AN within 20 years of initial diagnosis (Supplementary Figure 7).At least 38% of individuals were admitted due to any psychiatric disorder, regardless of combination of prior disorders (Supplementary Figure 8); and at least 72% had been admitted due to any somatic disorder (Supplementary Figure 9).

| Sensitivity analyses
Restricting the population to those who had been living in Denmark for at least 2 years prior to their initial AN diagnosis resulted in little to no change in findings (Supplementary Table 4).Restricting to those born from 1989 onwards resulted in slightly larger HRs for all outcomes except mortality, for which no changes or small reductions were observed across most groups (Supplementary Table 5).Including diagnoses made both before and after the AN diagnosis resulted in similar HRs for later AN admissions for most combinations of disorders (Supplementary Figure 10).Stratifying analyses by setting of first AN diagnosis indicated that number of prior diagnoses had less impact on risk of future AN admissions among those diagnosed as an inpatient than as an outpatient (Supplementary Table 6), but 95% CIs suggested results were in the same range.In a common model for the patient type, the HR of first contact being outpatient compared to inpatient was 0.47 (0.40-0.56).Additionally, there were similar results for the two patient groups for most combinations of other disorders, however  Categories combined due to small number of cases.
differences were observed for anxiety disorders, and for anxiety disorders and OED (Supplementary Figure 11).

| DISCUSSION
This population-based register study of 11,489 individuals with AN in Denmark suggests that the most common diagnoses other than AN in this group were OED and anxiety disorders.Our results show that individuals with AN who have had a larger number of other diagnoses were at increased risk of admission-due to AN, other psychiatric conditions or somatic disorders-and mortality.The increased risks were, however, more marked for mortality than for inpatient admissions: having ≥6 prior diagnoses resulted in a HR of 2.07 for AN admission, 3.03 for somatic admission and 3.60 for psychiatric admission; whereas it was associated with a HR of 7.09 for mortality.Similarly, while only 1% of individuals with no additional conditions had died within 20 years of their AN diagnosis, this proportion was greatly elevated (21.6%) among those with ≥6 prior diagnoses; cumulative incidence did not differ to the same extent for admissions (being relatively flat for admissions due to AN and somatic conditions).Additionally, the risk was greater for individuals with most combinations of other diagnoses (although 95% CIs often included 1).However, for AN patients with a prior diagnosis of an endocrine disorder, risk of inpatient admission with any psychiatric diagnosis was reduced, as was risk of somatic admissions among AN patients with congenital disorders (95% CIs included 1).Endocrine disorders may require regular medical monitoring, which may lead to early diagnosis of AN.As congenital disorders are often identified early in life, it could be that treatment needed for these is complete by the age that AN is typically diagnosed.It should be noted that the combinations of other disorders included in this analysis (those present in at least 100 persons at the date of first AN diagnosis) comprised a maximum of two conditions.Therefore, a large proportion of the study population were not included in the analysis and we did not capture individuals with more complex diagnosis profiles.
In line with the literature, 7,9,[31][32][33][34][35][36] our results show that it is common for those with AN to have other diagnosed conditions, with the majority of the study population having ≥1 other condition at the time of diagnosis.In our study population, 0.9% died during follow-up.8][39] Previous studies have consistently reported higher mortality among AN cases compared to those without AN. 12,14,15It has also been shown that comorbidity in patients with AN is associated with higher mortality rates. 40Of note, there are only a few published studies with comparable analysis to our study (i.e., assessing outcomes in people with AN according to the combinations of other disorders they have been diagnosed with), so it is not possible to directly compare our results to previous work.However, Kask et al. 18 found higher mortality rates in women with AN and a psychiatric comorbidity than in women with AN but no psychiatric comorbidity (crude HR of 4.3) using Swedish data.Another Swedish study indicated that both psychiatric and somatic disorder comorbidity increased risk of mortality, 17 reporting that somatic and psychiatric comorbidity prior to AN diagnosis were each associated with a 10% increased risk in mortality compared to those with no comorbidity; in individuals with both types of comorbidity, the risk was increased by 90%.For somatic and psychiatric comorbidity diagnosed after the AN diagnosis, increases of 20% and 100% were reported, respectively, compared to those with no subsequent diagnoses; those subsequently diagnosed with both physical and psychiatric conditions had a 170% increase in risk.Although number of other diagnoses was not examined, higher relative risks were observed with higher number of hospitalizations.A 2011 systematic review also concluded that psychiatric comorbidities were associated with higher mortality in AN. 14 A further study 4 did not find differences in psychiatric comorbidity at baseline between AN patients requiring rapid rehospitalization after diagnosis and those who did not.
We do not assume that AN directly the investigated outcomes or other disorders There is some evidence suggesting malnutrition may lead to comorbid psychiatric disorders through altered neurotransmitter metabolism or endocrine changes, [41][42][43] and prolonged caloric deprivation may also affect other systems resulting in somatic complications/comorbidity. 43,44 However, we did not observe universally increased risks, pointing to other possible links and pathways.For instance, both AN and other conditions may be associated with prior exposures (e.g., substance use; childhood abuse), shared genetic factors, or other complex biological or environmental interactions unique to the individual.Likewise, these factors could also impact the associations observed in our study.Persistent low weight and prolonged starvation have detrimental effects, especially in the presence of other somatic conditions, and the medical complications associated with AN can potentially modify the severity and outcome of other conditions, including psychiatric disorders. 45Furthermore, polypharmacy could also play a role in the outcomes.Our extended group has previously shown that AN cases in Denmark have a higher likelihood of prescription medications compared to population-matched controls, 46 including psychotropic medications, despite limited evidence of their efficacy in addressing core AN symptoms. 47However, determining the safety and optimal dosage of medications poses challenges for individuals with acute AN due to low body mass index and frail physical condition. 48Additionally, purging behaviors can lead to medical instability, increasing the risk of severe medication side effects. 49ur study uses data from Danish national registers, allowing for the inclusion of all individuals with a hospital diagnosis of AN while limiting recall or self-reporting bias.Additionally, with data available on the entire population and free and equal access to health care for Danish citizens, selection bias is minimized.
There are also important limitations to consider.First, diagnostic misclassification may exist.Our study considers diagnoses of psychiatric and somatic conditions made during hospital contacts.We have no data on individuals who do not seek medical advice for their conditions or who are treated solely during general practitioner visits.Therefore, there may be underdetection of some conditions, particularly those that are often treated in primary care, such as depression.The national registers also lack data on diagnoses conducted in private healthcare settings, which may lead to some degree of misclassification.However, the impact of this is presumed to be minimal, considering the availability of free healthcare services, which may reduce the reliance on private healthcare for diagnoses.Second, Danish registers do not include codes or data on remission or recovery.Here we considered other diagnoses throughout life.It may be that some other disorders would no longer be "active" at the time of AN diagnosis.Third, numbers of cases were too small to consider some the associations or provide some of the results we would have liked, for example, providing results for a larger number of combinations of disorders (the majority of individuals with other diagnoses were not included in the most common combinations), carrying out sexspecific analyses, or considering cause of admission or death.Fourth, while the study included a long followup period, some individuals were relatively young when follow-up finished.With access to longer follow-up times, results might differ.Finally, some of the most elevated point estimates for admissions were observed for disorder combinations including OED.Due to the challenges associated with diagnosing AN, we cannot rule out a contribution from lifetime diagnostic instability or more complicated disease trajectories.Diagnostic crossover among EDs is not uncommon. 50This may be due to the presence of different disorders over the lifetime, or due to appropriate diagnostic uncertainty during the course of disorders.
As seen in other studies and clinical practice, our study highlights that other diagnoses are common among individuals with AN: almost 70% had received at least one other diagnosis prior to AN diagnosis; the most common being OED and anxiety disorders.Increasing number of other conditions were associated with higher risks of admissions and mortality; higher risks were observed for most, but not all, disorder combinations.AN is a serious ED, which requires care to attend to the numerous medical complications, nutritionally rehabilitate the patient, and provide psychotherapy to normalize eating behaviors.Our findings highlight the need for comprehensive, multidisciplinary care of patients with AN considering the spectrum of other diagnosed conditions to improve health outcomes. a

F I G U R E 1
Risk of inpatient admission due to AN, among AN patients according to combinations of other conditions diagnosed before first AN diagnosis.Somatic conditions and psychiatric disorders are diagnosed prior to the AN diagnosis.These disorders may not necessarily be active at the time of AN diagnosis.F I G U R E 2 Cumulative incidence of inpatient admission (for AN and for any other psychiatric disorder) and death.Cumulative incidence of inpatient admissions for AN, any psychiatric disorder and any somatic disorder are shown by the red, green, and blue lines, respectively.Cumulative incidence of mortality is shown by the purple line.95% CIs are indicated by the gray shading.AN, anorexia nervosa; CI, confidence interval.
Characteristics of the study population.
T A B L E 1 Risk of outcomes among AN patients by number of other conditions diagnosed prior to the initial AN diagnosis.Prior somatic conditions and psychiatric disorders are diagnosed prior to the initial AN diagnosis.These disorders may not necessarily be active at the time of AN diagnosis.
T A B L E 2Note: