Lifetime Prevalence of Psychiatric Disorders in Adolescents with Unexplained Weight Loss, Underweight or Poor Appetite

Abstract

Background:

When adolescents present with symptoms of unexplained weight loss, underweight or poor appetite, eating disorders (EDs) are commonly on the list of differential diagnoses. However, the relationship of these symptoms to other psychiatric disorders is often less clear.

Methods:

Using the Rochester Epidemiology Project (REP) database, we conducted a retrospective cohort study of adolescents (13–18y) with billing diagnoses of weight loss, underweight or loss of appetite between January 2005 and December 2017. Patients were excluded who presented with conditions commonly associated with weight loss, underweight or poor appetite (e.g., cancer). We sought to examine the proportion of patients who received ED and psychiatric diagnoses within 5 years of the index visit, as well as patient characteristics associated with these diagnoses.

Results:

Of 884 patients diagnosed with symptoms of unexplained weight loss, underweight or poor appetite, 662 patients (M age = 15.8; SD=1.6; 66.0% female) met study criteria. Within 5 years of the index visit, the lifetime prevalence of all psychiatric disorders was 70% (n=461) and of EDs was 21% (n = 141). For both psychiatric disorders and EDs, gender and race were significantly associated with receiving a diagnosis within 5 years. Decrease in BMI percentile was associated with receiving an ED diagnosis, whereas highest historical BMI percentile was associated with receiving a psychiatric diagnosis.

Conclusion:

Patients presenting with symptoms of unexplained weight loss, underweight or poor appetite are at risk not only for EDs, but also other psychiatric disorders that may require further assessment and follow-up.

Unexplained weight loss, low weight or poor appetite in adolescents can occur for a host of environmental, medical, and/or psychological reasons ^1–3. When adolescents present to medical settings with any of these symptoms, eating disorders are commonly on the list of differential diagnoses ^1–3. Even when other causes for weight loss, underweight or poor appetite are identified, these symptoms can be associated with malnourishment and poor physical health, as well as risk for eating disorders which have a 2.7% lifetime prevalence in adolescents and one of the highest mortality rates of any mental illness ^4–6.

In addition to eating disorders, loss of appetite and associated weight loss commonly co-occur with other psychiatric conditions (e.g., depression, anxiety, substance abuse)⁷. Not only do mood and anxiety have a powerful effect on appetite ^8,9, loss of appetite and/or significant weight loss represent diagnostic criteria of depression. In fact, research suggests that loss of appetite and body weight are symptoms associated with more severe depressive illness¹⁰. One study examining a large sample of adolescents found loss of appetite and body weight were robustly associated with suicidal ideation and self-harm ¹¹. Another study found adults who completed suicide were significantly more likely to experience loss of appetite and body weight as prominent features of their depressive illness compared to adults with major depressive disorder without a history of suicide attempts¹².

Not only do changes in appetite and weight co-occur in psychiatric illness, seminal research on human starvation demonstrates the extensive impact of weight loss on psychological adjustment^13,14. In particular, this research highlights the association of restrictive eating and weight loss with serious psychological and behavioral disturbance including anhedonia, isolation, depressed mood, irritability, poor concentration, and fatigue ^13,14. Despite these associations, beyond eating disorders, there has been little attention paid to the relationship between symptoms of unexplained weight loss, underweight or poor appetite and other forms of psychiatric disturbance. Understanding whether unexplained weight loss, underweight or poor appetite in adolescents represents not only early warning signs of eating disorders, but also of other psychiatric conditions can prompt timely intervention to prevent these adverse consequences.

Using the Rochester Epidemiological Project (REP) database, the goal of this retrospective, observational study was to examine the lifetime prevalence of all psychiatric disorders and eating disorders in a large cohort of adolescents diagnosed with symptoms of unexplained weight loss, underweight or poor appetite. The REP is a database infrastructure funded by the National Institutes of Health, linking medical events and health-care providers for 95% of Olmsted County, Minnesota, USA residents throughout Southeast Minnesota and Wisconsin ^15,16. Because over 90% of Olmsted County residents return for follow up medical visits within 3 years and show low rates of attrition, the REP demonstrates excellent longitudinal data characterizing the development of medical and psychiatric conditions ^16,17. The population of the REP matches the general Minnesota population in terms of age, sex, and race¹⁷. A secondary aim of this study was to determine patient demographic and weight-related characteristics (i.e., BMI percentile at index visit, decrease in BMI percentile and highest historical BMI percentile) of this sample that are associated with receiving a diagnosis of a psychiatric disorder, as well as an eating disorder diagnosis within this 5-year period.

Methods

Participants

We used the REP database to identify all adolescent patients (13–18y) seen in any medical setting with a billing code diagnosis of weight loss or abnormal weight loss (r63.4), pediatric underweight (r63.6), and loss of appetite (r63.0) between January 1, 2005 and December 31, 2017 (n=883). Although family history and genetics play a strong role in a young person’s presenting weight such that for some youth, a low weight relative to population norms might be appropriate, we chose to include billing code diagnoses of underweight to capture adolescents’ whose presenting weight was a focus of the visit and presumably viewed with concern by the medical provider. Patients were excluded if they did not provide research authorization (n=15), were pregnant (n=8), or received a diagnosis of weight loss based on a medical recommendation to lose weight, as this diagnosis refers to weight loss counseling rather than actual weight loss (n=38). To focus exclusively on weight loss, underweight or poor appetite that is not explained by other conditions or circumstances, as well as to eliminate the confounding association of many medical conditions with psychiatric disorders, we excluded adolescents diagnosed with conditions/interventions that are commonly associated with weight loss. In particular, we excluded adolescents diagnosed with acute infection (n = 123), complex medical, metabolic or genetic issue such as cystic fibrosis (n = 37), diabetes mellitus (n = 25), gastrostomy/ileostomy (n = 16), cancer (n = 16), and developmental delay/cerebral palsy (n = 13). We chose not exclude patients with medical diagnoses that are also sequalae of eating disorders such as nutritional deficiencies, amenorrhea, fatigue, constipation or abdominal pain, as we wanted to capture potential cases of eating disorders. In total, 221 patients were excluded (some had multiple exclusion factors).

This study was approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards who review all REP research proposals. All REP studies must comply with the Mayo Clinic Research Authorization (Minnesota State privacy law-statute 144.335, 1997) that requires that individuals provide permission for their medical records to be used in research studies.

Medical record review

To access data from the REP database, a series of data macros developed from computer algorithms were created to assist in the retrieval of information from various REP sources. Macros pull a variety of data points including demographics and medical codes (e.g., demographics, diagnoses, labs, prescriptions). Using these data macros, we pulled baseline and outcome data on psychiatric billing codes for included patients. These billing codes were later collapsed into the following categories: eating disorders (anorexia nervosa, binge eating disorder, bulimia nervosa, eating disorder not otherwise specified/other specified feeding and eating disorder, avoidant/restrictive food intake disorder, feeding/eating disturbance), mood disorders/self-harm, anxiety disorders/obsessive compulsive disorder/post-traumatic stress disorder, substance use disorders, and other psychiatric conditions (e.g., attention deficit hyperactivity disorder, autism, conduct disorder, dissociative disorder, gender dysphoria, body dysmorphic disorder, factious disorder, personality disorder).

To gather information to describe our sample, as well as to use these patient characteristics as predictors of eating disorder and other psychiatric disorders, we also used data macros to pull data from the five-year period following the index visit. Our a priori reason for selecting this timeframe was that we believed it would be an adequate period for eating disorders and other psychiatric conditions to come to medical attention and also that it would span the majority of the adolescent developmental period for most of our sample. We extracted patient demographics, highest historical BMI percentile and BMI percentile at presentation. Data macros were also used to pull information related to age, sex, race, height, weight, BMI, and diagnoses. Height, weight, and age values at each visit for adolescents were later converted into BMI percentile using the Center for Disease Control calculator¹⁸. We then identified highest historical BMI percentile and the nearest BMI percentile within 1 month of the index date. Decrease in BMI percentile was calculated as the difference between the highest historical BMI percentile and the BMI percentile at index visit.

Statistical Analysis

Patient characteristics were summarized using median (interquartile range [IQR]) for continuous variables, and number (percentage) for categorical variables. Univariate and multivariable analyses were performed using Cox proportional hazards models to assess the relationship between baseline diagnoses, and the development of psychiatric diagnoses and eating disorders (within five years of index visit). Additionally, univariate Cox proportional hazard models assessed the relationship between patient characteristics and the development of psychiatric diagnoses and eating disorders. Variables selected for the multivariable Cox hazard models were chosen based on significant results from the univariate analysis, in addition to index diagnosis which was included in all multivariable models. Patient characteristics included age, sex, race, BMI percentile at index visit, change in BMI percentile, and highest historical BMI percentile. The time variable was the number of days from the index visit until last follow-up (i.e., last medical visit in 5 year follow-up period), date of diagnosis, or censored after five years, whichever came first. Patients were excluded from the Cox models if a last follow-up date was unavailable (n = 16). Kaplan-Meier curves were generated to show the relationship between time since index visit (plus one week) and the time to diagnosis of non-ED psychiatric disorders and eating disorders. Considering some patients had multiple visits in the first week following the index visit, we made the decision to consider any diagnosis made within 7 days of the index visit to be the same as a diagnosis at index, as these were highly likely to be related diagnoses.

For all analyses, p-values < 0.05 were used to signify statistical significance without correction for multiple testing given the exploratory nature of non-descriptive analyses. All analyses were performed in SAS Studio 3.8 (SAS Institute Inc, Cary, North Carolina) and RStudio (RStudio Team 2021, Boston, Massachusetts).

Results

Participant Characteristics

In total, 662 adolescent patients (median [IQR] age = 16 years [15, 17]) met our inclusion criteria. The demographics of our cohort are shown in Table 1. While the majority of the sample identified as female (n = 437; 66.0%), one-third (n = 225; 34.0%) identified as male. Patients presented with a median BMI percentile of 32.3 (IQR = 9.4, 62.1) and had a median decrease of 22.9 (IQR = 9.3, 41.7) BMI percentile points. Prior to the weight/appetite diagnosis, the median highest historical BMI percentile was 70.6 (IQR = 40.3, 89.1).

Table 1.

Participant Characteristics at Index Visit

Participant characteristic	Median	Interquartile Range
Age (years)	16.0	15.0, 17.0
Highest historical BMI %	70.6	40.3, 89.1
BMI % at index visit	32.3	9.4, 62.1
BMI % change	22.9	9.3, 41.7
Participant characteristic	N	%
Female	437	66
Asian	30	4.5
American Indian	2	0.3
Black	67	10.1
White	505	76.3
Other	35	5.3
Unknown	18	2.7
Did not report	5	0.8
Current or past eating disorder diagnosis	76	11.5
Current or past psychiatric diagnosis	302	45.6

BMI % = body mass index percentile

Prevalence of Psychiatric Disorders

The lifetime prevalence of any psychiatric disorder (including eating disorders) was 69.6% (n = 461). Approximately 50% of our sample received a psychiatric diagnosis at or before the index visit. Over the five years following index visit, 59.5% (n = 394) were diagnosed with a psychiatric condition. Of these patients, 45.6% (n=302) of these patients were diagnosed earlier (at or before the index visit) and 23.0% (n=152) were newly diagnosed during the subsequent 5 years (n=152).

Non-eating disorder psychiatric disorders received at or before the index visit included mood disorders/self-harm (n = 183; 27.6%), anxiety disorders (n = 215; 32.5%), other (n = 75; 11.3%), and substance abuse (n = 54; 8.2%). These diagnoses were not mutually exclusive, as patients could have one or more of these diagnoses. Over the five years following index visit, non-eating disorder psychiatric disorders included non-mutually exclusive diagnoses of mood disorders/self-harm (n = 284; 42.9%), anxiety disorders (n = 239; 36.1%), substance abuse (n = 89; 13.4%); and other (n = 72; 10.9%). The smoothed hazard curve indicating the time until psychiatric diagnosis (from 7 days after index visit) is represented in Figure 1. The median time to psychiatric diagnosis was 1018 days or 2.7 years.

Figure 1.

Kaplan-Meier curve reflecting time to psychiatric diagnosis.

Patient Characteristics Associated with Future Psychiatric Disorders

In terms of patient characteristics associated with receiving a non-eating disorder psychiatric disorder diagnosis in the five years following their index visit, females were 1.4 times more likely to receive a psychiatric diagnosis than males (HR: 1.369, 95% CI [1.090,1.721]; p = 0.007). Compared to white adolescents, Asian (HR: 0.523, 95% CI [0.287,0.955]; p = 0.035) and Black adolescents (HR: 0.538, 95% CI [0.345,0.807]; p = 0.003) and those of unknown race (HR: 0.084, 95% CI [0.012,0.597]; p = 0.013) were significantly less likely to be diagnosed with a psychiatric disorder. A patient’s historical highest BMI percentile was associated with receiving a psychiatric diagnosis within 5 years (HR: 1.004, 95% CI [1.000, 1.008]; p = 0.042) with a 10% increase in the chance of receiving a psychiatric diagnosis for each 10% increase in the highest historical BMI percentile. Patients who received a diagnosis of underweight, were less likely than those who received a diagnosis of weight loss to be diagnosed with a psychiatric disorder (HR: 0.645, 95% CI [0.449,0.926]; p = 0.018). Those who received a diagnosis of loss of appetite did not differ from those diagnosed with weight loss in the likelihood of receiving a psychiatric diagnosis (HR: 1.233, 95% CI [0.912,1.668]; p = 0.174). See Table 2.

Table 2.

Hazard Ratios of Patient Characteristics Associated with a Psychiatric Diagnosis (Excluding Eating Disorders) within 5 years.

	N	Hazard Ratio	95% CI	P-value
Age (years)	645	1.062	(0.995, 1.134)	0.0704
Gender	645	1.369	(1.090, 1.721)	0.0070
Race a	643			0.0025
American Indian		0.901	(0.127, 6.423)	0.9176
Asian		0.523	(0.287, 0.955)	0.0349
Black		0.528	(0.345, 0.807)	0.0032
Other		0.679	(0.41, 1.124)	0.1323
Did not report		1.341	(0.430, 4.183)	0.6134
Unknown		0.084	(0.012, 0.597)	0.0133
BMI % at index visit	631	1.002	(0.999, 1.006)	0.1898
Change in BMI %	493	1.003	(0.998, 1.009)	0.2228
Highest historical BMI %	494	1.004	(1.000, 1.008)	0.0423
Index diagnosis	645			0.0144
Lack of appetite		1.233	(0.912, 1.668)	0.1740
Underweight		0.645	(0.449, 0.926)	0.0175

Note. BMI % = body mass index percentile

^aWhite is the reference group due to the higher number of participants in this group.

Prevalence of Eating Disorders

The lifetime prevalence of any eating disorder was 21.3% (n =141). Approximately, 11.5% (n=76) of the sample received an eating disorder diagnosis prior to or at the index visit. Over the five years following the index visit, the prevalence of eating disorders was 17.8% (n=118). As with psychiatric disorders, some of these patients were diagnosed earlier (at or before the index visit; n=76, 11.5%) and some of these patients were newly diagnosed during the subsequent 5 years (n=67, 10.1%).

At the index visit, patients had received a prior or current eating disorder diagnoses of eating disorder not otherwise specified (n = 47), feeding or eating disturbance (n = 31), anorexia nervosa (n = 25) and bulimia nervosa (n = 6). Over the five years following index visit, patients received diagnoses of eating disorder not otherwise specified (n = 78), anorexia nervosa (n = 66), feeding or eating disturbance (n = 31), bulimia nervosa (n = 17), avoidant/restrictive food intake disorder (n=4), and binge eating disorder (n = 1). At both measurement times (at presentation and five years following index visit), the percentages sum to > 100 due to a portion of patients receiving more than one eating disorder diagnosis.

Patient Characteristics Associated with Future Eating Disorders

As shown in Table 3, female patients were 4.27 times more likely to be diagnosed with an eating disorder than males (HR: 4.271, 95% CI [2.395,7.615]; p < 0.001) over the five years following index visit. While overall race was not associated with eating disorder diagnosis (p = 0.176), Black adolescents were significantly less likely than adolescents of other racial/ethnic backgrounds to be diagnosed with an eating disorder (HR: 0.291, 95% CI [0.107, 0.789]; p = 0.015) as compared to white adolescents. Adolescents with a greater change in BMI percentile were more likely to be diagnosed with an eating disorder during the five years following their index visit (HR: 1.010, 95% CI [1.001, 1.020]; p = 0.032). The Kaplan-Meier curve (Figure 2) demonstrates that the majority of patients with symptoms of unexplained weight loss, underweight or poor appetite who received an eating disorder diagnosis were diagnosed during the first year following index visit (plus 7 days).

Table 3.

Hazard Ratios of Patient Characteristics Associated with an Eating Disorder Diagnosis within 5 years.

Participant Characteristic	N	Hazard Ratio	95% CI	p-value
Age (years)	646	0.985	(0.878, 1.106)	0.8019
Gender	646	4.271	(2.395, 7.615)	<0.0001
Race a	644			0.1763
American Indian		-	-	-
Asian		0.312	(0.077, 1.265)	0.1030
Black		0.291	(0.107, 0.789)	0.0153
Other		0.651	(0.265, 1.598)	0.3491
Did not report		-	-	-
Unknown		-	-	-
BMI % at index visit	632	0.994	(0.988, 1.001)	0.0808
Change in BMI %	494	1.010	(1.001, 1.020)	0.0323
Highest historical BMI %	495	1.000	(0.993, 1.007)	0.9686
Index diagnosis	646			0.1953
Lack of appetite		1.399	(0.849, 2.303)	0.1873
Underweight		0.707	(0.367, 1.362)	0.2996

Note. BMI % = body mass index percentile

^aWhite is the reference group due to the higher number of participants in this group.

Figure 2.

Kaplan-Meier curve reflecting time to eating disorder diagnosis.

Discussion

This study provides new insight into adolescent patients whose symptoms of unexplained weight loss, underweight or poor appetite come to the attention of a medical provider. In particular, the findings reveal that these youth not only have a markedly high lifetime prevalence of eating disorders (1 in 5 adolescents as compared to approximately 1 in 50 of the general population), but also of psychiatric disorders in general (nearly 3 in 4 adolescents). The high lifetime prevalence of psychiatric disorders in youth with symptoms of unexplained weight loss, underweight or poor appetite far exceeds that observed in the general adolescent population (70% vs. 20%) ¹⁹. Consequently, the symptom presentation of unexplained weight loss, underweight or poor appetite in adolescents should not be seen as benign and transitory, but instead may reflect a high-yield indicator of current or future psychiatric disorders with implications for screening, frequent follow-up visits, as well as for nutritional and weight restoration interventions. It should be noted that this sample included adolescents at any presenting BMI whose status of underweight or weight loss was determined relative to their personal historical growth trajectory, not simply those who were underweight based on age and sex norms for the population. As such, our findings highlight that weight loss and restrictive eating, no matter the presenting BMI, can lead to untoward psychiatric outcomes.

In addition to the high lifetime prevalence of all psychiatric disorders among these youth, the rate of eating disorders in this sample was nearly 8 times greater than the lifetime prevalence rate of adolescent eating disorders in the general population⁶. Given that eating disorders often go unrecognized in health care settings ^20–23, it is possible the high prevalence of eating disorders identified among these adolescents with unexplained weight loss, underweight or poor appetite represents a vast underestimate. Because symptoms of weight loss, underweight or poor appetite often co-occur with many eating disorders ²⁴, the fact that 21% of these youth will receive an eating disorder diagnosis at some point in their adolescence might not seem that surprising. Nonetheless, this finding has significant implications for detection of eating disorders in young patients presenting with unexplained weight loss, underweight or poor appetite and underscores the importance of including eating disorders, as well as other psychiatric diagnoses on the list of differential diagnoses regardless of the absence of patient-reported weight or shape concerns or other contributing factors.

In terms of specific characteristics of adolescents presenting with unexplained weight loss, underweight or poor appetite that are associated with receiving a diagnosis of an eating disorder or other psychiatric condition within 5 years, girls were more than 4.5 times more likely to be diagnosed with eating disorders and 1.3 times more likely to be diagnosed with other psychiatric disorders than boys. These figures appear to be higher than the proportion of females to males experiencing these disorders in the general population^6,19,25. In terms of race, white adolescents were more likely than Black adolescents to be diagnosed with an eating disorder and were also more likely than adolescents of Black, Asian, or unknown race to be diagnosed with another psychiatric condition. It is important to note that these findings are likely to reflect gender and racial disparities in detection of eating disorders and other psychiatric disorders, rather than specific factors associated with the development of these conditions. Specifically, pervasive stereotypes regarding patients thought to be at risk for eating disorders (e.g., white, cisgender girls) may deter clinicians from diagnosing eating disorders in adolescent boys and in persons of color²⁶.

In addition to race and gender, the extent of an adolescents’ weight loss was associated with receiving an eating disorder diagnosis within 5 years of the index visit with a 10% increase in the chance of receiving an eating disorder diagnosis for each 10% decrease in BMI percentile. Certainly, eating disorders may be easier to detect in adolescents presenting with a large degree of weight loss ²⁷. For psychiatric disorders, the higher an adolescent’s weight before the index visit, the greater the chance of receiving a psychiatric diagnosis. In the context of stereotypes that eating disorders involving symptoms of restrictive eating and weight loss are typified by anorexia nervosa and associated with low weight²⁶, such symptoms in adolescents of higher weight status may have been attributed to a general psychiatric disorder rather than to an eating disorder. In this sample, adolescents diagnosed with weight loss were more likely than patients who were diagnosed with underweight to be diagnosed with a psychiatric disorder. This finding may highlight the relevance of decreases in weight and insufficient caloric intake to psychiatric sequelae, rather than the impact of absolute body weight on psychiatric symptoms.

Future longitudinal research is needed to elucidate the direction of the associations between presentations of unexplained weight loss, underweight or poor appetite and psychiatric/eating disorder diagnoses. Because the focus of our study was on adolescents who were specifically diagnosed with unexplained weight loss, underweight or loss of appetite, the findings are likely to be biased to more severely symptomatic individuals. As such, future research examining symptoms of unexplained weight loss, underweight or poor appetite, whether or not they are identified by a medical provider, will be important for understanding how these symptoms may predispose to adverse outcomes. Future research should also incorporate behavioral measures to determine if there are specific correlates or questions on self-report measures that predict which adolescents will go on to develop eating disorders and which will develop other psychiatric conditions. Finally, studies on psychiatric outcomes related to provider management of adolescent unexplained weight loss, underweight or poor appetite are crucial to inform treatment guidelines to mitigate the negative psychological effects of weight loss, underweight status, and insufficient caloric intake.

The study has several strengths including the large population-based sample of adolescents with weight loss, underweight or poor appetite presenting to any medical setting across diverse health care systems which eliminates selection bias associated with patients presenting to specific settings. However, it is important to consider the findings of this study in the context of several limitations. Our study focused on a single population in the Midwestern United States which may not generalize to other populations, though the socioeconomic and demographic characteristics of the REP have been demonstrated to reflect a large segment of the entire United States population¹⁷. Although our findings suggest a much greater prevalence of eating disorders and other psychiatric conditions than those identified in prevalence studies of the general adolescent population, we do not have a comparison group of adolescents presenting to health care settings with which to place our findings in context^6,19,25,28. Of note, findings only generalize to patients whose weight loss, underweight or poor appetite symptoms were identified by a health care provider, and not to those whose symptoms did not garner medical attention. Another important limitation is that the participants did not have their psychiatric or eating disorder confirmed by a structured clinical interview. Without such diagnostic information, it is possible that psychiatric sequalae of eating disorders were attributed to psychiatric diagnoses, thereby underestimating the overall prevalence of eating disorders and overestimating the prevalence of other psychiatric conditions. However, given that eating disorders and other psychiatric conditions are commonly underdiagnosed in medical settings ²⁹,the prevalence of eating disorders and other psychiatric disturbance in adolescents presenting with symptoms of weight loss, underweight and loss of appetite may be even higher than identified in this study. Despite these limitations, this study provides important information on the characteristics of an understudied population to guide future research.

In summary, this is the first study to examine psychiatric outcomes in adolescents identified in a large population database presenting to all health care settings with unexplained weight loss, underweight or poor appetite. The findings of this study identified a remarkably high lifetime prevalence of psychiatric disorders among these youth. The data suggest that unexplained weight loss, underweight or poor appetite that comes to the attention of a medical provider should serve as a potentially high yield cue for further psychiatric assessment and close follow up to prevent and treat these adverse psychiatric outcomes. Future research is needed to further delineate the relationship of unexplained weight loss, underweight or poor appetite with psychiatric outcomes so that treatment guidelines can be expanded beyond the prevention and treatment of eating disorders to the full spectrum of mental health concerns.