Effects of malnutrition on length of stay in patients hospitalized in an acute psychiatric ward

Psychiatric patients are considered at risk for malnutrition due to pharmacological treatments, lifestyle habits and the mental illness by itself. Even though metabolic risk factors have been related to worse outcomes in certain conditions, the evidence regarding the nutritional status and its impact on the length of stay in psychiatric inpatients is scarce. This study aims to characterize the nutritional status in acute psychiatric patients, to correlate it with the length of stay, and to find specific potential indicators of malnutrition.

the nutritional status were found in: BMI, cholesterol, triglycerides, albumin, total proteins, prealbumin, iron, lymphocytes and zinc levels, and transferrin saturation.The multivariate analysis showed a significant association for cholesterol and zinc levels, lymphocyte count, and BMI.Conclusions: Our results suggest that nutritional status might influence the course of psychiatric admissions.Cholesterol and zinc levels, lymphocyte count, and BMI might be factors strongly associated with malnutrition.This consideration might allow the identification of profiles in which lifestyle interventions could be implemented.

| INTRODUCTION
Hospital malnutrition is a highly prevalent problem that affects up to 50% of both medical and surgical patients. 1,2linical malnutrition can be defined as a state resulting from lack of intake or uptake of nutrition that leads to altered body composition, such as decreased fat free mass, and body cell mass leading to diminished physical and mental function and impaired clinical outcomes. 3ndeed, some studies have shown that a poor nutritional status at admission is associated with increased patient morbidity and mortality, resulting in longer hospital stays and increased healthcare costs. 1,2,4,5In patients with psychiatric disorders there are some variables that might be considered risk factors for a poor nutritional status, such as pharmacological treatment side effects, lifestyle habits as inactivity, poor diet, and the illness by itself, depending on the phases of the disease.For example, during depressive phases patients can present reduced appetite and weight loss. 6There are also other barriers for an adequate diet, such as limited availability of healthy food, inadequate social support, a low socio-economic status, or a limited budget among others. 7,8Despite this, little data are available on malnutrition in psychiatric populations except for studies related to eating disorders. 9Another limitation is the lack of malnutrition risk screening tools for people with mental disorders. 10However, there is an increasing interest to study the nutritional status in patients with severe mental disorders, 11,12 given the important burden that diet, lifestyle factors, and their consequences, may have in this population where physical health is often compromised, presenting a higher prevalence of cardiovascular diseases and a reduced life expectancy compared to the general population. 6,13,14o far, various studies have reported nutritional status as a prognostic marker in a number of diseases.In order to assess the nutritional status, some studies use the controlling nutritional status score (CONUT), which takes into account the total lymphocyte count, serum albumin, and total cholesterol levels.6][17][18] Despite the available data, the studies performed in this field do not include psychiatric population, being psychiatric patients often excluded from the studies or not considered among the reference population. 4,5,19,20ome studies have reported metabolic disturbances are related to worse psychiatric outcomes.For instance,

Significant outcomes
• Malnutrition was observed in 42.5% of inpatients admitted to our acute psychiatric unit.• Plasmatic transferrin saturation, protein and iron levels were inversely correlated with length of stay, having low iron levels an association with longer hospitalizations.• Negative correlations with the nutritional status were found especially for cholesterol and zinc levels, lymphocyte count, and BMI.

Limitations
• There is a lack of previous evidence about the global nutritional status of psychiatric inpatients, with the exception of patients with eating disorders.• The cross-sectional design of this study cannot reveal causal dynamics between variables.
metabolic syndrome, obesity, and impaired glucose metabolism have been considered risk factors for worse outcomes in patients with bipolar disorder, since they have been related to a chronic course of the illness, worse global functioning, and rapid cycling. 21uring acute psychiatric hospitalizations, clinical practice tends to include the assessment of different analytical parameters, including cholesterol, total protein levels, and lymphocyte count.Nevertheless, to the best of our knowledge, there is a lack of evidence regarding the impact of the nutritional status in psychiatric inpatients on the length of stay, medical comorbidities, and other prognostic factors.
Considering the missing data on psychiatric population, this study aims to provide a first approach to evaluate whether a deficient nutritional status is correlated with specific outcomes, such as the length of stay in an acute hospitalization unit.Secondary aims of this study include the characterization of the nutritional status in patients admitted to an acute psychiatric ward, the correlation between different nutritional variables and the assessment of specific nutritional variables that might better predict a poor nutritional status in this population.

| Participants
The data analyzed derived from records corresponding to patients aged 18 years or older admitted to the Hospital Clínic of Barcelona acute psychiatric ward.The records included a 1-year period.Ethical approval was provided by the Hospital Clínic Research Ethics Committee (protocol code: D2017).
Patients from which no metabolic parameters were available at the beginning of the hospitalization were excluded from the study.Patients with more than one admission in the same year were included only once, considering the first admission.No restrictive criteria were established in terms of diagnosis, being all patients admitted in the acute psychiatric ward considered for inclusion.

| Sociodemographic and clinical measures
Some of the variables collected for analysis included age, gender, length of stay, the presence of medical conditions and primary and secondary diagnoses.Variables regarding treatment included use of benzodiazepines, antipsychotics, antidepressants, mood stabilizers or electroconvulsive therapy (ECT) during the admission.

| Nutritional measures
The nutritional status was assessed at admission with the CONUT score, which takes into account the total lymphocyte count, serum albumin, and total cholesterol levels.Total scores between 0 and 1 were considered normal, 2-4 mild malnutrition, 5-8 moderate malnutrition, and 9-12 severe malnutrition. 16ther metabolic variables included body mass index and triglycerides, proteins, transferrin, prealbumin, iron, zinc, and transferrin saturation.

| Statistical analyses
In the descriptive analysis, quantitative variables were expressed by mean and standard deviation (SD), or median and interquartile range (IQR).Results from categorical variables were shown as frequencies (number and percentage).
T-test and Mann Whitney U tests were used for mean comparisons between two groups, using t-test for data that were normally distributed.Chi-square tests were used for the comparison of categorical variables, and Fisher exact tests were performed when the expected cells were lower than five.
For the correlation of nutritional parameters with length of stay or nutritional status and for the identification of predictive parameters, linear regression was used after correlation analyses.To calculate the association of malnutrition with nutritional parameters, psychiatric diagnosis, sociodemographic characteristics and pharmacological treatment, logistic binary regression analyses were conducted.
Aiming to predict the influence in the length of stay of blood parameters or nutritional status logistic regression analyses were used.Univariate logistic regression models were conducted for each of the potential associated factors.A p-value <0.05 was used for the screening covariates.Forward stepwise selection algorithms were used for selecting the covariates in the multivariate logistic regression model.At each step, the least significant variable was discarded from the model.Only covariates with a p-value <0.10 remained in the final model.These results were contrasted with backward stepwise selection in order to find potential changes in the results.The odds ratio and 95% confidence limit were calculated too.The area under the ROC curve (AUC) was measured to assess the goodness-of-fit.Finally, we checked collinearity among the different variables assessed according to the variation inflation factor (VIF).All statistical analyses were performed by the use of a confidence interval of 95%, and significance was set at p < 0.05.Analyses were conducted with SPSS version 25.0.

| Characteristics of the sample by nutritional status
Among the 357 individuals admitted to the acute psychiatric ward who were recruited, 254 subjects were suitable for analysis.Socio-demographic and clinical characteristics of the whole sample are available in Table 1.
According to the CONUT scores, 57.5% showed a normal nutritional status (scores 0-1), 39.4% showed mild malnutrition (scores 2-4), and 3.1% moderate or severe malnutrition (scores >4).No significant differences were found in gender and mean age between groups.Treatments received during the hospitalization and frequency of each diagnosis did not reveal significant differences between groups.

| Length of stay
Grade of malnutrition calculated by CONUT score did not show a significant correlation with the length of stay.
The only parameters that showed a significant inverse correlation with length of stay were plasmatic protein levels, transferrin saturation, and iron levels, being these differences significant at p < 0.01.
In the lineal regression analyses, only low total proteins, iron levels, and transferrin saturation showed an association with greater length of stay (p < 0.01 and p < 0.05, respectively), as represented in Table 2.However, the multiple regression analysis did not show a significant association with the length of stay (R 2 = 0.03, p = 0.053).No significant differences were found when length of stay was compared between patients with and without malnutrition according to the CONUT score (median 20.00 days, IQR {13.0-26.8};median 19.50 days, IQR {13.8-25.3},respectively).
No significant differences were found in terms of days of hospitalization according to the treatment with ECT, antidepressants, mood stabilizers, antipsychotics, benzodiazepines, nor in patients treated with 3 or more of these drug families at the same time.When length of stay was compared according to the presence of bipolar disorder, major depressive disorder, schizophrenia, schizoaffective disorder, substance use disorder, medical comorbidity or organic mental disorder, no differences were found between groups.

| Nutritional status
Negative correlations with the nutritional status assessed with the CONUT total score were found with BMI and all the analytical parameters, including total cholesterol, triglycerides, albumin, total proteins, prealbumin, iron, lymphocytes, zinc, and transferrin saturation, being all these negative correlations statistically significant (p < 0.01) except for transferrin, and also not observed with length of stay or age.
In addition, most of the blood parameters assessed showed to be correlated with other nutritional factors, as shown in Supplementary Table 1.
In the assessment of potential variables associated with malnutrition, univariate analyses through binary logistic regression showed a significant association for cholesterol levels, triglycerides, albumin, total proteins, transferrin, prealbumin, transferrin saturation, iron, lymphocytes, zinc, and BMI.However, in the multivariate analysis, only cholesterol levels, lymphocyte count, zinc levels and BMI showed significant differences according to nutritional status (Figure 1).These results did not vary when backward and forward stepwise selection algorithms were used.The AUC value was obtained to assess the goodness-of-fit, which resulted 0.955 (95% CI 0.92-0.99).No evidence of collinearity was observed among cholesterol levels, lymphocyte count, zinc levels, and BMI, since the VIF resulted lower than 5 for all of them.Univariate and multivariate logistic regression is shown in Table 4.
Regarding pharmacological treatment, no significant differences were observed in the proportion of individuals treated with ECT, antidepressants, mood stabilizers, antipsychotics, benzodiazepines, nor in patients treated with 3 or more of these drug families at the same time, when patients with and without malnutrition were compared.
T A B L E 3 Differences in blood parameters between patients according to their nutritional status.In addition, no significant differences were found in the proportion of patients with bipolar disorder, major depressive disorder, schizophrenia, schizoaffective disorder, substance use disorder, medical comorbidity or organic mental disorder between the two groups.

| Body mass index
Patients were grouped into weight categories using BMI classification as defined by the World Health Organization (WHO) (underweight, BMI < 18.50; normal weight, BMI 18.50-24.99;overweight, BMI 25.00-29.99;and obese, BMI ≥ 30.00). 22onsidering the whole sample, the percentage of underweight patients was 6.1%, normal weight 47.5%, overweight 26.3%, and obese subsample of well-nourished patients according to the score, the percentage of underweight people was normal weight 40.9%, overweight 29.1%, and obesity 24.5%.Among patients with a poorer nutritional status, the frequency of underweight was 5.7%, normal weight 58.6%, overweight 21.8%, and obesity 13.8%.Differences between groups did not reach statistical significance.

| DISCUSSION
It is widely known that several psychiatric disorders and pharmacological treatments commonly used in psychiatric clinical practice have an impact on metabolism and medical comorbidities. 23Moreover, the available evidence indicates that somatic conditions, such as impaired glucose metabolism, may have an impact on the course of the psychiatric illness. 21However, scarce data are available regarding nutritional status in acute psychiatric patients, being most of the evidence related to patients with eating disorders. 9,24,25he present investigation shows through a crosssectional study a considerable rate of malnutrition (defined by CONUT score ≥ 2) in a sample of patients admitted in an acute psychiatric unit (42.5%), which is consistent with other studies using the same nutritional assessment tool. 26,27Considering the correlation of CONUT scores with a worse clinical prognosis in somatic disorders, including longer hospitalizations, 4,28 this study aimed to evaluate the association of malnutrition with the length of stay and also to find nutritional parameters that might better predict the nutritional status of patients.
The analyses performed did not reveal a significant association between malnutrition and length of stay when nutritional status was assessed with the CONUT score.However, plasmatic protein levels, transferrin saturation and iron levels showed a significant inverse correlation with the days of hospitalization, indicating a potential effect of nutritional status on the course of acute psychiatric admissions.Additionally, low iron levels showed an association with greater length of stay.These results suggest that, although specific nutritional parameters have shown to be related to greater length of stay, the CONUT score might not be the best short-term predictor in psychiatric hospitalizations.This could be associated with the fact that, in the context of an acute psychiatric ward, behavioral abnormalities and psychopathology are treated in an individualized manner, with some patients with severe episodes being treated with higher doses of the indicated treatment or with different drugs.Thus, the length of stay might not be reflecting the severity of the acute episodes in all cases.Based on these findings, further studies using prospective models should evaluate if this score is associated with different course of the psychiatric illnesses in the long term, and if the nutritional parameters associated with a poorer short-term prognosis should be considered risk factors also for an adverse course of illness.
In this study, the correlation of specific nutritional parameters with nutritional status found that, apart from those included in the CONUT screening tool, low zinc, iron, total proteins, prealbumin, transferrin, triglyceride levels, a low transferrin saturation index, and BMI, were indicators of a poorer nutritional status.Most parameters showed a correlation between them.Results indicating that lower BMI, contrary to the expected, is associated with a poorer nutritional status might suggest that this subgroup of patients might present reduced food intake or higher tendency to a poorer self-care, but further evidence should be conducted in order to confirm this hypothesis.However, as also seen in our results, it is not uncommon to find a significant proportion of patients with overweight, which can be due to several factors, such as lifestyle and habits, pharmacological treatments and mental illnesses by themselves.
Our results demonstrate a strong relation of several analytical parameters assessed with the nutritional status and allow the identification of specific subgroups, such as those with greater alteration of nutritional parameters, that may present longer courses of disease, having also demonstrated in previous evidence worse medical prognosis. 4ven though there was an important correlation between different nutritional parameters and all of them were correlated with malnutrition, this study identified cholesterol levels, lymphocyte count, zinc levels, and BMI as independent factors associated with malnutrition, which highlights the importance of their assessment in clinical practice.
Psychiatric diagnoses or treatment received did not seem to have influence over status or days of hospitalization.Thus, this study does not identify specific psychiatric diagnoses or psychotropic drugs that might increase the risk for malnutrition or for a worse course of acute psychiatric admissions, basing the associated findings on analytical parameters.
In 13.8% of the patients, obesity with nutritional deficiency according to CONUT score was found, which emphasizes the importance to characterize the different nutritional profiles of these patients according to their nutritional status. 29,30This might allow clinicians to develop adequate nutritional risk screening tools considering specificity of this population 10,11,31 and determined approaches for different clinical situations. 32,33ccording to our findings, clinicians should include nutritional status-related analytical parameters for a global assessment of the nutritional status, and to consider them as factors that might influence the organic and mental well-being of patients.The authors recommend a periodical study of patients' nutritional status from the outpatient unit and not only when they are hospitalized, since this might allow the identification of determined subgroups that could benefit from psychosocial and lifestyle interventions, the prescription of nutritional supplementation in specific cases, 32 as well as the implementation of individual or group sessions aimed to the assessment of daily habits and to the promotion of healthy routines.These strategies might provide patients some skills that lead to an improvement of their nutritional status, their medical and psychiatric prognosis and their quality of life.Specifically, physicians should pay attention to those patients with poor social support and severe mental illnesses in order to detect and manage a potential malnutrition status.
The admission in an acute psychiatric ward should serve as an opportunity for clinicians to identify existing abnormalities, since a poor nutritional status has been correlated with higher adverse prognostic factors. 4,28As found in this work, altered levels of cholesterol, lymphocyte count, zinc, and low BMI should be considered as potential indicators of a poor nutritional status.Patients in which different nutritional parameters were found at insufficient levels might be a target for nutritional and lifestyle interventions aimed to improve their physical and mental health.
The results obtained from this study aim to drive a line of research in which nutritional and metabolic data may serve clinicians for wider interventions beyond the improvement of the psychiatric episodes and medical comorbidities.Since periodical blood tests and anthropometric measures are performed in outpatient units, the design of new studies targeting nutritional and metabolic parameters in psychiatric patients might be also useful in order to distinguish in the future profile of patients who are at higher risk for either suffering medical comorbidities or presenting with adverse circumstances related to their psychiatric disorder.Further studies will also elucidate the influence of pharmacological treatment and specific psychiatric disorders in patients' nutritional status.In addition, future evidence will allow to study the relationship of poorer nutritional status and prognostic factors related to the psychiatric illness, such as suicidal behavior, frequency of episodes or number of admissions.
Our study comes with some strengths and limitations.To the authors' knowledge, there is a lack of evidence about the global nutritional status of psychiatric inpatients, with the exception of patients with eating disorders.The description of the nutritional status in psychiatric population might help determine the influence of pharmacological treatments, lifestyle habits, diet, and the different psychiatric illnesses on somatic health, and might allow the implementation of preventive or therapeutic interventions in specific profiles.The use of an objective tool to determine the nutritional status confers higher reliability to the results obtained.Regarding the limitations of this study, its cross-sectional design cannot reveal causal dynamics between considered variables, and cannot allow the inclusion of other nutritional factors.However, this cross-sectional study can represent a picture of the nutritional status of patients admitted in a psychiatric ward and is a first approach to study the interaction between specific factors related with their disease and malnutrition, which may in the future open new opportunities for individualized interventions.Studies in the recent future with larger samples and longitudinal data would be useful in order to progress in this field.
To conclude, our results showed that plasmatic protein levels, transferrin saturation and iron levels were inversely correlated with length of stay in patients admitted to an acute psychiatric ward, having low iron levels an association with greater length of stay.They suggest that nutritional status might influence the course of psychiatric admissions.
In acute psychiatric hospitalizations, the nutritional assessment should include the study of different parameters, such as cholesterol, albumin, lymphocyte count, zinc, iron, prealbumin, transferrin, triglycerides, transferrin saturation, and BMI, which might allow the identification of patients with a poor nutritional status in which lifestyle interventions might be promoted with a special focus.
Future studies will allow to elucidate the relationship between nutritional status and other specific psychiatric prognostic factors apart from length of stay, such as suicidal behaviors, frequency of episodes or number of admissions.
Socio-demographic and clinical characteristics of the sample according to malnutrition grade based on CONUT severity score.
Factors associated with length of stay.
T A B L E 2Note: Linear regression analyses were performed.Statistical significance was set at p < 0.05.*Significant at p < 0.05.**Significant at p < 0.01.