The impact of psychiatric comorbidity on length of stay of medical inpatients☆

1. Introduction 

The coexistence of psychiatric comorbidity (PC) in patients admitted to general medical wards is common [1], [2] and the PC is usually not recognized or is misdiagnosed [3], [4]. Furthermore, recent studies have shown that patients with physical illnesses and emotional dysfunction have low adherence [5], a worse prognosis [6], [7], and, consequently, increased costs [8], [9]. One of the reasons for these higher costs might be related to longer length of stay (LOS) in general wards when these patients are hospitalized due to a physical illness [8], [10].

Several studies conducted in general hospitals have tried to determine the impact of PC on LOS with conflicting results [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32]. The main difficulty is to evaluate if having a PC is an independent risk factor for increased LOS, and if so, which specific mental disorders can influence LOS. The first studies found an association of cognitive impairment (delirium and/or dementia) with increased LOS but they did not control for severity of physical illness, leaving the question as to whether these were biased results, because the more severely ill might have longer stays and also more cognitive impairment [11], [12], [14], [16], [17], [18], [19], [20], [24]. More recent prospective studies that addressed this issue, controlling for several variables, including the physical severity, have found similar results, with cognitive impairment increasing LOS [22], [23], [26], [29], [31], [32], [33] and others have not [25]. The association of other specific PCs such as depressive and anxiety disorders is less well established, with some studies showing depression increasing LOS [23], [30], [34] and others not [24], [31]. A recent study has also shown association of substance related diagnoses with increased LOS [32]. A critical review of 26 outcome studies that examined the association of PC and LOS in the general hospital showed that impaired cognition associated with delirium and dementia, depressed mood, and other personality variables contribute to prolonged hospital stays [28].

The reasons for these differences are probably due to the way patients were selected (consecutively admitted or consultation liaison patients), how the diagnoses were made (clinical diagnosis, scales, interviews), characteristics of patients (physical illnesses, age, sociodemographic variables), characteristics of their health insurances and other variables. We sought to improve this evaluation by using a prospective controlled observational design in which a psychiatrist used a well-validated structured interview [35], to diagnose all consecutive admissions to the general medical wards of a hospital funded by the Brazilian government and without any restraints concerning the LOS. The main goal of this study was to determine the impact of psychiatric comorbidity on LOS of medical inpatients.

2. Materials and methods 

3. Results 

Subjects (N=317) were predominantly male (65%), white (93%), married (67%). The age ranged from 18 to 64 years; being 70% <65 years old and the majority had less than 5 years of education (53%). The sample had a mean LOS±SD of 13±12 days and the main reasons for hospitalization were gastrointestinal and cardiovascular disorders (16.8% and 17.4%, respectively). When comparing patients with and without PC there were not any differences on sociodemographic variables. There were, however, statistically significant increases in the death rates (Odds Ratio [OR]=2.7; Confidence Interval [CI]=1.33–5.6) and LOS (F=6.9; p=0.05) of patients with PC. Table 1 shows detailed information on sociodemographic and medical variables and the comparison of both groups. Table 2 shows the frequencies of patients with any PC and according to each psychiatric diagnosis.

Table 1. Demographic and medical data for 317 medical inpatients with and without psychiatric comorbidity

	Variable*	All inpatients (N = 317)	Psychiatric comorbidity?		P**
			No (n = 161)	Yes (n = 156)
	Sociodemographic
	Age, years	53 ± 18.3	50.9 ± 17.5	55.1 ± 18.9	0.38
	Male, n (%)	205 (65)	110 (68.3)	95 (60.9)	0.17
	White, n (%)	294 (93)	149 (92.5)	145 (92.9)	0.56
	Married, n (%)	211 (67)	113 (70.2)	98 (62.8)	0.17
	Education, years	5.5 ± 4.3	5.7 ± 4	5.2 ± 4.7	0.33
	Medical
	Physical comorbidity†	2.4 ± 1.7	2.3 ± 1.6	2.6 ± 1.8	0.14
	Length of Stay, days	13.3 ± 12	12.1 ± 9.9	14.7 ± 13.8	0.05
	Inhospital Death, n (%)	40 (12.6)	12 (7.5)	28 (17.9)	<0.01
	Reason for hospitalization, n (%)‡
	Cardiovascular	50 (16.8)	32 (19.9)	18 (11.5)	0.06
	Gastrointestinal	55 (17.4)	27 (16.8)	28 (17.9)	0.78
	Neoplasms	37 (11.7)	24 (14.9)	13 (8.3)	0.68
	Pulmonary	35 (11)	13 (8.1)	22 (14.1)	0.09
	Infectious	38 (12)	15 (9.3)	23 (14.3)	0.14
	Other	102 (32.2)	50 (31.1)	52 (33.3)	0.66

* Values are mean ± Standard Deviations, unless otherwise noted ** t test, Chi-square test or Fisher’s exact test, as appropriate. † Physical comorbidity measured by the Charlson comorbidity index score [36]. ‡ According to the International Classification of Diseases (10th version).

Table 2. Psychiatric comorbidity among 317 medical inpatients

	Psychiatric comorbidity*	n	%
	Any psychiatric comorbidity**	156	49.2
	Cognitive impairment†	64	20.2
	Depressive disorders	52	16.4
	Substance related disorders	33	10.4
	Adjustment disorders	14	4.4
	Anxiety disorders	6	1.9

* According to the DSM-IV criteria [37]. ** Categories are not mutually exclusive. † Cognitive impairment = DSM-IV delirium and/or dementia [37].

In the univariate analysis of variance, only age (F=1.63; P<.01) and the Charlson Comorbidity Index (F=2.94; P<.01) were significantly associated with increased LOS in the PC group. In the multivariate analysis of variance for all sample (controlling for age and physical comorbidity), only patients with cognitive impairment had increased LOS (F=17.8; P<.01) (Table 3). To check if death was altering the results, this analysis was conducted, then, excluding patients who died during hospitalization, the results remained the same with cognitive impairment independently increasing LOS (F=26.2; P<.01) (Table 4). These analyses were also conducted excluding 2 patients who had longer LOS (outliers): one with delirium who stayed 120 days and another with no PC who stayed 91 days. The results remained the same.

Table 3. Multivariate analysis* of length of hospital stay by psychiatric comorbidity, controlling for age and physical severity for the total sample (N = 317).

	Psyciatric comorbidity†	Covariates	F	p-value*
	Any psychiatric diagnosis	Age	3.62	0.057
		Physical severity**	2.27	0.097
		Both age and physical severity	2.93	0.087
	Cognitive impairment	Age	22.1	<0.01
		Physical severity	15.3	<0.01
		Both age and physical severity	17.8	<0.01
	Depressive disorder	Age	0.54	0.46
		Physical severity	0.41	0.52
		Both age and physical severity	0.36	0.55
	Anxiety disorder	Age	1.93	0.16
		Physical severity	1.38	0.24
		Both age and physical severity	1.48	0.22
	Substance related disorder	Age	0.81	0.36
		Physical severity	0.67	0.41
		Both age and physical severity	1.05	0.30

* Analysis of covariance. ** Physical severity measured by the Charlson Comorbidity Index [36]. † Psychiatric comorbidity = DSM-IV [37] diagnosed by SADS [35]; depressive disorder (Major depression, dysthymia, minor depression); cognitive impairment (delirium and/or dementia); anxiety disorder (panic, generalized anxiety, phobia); substance related disorder (abuse or dependence of alcohol and other substances).

Table 4. Multivariate analysis* of length of hospital stay by psychiatric comorbidity, controlling for age and physical severity** excluding deaths during hospitalization (N = 277)

	Psychiatric comorbidity†	Covariates	F	p-value*
	Any psychiatric diagnosis	Age	3.99	0.046
		Physical severity	3.80	0.050
		Both age and physical severity	3.95	0.047
	Cognitive impairment	Age	29.4	<0.01
		Physical severity	21.8	<0.01
		Both age and physical severity	26.2	<0.01
	Depressive disorder	Age	0.49	0.48
		Physical severity	0.28	0.59
		Both age and physical severity	0.24	0.62
	Anxiety disorder	Age	1.83	0.17
		Physical severity	1.32	0.25
		Both age and physical severity	1.45	0.22
	Substance related disorder	Age	1.04	0.30
		Physical severity	1.06	0.30
		Both age and physical severity	1.70	0.19

* Analysis of covariance. ** Physical severity measured by the Charlson Comorbidity Index [36]. † Psychiatric comorbidity = DSM-IV [37] diagnosed by SADS [35]: depressive disorder (Major depression, dysthymia, minor depression); cognitive impairment (delirium and/or dementia); anxiety disorder (panic, generalized anxiety, phobia); substance related disorder (abuse or dependence of alcohol and other substances).

4. Discussion 

These results show that cognitive impairment measured early in the course of a hospital admission, predicts increased LOS in medical inpatients. The impact of this comorbidity on LOS remained statistically significant even after controlling for covariates such as age and physical disease morbidity, separately and simultaneously. When patients who died during the hospitalization were excluded from the analysis, the results for cognitive impairment were the same and an association of the category “any psychiatric comorbidity” with increased LOS became significant (F=3.95; P=.047). Other PCs, such as depressive, adjustment, anxiety, substance related disorders did not increase LOS.

This finding is consistent with several prospective studies that showed that cognitive impairment is associated with an increase in LOS [21], [23], [24], [26], [28], [32]. However, Incalzi et al. [25] did not find this association in a sample of patients older than 70 years. The lack of association of depressive disorders with increased LOS is in agreement with three prospective well controlled studies [25], [31], [32], but is in disagreement with others [23], [27]. These discrepancies probably are due to differences in the methodology of studies (selection of sample and how the PCs were diagnosed) and administrative issues (e.g., health insurance policies and hospital rules). For example, if only consultation-liaison (C-L) patients are evaluated, probably different results are found, because they have some characteristics (e.g., personality traits?) that motivate their attending physicians to ask for a consult as opposed to other patients with PCs that are not even detected. Also, if only elderly patients are evaluated, there may be so many potential competing factors that influence LOS that it decreases the power of PCs.

Possibly, the reasons for the association of cognitive impairment with increased LOS is because delirious patients do not understand, do not cooperate, do not explain their needs, do not sleep, and are frequently agitated withdrawing their iv lines and other devices which postpones their improvement.

These findings highlight the importance of detecting cognitive impairment in medical inpatients since it worsens their outcomes. Furthermore, the diagnosis can be done in a clinical interview and practical things can be done to prevent/treat delirium: withdrawing (when possible) medications with anticholinergic effects, controlling pain, early detecting and treating possible causes, such as infections and electrolyte imbalance [39], [40], [41].

There are some limitations in our study that deserve mention. First, the determinants of LOS in a public hospital in Brazil may be different from the factors that influence this variable in other countries. On the other hand, it is a public hospital and the Brazilian Government funds all costs. Therefore, there is not any pressure to reduce the LOS.

Second, we did not use a scale to measure cognitive impairment like the Mini-Mental State Exam (MMSE) [42] as other authors have done [23], [31]. The reason is because it would not be valid since it has already been shown that subjects with less than 8 years of education have inflated rates of “cognitive impairment” when this instrument is used [43], [44]. We decided to use clinical diagnosis, based on the delirium/dementia DSM-IV criteria [37] because we believed that the clinical judgment of a psychiatrist using standardized criteria would be more valid to diagnose cognitive impairment in this low educated population than a nonvalid scale.

5. Conclusions 

Medical inpatients with cognitive impairment have increased LOS after controlling for possible confounders, such as physical severity and age. Further studies should address the question as to whether measures to prevent, increase recognition and improve management of delirium and dementia in this population improve their outcomes and reduce costs.