Clinical and care intensity differences among patients suffering from somatic diseases electively referred to a consultation and liaison service

Introduction

Consultation and liaison psychiatry (CLP) became an acknowledged subspecialty because of its importance in the early diagnosing of mental conditions in general medicine, the integrated treatment when co-occurrence of somatic and psychiatric comorbidity, and not least because of its importance in secondary as well as tertiary prevention in the provision of care to inpatients under nonpsychiatric care. According to Ajiboye, ¹ CLP could be considered as “the guardian of the holistic approach to the patient.” The assumption of the psychosomatic perspective within the biopsychosocial model is responsible for this holistic approach beyond a merely co-treatment of co-occurrence of mental disorders in hospitalized patients because of primarily medical conditions. ² However, Nakao and Takeuchi ³ noted that the definition of psychosomatic medicine is not consistent across countries.

The scope of psychiatric and psychosomatic interventions as well as the tasks of consultation and liaison service (CLS) enlarged considerably in the two last decades: from merely diagnosing and prescription of psychopharmacological drugs or assessment of mental capacity to a broad scope of tasks. These tasks include counseling for both patients and relatives; multi-professional decision-sharing, especially when mental conditions do influence course of disease, quality of life, or adherence to treatment; decision-making when suspicion of lacking of capacity to consent; universal assessment respectively screening measures when early diagnosis of mental problems become important for further treatment, for example, after suicide attempt, before and after organ transplantation, in oncology centers, in pain management units or palliative care.^4–6

In addition to the increasing tasks, CLS enlarged their settings according to the acknowledgment and the psychosomatic alignment, ⁷ including an on-demand or consultation model, a liaison model with integration in general hospital teams, ⁸ hybrid models in emergency services,^9,10 and support for general practitioners (GPs) as well as provisions of follow-up by outpatient services. Lücke et al. ⁸ demonstrated that “a quasi-liaison model is recommended to be the most suitable and cost-effective way of providing psychiatric care to somatically ill patients with psychiatric comorbidities.” An important aim of CLS is to ensure adherence to psychiatrists’ recommendations after discharge. Burian et al. ¹¹ compared three different communication pathways CLS and GPs, demonstrating that six weeks after discharge, depression symptoms decreased more intensively when the information transference was made by a telephone call. In general, the effectiveness of CLS depends on many structural, functional, transference, and long-term treatment plans after hospital discharge, including adherence to recommendations. Wood and Wand ¹² reviewed 40 articles in order to evaluate the evidence for effectiveness. They grouped the measurements of effectiveness as follows: (i) cost-effectiveness including length of stay; (ii) concordance with management recommendations; (iii) staff and patient feedback; and (iv) follow-up outcome studies. The authors found that “effectiveness of CLS was demonstrated by cost-effectiveness, earlier referrals to CLP predicting shorter length of stay, and concordance with certain management recommendations.”^12,13 Although cost-effectiveness became a prominent legitimating factor for CLS, communicative issues, distress-reducing and confidence-enhancing interventions, multi-professional decision-making, a prompt interpersonal response for existential threats, motivational approaches, and an effective transfer of information to GPs with a realistic middle-term therapeutic plan could be even more important factors from a holistic point of view and on the long run.

The most frequent CLP diagnoses are affective, anxiety, somatoform, and adjustment disorders,^14,15 but also alcohol-related disorders, delirium, dementia, and risk management.^16,17 The distribution of diagnoses varies with the size of general hospital, CLS structure, statutory duties, care priorities, key areas in the general hospital, accreditation issues, professional experience of referring physicians, and depend on CLS competences. According to our research topic, diagnoses distribution, the resources implemented, and treatment time spent depend on the structure and the objectives of the CLS. Only on-demand assessment and treatment of patients electively referred or universal intervention of all admitted individuals for specific patient or diagnostic groups were carried out. Some researchers stress the importance of an accurate referral according to the presence of a psychiatric diagnosis ¹⁸ and screening methods. ¹⁹ Therefore, tools to gather accurately relevant referral information across different care levels ²⁰ as well as for some diagnostic groups like Basic Documentation for Psycho-Oncology for oncologic diseases ²¹ were developed. Nevertheless, prior research of the authors demonstrated the importance of distress, independently of the presence of a mental disorder. ²²

In prior investigations, the authors demonstrated important clinical and care intensity differences between patients showing a current mental disorder and patients without psychiatric diagnosis as well as between electively referred and universally attended patients who were admitted because of nonpsychiatric reasons in a general hospital.^23,24 In this study, the authors reduced the attended population of a hybrid CLS model only to the electively referred patients in order to make the sample more homogeneous according to the referral decision and focused on the kind of medical conditions in order to assess clinical and care differences from a psychosomatic perspective.

Objectives

The aim of this investigation was to compare clinical variables and the intensity of psychiatric care given to patients suffering from different medical conditions who were electively referred to a psychiatric and psychosomatic CLS during a period of five years in a German general hospital with 520 beds. This general aim encompasses three specific objectives:

Comparison of clinical and treatment variables in patients suffering from 11 different diagnoses by means of bivariate tests.

Material and methods

Results

All differences among diagnostic groups were statistically significant (p<0.001) for all measured variables. Patients suffering from cardiovascular diseases are oldest, and women suffering from gynecological diseases are youngest. Children whose mothers were attended to because of relevant psychological strains are 4.6 year old on average. Aside from gender-related diseases, women are overrepresented in all categories, especially pain (64%), with the exception of cardiovascular diseases (50%). Current psychiatric comorbidity amounted to exactly three quarters of the sample; the lowest percentages have been found in patients suffering from cancer (31%). Psychiatric comorbidity prior to admission was assessed in 33% of the sample, in contrast to 75% currently on the occasion of psychiatric assessment by CLS. The GAF score averaged 59.4 (SD = 24.6) and was especially low in patients suffering from unclear medical conditions (M = 39.4; SD = 21.2) and cancer (M = 46.9; SD = 21.2) and high in patients suffering from gynecological diseases and pain as well as mothers of hospitalized children who have shown psychological strain. The average performance status (ECOG) amounted to 1.63 (SD = 1.49) out of a maximum of 4 and was higher in patients suffering from orthopedic conditions (2.62), severe infections (2.09), and cancer (2.09) and was especially low in stressed mothers of hospitalized children and women suffering from gynecological diseases (see Table 1).

OPEN IN VIEWER

Table 1.

Descriptive statistics and differences among the main disease groups according to clinical variables.

	N	Age (N = 2514)M (SD)	Proportion women (N = 2518)(%)	Psychiatric comorbidity prior to hospitalization (N = 2478) (%)	Proportion psychiatric comorbidity (N = 2470) (%)	GAF (N = 1217) M (SD)	ECOG (N = 1780) M (SD)
0. Unclear medical conditions	101	61.4 (21.8)	56%	36%	94%	39.4 (21.2)	1.61 (1.51)
1. Cardiovascular diseases	277	67.7 (14.9)	50%	36%	93%	51.6 (23.1)	1.83 (1.41)
2. Gastrointestinal diseases	173	51.9 (19.9)	61%	40%	87%	58.7 (24.2)	1.42 (1.25)
3. Neurological diseases	174	58.9 (20.6)	58%	41%	93%	49.5 (21.5)	1.70 (1.52)
4. Orthopedic conditions	262	59.8 (22.0)	56%	30%	91%	50.9 (19.7)	2.62 (1.39)
5. Pain	338	60.0 (15.6)	64%	47%	93%	78.8 (15.1)	0.70 (0.93)
6. Gynecological diseases and obstetrics	117	31.2 (11.7)	100%	43%	93%	76.9 (16.5)	0.51 (1.03)
7. Pediatrics (data related to mothers)	74	4.6 (8.3) (age of children)	54% (children)	16%	81%	87.4 (8.2)	0.10 (0.30)
8. Infections	106	61.9 (22.2)	57%	44%	89%	53.6 (23.2)	2.09 (1.52)
9. Other specific physical illnesses	225	57.1 (19.1)	56%	42%	87%	51.2 (24.2)	1.74 (1.58)
10. Cancer	671	64.5 (13.1)	57%	15%	31%	46.9 (21.2)	2.09 (1.32)
Total	2518	58.3 (20.8)	59%	33%	75%	59.4 (24.6)	1.63 (1.49)
DifferencesFor metric variables: ANOVA and SchefféFor dichotomous variables: Chi-square		F = 123; p<0.0011>all except 0,8,106<all adults2>4,5,8,10	Chi2 (10) = 82p<0.001Cramer’s V = 0.18	Chi2 (10) = 162p<0.001Cramer’s V = 0.26	Chi2 (10) = 922p< 0.001Cramer’s V = 0.61	F = 65.7;p<0.0010<2,5,610<2,5,6,75,6,7>all others	F = 45.8;p<0.00110>2,5,6,78>5,6,74>all, except 0,8,106<all, except 5,77<all others

DT: distress thermometer (Range: 0–10); ECOG: Eastern Cooperative Oncology Group Scale; GAF: Global Assessment of Functioning Scale; N: sample size; M: mean; SD: standard deviation; Range: smallest value–highest value; ANOVA: bivariate analysis of variance; b: regression coefficient; p: level of significance; F: value at a F-distribution for variance; Scheffé: post hoc test for unequal variances among diagnoses; Chi²( ): value at a Chi-square distribution with ( ) degrees of freedom.

In a second step, physical illnesses were compared with each other illnesses according to treatment variables related to the supporting spectrum of CLS. For all measured variables, differences were statistically significant (p<0.001). Every patient received on average 2.26 (SD = 2.81) interventions, and 37% of them received two or more interventions (range: 2–47). Patients suffering from pain as the leading diagnosis for treatment decisions had received the most interventions on average (4.49), followed by stressed mothers of hospitalized children (2.93) and cancer patients. Patients had received on average 111 minutes of care (SD = 160), with higher figures for patients suffering from pain (234), stressed mothers of hospitalized children (221), and patients suffering from digestive diseases (118). Less than a third of the sample (30%) received any psychopharmacological intervention, whereas 43% received any psychotherapeutic intervention. A post-discharge recommendation for future psychosocial support was received by 54% of the sample; the proportion was lowest for oncologic patients (see Table 2).

OPEN IN VIEWER

Table 2.

Descriptive statistics and differences among the main disease groups according to treatment variables.

	Number of interventions(N = 2517) M (SD)	Proportion ≥2 contacts (N = 2517) (%)	Total treatment time (min.) (N = 2517) M (SD)	Proportion patients receiving pharmacological treatment (N = 2276) (%)	Proportion patients receiving psychotherapeutic treatment (N = 2280) (%)	Proportion post-treatment recommendation (N = 2271) (%)
0. Unclear medical conditions	1.13 (0.54)	8%	59 (33)	32%	33%	72%
1. Cardiovascular diseases	1.31 (1.15)	16%	72 (76)	46%	34%	67%
2. Gastrointestinal diseases	2.38 (3.33)	34%	118 (160)	34%	50%	60%
3. Neurological diseases	1.31 (0.88)	18%	66 (41)	40%	32%	61%
4. Orthopedic conditions	1.93 (2.29)	36%	93 (113)	44%	35%	58%
5. Pain	4.49 (3.08)	76%	234 (165)	42%	34%	50%
6. Gynecological diseases and obstetrics	1.73 (1.59)	35%	96 (92)	20%	62%	79%
7. Pediatrics (data related to mothers)	2.93 (3.80)	42%	221 (273)	73%	46%	71%
8. Infections	2.08 (2.54)	33%	107 (110)	41%	32%	61%
9. Other specific physical illnesses	1.48 (1.24)	26%	72 (62)	37%	43%	63%
10. Cancer	2.35 (3.47)	41%	93 (209)	13%	52%	36%
Total	2.26 (2.81)	37%	111 (160)	30%	43%	54%
DifferencesFor metric variables: ANOVA and SchefféFor dichotomous variables: Chi-square	F = 35.3; p<0.0015 > all others7,10>0,1,3	Chi2 (10) = 358p < 0.001Cramer’s V = 0.38	F = 34.3; p<0.0015,7>all others	Chi2 (10) = 210p < 0.001Cramer’s V = 0.30	Chi2 (10) = 79.1p < 0.001Cramer’s V = 0.19	Chi2 (10) = 167p < 0.001Cramer’s V = 0.27

N: sample size; M: mean; SD: standard deviation; Range: smallest value–highest value; ANOVA: bivariate analysis of variance; b: regression coefficient; p: level of significance; F: value at a F-distribution for variance; Scheffé: post hoc test for unequal variances among diagnoses; Chi²( ): value at a Chi-square distribution with ( ) degrees of freedom.

Differences among physical illnesses with respect to main psychiatric categories were significant (p < 0.001). The most frequent diagnoses were adjustment and anxiety (psychosomatic) disorders (F4 = 45%), followed by mood disorders (F3 = 22%) and organic mental disorders (F0 = 12%), while a quarter of the sample did not show a current psychiatric diagnosis; 69% of oncologic patients did not show any psychiatric disorder according to ICD-10. Adjustment and anxiety disorders were diagnosed especially in strained mothers of hospitalized children (67%) and patients suffering from pain (65%); organic mental disorders (F0) had been more frequently diagnosed in patients suffering from cardiovascular and neurological diseases (21% each) and unclear medical conditions (20%) (19.2%) (see Table 3).

OPEN IN VIEWER

Table 3.

Distribution of psychiatric diagnoses according to main disease groups (ICD-10, Chapter F).

	N	% Within each medical diagnostic Group 0–10									DifferencesF0 until Z
		Any psychiatric diagnosis	F0	F1	F2	F3	F4	F5	F6	Z-Codes
0. Under medical conditions	100	94.0%	20%	1%	6%	21%	39%	4%	3%	6%
1. Cardiovascular diseases	272	92.7%	21%	3%	3%	26%	38%	0.7%	1%	7%
2. Gastrointestinal diseases	169	87.0%	11%	8%	4%	24%	37%	2%	2%	13%
3. Neurological diseases	170	93.0%	21%	5%	5%	17%	42%	0.6%	2%	7%
4. Orthopedic conditions	255	91.0%	19%	5%	4%	16%	44%	1%	3%	9%	Chi2 (70) = 1.3e3
5. Pain	355	92.5%	1%	1%	0.3%	25%	65%	0%	1%	8%	p<0.001
6. Gynecological diseases and obstetrics	107	92.5%	0%	2%	7%	24%	53%	2%	5%	8%	Cramer’s V = 0.27
7. Pediatrics	73	80.8%	0%	1%	1%	4%	67%	1%	6%	19%
8. Infections	103	89.3%	17%	4%	10%	22%	32%	1.0%	4%	11%
9. Other specific physical illnesses	223	87.4%	9%	12%	4%	26%	33%	2%	3%	13%
10. Cancer	663	31.2%	6%	0.3	2%	8%	15%	0%	0.2%	69%
Total	2470	2470	221	80	68	390	818	20	41	169
Proportion within all psychiatric diagnosis	100%	72.7%	12%	4%	4%	22%	45%	1%	2%	25.3%

For each diagnostic group: distribution of psychiatric diagnoses within the medical diagnostic group.

Psychiatric diagnoses according to International Classification of Diseases, 10th Revision: F0: organic mental disorders; F1: use disorders; F2: schizophrenia; F3: mood disorders; F4: psychosomatic disorders; F5: eating disorders; F6: personality disorders; Z: factors influencing health status.

Chi²( ): value at a Chi-square distribution with ( ) degrees of freedom; p: level of significance.

In multivariate regression as well as logistic analyses, differences among somatic illnesses almost disappear for psychopharmacological and psychotherapeutic treatment as well as post-discharge psychosocial treatment recommendations, whereas differences in care intensity (number of interventions and total treatment time) remained. Therefore, patients suffering from pain, oncologic patients, and those suffering from gastrointestinal, orthopedic, and infectious diseases received more intensive treatment. The main differences have been found for clinical variables, independently of medical condition: older patients and men have obtained less care intensity and fewer psychotherapeutic interventions (older people more often psychopharmacology). On the other hand, more impaired patients (lower functionality levels and performance status as well as presence of psychiatric comorbidity) have obtained more intensive psychological support by CLS (see Table 4).

OPEN IN VIEWER

Table 4.

Multivariate linear and logistic regression models assessing associations between clinical and treatment variables.

	Number interventions			Treatment time			Psycho-pharmacology			Psychotherapy			Post-discharge recommendation
	b	t	p	b	t	p	OR	z	p	OR	z	p	OR	z	p
Age	−0.03	−3.62	<0.001	−1.88	−3.52	<0.001	1.01	3.02	0.003	0.98	−3.13	0.002	0.98	−3.72	<0.001
Gender	−0.76	−3.70	<0.001	−46.1	−3.91	<0.001		n.s.		0.72	−2.41	0.016	0.70	−2.51	0.012
GAF	0.02	3.31	0.001	1.82	4.27	<0.001		n.s.			n.s.			n.s.
ECOG	0.29	2.12	0.034	17.6	2.02	0.043		n.s		1.22	2.49	0.013	0.78	−2.98	0.003
Psychiatric comorbidity	0.83	2.43	0.015	65.2	3.19	0.001	9.90	6.83	<0.001		n.s.		4.79	7.81	<0.001
Physical diseases (Base outcome:unclear medical conditions)
Cardiovascular		n.s.			n.s.			n.s.			n.s.			n.s.
Gastrointestinal	1.91	3.09	0.002	95.7	2.99	0.003		n.s.			n.s.			n.s.
Neurological		n.s.			n.s.			n.s.			n.s.			n.s.
Orthopedic	1.26	3.23	0.001	47.3	2.35	0.019		n.s.			n.s.		0.15	−3.91	<0.001
Pain	3.61	12.0	<0.001	165	9.69	<0.001	2.60	2.09	0.036		n.s.			n.s.
Gynecological and obstetrics		n.s.			n.s.			n.s.			n.s.			n.s.
Pediatrics		n.s.			n.s.			n.s.			n.s.		0.22	−2.60	0.009
Infections	0.94	2.47	0.014	50.8	2.41	0.016		n.s.			n.s.			n.s.
Others pecific		n.s.			n.s.			n.s.			n.s.			n.s.
Cancer (electively referred)	2.53	6.03	<0.001	105	4.13	<0.001		n.s.			n.s.			n.s.
Constant	0.82	1.31	0.192	27.3	0.77	0.440	0.03	−4.94	<0.001	0.88	−0.21	0.836	2.59	−1.12	0.262
N	1064			1065			1061			1063			1058
F/prob>F	29.0/<0.001			30.0/<0.001			191/<0.001			46.8/<0.001			190/<0.001
R2/pseudo R2	0.17			0.16			0.14			0.03			0.13
Power (1−β post hoc, f2 = 0.15)	1.00			1.00			0.96			0.96			0.96

GAF: Global Assessment of Functioning; ECOG: Eastern Cooperative Oncology Group Scale; Gender: women = 0; men = 1; psychiatric diagnosis: 0 = not given; 1 = given; b: regression coefficient; p: level of significance; N: sample size; F: value at a F-distribution for variance; R²: multiple correlation coefficient (determination coefficient); OR: odds ratio, similar to relative risk; t and z: value at normalized t and z distributions; LR: likelihood ratio; n.s.: not significant at 0.005 level.

Discussion

The first most important result of this investigation was the small differences in psychological support by a CLS among electively referred patients according to medical illnesses with the exception of pain and cancer. Patients suffering from pain have more frequently shown adjustment disorders and have received more intensive care by CLS, and cancer patients suffered less frequently from psychiatric comorbidity. After all, 25% of the sample did not suffer from any psychiatric comorbidity. The second most important finding consists of the importance of general clinical variables as predictors for care intensity held by CLS; therefore, less functional, more impaired, and comorbid patients received much more care efforts by psychiatric CLS.

The assessed CLS treated approximately 3.4% of all admissions in the surveyed general hospital, which is an adequate level considering the international recommendations for CLS quotas. ²⁷ However, electively referred patients received 2.26 consultations on average and a total of 111 minutes treatment, which is less than the recommended 3.9 interventions and 2.6 hours total treatment time proposed by Holmes et al. ²⁸ This gap was due to the staffing level in the CLS, which was about 0.23 full-time equivalent clinicians per 100 beds rather than the recommended international standard of 1.0 full-time equivalent per 100 beds. This understaffing surely influences decisions under pressure about allocation of resources.

It is remarkable that even electively referred oncology patients have less frequently displayed a psychiatric comorbidity than patients suffering from other physical illnesses. This could indicate more awareness of the psychological support needs of cancer patients and could reflect care priorities in Western societies. The lower functionality and performance status as well as the presence of minor psychiatric disorders could be more important in oncologic patients than full psychiatric comorbidity. On the contrary, organic mental disorders appear to be important in diseases with an influence on brain physiology such as unclear medical conditions, infections, neurological, and cardiovascular diseases. Overall, adjustment disorders become more important than full mood disorders in CLP, especially concerning stressed mothers of hospitalized children, patients with pain as the main reason for treatment, and women with gynecological and obstetric conditions, who more frequently received psychotherapeutic interventions. These tendencies reflect priorities of assessed CLS, but also levels of distress as well as needs of the mentioned groups, and finally focal points of care in hospitals generally.

The fact that pain patients received significantly more contacts and treatment time but less frequent psychotherapeutic interventions is probably due to the implemented standardized psycho-educative program tailored for this group, which is not considered as psychotherapeutic intervention in the narrow sense explained in the Methods section above. Their higher levels of psychiatric comorbidity are due to adjustment disorders and indicate a higher level of psychological strain, probably because of a somatoform component that is hard to differentiate from strictly biological processes.

Special attention has to be paid to distressed mothers of hospitalized children. They do not suffer from a classical psychiatric disease and show a high functionality and a very good performance status, but they need above average number of contacts and treatment time, due to aroused distress during hospitalization of children (65% develop a depressive or adjustment disorder). A closer collaboration between psychosomatics and pediatrics as well as obstetrics could be derived from these results.

The gap between the proportion of patients who had a psychiatric diagnosis prior to their current hospitalization and the current proportions of psychiatric comorbidity (a difference of around 42%) indicates that hospitalization itself could trigger psychiatric comorbidity or identify hitherto undiagnosed psychiatric conditions. Nevertheless, the mental disorders found are not severe. They include mood, anxiety, and adjustment disorders, apart from organic mental disorders in specific medical conditions already discussed above as well as in a previous research. ²⁴ Hence, psychiatric comorbidity in general hospitals is mainly subthreshold, minor, or psychosomatic disorders rather than major disorders, which have to be taken into consideration in the allocation of treatment resources and care planning.^29,30 It can be considered good practice that patients who display psychiatric comorbidity, lower functionality, and low performance status received more intensive support by the CLS.

It is important to note that psychiatric support by CLS has been negatively associated with age and with male gender; this inequality has to be explained carefully, taking into consideration the expectations and real needs of these groups of patients as well as assumptions of the public and professionals about the allocation of resources depending on age, gender, and family status.^31,32 It is possible that psychogeriatric, nurse-based, or family-centered interventions are more effective for older patients than classical face-to-face psychotherapeutic interventions and that men benefit more from informal counseling or from family interventions. ²²

Finally, the importance of referral decisions has to be pointed out. The presence of a positive mental diagnosis is not necessarily the definite criterion for an accurate referral decision, but the patient’s distress, needs, and systemic factors, considering the limited resources available. Chen et al. ³³ reviewed 35 eligible articles on barriers for referral and grouped the reasons in three different categories: referrer (i.e. to be comfortable with CLS and to work in internal medicine units); patient (young, living in an urban setting, having a psychiatric history, especially a functional psychosis); and systemic factors (provision of a specialized CLS, active CLS consultant, collaborative screening of patients). Furthermore, concordance with CLS’ recommendations and ensuring of an adequate follow-up are of major relevance for prognosis; Chen et al. ³⁴ found that three-year all-cause mortality among suicide attempters in CLS was 20.4% with associate risks like older age, depressive disorders, and lack of psychiatric follow-up. In this light, improvement of outpatient services and implementation of telemedicine in CLS when resources are scarce ³⁵ could better meet patients’ care needs.

Clinical implications and limitations

This investigation analyzes the treatment provided by a CLS in a middle-sized general hospital on a naturalistic basis located in a rather rural region in southern Germany. The sample consists of only electively referred patients with psychosocial strains because of diagnostic and/or treatment reasons. Eleven main physical disease groups were identified and compared according to clinical and treatment variables. No relevant differences were found for assessed diseases with the exception of pain and cancer patients. More relevant were the differences according to psychiatric comorbidity, age, and gender, independently of medical conditions considered. These results have to be taken into consideration in referral and multi-professional care planning as well as in allocation of resources and implemented therapeutic approaches.

Finally, we summarize the most important limitations of this study: (i) this investigation was naturalistic and not a hypothesis-oriented investigation; (ii) this investigation reflects real treatment efforts but not necessarily the care needs of the patients surveyed; (iii) for some variables, there are important missing data, but the statistical program performed a list-wise deletion of missing data automatically in regression or logistic models; and (iv) the present study deals only with referred patients and not with all admissions in a period of time according to definite diagnoses; therefore, we cannot make claims in general for hospitalized patients suffering from certain diagnoses, but for the group of patients electively referred to a CLS.