Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

Sophia Stahl-Toyota; Christoph Nikendei; Ede Nagy; Stefan Bönsel; Ivo Rollmann; Inga Unger; Julia Szendrödi; Norbert Frey; Patrick Michl; Carsten Müller-Tidow; Dirk Jäger; Hans-Christoph Friederich; Achim Hochlehnert

doi:10.1371/journal.pone.0287234

. 2023 Jun 22;18(6):e0287234. doi: 10.1371/journal.pone.0287234

Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

Sophia Stahl-Toyota ^1,^*, Christoph Nikendei ¹, Ede Nagy ¹, Stefan Bönsel ², Ivo Rollmann ¹, Inga Unger ³, Julia Szendrödi ⁴, Norbert Frey ⁵, Patrick Michl ⁶, Carsten Müller-Tidow ⁷, Dirk Jäger ⁸, Hans-Christoph Friederich ¹, Achim Hochlehnert ²

Editor: Sebastien Kenmoe⁹

¹Department of General Internal Medicine and Psychosomatics, Medical University Hospital, Heidelberg, Germany

²Department of Medicine Controlling, Medical University Hospital, Heidelberg, Germany

³Nursing Management, Department of Internal Medicine, Medical University Hospital, Heidelberg, Germany

⁴Department of Endocrinology and Clinical Chemistry, Medical University Hospital, Heidelberg, Germany

⁵Department of Cardiology, Angiology and Pneumology, Medical University Hospital, Heidelberg, Germany

⁶Department of Gastroenterology, Hepatology and Infectious Diseases, Medical University Hospital, Heidelberg, Germany

⁷Department of Hematology, Oncology and Rheumatology, Medical University Hospital, Heidelberg, Germany

⁸Department of Medical Oncology, Medical University Hospital, National Center for Tumor Diseases, Heidelberg, Germany

⁹University of Buea, CAMEROON

Competing Interests: The authors have declared that no competing interests exist.

^✉

* E-mail: sophia.stahl-toyota@med.uni-heidelberg.de

Roles

Sophia Stahl-Toyota: Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing – original draft

Christoph Nikendei: Conceptualization, Project administration, Resources, Supervision, Writing – review & editing

Ede Nagy: Formal analysis, Methodology, Validation, Writing – review & editing

Stefan Bönsel: Data curation, Investigation, Resources, Writing – review & editing

Ivo Rollmann: Conceptualization, Formal analysis, Methodology, Writing – review & editing

Inga Unger: Resources, Writing – review & editing

Julia Szendrödi: Resources, Writing – review & editing

Norbert Frey: Resources, Writing – review & editing

Patrick Michl: Resources, Writing – review & editing

Carsten Müller-Tidow: Resources, Writing – review & editing

Dirk Jäger: Resources, Writing – review & editing

Hans-Christoph Friederich: Conceptualization, Funding acquisition, Resources, Supervision, Writing – review & editing

Achim Hochlehnert: Conceptualization, Data curation, Investigation, Writing – original draft

Sebastien Kenmoe: Editor

PMCID: PMC10287009 PMID: 37347745

Abstract

Background

Mental comorbidities of physically ill patients lead to higher morbidity, mortality, health-care utilization and costs.

Objective

The aim of the study was to investigate the impact of mental comorbidity and physical multimorbidity on the length-of-stay in medical inpatients at a maximum-care university hospital.

Design

The study follows a retrospective, quantitative cross-sectional analysis approach to investigate mental comorbidity and physical multimorbidity in internal medicine patients.

Patients

The study comprised a total of n = 28.553 inpatients treated in 2017, 2018 and 2019 at a German Medical University Hospital.

Main measures

Inpatients with a mental comorbidity showed a median length-of-stay of eight days that was two days longer compared to inpatients without a mental comorbidity. Neurotic and somatoform disorders (ICD-10 F4), behavioral syndromes (F5) and organic disorders (F0) were leading with respect to length-of-stay, followed by affective disorders (F3), schizophrenia and delusional disorders (F2), and substance use (F1), all above the sample mean length-of-stay. The impact of mental comorbidity on length-of-stay was greatest for middle-aged patients. Mental comorbidity and Elixhauser score as a measure for physical multimorbidity showed a significant interaction effect indicating that the impact of mental comorbidity on length-of-stay was greater in patients with higher Elixhauser scores.

Conclusions

The findings provide new insights in medical inpatients how mental comorbidity and physical multimorbidity interact with respect to length-of-stay. Mental comorbidity had a large effect on length-of-stay, especially in patients with high levels of physical multimorbidity. Thus, there is an urgent need for new service models to especially care for multimorbid inpatients with mental comorbidity.

Introduction

The analysis of the influence of mental comorbidity and physical multimorbidity in hospitalized patients on health-economic parameters is of great interest in health care systems [1].

The course and prognoses of physical diseases is determined decisively by the concurrent presence of mental comorbidities, such as depression or anxiety disorder [2, 3]. In addition, physical diseases are often accompanied by a pronounced psychosocial strain [4]. Regarding socio-economic parameters, it is well known that mental comorbidities of patients with physical diseases lead to prolonged length-of-stay in hospitals [5–7], higher morbidity and mortality rates [2–5], as well as lower quality of life [6]. It is therefore an imperative that comorbid mental disorders are diagnosed in somatic hospitals and treated in a timely manner. Cardiovascular patients with mental comorbidity, for example, show a significant increase of the average length-of-stay from 8.9 (± 0.3) to 13.2 (± 0.7) days [8]. The physical-mental interplay deteriorates dramatically the physical health condition associated with significant higher hospital costs. However, regarding the insurance payment systems, extra resources engaged in such cases are not adequately represented [8–10].

Besides potential mental comorbidities, the presence of additional physical conditions in somatically ill patients also shows a negative influence on the length-of-stay at the hospital [11, 12]. For in-patients the prevalence of a multimorbidity status, defined as the presence of at least two chronic conditions by the World-Health-Organisation, is about 80% [13, 14]. There do exist further definitions of multimorbidity, that are relevant to address the present research question. Aubert et al. have investigated eight different definitions for multimorbidity, which all had a moderate to medium separative power in their ability to predict 30-day-hospital-readmission and prolonged length-of-stay [15]. Among others, they defined multimorbidity using the Elixhauser-van Walraven comorbidity index and the absolute number of conditions. Mueller et al. described an increased length-of-stay of 4.7 (± 10.7) days for multimorbid cases in comparison to non multimorbid cases [14].

Taken together, these findings emphasize that the incorporation of physical multimorbidity and mental comorbidity is important in models of socio-economic target parameters such as length-of-stay. A recent review on length-of-stay prediction lists only few studies that actually take comorbidities into account [16]. And those current studies that do discuss length-of-stay in the context of multimorbidity, should be interpreted with caution due to potential contortion of results due to the fact that physical multimorbidity and mental comorbidity might interact with each other [1]. Thus the relationship between mental comorbidity and length-of-stay might be changing depending on the level of physical multimorbidity. This is of great importance for a more realistic and valid prediction of the influence of mental comorbidity on the length-of-stay in medical inpatients. In addition, a limiting factor can be seen in the restricted sample sizes [8].

Therefore, the aim of the presented study was to investigate the influence of (1) mental comorbidity, (2) physical multimorbidity and (3) the interaction of physical multimorbidity and mental comorbidity on the length-of-stay using a large database of a center for internal medicine at a German university hospital. The University clinic of Heidelberg belongs to the publicly funded hospitals, which is the group, next to charitable and private hospitals, that provides nearly 50% of all the available beds in Germany [17]. University hospitals treat around 2 million patients annually in a stationary setting [18]. Another way to classify hospitals is by four levels of care, ranging from basic to regular, specialized and maximum-care. University clinics usually cover all medical disciplines and are classified as maximum-care. Concerning mental comorbidities in medically ill patients, the “consultation-liasion services” by psychiatric and psychosomatic specialists is the model offered most commonly to patients in the medical departments throughout Germany [19]. According to the typology by Kathol et al., the cohort described here was treated on a Type II medical consultation-liaison unit for low mental and medium to high medical acuity [20].

The analyses presented here may help to allocate resources more adequately at medical hospitals and to provide early psychosocial interventions for patients with additional needs. Moreover, it could be imagined that a future predictive model estimates the expected length-of-stay at the beginning of a hospital stay and supports clinical decisions for early interventions, such as a proactive psychosomatic and psychiatric consultation service. Since resources are limited, targeted interventions should be possible that are of economic and medical importance by early identification of especially critical cases.

Materials and methods

Study design and participants

The study comprised a retrospective data analysis of all inpatient cases of the years 2017, 2018 and 2019 at the Center of Internal Medicine of the University Hospital in Heidelberg, Germany. After data preprocessing and removal of one patient for whom the gender was unknown, N = 28,553 cases had no missing values for the variables of interest and met the inclusion criteria, which included cases of 20,193 patients who (1) were at least 18 years old, (2) were admitted to one of the internal medicine units, (3) had no main diagnosis of the ICD-10 (International Statistical Classification Of Diseases And Related Health Problems, 10th revision, German Modification) code chapter V for psychiatric diseases, (4) and had a length-of-stay of at least two days. These criteria were selected so that only stationary patients who stayed at least one night and who were admitted with a primarily somatic main diagnosis were included. The source of the data were the medical records that entailed use of resources via diagnostic and therapeutic effort, so only diagnostic ICD-10 codes that were relevant to the respective hospital stay were available. Data preprocessing involved merging cases that were direct follow-up admissions. Hospital stays that were medically associated, e.g. due to complications, and occurred within a 30 days interval were thus counted as a single case.

Measures

Outcome length-of-stay

The outcome variable of interest was the length-of-stay. It was extracted from clinical routine data directly, representing the number of days within one case that the patient stayed at the hospital.

Physical multimorbidity

For representing physical multimorbidity, the Elixhauser score was retrieved via the R package comorbidity scores for all comorbidity diagnoses, not including the main diagnosis [21]. As Bartlett et al. [22] did, the Elixhauser score was modified to exclude the four groups related to mental diseases (psychoses, depression, drug and alcohol abuse) to separate physical from mental comorbidities, leaving 27 of the 31 groups represented by the score.

Mental comorbidity

We defined mental comorbidity according to Wolff et al. [23] as any secondary diagnosis code from Chapter V (F0-F9) of the ICD-10. Several previous studies omitted codes representing diseases such as dementia, delirium or nicotine abuse [8, 11, 23, 24]. However, since there was no consensus on the reasons for omitting particular diagnosis codes, all codes were kept in the definition for this analysis. The mental comorbidity count was the number of ICD-10 codes that fulfilled the above definition for mental comorbidity.

Other variables

As potential confounders, the patients’ gender, age at hospitalization and main diagnosis ICD-10 chapter were included in the analysis.

Statistical analyses

To describe the characteristics of the study population, we report the mean, standard deviation, median, and range for continuous variables. For categorical variables, the absolute number and the ratio of each value is presented. All characteristics were computed for the total population as well as for two groups separated by presence or absence of mental comorbidity. To evaluate the difference of the variable distributions between these groups, the Pearson’s chi-squared test was computed for categorical variables and the Mann-Whitney-U test for numerical variables. In addition, each variable’s correlation with the outcome parameter length-of-stay was computed as either the robust correlation coefficient with percentage bend (r) or correlation ratio (η) for exploratory data analysis to aid in the model selection process [25].

For the analysis of length-of-stay differentiated by mental comorbidity count, F-category of mental comorbidity, age groups with and without mental comorbidity, and increasing Elixhauser score, the mean and 95% confidence interval in the respective subgroups were computed and plotted. To further assess the relationship between the explanatory variables with length-of-stay, we realized single variate as well as multivariate analysis to analyse main and interaction effects with negative binomial regression. Since this type of model has a parameter to control for overdispersion, it is well-suited for the characteristic long-tailed distribution of length-of-stay [26, 27]. As in those studies, the analysis was performed on the basis of hospitalizations, so that one patient may be represented by several cases. Model performance was evaluated mainly based on the Akaike Information Criterion (AIC), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) [28, 29]. The derivation of the model with R code is portrayed in S1 File. The choice of methods put a high focus on intuitive interpretability, as reliable reasoning is essential for trusting the results [16].

The data preparation and characteristics description was implemented in Python 3.11 (scipy 1.8.0 [30], pingouin 0.5.0 for statistical tests [31]). All remaining steps were computed with R 4.1.2 (comorbidity 1.0.0 [21] for Elixhauser score determination, mass 7.3.54 [32] for negative binomial regression).

Ethical approval

This study was approved by the Ethics committee of the Medical faculty of the University of Heidelberg (No. S-690/2021). As only clinical routine data were used, the need for consent was waived by the Ethics committee.

Results

Characteristics of study participants

The characteristics of the 28.553 cases that met the inclusion criteria are displayed in Table 1. Out of all cases, 15.2% were diagnosed with at least one mental comorbidity. These patients with mental comorbidity are younger and have significantly higher Elixhauser scores. The mean and median length-of-stay in the different main diagnosis ICD10 chapter groups are included in S1 Table. The correlation ratio for the main diagnosis chapter and length-of-stay was η = 0.19, and for gender η = 0.002.

Table 1. Characteristics of the sample of internal medicine inpatients with and without mental comorbidity.

Variable	Measure	Total	Without mental comorbidity	With mental comorbity
Cases	N (%)	28553 (100%)	24225 (85%)	4328 (15%)
Gender	female (%)	11334 (39.7%)	9540 (39.4%)	1794 (41.5%)
Age at hospitalization (y)	mean [SD]	64.0 [16.9]	64.2 [16.9]	62.9 [17.2]
	median	66	67	64
	min-max	18–108	18–108	18–98
Physical comorbidities Elixhauser score	mean [SD]	2.8 [2.0]	2.7 [2.0]	3.4 [2.2]
	median	3	2	3
	min-max	0–15	0–13	0–15
Mental comorbidity count	mean [SD]	0.2 [0.5]	0.0 [0.0]	1.3 [0.7]
	median	0	0	1
	min-max	0–7	0–0	1–7
Length-of-stay (d)	mean [SD]	9.8 [13.6]	8.8 [9.8]	15.2 [25.3]
	median	6	6	8
	min-max	2–463	2–274	2–463
	90th percentile	20	19	32

Open in a new tab

SD: standard deviation, CI95: 95% confidence interval, 90^th percentile with interpolation = ‘nearest’.

Length-of-stay with vs. without mental comorbidity

The most frequent mental diagnoses belonged to the categories F0 organic diseases (n = 1394), F1 substance use (n = 1319), F4 neurotic, stress and somatoform disorders (n = 1042) and F3 affective disorders (n = 879). The number of cases in the other F-categories were less than 200 cases per category.

Mean length-of-stay of patients without a mental comorbidity was 8.8 days (95% confidence interval, CI95 = [8.7, 8.9]; median = 6), while mean length-of-stay of patients with one or more mental comorbidities was 15.2 days (CI95 = [14.4, 15.9]; median = 8). Thus, on average patients with mental comorbidities stayed 6.4 days longer. The difference of the median was 2 days. The common language effect size, for which deviation from 0.5 expresses significance, was f = 0.4.

Length-of-stay by number of mental comorbidities

Fig 1 shows a more detailed view of the relationship between length-of-stay and number of mental comorbidities (correlation coefficent r = 0.09, CI95 = [0.08, 0.1]). With increasing number of mental comorbidities, the average length-of-stay increased. Out of all n = 4328 cases with mental comorbidities, the largest group (n = 3411) with one mental comorbidity diagnosis had an average length-of-stay of 13.8 days, which is 5.0 days longer than cases without any mental comorbidity and without overlapping confidence intervals.

Length of stay differentiated for the F-category of mental comorbidity

The length-of-stay of patients with mental comoribities is further differentiated by the particular type of mental comorbidity in Fig 2. Patients with neurotic disorders (F4) had the highest mean length-of-stay with 23.01 days. Patients with behavioral syndromes (F5) stayed on average 19.28 days, but the group was small with 78 cases and had high variance, followed by organic disorders (F0) with 18.94 days and affective disorders (F3) with 14.97 days on average. Patients with substance use (F1) had rather short length-of-stay with on average 10.93 days. This was slightly above the sample mean of 9.8 days.

Length-of-stay for different age groups with and without mental comorbidity

Fig 3 displays the mean length-of-stay for different age groups with and without mental comorbidity, S2 Table shows the corresponding numerical values. There was a negative correlation (r = -0.05, CI95 = [-0.06, -0.04]) of age at hospitalization with length-of-stay. The ratio of patients with mental comorbidity was highest in the age groups that encompass 40 to 59 years at hospitalization (18%). Length-of-stay without mental comorbidity was around the population mean (9.8 days) for all age groups (group mean ranging from 7.03 to 10.19 days). Mean length-of-stay for cases with mental comorbidity ranged from 7.58 to 20 days and was higher within each age group than for those without mental comorbidity. Mean length-of-stay was highest for middle-aged groups with mental comorbidity: patients in their thirties (mean 20 days, 11 days longer than without mental comorbidity), were followed by those in their fifties (mean 19.12, 9 days longer than without mental comorbidity).

Length-of-stay in relation to physical multimorbidity with and without mental comorbidity

Fig 4 shows the relationship between physical multimorbidty, represented by the Elixhauser score, and length-of-stay separately for cases with and without mental comorbidty; S3 Table shows the corresponding numerical values. The overall correlation coefficient for the variable Elixhauser score was r = 0.17 (CI95 = [0.16, 0.18]). Increasing Elixhauser score without mental comorbidity was associated with moderate increase in length-of-stay. Additional mental comorbidity was associated with stronger increases in length-of-stay dependent on the level of physical multimorbidity. The lines of mental comorbidity and Elixhauser Score in Fig 4 cross at about Elixhauser score 1 indicating that mental comorbidity has an increasingly additional effect on length-of-stay starting from an Elixhauser score above 1. This relationship indicates an interaction between mental comorbidity and physical multimorbidity and was therefore included as an interaction term in the regression model.

Negative binomial regression model with Elixhauser score and mental comorbidity presence

Table 2 displays the results of negative binomial regression analysis for length-of-stay. According to multivariate analysis, each additional year of age at hospitalization reduced the average length-of-stay by 0.4%. Presence of mental comorbidity increased average length-of-stay by 13.1%. An increase in Elixhauser score (which may cover many comorbidity codes, so not directly comparable to mental comorbidity presence) increased average length-of-stay by 13.8%. In interaction with a mental comorbidity, it increased by an additional 8.5%. This interaction between presence of mental comorbidity and Elixhauser score was significant. S2 File is a case simulation tool where individual predictions can be simulated and the contribution of each variable to the prediction can be retraced according to the model’s coefficients as derived in S3 File.

Table 2. Negative binomial regression analysis of length-of-stay.

	Univariate analysis			Multivariate analysis
Variable	IRR	CI95	p-value	IRR	CI95	p-value
Gender
Male	/	/	/	/	/	/
Female	1.005	0.985–1.026	0.62	1.019	1–1.038	0.05
Age at hospitalization	0.994	0.994–0.995	< = 0.001***	0.996	0.995–0.997	< = 0.001***
Mental comorbidity present	1.722	1.676–1.77	< = 0.001***	1.131	1.081–1.183	< = 0.001***
Somatic comorbidities Elixhauser score	1.11	1.105–1.116	< = 0.001***	1.138	1.131–1.144	< = 0.001***
Main Diagnosis ICD-10 chapter
I Infectious and parasitic diseases	/	/	/	/	/	/
II Neoplasms	1.377	1.311–1.446	< = 0.001***	1.486	1.42–1.554	< = 0.001***
XIV Genitourinary system	1.179	1.1–1.264	< = 0.001***	1.132	1.063–1.207	< = 0.001***
XIX Pregnancy, childbirth and puerperium	1.218	1.132–1.31	< = 0.001***	1.053	0.985–1.126	0.13
VI Nervous system	1.081	0.902–1.305	0.41	1.014	0.859–1.204	0.87
XIII Musculoskeletal and connective tissue	0.886	0.803–0.978	< = 0.05*	0.993	0.907–1.088	0.88
XI Digestive system	0.946	0.899–0.995	< = 0.05*	0.905	0.864–0.947	< = 0.001***
IV Endocrine, nutritional and metabolic dis.	0.879	0.827–0.934	< = 0.001***	0.888	0.84–0.939	< = 0.001***
VII Eye and adnexa	0.822	0.47–1.558	0.52	0.858	0.512–1.53	0.58
III Blood and immune mechanisms	0.852	0.763–0.953	< = 0.01**	0.856	0.773–0.949	< = 0.01**
XII Skin and subcutaneous tissue	0.825	0.651–1.059	0.12	0.825	0.662–1.038	0.09
X Respiratory system	0.797	0.748–0.849	< = 0.001***	0.78	0.735–0.826	< = 0.001***
XV Origin in perinatal period	0.769	0.543–1.121	0.15	0.702	0.507–0.991	< = 0.05*
VIII Ear and mastoid process	0.574	0.331–1.062	0.06	0.68	0.406–1.193	0.15
IX Circulatory system	0.748	0.715–0.783	< = 0.001***	0.633	0.606–0.66	< = 0.001***
XVII Findings not elsewhere classified	0.633	0.515–0.785	< = 0.001***	0.59	0.487–0.72	< = 0.001***
XVIII Injury, poisoning	0.578	0.534–0.626	< = 0.001***	0.588	0.547–0.634	< = 0.001***
XXI Factors influencing health status	0.412	0.37–0.458	< = 0.001***	0.391	0.354–0.432	< = 0.001***
Interaction of Mental comorbidity present and Somatic comorbidities Elixhauser score				1.085	1.073–1.097	< = 0.001***

Open in a new tab

IRR: incidence rate ratio (represents change in length-of-stay in terms of percentage, as determined by distance from 1, per unit increase for continuous variables and per presence of category for categorical variables); CI95: 95% confidence interval. Note that several main diagnosis groups encompass less than 100 cases, refer to S2 Table for exact case counts.

Discussion

The present study underlines the importance of mental comorbidity on health-economic outcomes such as length-of-stay. Besides organic diseases (F0), substance use (F1), neurotic, stress and somatoform disorders (F4) and affective disorders (F3) were the most frequent comorbid mental diagnoses at a maximum-care hospital of internal medicine. The greatest length-of-stay in the present sample was observed for middle-aged medical inpatients (30–59 years) and patients with neurotic, stress and somatoform disorders (F4). The main finding of the present study is that the influence of mental comorbidity on the length-of-stay interacts with the level of physical multimorbidity.

We found that the mean length-of-stay of patients with mental comorbidity was 6.4 days longer than the length-of-stay of patients without mental comorbidities. The median differed by 2 days. These findings match a review by Jansen et al., who reviewed studies that calculated the difference in length-of-stay with and without mental comorbidity [1]. Across 20 studies, the mean difference between controls and the group with mental comorbidities was on average 8.9 days (SD = 13.6). The median of the mean differences was 5 days. The individual studies showed great heterogeneity. The median among the mean length-of-stay reported by the 20 reviewed studies was 13.9 days with and 9.2 days without mental comorbidity, which is close to the mean length-of-stay of 15.2 and 8.8 days found in our study for the two groups, respectively. Another study by Beeler et al. showed that an additional non-depression diagnosis was independently associated with an increased length-of-stay by 10% and an ancillary depression even by 24% [33]. The analysis, however, excluded several mental comorbidities and the prevalence of depression was 4.9% [33].

These previous studies have not considered the physical-mental interplay with respect to length-of-stay, which have led to distorted results. For an average medical inpatient with an Elixhauser score of 2, our model associated the addition of mental comorbidity with a 3.0 days extension length-of-stay. When the Elixhauser score was 7, the length-of-stay difference between a patient with compared to without mental comorbidity was 17.2 days (see S2 File for interactive individual case simulations and the derivation of these numbers). Thus, our study provides evidence from a large sample of inpatients at a maximum-care university hospital that mental comorbidity predicts length-of-stay dependent on physical multimorbidity.

There was a tendency that younger patients had greater length-of-stay, with each additional year reducing the length-of-stay by 0.04% in the multivariate analysis, which is in accordance with previous studies [26, 27]. It is notable that the gap with respect to length-of-stay between the group with mental comorbidity in comparison to those without mental comorbidity is highest in middle-aged inpatients (30–59 years). This finding may also be characteristic for patients at a maximum-care hospital compared to a basic and regular care hospital with less elderly and frail patients suffering from organic diseases (F0). However, the findings also underline the impact of mental comorbidity on length-of-stay in patients younger than 60 years. These patients show greater psychosocial vulnerability to physical morbidity, as the consequences of illness appear in critical phases of career planing, starting a family and partnership. Higher levels of suffering in 30–59 year old patients may exceed own ressources resulting in a cascade of dysfunctions that contribute to a greater length-of-stay.

Regarding the mental comorbidity subcategories, we would have expected that especially organic disorders (F0) such as dementia and delirium in addition to physical comorbidities would be the main comorbidity-related factors increasing length-of-stay. Our study, however, showed that cases with neurotic, stress-related and somatoform disorders (F4) had the highest length-of-stay. This ICD-10-chapter, however, includes several anxiety disorders like phobic, panic, and obsessive-compulsive disorders as well as disease processing and somatoform disorders that are a domain for psychotherapy.

The fact that different kinds of depression, which are summarized in ICD-10-chapter F3, are associated with a longer length-of-stay, as reported previously [33, 34], is corroborated by the present study, although the impact on length-of-stay was lower compared to mental diseases from ICD-10 chapters F0, F4, F5.

The present findings have several clincial implications. First of all, there is a great need to develop new concepts of integrated care for medical inpatients with mental comorbidity and asscociated complex care needs. Mental comorbidity has a prominent effect on the length-of-stay, especially in multimorbid inpatients. The interaction between mental comorbidity presence and Elixhauser score can be interpreted in the sense that for patients with mental comorbidity, the effect of increasing levels of multimorbidity on length-of-stay was stronger. This pertains only to the subgroup of patients with mental comorbidity, which in this cohort was 15%. The true ratio, however, is expected to be higher if systematic screening for mental comorbidities would be performed. Therefore, it is necessary to diagnose and treat mental comorbidity especially in physically multimorbid patients upon admission to hospital and along inpatient treatment. Knowing which comorbidities have the strongest effect on length-of-stay can help to allocate resources to address the specific needs of these patients early on.

Consultation-liasion services have the limitation that they traditionally become active only on request. Given the prevalence of mental comorbidity of about 35% in medical inpatients [6], the present findings of 15% show that physicans in internal medicine identify less than half of the patients with mental comorbidity and even less are seen by the psychosomatic or psychiatric consultation-liasion services. To address these limitations, for inpatients with mental comorbidity, a systematic screening for mental comorbidity of all patients should be implemented. Furthermore, a proactive consultation is necessary, so that more patients receive the complex care they need. Alternatively, the psychosomatic or psychiatrist should be part of the ward team. This means he or she is actively involved in patients’ ongoing inpatient care in the sense of a liaison service model. The latter expands and improves clincial care to a bio-psychosocial care model. This suggestion is similar to a behavioral intervention team that was shown to be a promising way of decreasing length-of-stay in general medical units [35] as well as the Proactive Integrated Consultation-Liaison Psychiatry (PICLP) for which effectiveness is currently under evaluation [36].

As for most studies on length-of-stay prediction [16], a limitation is that the generalizability to other settings may be questionable. The sample was limited to the Center of Internal Medicine, but more applicable conclusions could be drawn if data for the entire hospital or even other hospitals were included. S1 Table shows that the entire ICD-10 spectrum of internal disases was covered by the cohort, however, the cases might not be representative for all hospitals in Germany, as our data represents a maximum-care university hospital.

Even though the variables employed in this study, age, gender, primary diagnosis and comorbidity, are among the key features that appear in multiple length-of-stay studies listed by a recent review [16], information of potential confounders such as disease severity as well as social and economic status of the patients was missing.

Furthermore, the prevalence of mental comorbidity was lower than the previously reported 35% for a smaller cohort at the same hospital [6], indicating that approximately half of the mental comorbidities may not have been identified by the specialists in internal medicine. This may have influenced the findings as certain diagnoses were more easily detected by physicians.

In order to reach more clinically relevant conclusions, the model would also need to differentiate different subcategories of physical comorbidities and incorporate the different subcategories of mental comorbidities as their severity is not equal and impact on length-of-stay can be presumed to vary according to the descriptive results presented here.

All relevant diagnoses for the hospital stay counted equally towards the analysis, not distinguishing between diagnoses placed at the beginning or end of the stay, allowing no differentiation between preexisting or a newly developed mental comorbidity. In addition, diagnoses from the medical history of the patient were not available, as only those codes that were available in the medical records of the current hospital stay were included in this data set.

Although directly connected cases were merged and the remaining multiple hospitalizations for the same patient were due to independent admission reasons, the observations are not entirely independent of each other, since the same patients were evaluated more than once. Even though this kind of analysis on a hospitalization-level is consistent with similar studies [26, 27], future work should take the patient-level information into account to address this potential bias. Also a multi-level model with the hospital departments as clusters may be considered for model improvement, as their mean lengths-of-stay varied (S4 Table).

Besides improving the model by adapting the choice of features, the model building and evaluation process itself could also be performed with less risk of overfitting. Instead of one single iteration that uses the entire dataset for model building, methods such as cross-validation for model tuning and splitting the available data into training, validation and test sets could be employed [16].

Conclusions

In conclusion, this study confirmed that the length-of-stay of patients with a mental comorbidity is longer than of patients without such a comorbidity at a German university hospital.

With detailed differentiation among subgroups divided by age, mental comorbidity subcategories and physical multimorbidity, we were able to highlight those subgroups with especially high length-of-stay. Furthermore, the difference between length-of-stay in medical inpatients with and without mental comorbidity increases with the level of physical multimorbidity.

In the future, the context of these individual variables should be analyzed even further in multivariate predictive models that allow more precise prediction of increased hospital length-of-stay in order to deliver actionable insights regarding hospital resource allocation.

Supporting information

S1 Table. Length-of-stay (LOS, days) mean and median per main diagnosis chapter.

-P: without psychiatric comorbidity. +P: with psychiatric comorbidity. Sorted by LOS mean difference between cases with and without psychiatric comorbidity. Chapters with total number of cases less than 500 grouped into Z Other: ("III", "VI", "VII", "VIII", "XII", "XIII", "XV", "XVII", "XXI"). Test type U: Mann-Whitney-U. f: common language effect size (value of 0.5 means no significant difference, deviation from 0.5 expresses siginifance), p-value significance: *p ≤ 0.05, **p≤ 0.01, ***p≤ 0.001.

(DOCX)

Click here for additional data file.^{(33.8KB, docx)}

S2 Table. Length-of-stay by age group and presence of mental comorbidity.

These are the underlying numbers for Fig 3. N: number of cases; LOS: length-of-stay; CI95: 95% confidence interval.

(DOCX)

Click here for additional data file.^{(32.9KB, docx)}

S3 Table. Length-of-stay for increasing Elixhauser score with and without mental comorbidity.

These are the underlying numbers for Fig 4. N: number of cases; LOS: length-of-stay; CI95: 95% confidence interval.

(DOCX)

Click here for additional data file.^{(34.3KB, docx)}

S4 Table. Hospital departments.

Number of cases ratio and mean length-of-stay with and without mental comorbidity. -P: without psychiatric comorbidity. +P: with psychiatric comorbidity.

(XLSX)

Click here for additional data file.^{(17.2KB, xlsx)}

S1 File. Model selection.

R code and output as well as comments for model selection process.

(PDF)

Click here for additional data file.^{(169.7KB, pdf)}

S2 File. Case simulation.

In the sheet “case_simulation”, enter values of interest in columns C-G to see the predicted length-of-stay in column H. The details of the contribution of each variable to the prediction are in columns M-S. The other sheets contain the coefficients that are used for the calculation. Highlighted yellow are the two numbers for example predictions mentioned in the discussion.

(XLSX)

Click here for additional data file.^{(34.2KB, xlsx)}

S3 File. Derivation of case simulation calculations.

(DOCX)

Click here for additional data file.^{(14.7KB, docx)}

S1 Fig. Length-of-stay in relation to number of mental comorbidities with outliers.

μ: mean length-of-stay, plotted as diamonds; ci: 95% confidence interval, also shown by error bars; N: number of cases. The number of mental comorbidities on the x-axis is displayed up to 5, as the number of cases with 6 or 7 mental comorbidities is less than 10.

(TIF)

Click here for additional data file.^{(52.2KB, tif)}

S2 Fig. Length-of-stay differentiated by mental comorbidity spectrum with outliers.

μ: mean length-of-stay, plotted as diamonds; ci: 95% confidence interval, also shown by error bars; N: number of cases that have a diagnosis in the respective F-category (ICD-10 range starting with the characters F0-F6). Cases that have mental comorbidity diagnoses in several F-categories are counted in each one separately and are therefore represented in multiple barsboxes.

(TIF)

Click here for additional data file.^{(67.2KB, tif)}

Data Availability

Minimal data for this study cannot be shared publicly because of identifying personal patient information gathered in clinical routine that underlies personal data protection regulations imposed by Ethikkommission Medizinische Fakultät Heidelberg. Data will be made available upon request from Ethikkommission Medizinische Fakultät Heidelberg via email (ethikkommission-I@med.uni-heidelberg.de) for researchers who meet the criteria for access to confidential data.

Funding Statement

The authors received no specific funding for this work.

References

1.Jansen L, van Schijndel M, van Waarde J, van Busschbach J. Health-economic outcomes in hospital patients with medical-psychiatric comorbidity: A systematic review and meta-analysis. PloS one. 2018;13(3):e0194029. Epub 13.03.2018. doi: 10.1371/journal.pone.0194029 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Köhle K, Herzog W, Joraschky P, Kruse J, Langewitz W, Söllner W. Uexküll, Psychosomatische Medizin: Theoretische Modelle und klinische Praxis (Engl: Psychosomatic medicine: theoretical models and clinical practice). 8th ed. München: Elsevier; 2017. [Google Scholar]
3.Müller-Tasch T, Frankenstein L, Holzapfel N, Schellberg D, Löwe B, Nelles M, et al. Panic disorder in patients with chronic heart failure. Journal of Psychosomatic Research. 2008;64(3):299–303. doi: 10.1016/j.jpsychores.2007.09.002 [DOI] [PubMed] [Google Scholar]
4.Jünger J, Schellberg D, Müller-Tasch T, Raupp G, Zugck C, Haunstetter A, et al. Depression increasingly predicts mortality in the course of congestive heart failure. European Journal of Heart Failure. 2005;7(2):261–7. doi: 10.1016/j.ejheart.2004.05.011 [DOI] [PubMed] [Google Scholar]
5.Häuser W, Wilhelm-Schwenk R, Klein W, Zimmer C, Krause-Wichmann D. Einfluss psychischer Komorbidität auf die stationäre Verweildauer internistischer Patienten im G-DRG-System (Engl.: Influence of mental comorbidity on the stationary length-of-stay for internistic patients in the G-DRG-System). Psychotherapie, Psychosomatik, medizinische Psychologie. 2006;56(9–10):370–5. doi: 10.1055/s-2006-940179 [DOI] [PubMed] [Google Scholar]
6.Friederich H-C, Hartmann M, Bergmann G, Herzog W. Psychische Komorbidität bei internistischen Krankenhauspatienten ‐ Prävalenz und Einfluss auf die Liegedauer (Engl.: Mental comorbidity of internistic hospital patients ‐ prevalence and influence on length-of-stay). Psychotherapie, Psychosomatik, medizinische Psychologie. 2002;52(7):323–8. doi: 10.1055/s-2002-32865 [DOI] [PubMed] [Google Scholar]
7.Saravay SM, lavin M. Psychiatric Comorbidity and Length of Stay in the General Hospital. Psychosomatics. 1994;35(3):233–52. doi: 10.1016/s0033-3182(94)71772-2 [DOI] [PubMed] [Google Scholar]
8.Hochlehnert A, Niehoff D, Wild B, Jünger J, Herzog W, Löwe B. Psychiatric comorbidity in cardiovascular inpatients: costs, net gain, and length of hospitalization. Journal of Psychosomatic Research. 2011;70(2):135–9. Epub 18.11.2010. doi: 10.1016/j.jpsychores.2010.09.010 [DOI] [PubMed] [Google Scholar]
9.Brilleman SL, Purdy S, Salisbury C, Windmeijer F, Gravelle H, Hollinghurst S. Implications of comorbidity for primary care costs in the UK: a retrospective observational study. British Journal of General Practice. 2013;63(609):e274–82. doi: 10.3399/bjgp13X665242 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Hochlehnert A, Niehoff D, Herzog W, Löwe B. Höhere Kosten bei internistischen Krankenhauspatienten mit psychischer Komorbidität: Fehlende Abbildung im DRG-System (Engl.: Higher cost of internistic hospital patients with mental comorbidity: lacking representation in DRG-system). Psychotherapie, Psychosomatik, medizinische Psychologie. 2007;57(2):70–5. Epub 08.01.2007. doi: 10.1055/s-2006-951924 [DOI] [PubMed] [Google Scholar]
11.Siddiqui N, Dwyer M, Stankovich J, Peterson G, Greenfield D, Si L, et al. Hospital length of stay variation and comorbidity of mental illness: a retrospective study of five common chronic medical conditions. BMC health services research. 2018;18(1):498. Epub 27.06.2018. doi: 10.1186/s12913-018-3316-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Freitas A, Silva-Costa T, Lopes F, Garcia-Lema I, Teixeira-Pinto A, Brazdil P, et al. Factors influencing hospital high length of stay outliers. BMC health services research. 2012;12:265. Epub 20.08.2012. doi: 10.1186/1472-6963-12-265 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Aubert CE, Schnipper JL, Fankhauser N, Marques-Vidal P, Stirnemann J, Auerbach AD, et al. Association of patterns of multimorbidity with length of stay: A multinational observational study. Medicine. 2020;99(34):e21650. doi: 10.1097/MD.0000000000021650 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Müller M, Huembelin M, Baechli C, Wagner U, Schuetz P, Mueller B, et al. Association of in-hospital multimorbidity with healthcare outcomes in Swiss medical inpatients. Swiss medical weekly. 2021;151:w20405. Epub 05.02.2021. doi: 10.4414/smw.2021.20405 [DOI] [PubMed] [Google Scholar]
15.Aubert CE, Schnipper JL, Roumet M, Marques-Vidal P, Stirnemann J, Auerbach AD, et al. Best Definitions of Multimorbidity to Identify Patients With High Health Care Resource Utilization. Mayo Clinic Proceedings: Innovations, Quality & Outcomes. 2020;4(1):40–9. Epub 14.01.2020. doi: 10.1016/j.mayocpiqo.2019.09.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Stone K, Zwiggelaar R, Jones P, Mac Parthaláin N. A systematic review of the prediction of hospital length of stay: Towards a unified framework. PLOS Digital Health. 2022;1(4):e0000017. doi: 10.1371/journal.pdig.0000017 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Ratgeber Krankenhaus: Bundesministerium für Gesundheit; 2023. Available from: https://www.bundesgesundheitsministerium.de/service/publikationen/details/ratgeber-krankenhaus.html.
18.Der Beitrag der deutschen Universitätsklinika für das Gesundheitssystem. Verband der Universitätsklinika Deutschlands e.V. (VUD) [cited 24 April 2023]. [Internet]. Berlin. Available from: https://www.uniklinika.de/die-deutschenuniversitaetsklinika/zahlen-und-fakten/.
19.Diefenbacher A. Psychiatry and Psychosomatic Medicine in Germany: Lessons to be Learned? Australian & New Zealand Journal of Psychiatry. 2005;39(9):782–94. doi: 10.1080/j.1440-1614.2005.01683.x [DOI] [PubMed] [Google Scholar]
20.Kathol RG, Harsch HH, Hall RC, Shakespeare A, Cowart T. Categorization of types of medical/psychiatry units based on level of acuity. Psychosomatics. 1992;33(4):376–86. doi: 10.1016/S0033-3182(92)71942-2 [DOI] [PubMed] [Google Scholar]
21.Gasparini A. comorbidity: An R package for computing comorbidity scores. Journal of Open Source Software. 2018;3(23):648. doi: 10.21105/joss.00648 [DOI] [Google Scholar]
22.Bartlett VL, Ross JS, Balasuriya L, Rhee TG. Association of Psychiatric Diagnoses and Medicaid Coverage with Length of Stay Among Inpatients Discharged to Skilled Nursing Facilities. Journal of General Internal Medicine. 2022. doi: 10.1007/s11606-021-07320-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Wolff J, Heister T, Normann C, Kaier K. Hospital costs associated with psychiatric comorbidities: a retrospective study. BMC Health Services Research. 2018;18(1). doi: 10.1186/s12913-018-2892-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Siskind D, Harris M, Diminic S, Carstensen G, Robinson G, Whiteford H. Predictors of mental health-related acute service utilisation and treatment costs in the 12 months following an acute psychiatric admission. Aust N Z J Psychiatry. 2014;48(11):1048–58. Epub 2014/07/18. doi: 10.1177/0004867414543566 [DOI] [PubMed] [Google Scholar]
25.Wilcox RR. The percentage bend correlation coefficient. Psychometrika. 1994;59(4):601–16. doi: 10.1007/bf02294395 [DOI] [Google Scholar]
26.Feliciana Silva F, Macedo Da Silva Bonfante G, Reis IA, André Da Rocha H, Pereira Lana A, Leal Cherchiglia M. Hospitalizations and length of stay of cancer patients: A cohort study in the Brazilian Public Health System. PLOS ONE. 2020;15(5):e0233293. doi: 10.1371/journal.pone.0233293 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Rosa RG, Goldani LZ. Factors Associated with Hospital Length of Stay among Cancer Patients with Febrile Neutropenia. PLoS ONE. 2014;9(10):e108969. doi: 10.1371/journal.pone.0108969 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Cavanaugh JE, Neath AA. The Akaike information criterion: Background, derivation, properties, application, interpretation, and refinements. WIREs Computational Statistics. 2019;11(3):e1460. doi: 10.1002/wics.1460 [DOI] [Google Scholar]
29.Caetano N, Laureano R, Cortez P. A Data-driven Approach to Predict Hospital Length of Stay ‐ A Portuguese Case Study. In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014): p. 407–14; 2014. doi: 10.5220/0004892204070414 [DOI] [Google Scholar]
30.Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods. 2020;17(3):261–72. doi: 10.1038/s41592-019-0686-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Vallat R. Pingouin: statistics in Python. Journal of Open Source Software. 2018;3(31):1026. doi: 10.21105/joss.01026 [DOI] [Google Scholar]
32.Venables WN, Ripley BD. Modern Applied Statistics with S. 4th ed. New York: Springer; 2002. [Google Scholar]
33.Beeler PE, Cheetham M, Held U, Battegay E. Depression is independently associated with increased length of stay and readmissions in multimorbid inpatients. European Journal of Internal Medicine. 2020;73:59–66. Epub 30.11.2019. doi: 10.1016/j.ejim.2019.11.012 [DOI] [PubMed] [Google Scholar]
34.Prina AM, Cosco TD, Dening T, Beekman A, Brayne C, Huisman M. The association between depressive symptoms in the community, non-psychiatric hospital admission and hospital outcomes: a systematic review. J Psychosom Res. 2015;78(1):25–33. Epub doi: 10.1016/j.jpsychores.2014.11.002 . [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Sledge WH, Gueorguieva R, Desan P, Bozzo JE, Dorset J, Lee HB. Multidisciplinary Proactive Psychiatric Consultation Service: Impact on Length of Stay for Medical Inpatients. Psychother Psychosom. 2015;84(4):208–16. Epub doi: 10.1159/000379757 . [DOI] [PubMed] [Google Scholar]
36.Sharpe M, Toynbee M, Walker J. Proactive Integrated Consultation-Liaison Psychiatry: A new service model for the psychiatric care of general hospital inpatients. General Hospital Psychiatry. 2020;66:9–15. doi: 10.1016/j.genhosppsych.2020.06.005 [DOI] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0287234.r001

Decision Letter 0

Sebastien Kenmoe

13 Mar 2023

PONE-D-23-03643Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatientsPLOS ONE

Dear Dr. Stahl-Toyota,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Apr 27 2023 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Sebastien Kenmoe

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please provide additional details regarding participant consent. In the ethics statement in the Methods and online submission information, please ensure that you have specified (1) whether consent was informed and (2) what type you obtained (for instance, written or verbal, and if verbal, how it was documented and witnessed). If your study included minors, state whether you obtained consent from parents or guardians. If the need for consent was waived by the ethics committee, please include this information.

If you are reporting a retrospective study of medical records or archived samples, please ensure that you have discussed whether all data were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

3. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

4. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

Reviewer #3: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

This observational study explores the impact of mental comorbidity and physical multimorbidity on the length-of-stay (LoS) in medical inpatients. Hospital LoS is an important outcome in the planning of healthcare services and control of medical costs. The authors focused on the effects of mental and physical multimorbidity instead of age, gender and primary diagnosis as the ‘usual suspects’ in studies that aim to predict LoS. Like most studies in this field, a limitation of this paper is that data collection and modelling are restricted to one hospital: the internal medicine department of the University Hospital in Heidelberg, Germany. However, there are other limitations and methodological issues.

1. STROBE-guidelines state that “readers need information on setting and locations to assess the context and generalizability of a study’s results”. Information is lacking on special features of the university hospital in the region and the healthcare system in Germany in general. The authors conclude that “there is a great need to develop new concepts of integrated care for medical inpatients with mental comorbidity”, but they are not clear about existing concepts other then consultation-liasion services. The typology of Kathol et al. (1992) could be helpful to clarify the regional and national context.

2. The STROBE-statement also points at the importance of clearly defining all variables and reporting missing values for each variable of interest and for each step in the analysis. However, in this study “no missing values” was one of the inclusion criteria. Other criteria are not clarified: why were cases with less than two hospital days and with main diagnosis of the ICD-10 code chapter V for psychiatric diseases excluded? In the discussion section, the authors mention that for patients 60+ years old early transfer in geriatric or general hospitals occurs regularly. Should the analyses not have been restricted to the age group 18-60? Organic diseases (n=1394), substance use (n=1319), neurotic and somatoform disorders (n=1042) and affective disorders (n=879) were the most frequent mental diagnoses, but these are not main diagnoses of the ICD-10 code chapter V?

Also, information is lacking on how diagnoses were recorded. The limitations section makes clear that no differentiation could be made between preexisting or newly developed mental comorbidity. But in 911 cases two up to five diagnoses were registered, including (unlike other studies) dementia, delirium or nicotine abuse. In the final analysis, only “Mental comorbidity present” is used, which in terms of severity and impact puts very different mental diagnoses on an equal footing. Sensitivity analyses could help to clarify the effects of this approach.

3. In total, N=28,553 cases met the inclusion criteria of which 8.360 admissions (29% of all cases) were follow-up admissions. The authors acknowledge that this is a violation of the assumption in regression analysis that data should be independent, but make no effort to explore this possible source of bias. It could have been helpful to compare characteristics in Table 1 of patients with one or multiple admissions (although single admissions in the beginning of 2017 and at the end of 2019 could be part of multiple admissions that fall outside the observation period).

4. In the statistical analyses section, the authors state that robust correlation coefficients with percentage bend (r) or correlation ratios (η) were calculated, but it is not clear what these un-directional correlation coefficients would add to the univariate and multivariate analyses regressing LoS on mental comorbidity and physical multimorbidity.

Gender, age and main diagnosis ICD-10 are seen as potential confounders, but main diagnosis is not included in the negative binomial regression model. Supplementary material 1 shows large LoS differences between cases with and without psychiatric comorbidity per ICD-10 chapter, which suggest relevant co-occurrence of main diagnosis and prevalence of mental problems: injury and poisoning (suicide risk?) or neoplasms (depression and anxiety?). Moreover, the statistical analysis plan does not include model comparison, sensitivity analyses and methods to check model fit.

5. Table 1 in this paper reports characteristics of the sample of internal medicine inpatients and tests differences between cases with and without mental comorbidity. However, STROBE-guidelines state that inferential measures and significance tests should be avoided in descriptive tables.

6. Table 2 presents the main analyses in terms of incidence rate ratios. The model includes an interaction-effect of mental comorbidity and physical multimorbidity, which concerns the primary hypothesis. Yet in the text (line 268) this is based on figure 4 (which shows different LoS-values per Elixhauser score for patients with and without mental comorbidity), but no interaction-effect of mental comorbidity and age although figure 3 shows different LoS-values per age-group for patients with and without mental comorbidity. And probably there is also an interaction effect of age and Elixhauser score. Therefor it is unclear how model selection came about and how models were compared. The fit of the final model or predictive power is not discussed.

In the discussion section the authors calculated some LoS-estimates based on the model (3 days extension for someone with mental comorbidity and an Elixhauser score of 2 and 17.2 days for an Elixhauser score of 7), but these calculations are difficult to follow and are presented as LoS-values instead of expected averages with confidence intervals or estimates with prediction intervals. The actual number of predicted hospital days can be calculated from the unexponentiated model coefficients, but these values are not reported.

7. The authors conclude that “increasing levels of multimorbidity are associated with a growing positive influence of mental comorbidity on the length-of-stay.” But this is not what the interaction-term implies. The 8.5% increase per additional physical morbidity concerns only patients with mental comorbidity. For this group the effect of physical multi morbidities increases somewhat more compared to patients without mental comorbidity. In the discussion section the authors fail to give a substantive interpretation of this complex interaction effect. This interpretation should support the conclusion that new concepts of integrated care need to be developed.

8. In the discussion section, the authors point at the limitations of this study. The recent review of studies on the prediction of hospital length of stay by Stone et al. (2022) could be a useful reference in this regard.

References

R.G. Kathol, H.H. Harsch, R.C.W. Hall, et al. Categorization of types of medical/psychiatry units based on level of acuity. Psychosomatics, 33 (1992), pp. 376-386

Stone K, Zwiggelaar R, Jones P, Mac Parthaláin N (2022) A systematic review of the prediction of hospital length of stay: Towards a unified framework. PLOS Digit Health 1(4): e0000017. https://doi.org/10.1371/journal.pdig.0000017

Reviewer #2: The authors aimed to investigate the impact of concurrent mental comorbidity and physical multimorbidity on the length of stay in medical inpatients at a maximum-care university hospital. This cross-sectional study has 28,553 inpatients treated in Germany between 2017 and 2019. The authors identified mental health conditions using the ICD-10 chapter and physical comorbidity using the Elixhauser score. Although the research presents an important issue, the authors need to address some issues.

Reviewer #3: Dear Authors,

You investigate the influence of comorbid mental disorders on the length of inpatient treatment in somatic hospitals. In a retrospective design,routine clinical data was evaluated and a correlation between comorbidity and length of stay was found. The conclusion from from the results is that new care models are needed for multimorbid patients with concomitant mental illness.

The question of the present study is undoubtedly of great relevance and the the analysis is based on an impressive number of cases collected across several units. The manuscript is also clearly and comprehensibly written.

However, some questions or points of discussion arise for me.

(1) It seems to me that the literature in the introduction is not up to date. In addition, it partly refers to other health systems that may only be comparable to a limited extent.

(2) How exactly were the mental comorbidities diagnosed? The diagnosis of mental illness is often quite difficult. If the diagnosis is not made by specialists, misdiagnoses cannot be ruled out.

(3) The statistical analysis seems to me to have room for improvement. Since the data were collected across different departments with probably different mean lengths of stay, I think a multi-level model with the departments as clusters could be considered. If varying slopes were allowed, the effects could be estimated in a more differentiated way.

(4) It would be helpful for the reader to have more information on the departments involved. Could the lengths of stay and the proportions of cases with comorbidities be broken down by department?

Overall, a publication of the study is very desirable, but I would advise a revision beforehand.

Yours sincerely

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: A.I. Wierdsma

Reviewer #2: No

Reviewer #3: No

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Attachment

Submitted filename: PONE-D-23-03643 review.docx

Click here for additional data file.^{(14.2KB, docx)}

PLoS One. 2023 Jun 22;18(6):e0287234. doi: 10.1371/journal.pone.0287234.r002

Author response to Decision Letter 0

12 May 2023

Please refer to the file Response_to_Reviewers for a formatted version of this text:

Journal Requirements

Thank you for giving us the chance to specify this important aspect. The need for consent was waived by the ethics committee. We added this information in the Methods section:

‘This study was approved by the Ethics committee of the Medical faculty of the University of Heidelberg (No. S-690/2021). As only clinical routine data were used, the need for consent was waived by the ethics committee.’

In your revised cover letter, please address the following prompts:

Thank you for this comment and request. However, the de-identified data set still includes the variables gender, age, main diagnosis and comorbidities. Together with the information in which time frame and at which hospital the patients were treated, the data set was deemed as potentially sensitive by the ethics committee. We agree with the suggestion for the Data Availability statement: "Minimal data for this study cannot be shared publicly because of identifying personal patient information gathered in clinical routine that underlies personal data protection regulations imposed by Ethikkommission Medizinische Fakultät Heidelberg. Data will be made available upon request from Ethikkommission Medizinische Fakultät Heidelberg via email (ethikkommission-I@med.uni-heidelberg.de) for researchers who meet the criteria for access to confidential data."

Reviewer 1:

Dear Reviewer #1,

Thank you very much for reviewing this manuscript. We have considered each of your important indications thoroughly.

Comments:

1) STROBE-guidelines state that “readers need information on setting and locations to assess the context and generalizability of a study’s results”. Information is lacking on special features of the university hospital in the region and the healthcare system in Germany in general. The authors conclude that “there is a great need to develop new concepts of integrated care for medical inpatients with mental comorbidity”, but they are not clear about existing concepts other then consultation-liasion services. The typology of Kathol et al. (1992) could be helpful to clarify the regional and national context.

Thank you for setting the perspective of the international readership into focus. We provided more information about the German health care system in the introduction. According to the typology of Kathol et al., the type of care has been stated more precisely to improve the understanding of the setting.

2) The STROBE-statement also points at the importance of clearly defining all variables and reporting missing values for each variable of interest and for each step in the analysis. However, in this study “no missing values” was one of the inclusion criteria. Other criteria are not clarified: why were cases with less than two hospital days and with main diagnosis of the ICD-10 code chapter V for psychiatric diseases excluded? In the discussion section, the authors mention that for patients 60+ years old early transfer in geriatric or general hospitals occurs regularly. Should the analyses not have been restricted to the age group 18-60? Organic diseases (n=1394), substance use (n=1319), neurotic and somatoform disorders (n=1042) and affective disorders (n=879) were the most frequent mental diagnoses, but these are not main diagnoses of the ICD-10 code chapter V?

Thank you very much for your feedback. We agree that the definition of the cohort needs more detailed clarification and would like to address your suggestions in these ways:

a) Missing values: Indeed, the data completeness was high in this data set and only one patient had a missing value for gender. Please refer to the R code in S4 Supporting Information for the check for missing values. We reworded the passage about inclusion criteria to state more clearly that there were indeed no otherwise missing values, and we did not remove any further data for that reason.

b) Cases with only one hospital day were excluded: Cases with a value for length-of-stay less than two days meant that they were treated as ambulant cases and not as stationary, as was the cohort of interest for this study. We added this clarification to the inclusion criteria reasoning.

c) Chapter V for psychiatric diseases excluded for main diagnosis: This study was not about psychiatric patients with additional mental comorbidities. They are taken care of in a psychiatric department with a different reimbursement system. Our study is about somatic patients with mental comorbidity for whom the potential need for psychiatric support is hypothesized to be underestimated. We also added this reason to the inclusion criteria.

d) Transfer of patients 60+ years old: The sentence in the discussion seems to be misleading as the number of patients who are transferred is not as high as was suggested here. The hospital is a maximum care hospital, so it is also representative for patients aged 60+. The sentence was removed from the discussion.

e) Chapter V for psychiatric diseases included for comorbidity diagnoses: The raw data set had two different variables for the main diagnosis and comorbidities. Both variables contain ICD-10 codes. Therefore, we could use the main diagnosis variable to filter out mental main diagnoses, while determining mental comorbidities for all remaining main diagnoses from the comorbidity variable.

f) Information on how diagnoses were recorded: Only diagnoses were recorded that have entailed use of resources / medication / diagnostic and therapeutic effort, so diagnoses that were relevant for the focus of treatment for the respective hospital stay. This information was added after the inclusion criteria and to a note regarding this limitation in the discussion.

g) Severity of different mental diagnoses: Of course, the impact of schizophrenia is probably more severe on length-of-stay than an addiction to tobacco, for example. We are aware that the model we present puts all mental diseases on the same level, and future models should account for the differences in severity. We added this thought explicitly to the limitations and would declare the presented model as a kind of benchmark model for which intuitive interpretation was the main aim.

3) In total, N=28,553 cases met the inclusion criteria of which 8.360 admissions (29% of all cases) were follow-up admissions. The authors acknowledge that this is a violation of the assumption in regression analysis that data should be independent, but make no effort to explore this possible source of bias. It could have been helpful to compare characteristics in Table 1 of patients with one or multiple admissions (although single admissions in the beginning of 2017 and at the end of 2019 could be part of multiple admissions that fall outside the observation period).

In the study design and the discussion, we added a note to distinguish between follow-up admissions that were already merged during data pre-processing and truly different cases. We absolutely agree that the topic of readmissions needs to be addressed. This is an important aspect which we plan to address in dedicated future studies as a prediction target variable in the sense of readmission within a particular time frame but consider it out of scope for the present study.

4) In the statistical analyses section, the authors state that robust correlation coefficients with percentage bend (r) or correlation ratios (η) were calculated, but it is not clear what these un-directional correlation coefficients would add to the univariate and multivariate analyses regressing LoS on mental comorbidity and physical multimorbidity.

Thank you for requesting more details on the model selection process, as of course, more effort than presented initially took place. We show the derivation of the model with the corresponding R code and output in the new S4 Supporting Information where we also portray the measures used for checking model fit (AIC, RMSE, MAE).

The main diagnosis was included in the negative binomial regression model, but the coefficients had been missing in the table and were now added.

The correlation coefficients and correlation ratios were calculated as part of the exploratory analysis to determine which variables to include in the model. We also assume that it is advantageous to present these values for which the interpretation should be familiar to the readership.

5) Table 1 in this paper reports characteristics of the sample of internal medicine inpatients and tests differences between cases with and without mental comorbidity. However, STROBE-guidelines state that inferential measures and significance tests should be avoided in descriptive tables.

We removed the respective column from the table and transferred the most relevant test statistic to the paragraph about length-of-stay with vs. without mental comorbidity.

6) Table 2 presents the main analyses in terms of incidence rate ratios. The model includes an interaction-effect of mental comorbidity and physical multimorbidity, which concerns the primary hypothesis. Yet in the text (line 268) this is based on figure 4 (which shows different LoS-values per Elixhauser score for patients with and without mental comorbidity), but no interaction-effect of mental comorbidity and age although figure 3 shows different LoS-values per age-group for patients with and without mental comorbidity. And probably there is also an interaction effect of age and Elixhauser score. Therefor it is unclear how model selection came about and how models were compared. The fit of the final model or predictive power is not discussed.

We are glad to provide more details regarding the model selection, please refer to S4 Supporting Information. The effect of interaction between mental comorbidity and age is explored there, as well. The model improvement was not as strong as the effect of mental comorbidity and Elixhauser, however. A potential interaction between age and Elixhauser may be incorporated in future models that aim to optimize model prediction. Here, we wanted to focus on the effect of mental comorbidity and therefore tried to minimize model complexity.

As interpretability was among our highest priorities, we had prepared a case simulation excel file to perform those calculations. Since these calculations appear to be of interest to the reader, we naturally can include them as S5 Supporting Information. We also added a note in the discussion regarding the usage of this file.

7) The authors conclude that “increasing levels of multimorbidity are associated with a growing positive influence of mental comorbidity on the length-of-stay.” But this is not what the interaction-term implies. The 8.5% increase per additional physical morbidity concerns only patients with mental comorbidity. For this group the effect of physical multi morbidities increases somewhat more compared to patients without mental comorbidity. In the discussion section the authors fail to give a substantive interpretation of this complex interaction effect. This interpretation should support the conclusion that new concepts of integrated care need to be developed.

Thank you for pointing out the need for clearer communication of the interaction implications. We reworded the quoted sentence to stress the fact that the interaction is only relevant for the sub-group with mental comorbidity and moved it to the paragraph about the clinical implications.

8) In the discussion section, the authors point at the limitations of this study. The recent review of studies on the prediction of hospital length of stay by Stone et al. (2022) could be a useful reference in this regard.

That is a very valuable resource. We have incorporated a few of the shortcomings of current LOS prediction research listed by the authors that also apply to our study.

Reviewer 2:

Dear Reviewer #2,

Thank you very much for reviewing our manuscript. Your advice was very helpful and pointed out important issues of our study.

Comments:

1) Introduction: The authors wrote, “Moreover, it could be imagined that a future predictive model estimates the expected length-of-stay at the beginning of a hospital stay and supports clinical decisions for early interventions, such as a proactive psychosomatic and psychiatric consultation service.”, I understand this idea is attractive and agree that it is crucial to allocate resources appropriately. However, in practice, principal and additional diagnoses are only recorded after the care is complete, i.e. at the end, so I’m unsure how the model suggested by the authors can be fitted.

It is true that the final entry of diagnoses is required by law for reimbursement purposes only at the end of the hospital stay when care is completed. In practice, however, diagnoses are often recorded as soon as they are identified. Most diagnoses are identified at admission. If further diagnoses are placed, they are available as soon as the diagnosis is made. The challenge for a would be to make the diagnoses available to the model as soon as they are identified.

2) Method: The authors should provide more information on the health care system in Germany and, if they need more space, shorten the results section as people from different jurisdictions will need help understanding the system in Germany.

Thank you for setting the perspective of the international readership into focus. We provided more information about the German health care system in the introduction. The setting of the psychiatric/medical unit has also been stated more precisely.

3) Method: Physical multimorbidity -> need more details on which diagnosis (primary, additional, or all) was used to determine this. I’m guessing it’s all diagnoses, but I need clarification.

Only the additional diagnoses were used to determine the Elixhauser score, the primary diagnosis was not included. We added this indeed important detail to the respective section in the methods.

4) Method: Mental comorbidity -> Authors should consider not including F7 in the mental comorbidity as it is an intellectual disability, which is not considered a mental disorder. Generally, people with intellectual disability are known to have a longer stay in the hospital.

It is true that an intellectual disability is not to be set equal to a mental disorder. In order to consider not including F7, we checked how many cases would be affected by this change. 36 cases have an F7 comorbidity, so the expected effect on the result of the present study’s results would be negligible. Therefore, we would stick to the current data pre-processing decisions, also as other similar studies list very heterogeneous filtering criteria for mental disorders. In future work we would definitely reconsider this decision, as more detailed analysis of the mental disorder subcategories is desired anyhow.

5) Statistical analysis: The authors mentioned that the analysis was performed at the hospitalisation level. Can you please clarify whether you can identify patients from your data? While I agree that the analysis should be performed at the hospitalisation level, I want to know how the authors control for within-patient variation. Records from the same patients could share some characteristics and have a similar outcome; not accounting for it could produce biased results. Also, the authors’ statistical methods assumed data independence to be valid.

Yes, we can identify patients from our data. In the study design and the discussion, we added a note to distinguish between follow-up admissions that were already merged during data pre-processing and truly different cases. This merge of case records should prevent bias through complications from the same patients with the same admission reason resulting in different case numbers in the raw data. Only age and gender are expected to be shared among the remaining different cases of the same patients, but main diagnosis, Elixhauser score and mental comorbidity presence can vary. In the future, however, we plan to examine the topic of readmission more closely, as information such as readmission within a particular timeframe would also be an interesting target variable.

6) Results: Can you please add the 90th percentile for the length of stay in Table 1? The data is skew, so the 90th percentile will help the reader understand the distribution better.

To emphasize the overdispersion, that is typical for length-of-stay distributions, is a very good idea; an additional row was added at the bottom of Table 1 for the 90th percentile which was derived with the interpolation option set to ‘nearest’.

7) Results: Figures 1 & 2: I believe the box plot with 6 boxes for each mental health group will be more informative than the current graphs.

Thank you for this good idea, we have replaced the figures with box plots and included the measures plotted before in the x axis labels.

Reviewer 3:

Dear Reviewer #3,

Thank you very much for reviewing our manuscript. Your comments were very valuable and helped to improve the description of our study.

Comments:

1) It seems to me that the literature in the introduction is not up to date. In addition, it partly refers to other health systems that may only be comparable to a limited extent.

Thank you for taking the international perspective of the readership into focus. We added a short description of the hospital landscape in Germany and how the presented cohort is classified. We also added Freitas et al. (12) and Stone et al. (16) to the literature.

2) How exactly were the mental comorbidities diagnosed? The diagnosis of mental illness is often quite difficult. If the diagnosis is not made by specialists, misdiagnoses cannot be ruled out.

We agree that the diagnosis of mental illness should be performed by specialists. That is why most mental comorbidities that are reported here are diagnosed by a psychiatric specialist who is called in by the treating physician if a mental comorbidity is suspected for a patient. Misdiagnoses of course cannot be ruled out in principle. But the risk for this cohort is reduced due to application of the existing diagnostic frameworks ICD-10 (International Classification of Diseases, Tenth Revision) and DSM-V (Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition) by specialists. What is more probable is that diagnoses are entirely missing because no systematic screening is taking place, but a specialist is only called at the internal medicine department physician’s suspicion. Diagnoses from the medical history of the patient are also not included but could have included mental illness. That is why this article is aiming to recommend early consideration of possible mental illness diagnoses by specialists. We added a note in this regard to the discussion.

3) The statistical analysis seems to me to have room for improvement. Since the data were collected across different departments with probably different mean lengths of stay, I think a multi-level model with the departments as clusters could be considered. If varying slopes were allowed, the effects could be estimated in a more differentiated way.

Yes, the data were collected across six different departments. As suggested in the next comment, the varying mean lengths of stay were added in S7 Table. We agree that the detectable variation could be represented more accurately in a multi-level model. However, we decided on the current model architecture for the following reasons: (i) The variable hospital department and main diagnosis correlate strongly (Cramer’s V=0.52), (ii) the aim for this study was a simple, intuitive model to serve as a benchmark for future studies. We nevertheless added this idea to the limitations in the discussion. Please also refer to the new S4 Supplemental Information for more details on the model selection process.

4) It would be helpful for the reader to have more information on the departments involved. Could the lengths of stay and the proportions of cases with comorbidities be broken down by department?

We added the supplemental S7 Table that shows the average length-of-stay and proportions of cases with mental comorbidity broken down by the 6 hospital departments that were part of this data set.

Attachment

Submitted filename: Response_to_Reviewers.docx

Click here for additional data file.^{(547.5KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0287234.r003

Decision Letter 1

Sebastien Kenmoe

2 Jun 2023

Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

PONE-D-23-03643R1

Dear Dr. Stahl-Toyota,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Sebastien Kenmoe

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

Reviewer #2: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Partly

Reviewer #2: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: No

Reviewer #2: (No Response)

Reviewer #3: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: (No Response)

Reviewer #3: Yes

**********

6. Review Comments to the Author

Reviewer #1: Th authors adequately responded on several issues that were pointed out in the text, but for some important topics readers are referred to supplementary material and future studies. This makes the review process something of a quest for information. Here results are mostly expressed as means and SD, whereas the negative binomial models implicate that these are not very helpful descriptive measures. Minor note: in S4_Supporting_information (page 6) Glm function in R is referred to as general linear model, whereas all models compared are generalized linear model with different distribution families.

Furthermore, the authors show a very pragmatic perspective when reporting statistical analyses: “the aim for this study was a simple, intuitive model to serve as a benchmark for future studies”. But the model presented is not that simple or intuitive. As the variable hospital department and main diagnosis correlated strongly, it is unclear how to distinguish between department-effect and diagnosis-effect. LOS-estimates based on the model are presented as point estimates instead of expected LOS-values with prediction intervals.

In addition, interpretation of the interaction effect is still confusing. The authors conclude that “our study provides evidence … that mental comorbidity predicts length-of-stay dependent on physical multimorbidity.” But this is reversed: for low Elixhauser scores there is no difference in LOS, as this score increase the LOS increases, but in the group with mental comorbidity we see that higher Elixhauser have a stronger impact on LOS. Or as the authors rephrase it: “The interaction between mental comorbidity presence and Elixhauser score can be interpreted in the sense that for patients with mental comorbidity, the effect of increasing levels of multimorbidity on length-of-stay was stronger.” But this is descriptive and not really an interpretation of the interaction effect. Supplementary material 1 shows that the largest differences in median LOS-values between Yes/No mental comorbidity are found for ‘Neoplasms’ and ‘Injury, poisoning and certain other consequences of external causes’. Ideas are lacking on why these associations were found and how they could be linked to physical multimorbidity.

It is up to the editors to decide if this ‘benchmark’ for future studies is suitable for publication. In my view this kind of theory-poor applied research only adds to the body of not replicable findings.

Reviewer #2: The authors addressed all queries and amended the manuscript according to the comments. I have no further comments

Reviewer #3: Dear Authors,

my comments have been duly taken into account and I support the publication of the manuscript.

With best regards

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: André Wierdsma

Reviewer #2: No

Reviewer #3: No

**********

PLoS One. doi: 10.1371/journal.pone.0287234.r004

Acceptance letter

Sebastien Kenmoe

6 Jun 2023

PONE-D-23-03643R1

Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

Dear Dr. Stahl-Toyota:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Sebastien Kenmoe

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Table. Length-of-stay (LOS, days) mean and median per main diagnosis chapter.

(DOCX)

Click here for additional data file.^{(33.8KB, docx)}

S2 Table. Length-of-stay by age group and presence of mental comorbidity.

These are the underlying numbers for Fig 3. N: number of cases; LOS: length-of-stay; CI95: 95% confidence interval.

(DOCX)

Click here for additional data file.^{(32.9KB, docx)}

S3 Table. Length-of-stay for increasing Elixhauser score with and without mental comorbidity.

These are the underlying numbers for Fig 4. N: number of cases; LOS: length-of-stay; CI95: 95% confidence interval.

(DOCX)

Click here for additional data file.^{(34.3KB, docx)}

S4 Table. Hospital departments.

Number of cases ratio and mean length-of-stay with and without mental comorbidity. -P: without psychiatric comorbidity. +P: with psychiatric comorbidity.

(XLSX)

Click here for additional data file.^{(17.2KB, xlsx)}

S1 File. Model selection.

R code and output as well as comments for model selection process.

(PDF)

Click here for additional data file.^{(169.7KB, pdf)}

S2 File. Case simulation.

(XLSX)

Click here for additional data file.^{(34.2KB, xlsx)}

S3 File. Derivation of case simulation calculations.

(DOCX)

Click here for additional data file.^{(14.7KB, docx)}

S1 Fig. Length-of-stay in relation to number of mental comorbidities with outliers.

(TIF)

Click here for additional data file.^{(52.2KB, tif)}

S2 Fig. Length-of-stay differentiated by mental comorbidity spectrum with outliers.

(TIF)

Click here for additional data file.^{(67.2KB, tif)}

Attachment

Submitted filename: PONE-D-23-03643 review.docx

Click here for additional data file.^{(14.2KB, docx)}

Attachment

Submitted filename: Response_to_Reviewers.docx

Click here for additional data file.^{(547.5KB, docx)}

Data Availability Statement

[pone.0287234.ref001] 1.Jansen L, van Schijndel M, van Waarde J, van Busschbach J. Health-economic outcomes in hospital patients with medical-psychiatric comorbidity: A systematic review and meta-analysis. PloS one. 2018;13(3):e0194029. Epub 13.03.2018. doi: 10.1371/journal.pone.0194029 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref002] 2.Köhle K, Herzog W, Joraschky P, Kruse J, Langewitz W, Söllner W. Uexküll, Psychosomatische Medizin: Theoretische Modelle und klinische Praxis (Engl: Psychosomatic medicine: theoretical models and clinical practice). 8th ed. München: Elsevier; 2017. [Google Scholar]

[pone.0287234.ref003] 3.Müller-Tasch T, Frankenstein L, Holzapfel N, Schellberg D, Löwe B, Nelles M, et al. Panic disorder in patients with chronic heart failure. Journal of Psychosomatic Research. 2008;64(3):299–303. doi: 10.1016/j.jpsychores.2007.09.002 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref004] 4.Jünger J, Schellberg D, Müller-Tasch T, Raupp G, Zugck C, Haunstetter A, et al. Depression increasingly predicts mortality in the course of congestive heart failure. European Journal of Heart Failure. 2005;7(2):261–7. doi: 10.1016/j.ejheart.2004.05.011 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref005] 5.Häuser W, Wilhelm-Schwenk R, Klein W, Zimmer C, Krause-Wichmann D. Einfluss psychischer Komorbidität auf die stationäre Verweildauer internistischer Patienten im G-DRG-System (Engl.: Influence of mental comorbidity on the stationary length-of-stay for internistic patients in the G-DRG-System). Psychotherapie, Psychosomatik, medizinische Psychologie. 2006;56(9–10):370–5. doi: 10.1055/s-2006-940179 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref006] 6.Friederich H-C, Hartmann M, Bergmann G, Herzog W. Psychische Komorbidität bei internistischen Krankenhauspatienten ‐ Prävalenz und Einfluss auf die Liegedauer (Engl.: Mental comorbidity of internistic hospital patients ‐ prevalence and influence on length-of-stay). Psychotherapie, Psychosomatik, medizinische Psychologie. 2002;52(7):323–8. doi: 10.1055/s-2002-32865 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref007] 7.Saravay SM, lavin M. Psychiatric Comorbidity and Length of Stay in the General Hospital. Psychosomatics. 1994;35(3):233–52. doi: 10.1016/s0033-3182(94)71772-2 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref008] 8.Hochlehnert A, Niehoff D, Wild B, Jünger J, Herzog W, Löwe B. Psychiatric comorbidity in cardiovascular inpatients: costs, net gain, and length of hospitalization. Journal of Psychosomatic Research. 2011;70(2):135–9. Epub 18.11.2010. doi: 10.1016/j.jpsychores.2010.09.010 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref009] 9.Brilleman SL, Purdy S, Salisbury C, Windmeijer F, Gravelle H, Hollinghurst S. Implications of comorbidity for primary care costs in the UK: a retrospective observational study. British Journal of General Practice. 2013;63(609):e274–82. doi: 10.3399/bjgp13X665242 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref010] 10.Hochlehnert A, Niehoff D, Herzog W, Löwe B. Höhere Kosten bei internistischen Krankenhauspatienten mit psychischer Komorbidität: Fehlende Abbildung im DRG-System (Engl.: Higher cost of internistic hospital patients with mental comorbidity: lacking representation in DRG-system). Psychotherapie, Psychosomatik, medizinische Psychologie. 2007;57(2):70–5. Epub 08.01.2007. doi: 10.1055/s-2006-951924 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref011] 11.Siddiqui N, Dwyer M, Stankovich J, Peterson G, Greenfield D, Si L, et al. Hospital length of stay variation and comorbidity of mental illness: a retrospective study of five common chronic medical conditions. BMC health services research. 2018;18(1):498. Epub 27.06.2018. doi: 10.1186/s12913-018-3316-2 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref012] 12.Freitas A, Silva-Costa T, Lopes F, Garcia-Lema I, Teixeira-Pinto A, Brazdil P, et al. Factors influencing hospital high length of stay outliers. BMC health services research. 2012;12:265. Epub 20.08.2012. doi: 10.1186/1472-6963-12-265 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref013] 13.Aubert CE, Schnipper JL, Fankhauser N, Marques-Vidal P, Stirnemann J, Auerbach AD, et al. Association of patterns of multimorbidity with length of stay: A multinational observational study. Medicine. 2020;99(34):e21650. doi: 10.1097/MD.0000000000021650 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref014] 14.Müller M, Huembelin M, Baechli C, Wagner U, Schuetz P, Mueller B, et al. Association of in-hospital multimorbidity with healthcare outcomes in Swiss medical inpatients. Swiss medical weekly. 2021;151:w20405. Epub 05.02.2021. doi: 10.4414/smw.2021.20405 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref015] 15.Aubert CE, Schnipper JL, Roumet M, Marques-Vidal P, Stirnemann J, Auerbach AD, et al. Best Definitions of Multimorbidity to Identify Patients With High Health Care Resource Utilization. Mayo Clinic Proceedings: Innovations, Quality & Outcomes. 2020;4(1):40–9. Epub 14.01.2020. doi: 10.1016/j.mayocpiqo.2019.09.002 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref016] 16.Stone K, Zwiggelaar R, Jones P, Mac Parthaláin N. A systematic review of the prediction of hospital length of stay: Towards a unified framework. PLOS Digital Health. 2022;1(4):e0000017. doi: 10.1371/journal.pdig.0000017 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref017] 17.Ratgeber Krankenhaus: Bundesministerium für Gesundheit; 2023. Available from: https://www.bundesgesundheitsministerium.de/service/publikationen/details/ratgeber-krankenhaus.html.

[pone.0287234.ref018] 18.Der Beitrag der deutschen Universitätsklinika für das Gesundheitssystem. Verband der Universitätsklinika Deutschlands e.V. (VUD) [cited 24 April 2023]. [Internet]. Berlin. Available from: https://www.uniklinika.de/die-deutschenuniversitaetsklinika/zahlen-und-fakten/.

[pone.0287234.ref019] 19.Diefenbacher A. Psychiatry and Psychosomatic Medicine in Germany: Lessons to be Learned? Australian & New Zealand Journal of Psychiatry. 2005;39(9):782–94. doi: 10.1080/j.1440-1614.2005.01683.x [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref020] 20.Kathol RG, Harsch HH, Hall RC, Shakespeare A, Cowart T. Categorization of types of medical/psychiatry units based on level of acuity. Psychosomatics. 1992;33(4):376–86. doi: 10.1016/S0033-3182(92)71942-2 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref021] 21.Gasparini A. comorbidity: An R package for computing comorbidity scores. Journal of Open Source Software. 2018;3(23):648. doi: 10.21105/joss.00648 [DOI] [Google Scholar]

[pone.0287234.ref022] 22.Bartlett VL, Ross JS, Balasuriya L, Rhee TG. Association of Psychiatric Diagnoses and Medicaid Coverage with Length of Stay Among Inpatients Discharged to Skilled Nursing Facilities. Journal of General Internal Medicine. 2022. doi: 10.1007/s11606-021-07320-4 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref023] 23.Wolff J, Heister T, Normann C, Kaier K. Hospital costs associated with psychiatric comorbidities: a retrospective study. BMC Health Services Research. 2018;18(1). doi: 10.1186/s12913-018-2892-5 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref024] 24.Siskind D, Harris M, Diminic S, Carstensen G, Robinson G, Whiteford H. Predictors of mental health-related acute service utilisation and treatment costs in the 12 months following an acute psychiatric admission. Aust N Z J Psychiatry. 2014;48(11):1048–58. Epub 2014/07/18. doi: 10.1177/0004867414543566 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref025] 25.Wilcox RR. The percentage bend correlation coefficient. Psychometrika. 1994;59(4):601–16. doi: 10.1007/bf02294395 [DOI] [Google Scholar]

[pone.0287234.ref026] 26.Feliciana Silva F, Macedo Da Silva Bonfante G, Reis IA, André Da Rocha H, Pereira Lana A, Leal Cherchiglia M. Hospitalizations and length of stay of cancer patients: A cohort study in the Brazilian Public Health System. PLOS ONE. 2020;15(5):e0233293. doi: 10.1371/journal.pone.0233293 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref027] 27.Rosa RG, Goldani LZ. Factors Associated with Hospital Length of Stay among Cancer Patients with Febrile Neutropenia. PLoS ONE. 2014;9(10):e108969. doi: 10.1371/journal.pone.0108969 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref028] 28.Cavanaugh JE, Neath AA. The Akaike information criterion: Background, derivation, properties, application, interpretation, and refinements. WIREs Computational Statistics. 2019;11(3):e1460. doi: 10.1002/wics.1460 [DOI] [Google Scholar]

[pone.0287234.ref029] 29.Caetano N, Laureano R, Cortez P. A Data-driven Approach to Predict Hospital Length of Stay ‐ A Portuguese Case Study. In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014): p. 407–14; 2014. doi: 10.5220/0004892204070414 [DOI] [Google Scholar]

[pone.0287234.ref030] 30.Virtanen P, Gommers R, Oliphant TE, Haberland M, Reddy T, Cournapeau D, et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature Methods. 2020;17(3):261–72. doi: 10.1038/s41592-019-0686-2 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref031] 31.Vallat R. Pingouin: statistics in Python. Journal of Open Source Software. 2018;3(31):1026. doi: 10.21105/joss.01026 [DOI] [Google Scholar]

[pone.0287234.ref032] 32.Venables WN, Ripley BD. Modern Applied Statistics with S. 4th ed. New York: Springer; 2002. [Google Scholar]

[pone.0287234.ref033] 33.Beeler PE, Cheetham M, Held U, Battegay E. Depression is independently associated with increased length of stay and readmissions in multimorbid inpatients. European Journal of Internal Medicine. 2020;73:59–66. Epub 30.11.2019. doi: 10.1016/j.ejim.2019.11.012 [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref034] 34.Prina AM, Cosco TD, Dening T, Beekman A, Brayne C, Huisman M. The association between depressive symptoms in the community, non-psychiatric hospital admission and hospital outcomes: a systematic review. J Psychosom Res. 2015;78(1):25–33. Epub doi: 10.1016/j.jpsychores.2014.11.002 . [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0287234.ref035] 35.Sledge WH, Gueorguieva R, Desan P, Bozzo JE, Dorset J, Lee HB. Multidisciplinary Proactive Psychiatric Consultation Service: Impact on Length of Stay for Medical Inpatients. Psychother Psychosom. 2015;84(4):208–16. Epub doi: 10.1159/000379757 . [DOI] [PubMed] [Google Scholar]

[pone.0287234.ref036] 36.Sharpe M, Toynbee M, Walker J. Proactive Integrated Consultation-Liaison Psychiatry: A new service model for the psychiatric care of general hospital inpatients. General Hospital Psychiatry. 2020;66:9–15. doi: 10.1016/j.genhosppsych.2020.06.005 [DOI] [PubMed] [Google Scholar]

PERMALINK

Interaction of mental comorbidity and physical multimorbidity predicts length-of-stay in medical inpatients

Sophia Stahl-Toyota

Christoph Nikendei

Ede Nagy

Stefan Bönsel

Ivo Rollmann

Inga Unger

Julia Szendrödi

Norbert Frey

Patrick Michl

Carsten Müller-Tidow

Dirk Jäger

Hans-Christoph Friederich

Achim Hochlehnert

Roles

Abstract

Background

Objective

Design

Patients

Main measures

Conclusions

Introduction

Materials and methods

Study design and participants

Measures

Outcome length-of-stay

Physical multimorbidity

Mental comorbidity

Other variables

Statistical analyses

Ethical approval

Results

Characteristics of study participants

Table 1. Characteristics of the sample of internal medicine inpatients with and without mental comorbidity.

Length-of-stay with vs. without mental comorbidity

Length-of-stay by number of mental comorbidities

Fig 1. Length-of-stay in relation to number of mental comorbidities.

Length of stay differentiated for the F-category of mental comorbidity

Fig 2. Length-of-stay differentiated by mental comorbidity spectrum.

Length-of-stay for different age groups with and without mental comorbidity

Fig 3. Length-of-stay differentiated by age group and presence of mental comorbidity.

Length-of-stay in relation to physical multimorbidity with and without mental comorbidity

Fig 4. Length-of-stay for increasing Elixhauser score with and without mental comorbidity.

Negative binomial regression model with Elixhauser score and mental comorbidity presence

Table 2. Negative binomial regression analysis of length-of-stay.

Discussion

Conclusions

Supporting information

Data Availability

Funding Statement

References

Decision Letter 0

Sebastien Kenmoe

Roles

Author response to Decision Letter 0

Decision Letter 1

Sebastien Kenmoe

Roles

Acceptance letter

Sebastien Kenmoe

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases