Impact of early detection and management of emotional distress on length of stay in non-psychiatric inpatients: A retrospective hospital-based cohort study

Abstract

Background:

While emotional distress, encompassing anxiety and depression, has been associated with negative clinical outcomes, its impact across various clinical departments and general hospitals has been less explored. Previous studies with limited sample sizes have examined the effectiveness of specific treatments (e.g., antidepressants) rather than a systemic management strategy for outcome improvement in non-psychiatric inpatients. To enhance the understanding of the importance of addressing mental health care needs among non-psychiatric patients in general hospitals, this study retrospectively investigated the impacts of emotional distress and the effects of early detection and management of depression and anxiety on hospital length of stay (LOS) and rate of long LOS (LLOS, i.e., LOS >30 days) in a large sample of non-psychiatric inpatients.

Methods:

This retrospective cohort study included 487,871 inpatients from 20 non-psychiatric departments of a general hospital. They were divided, according to whether they underwent a novel strategy to manage emotional distress which deployed the Huaxi Emotional Distress Index (HEI) for brief screening with grading psychological services (BS-GPS), into BS-GPS (n = 178,883) and non-BS-GPS (n = 308,988) cohorts. The LOS and rate of LLOS between the BS-GPS and non-BS-GPS cohorts and between subcohorts with and without clinically significant anxiety and/or depression (CSAD, i.e., HEI score ≥11 on admission to the hospital) in the BS-GPS cohort were compared using univariable analyses, multilevel analyses, and/or propensity score-matched analyses, respectively.

Results:

The detection rate of CSAD in the BS-GPS cohort varied from 2.64% (95% confidence interval [CI]: 2.49%–2.81%) to 20.50% (95% CI: 19.43%–21.62%) across the 20 departments, with a average rate of 5.36%. Significant differences were observed in both the LOS and LLOS rates between the subcohorts with CSAD (12.7 days, 535/9590) and without CSAD (9.5 days, 3800/169,293) and between the BS-GPS (9.6 days, 4335/178,883) and non-BS-GPS (10.8 days, 11,483/308,988) cohorts. These differences remained significant after controlling for confounders using propensity score-matched comparisons. A multilevel analysis indicated that BS-GPS was negatively associated with both LOS and LLOS after controlling for sociodemographics and the departments of patient discharge and remained negatively associated with LLOS after controlling additionally for the year of patient discharge.

Conclusion:

Emotional distress significantly prolonged the LOS and increased the LLOS of non-psychiatric inpatients across most departments and general hospitals. These impacts were moderated by the implementation of BS-GPS. Thus, BS-GPS has the potential as an effective, resource-saving strategy for enhancing mental health care and optimizing medical resources in general hospitals.

Introduction

Numerous studies have demonstrated that emotional distress, including depression and anxiety, is common among non-psychiatric patients and is associated with negative mental health outcomes as well as adverse effects on the treatment, management, and prognosis of physical diseases.^[1–6] This results in greater financial burden and more severe functional impairment.^[7,8] Among non-psychiatric inpatients, these emotional disorders are linked to adverse hospital outcomes, such as prolonged hospital length of stay (LOS), readmission, and in-hospital and post-discharge mortality.^[9–13] However, it has been found that the recognition rate of anxiety and depression by non-psychiatric physicians, especially in China, is relatively low, with less than 20% of these disorders being identified,^[14–16] or even lower (4.0%) when physicians are not preinformed.^[16] This is significantly lower than the recognition rate in Western countries (32.5–64.3%).^[17–19]

Evidence from Western countries suggests that brief screening (BS) can enhance the accurate identification of emotional disorders and should be coupled with robust systems to ensure accurate diagnosis, effective treatment, and appropriate follow-up.^[20–23] In response, we developed the “Sunshine Hospital Project (SHP)” at the West China Hospital (WCH) of Sichuan University, China, to improve mental health services since 2015. This program, aimed at all non-psychiatric patients in the hospital, utilizes BS followed by a three-tier (i.e., primary, secondary, and tertiary) psychological service (GPS) strategy for the early detection and management of emotional distress in non-psychiatric inpatients during routine clinical practice [Supplementary Figure 1, http://links.lww.com/CM9/C417]. For the implementation of BS-GPS, staff from various departments who had completed basic counselor training or specialized psychiatric nurse training were recruited. They were willing and had the necessary skills to perform the required tasks. After recruitment, they received six weeks of training from professional psychotherapists and psychiatrists. The training consisted of two theoretical sessions and one practical session per week. The training covered the identification, assessment, and initial management of common psychiatric issues in clinical patients at general hospitals. It also included instruction on doctor–patient communication skills and crisis intervention. Upon completion of the training and passing the assessment, individuals were qualified to perform this work. Professional psychotherapists and psychiatrists provided monthly monitoring to improve the implementation of BS-GPS. Due to its practicability, the BS-GPS strategy has attracted great interest among mental health professionals in general hospitals in China, leading to the establishment of the Chinese League of Sunshine Hospitals (CLOSH). By 2018, more than 30 tertiary general hospitals from 14 provincial-level administrative regions had joined the League.

However, the practical BS-GPS strategy has not been widely adopted by other Chinese hospitals. Even within some CLOSH member hospitals, non-psychiatric medical staff and policy-makers are less motivated to promote mental health services across a broad range of clinical departments. This reluctance could be due to two factors. First, previous studies focusing on the association of emotional distress with negative hospital outcomes in non-psychiatric inpatients were primarily based on patients with specific conditions, such as cancer, peritoneal dialysis, and certain neurological and cardiovascular diseases,^[9–13] making broad generalizations to most clinical departments in a general hospital challenging. Second, only a handful of studies with small samples have examined the effectiveness of anxiety/depression interventions in improving hospital outcomes. These studies focused solely on specific methods used to treat depression among patients with particular physical diseases rather than a practicable systemic service strategy.^[9,10]

Given that both the LOS and the rate of long LOS (LLOS) are key clinical outcomes relevant to health costs and service utilization by patients and the public, policy-makers, and hospital staff,^[24–26] this study aimed to: (1) investigate the impacts of emotional distress and the effects of early detection and management of depression and anxiety on the LOS and LLOS rate in non-psychiatric inpatients and (2) provide better insights for the provision of medical care to meet the needs of non-psychiatric patients receiving general hospital care.

Methods

Subjects

This retrospective hospital-based cohort study was based on a real-world setting and was approved by the Ethics Committee on Biomedical Research, West China Hospital (WCH) of Sichuan University, Chengdu, Sichuan, China (No. 2019-364). The requirement for consent was waived because the study was a retrospective analysis of routinely collected administrative data. We extracted demographic data, including sex, age, and marital status, as well as department and the year of discharge for each subject, from the hospital information system (HIS). The study included inpatients aged ≥15 years who were admitted to and discharged from non-psychiatric departments of WCH between January 2015 and December 2018. We excluded patients from certain departments for the following reasons: fewer than 2000 person-times were screened by the Huaxi Emotional Distress Index (HEI) at admission (departments including gastroenterology, nuclear medicine, and intensive care units), the average LOS was less than 3 days (ophthalmology), and the rate of readmission was greater than 60% (oncology and hematology) during the study period. Consequently, our study included 487,871 subjects from 20 departments, including 12 non-surgical and 8 surgical departments [Figure 1].

Figure 1.

Flow diagram of subjects screening and grouping. BS: Brief screening; CSAD: Clinically significant anxiety/depression; GPS: Grading psychological service; HEI: Huaxi Emotional Distress Index; LLOS: Long hospital length of stay; LOS: Hospital length of stay.

Cohorts and measurements

Subjects were grouped into BS-GPS and non-BS-GPS cohorts depending on whether they had undergone BS-GPS. Subjects in the BS-GPS cohort were screened with the HEI on admission to the hospital, and their HEI screening scores were obtained through the HIS for this study. The HEI was developed for rapid screening of emotional distress, specifically anxiety and/or depression, and can be conveniently conducted as a routine assessment in non-psychiatric clinical settings. The scale includes nine items that assess anxiety and depression experienced in the last month and demonstrated good reliability and validity for screening for anxiety and depressive disorders based on the Mini International Neuropsychiatric Interview (MINI) in our previous study.^[27] Most Chinese patients can complete the scale within 5 min. We defined clinically significant anxiety and/or depression (CSAD) based on its optimal cut-off score (≥11), with a sensitivity of 0.880 and specificity of 0.766. Subsequently, subjects in the BS-GPS cohort were classified into non-CSAD and CSAD cohorts accordingly [Figure 1].

Our first outcome measurement, namely, LOS (days), was derived from the cover pages of the patients’ medical records in the HIS. The second outcome measurement, LLOS, was defined as an LOS exceeding 30 days, according to the definition provided by the medical administrative department of the WCH.

Statistical analyses

We used descriptive methods to calculate the general demographic information, means of the LOS, rates of CSAD identified by the HEI, year of discharge, and departments of discharge, and the LLOS rate. Beyond univariable comparisons of the means and rates using independent t tests, or chi-squared tests, we used Pearson correlation analysis to explore the association of the rate of CSAD with the LOS and LLOS across all 20 departments. Since the subjects who had undergone BS-GPS were not randomly selected, we conducted propensity score matching (PSM) with a one-to-one matching process (greedy-matching algorithm) to generate a matched comparison. This approach was used to reduce the effects of confounding factors such as demographics, year of discharge and departments of discharge. We assessed the balance of covariates between the non-BS-GPS and BS-GPS cohorts using standardized differences, considering a negligible imbalance when less than 10% of a baseline covariate differed.^[28] A two-stage strategy was used to differentiate the effects of controlling or matching for the year of discharge when performing the PSM comparisons. Furthermore, to account for variations in patient characteristics and department clustering, we conducted a multilevel analysis with the department as level 1 and the patient as level 2. This approach helped to minimize bias resulting from departmental differences. Model 0 assessed the variance in the LOS/LLOS without explanatory variables. Model 1 incorporated individual characteristics other than year of discharge to determine their association with LOS/LLOS. Model 2 included the patient’s year of discharge as an explanatory variable in addition to Model 1. The goodness of fit of the multilevel analysis was determined using the intraclass correlation coefficient (ICC), which measures the variability between each level. A larger ICC indicates greater explanatory power of the department differences affecting the LOS/LLOS.

The statistical significance of the above analysis was evaluated based on an alpha level of 0.05 from a two-tailed test using Stata 15.0 statistical software (StataCorp LLC, College Station, Texas, USA). Additionally, we created charts in Excel (Microsoft Corporation, Redmond, Washington, USA) to visualize the variation in the rates of CSAD and LLOS as well as LOS. The associations of CSAD and BS-GPS with LOS and LLOS across departments were based on the results from the Stata analysis.

Results

Demographics

This study included 487,871 non-psychiatric inpatients, of whom 178,883 (36.67%) underwent BS-GPS (i.e., the BS-GPS cohort) and 308,988 (63.33%) did not (i.e., the non-BS-GPS cohort). There were significant differences in sex, age, marital status, year of discharge, and departments of discharge between subjects in the BS-GPS and non-BS-GPS cohorts [Table 1].

Table 1.

Numbers and percentages of subjects who underwent the BS-GPS in the overall sample (n = 487,871) and those with CSAD in the BS-GPS cohort (n = 178,883) by demographics, year of discharge, and department of discharge.

Variables	Numbers in overall sample	BS-GPS in the overall sample			CSAD in the BS-GPS cohort
		Percentages (95% CI)	χ2 (P) between cohorts*	Adjusted OR (95% CI)†	Percentages (95% CI)	χ2 (P) between subcohorts‡	Adjusted OR (95% CI)†
Sex			283.34 (<0.001)			492.24 (<0.001)
Female	225,256	37.92 (37.72–38.12)		Reference	6.60 (6.43–6.77)		Reference
Male	262,615	35.59 (35.41–35.77)		0.92 (0.90–0.93)	4.23 (4.10–4.36)		0.61 (0.59–0.64)
Age (years)			984.58 (<0.001)			305.12 (<0. 001)
15–24	30,527	32.35 (31.83–32.88)		Reference	7.41 (6.91–7.95)		Reference
25–34	50,608	36.15 (35.73–36.57)		1.10 (1.05–1.14)	6.94 (6.58–7.31)		1.10 (0.98–1.23)
35–44	68,133	36.28 (35.92–36.64)		1.15 (1.10–1.21)	6.11 (5.81–6.41)		1.06 (0.94–1.20)
45–54	112,525	39.08 (38.80–39.37)		1.21 (1.16–1.27)	5.34 (5.14–5.56)		0.94 (0.84–1.06)
55–64	91,901	38.82 (38.51–39.14)		1.27 (1.21–1.33)	4.57 (4.35–4.79)		0.81 (0.71–0.91)
≥65	134,177	34.54 (34.28–34.79)		1.05 (1.01–1.10)	4.54 (4.35–4.73)		0.74 (0.65–0.83)
Marital status			3.0e + 03 (<0.001)			236.08 (<0.001)
Married	402,747	37.73 (37.58–37.88)		Reference	5.04 (4.93–5.15)		Reference
Single	40,035	35.05 (34.59–35.52)		0.96 (0.92–0.99)	7.53 (7.11–7.98)		1.24 (1.12–1.37)
Divorced	8318	42.02 (40.96–43.08)		1.05 (1.00–1.11)	7.38 (6.56–8.30)		1.27 (1.12–1.45)
Widowed	15,735	32.25 (31.52–32.98)		0.82 (0.79–0.86)	5.91 (5.30–6.60)		1.04 (0.92–1.17)
Other	3568	28.42 (26.96–29.92)		0.91 (0.83–1.00)	5.52 (4.27–7.11)		0.88 (0.67–1.16)
Missing	17,468	18.88 (18.31–19.47)		0.55 (0.53–0.58)	7.94 (7.07–8.92)		1.44 (1.25–1.65)
Year of discharge			1.3e + 05 (<0.001)			125.98 (<0.001)
2015	106,234	5.38 (5.24–5.51)		Reference	8.04 (7.36–8.77)		Reference
2016	113,696	24.03 (23.78–24.28)		6.83 (6.61–7.05)	6.07 (5.79–6.36)		0.67 (0.60–0.75)
2017	122,649	33.08 (32.81–33.34)		10.44 (10.13–10.76)	5.27 (5.06–5.49)		0.55 (0.49–0.62)
2018	145,292	72.46 (72.23–72.69)		68.81 (66.71–70.98)	5.07 (4.94–5.20)		0.52 (0.47–0.58)
Departments			3.5e + 04 (<0.001)			4.5e + 03 (<0.001)
Non-surgical departments	240,171	31.70 (31.52–31.89)	2.4e + 04 (<0.001)		7.63 (7.44–7.82)	2.1e + 03 (<0.001)
Rehabilitation medicine	10,543	49.55 (48.60–50.50)		Reference	20.50 (19.43–21.62)		Reference
Geriatrics	18,298	20.94 (20.36–21.54)		0.25 (0.24–0.27)	6.94 (6.18–7.79)		0.42 (0.36–0.48)
Endocrinology and metabolism	8064	56.83 (55.75–57.91)		2.08 (1.94–2.24)	11.70 (10.80–12.66)		0.53 (0.48–0.60)
Nephrology	28,311	25.44 (24.93–25.95)		0.30 (0.28–0.31)	8.99 (8.35–9.67)		0.42 (0.37–0.46)
Respiratory and infectious medicine	43,259	21.54 (21.15–21.93)		0.21 (0.20–0.23)	3.57 (3.22–3.97)		0.17 (0.15–0.19)
General medicine	10,425	38.26 (37.34–39.20)		0.98 (0.92–1.05)	6.09 (5.39–6.88)		0.28 (0.24–0.32)
Dermatology and venerology	7247	74.78 (73.76–75.76)		6.82 (6.30–7.37)	6.33 (5.71–7.01)		0.24 (0.21–0.28)
Integrated TCM and Western medicine	13,968	28.83 (28.08–29.59)		0.36 (0.34–0.38)	8.02 (7.22–8.90)		0.41 (0.36–0.47)
Rheumatology and immunology	12,401	26.51 (25.74–27.29)		0.33 (0.30–0.35)	8.67 (7.76–9.68)		0.34 (0.29–0.39)
Pain management	6614	52.45 (51.24–53.65)		1.47 (1.37–1.59)	4.04 (3.43–4.74)		0.18 (0.15–0.21)
Neurology	35,727	14.41 (14.05–14.77)		0.12 (0.12–0.13)	12.59 (11.71–13.53)		0.63 (0.56–0.70)
Cardiology	45,314	45.57 (45.11–46.03)		0.95 (0.90–1.00)	4.72 (4.44–5.02)		0.23 (0.21–0.26)
Surgical departments	247,700	41.48 (41.28–41.67)	7.2e + 03 (<0.001)		3.68 (3.57–3.80)	301.65 (<0.001)
Thoracic and cardiac surgery	24,487	38.73 (38.12–39.34)		0.64 (0.61–0.68)	3.91 (3.54–4.32)		0.17 (0.15–0.19)
Orthopedic surgery	35,761	58.08 (57.57–58.59)		2.10 (1.99–2.21)	4.94 (4.65–5.24)		0.20 (0.18–0.22)
Vascular surgery	6651	54.44 (53.24–55.64)		1.26 (1.17–1.36)	3.87 (3.29–4.55)		0.17 (0.14–0.20)
Burns and plastic surgery	8006	38.57 (37.51–39.64)		0.75 (0.70–0.81)	6.83 (6.00–7.78)		0.26 (0.23–0.31)
Abdominal surgery	102,938	36.85 (36.56–37.15)		0.60 (0.57–0.63)	2.64 (2.49–2.81)		0.11 (0.10–0.13)
Neurosurgery	27,925	31.10 (30.56–31.64)		0.40 (0.38–0.42)	3.58 (3.21–3.99)		0.15 (0.13–0.17)
Thyroid and breast surgery	21,592	43.09 (42.44–43.75)		0.70 (0.66–0.74)	4.06 (3.68–4.48)		0.14 (0.12–0.16)
Otolaryngology head and neck surgery	20,340	48.44 (47.75–49.12)		1.29 (1.22–1.37)	3.46 (3.12–3.84)		0.13 (0.11–0.14)
Overall	487,871	36.67 (36.53–36.80)			5.36 (5.26–5.47)

^*χ² values for each variable was calculated for comparison between the BS-GPS group and the non-BS-GPS group. ^†These odds ratios were based on multivariable regression analysis using gender, group of age, type of marital status, year of discharge and department of discharge as independent variables. ^‡χ² values for each variable was calculated for comparison between the groups with CSAD and without CSAD in the BS-GPS cohort. BS: Brief screening; CI: Confidence Interval; CSAD: Clinically significant anxiety/depression; GPS: Grading psychological service; TCM: Traditional Chinese medicine.

LOS and LLOS rates in the overall sample

The average LOS (days) and LLOS rates were 10.34 days and 3.24%, respectively. Both the average LOS [Supplementary Table 1, http://links.lww.com/CM9/C417] and LLOS rates [Supplementary Tables 2 and 3, http://links.lww.com/CM9/C417] significantly differed among groups stratified by demographics and year of discharge and varied significantly across the 20 departments.

Rates of CSAD by HEI screening in the BS-GPS cohort

Within the BS-GPS cohort, the average detection rate of CSAD was 5.36%. The rates significantly varied among groups stratified by demographic characteristics and year of discharge and significantly differed across the 20 departments [Table 1 and Figure 2]. The top five departments in terms of CSAD detection rate, in descending order, were the departments of rehabilitation medicine (20.50%), neurology (12.59%), endocrinology and metabolism (11.70%), nephrology (8.98%), and rheumatology and immunology (8.67%), all of which were from nonsurgical departments. Among the surgical departments, the department of burns and plastic surgery had the highest rate of CSAD (6.83%). The average rate of CSAD in the non-surgical departments (7.63%) was almost double that in the surgical departments (3.68%). The pattern of differences in the unweighted rates of CSAD was similar to the pattern of its rates weighted according to demographics, year of discharge, and departments of discharge [Supplementary Table 1, http://links.lww.com/CM9/C417].

Figure 2.

Percentage of identified CSAD, average LOS, and percentage of LLOS in BS-GPS cohort. The 12 non-surgical and 8 surgical departments are presented in descending order of the average LOS. BS: Brief screening; CI: Confidence interval; CSAD: Clinically significant anxiety/depression; GPS: Grading psychological service; LLOS: Long hospital length of stay; LOS: Hospital length of stay; TCM: Traditional Chinese medicine.

Longitudinal associations of CSAD with LOS and LLOS

Within the BS-GPS cohort, the average LOS [Supplementary Table 2, http://links.lww.com/CM9/C417] in the CSAD subcohort (12.68 days) was approximately 1.5 times that in the non-CSAD subcohort (9.45 days), and the LLOS rate [Supplementary Table 3, http://links.lww.com/CM9/C417] in the CSAD subcohort (5.58%) was more than twice that in the non-CSAD subcohort (2.24%).

Further analysis revealed that the strong longitudinal association between CSAD and either LOS or LLOS at two levels. First, the departments with higher detection rates of CSAD tend to have longer LOS and a higher proportion of LLOS [Figure 2]. Correlation analysis showed that CSAD was significantly associated with both LOS (r = 0.749, P <0.001) and LLOS (r = 0.743, P <0.001) of the 20 departments. Second, upon further comparison, a longer average LOS and greater LLOS rate in the CSAD subcohort than in the non-CSAD subcohort were also observed in most individual departments [Figure 3].

Figure 3.

Average LOS and percentage of LLOS in subcohorts with and without CSAD in BS-GPS cohort. BS: Brief screening; CI: Confidence interval; CSAD: Clinically significant anxiety/depression; GPS: Grading psychological service; LLOS: Long hospital length of stay; LOS: Hospital length of stay; TCM: Traditional Chinese medicine.

Associations between BS-GPS and LOS and LLOS

Among all the subjects, the average LOS [Supplementary Table 2, http://links.lww.com/CM9/C417] in the BS-GPS cohort (9.63 days) was shorter than that in the non-BS-GPS cohort (10.76 days), and the LLOS rate [Supplementary Table 3, http://links.lww.com/CM9/C417] in the BS-GPS cohort (2.42%) was less than two-thirds that in the non-BS-GPS cohort (3.72%).

The results of the propensity score-matched comparisons are listed in Table 2. When only demographics and discharge departments were used as matching variables, 167,284 subjects were matched in the BS-GPS and non-BS-GPS cohorts respectively. The average LOS (9.58 days vs. 10.40 days]) and LLOS percentages (2.48% vs. 3.81%) in the matched BS-GPS cohort were significantly shorter/lower than those in the matched non-BS-GPS cohort (all P <0.001). When the year of discharge was additionally matched, 102,987 subjects were matched for the BS-GPS and non-BS-GPS cohorts respectively. The differences in the LOS (9.87 days vs. 10.00 days) and LLOS percentages (2.79% vs. 4.18%) between the matched cohorts also remained significant (P = 0.009 and <0.001, respectively).

Table 2.

Summary of subjects’ characteristics between BS-GPS and non-BS-GPS in PSM-matched cohorts with stepwise adjustment for demographics, department of discharge, and year of discharge.

Variables	Demographics and department of discharge matched (n = 334,568)			Year of discharge was additionally matched (n = 205,974)
	non-BS-GPS (n = 167,284) (n [%])	BS-GPS (n = 167,284) (n [%])	SD (%)*	non-BS-GPS (n = 102,987) (n [%])	BS-GPS (n = 102,987) (n [%])	SD (%)*
Sex
Female	78,857 (47.1)	77,993 (46.6)	–1.0	49,702 (48.3)	50,402 (48.9)	1.4
Male	88,427 (52.9)	89,291 (53.4)	1.0	53,285 (51.7)	52,585 (51.1)	–1.4
Age (years)
15–24	9314 (5.6)	9852 (5.9)	1.4	6055 (5.9)	6786 (6.6)	2.9
25–34	16,770 (10.0)	17,723 (10.6)	1.9	11,094 (10.8)	11,538 (11.2)	1.4
35–44	23,336 (13.9)	24,176 (14.5)	1.4	15,226 (14.8)	14,514 (14.1)	–2.0
45–54	41,239 (24.7)	37,762 (22.6)	–4.9	25,228 (24.5)	24,157 (23.5)	–2.4
55–64	32,829 (19.6)	31,922 (19.1)	–1.4	19,323 (18.8)	18,837 (18.3)	–1.2
≥65	43,796 (26.2)	45,849 (27.4)	2.8	26,061 (25.3)	27,155 (26.4)	2.4
Marital status
Married	142,127 (85.0)	141,372 (84.5)	–1.3	86,347 (83.8)	85,032 (82.6)	–3.4
Single	13,211 (7.9)	13,679 (8.2)	1.0	8408 (8.2)	9139 (8.9)	2.5
Divorced	3380 (2.0)	2902 (1.7)	–2.1	1903 (1.8)	1662 (1.6)	–1.8
Widowed	4991 (3.0)	5059 (3.0)	0.2	3459 (3.4)	3446 (3.3)	–0.1
Other	1091 (0.7)	974 (0.6)	–0.9	747 (0.7)	807 (0.8)	0.7
Missing	2484 (1.5)	3298 (2.0)	3.7	2123 (2.1)	2901 (2.8)	4.9
Year of discharge
2015				5782 (5.6)	5674 (5.5)	–0.5
2016				22,667 (22.0)	23,861 (23.2)	2.8
2017				36,869 (35.8)	35,128 (34.1)	–3.5
2018				37,669 (36.6)	38,324 (37.2)	1.3
Departments
Non-surgical departments
Rehabilitation medicine	5153 (3.1)	5224 (3.1)	0.2	3790 (3.7)	2877 (2.8)	–5.0
Geriatrics	3847 (2.3)	3832 (2.3)	–0.1	2939 (2.9)	2926 (2.8)	–0.1
Endocrinology and metabolism	3299 (2.0)	3603 (2.2)	1.3	1607 (1.6)	2439 (2.4)	5.8
Nephrology	7289 (4.4)	7201 (4.3)	–0.3	2303 (2.2)	2477 (2.4)	1.1
Respiratory and infectious medicine	9360 (5.6)	9317 (5.6)	–0.1	8169 (7.9)	8282 (8.0)	0.4
General medicine	4171 (2.5)	3989 (2.4)	–0.7	3291 (3.2)	3058 (3.0)	–1.3
Dermatology and venerology	1828 (1.1)	2031 (1.2)	1.1	1593 (1.5)	1579 (1.5)	–0.1
Integrated TCM and Western medicine	4066 (2.4)	4027 (2.4)	–0. 2	1049 (1.0)	1403 (1.4)	3.2
Rheumatology and immunology	3296 (2.0)	3287 (2.0)	0.0	3051 (3.0)	3094 (3.0)	0.2
Pain management	2957 (1.8)	3456 (2.1)	2.2	1388 (1.3)	1913 (1.9)	4.1
Neurology	5148 (3.1)	5147 (3.1)	0.0	5282 (5.1)	4998 (4.9)	–1.3
Cardiology	21,202 (12.7)	20,648 (12.3)	–1.0	10,724 (10.4)	11,846 (11.5)	3.5
Surgical departments
Thoracic and cardiac surgery	9614 (5.7)	9483 (5.7)	–0.3	5523 (5.4)	4980 (4.8)	–2.4
Orthopedic surgery	14,975 (9.0)	13,583 (8.1)	–3.0	8984 (8.7)	8057 (7.8)	–3.3
Vascular surgery	3014 (1.8)	3591 (2.1)	2.5	2396 (2.3)	1952 (1.9)	–3.0
Burns and plastic surgery	3318 (2.0)	3088 (1.8)	–1.0	1933 (1.9)	1729 (1.7)	–1.5
Abdominal surgery	38,060 (22.8)	37,936 (22.7)	–0.2	19,764 (19.2)	19,314 (18.8)	–1.1
Neurosurgery	8822 (5.3)	8684 (5.2)	–0.4	4601 (4.5)	4853 (4.7)	1.2
Thyroid and breast surgery	8289 (5.0)	9305 (5.6)	2.7	6316 (6.1)	7116 (6.9)	3.1
Otolaryngology head and neck surgery	9576 (5.7)	9852 (5.9)	0.7	8284 (8.0)	8094 (7.9)	–0.7

^*These SDs (non-BS-GPS vs. BS-GPS) are reported as percentages, for which a difference of less than 10% reveals a negligible imbalance. BS: Brief screening; CI: Confidence interval; GPS: Grading psychological service; LLOS: Long LOS (LOS more than 30 days); LOS: Hospital length of stay; PSM: Propensity score matching; SD: Standardized difference; TCM: Traditional Chinese medicine.

Multilevel analysis [Supplementary Table 4, http://links.lww.com/CM9/C417] indicated that the implementation of BS-GPS was negatively associated with the LOS and LLOS after controlling for the confounding effects of demographics and departments of discharge (P <0.001). However, after further adjusting for the year of discharge, this association of LOS became non-significant, while the association with LLOS remained significant (P <0.001). The shorter average LOS and lower LLOS rates in the BS-GPS cohort than in the non-BS-GPS cohort were also observed in the majority of departments, though a significantly longer LOS in the BS-GPS cohort relative to the non-BS-GPS cohort was observed in five departments (Nephrology, Neurology, Thyroid & Breast Surgery, Otolaryngology Head & Neck Surgery, and Vascular Surgery) [Figure 4].

Figure 4.

Average LOS and percentage of LLOS in BS-GPS and non-BS-GPS cohorts. BS: Brief screening; CI: Confidence interval; GPS: Grading psychological service; LLOS: Long hospital length of stay; LOS: Hospital length of stay; TCM: Traditional Chinese medicine.

Discussion

This study represents a pioneering investigation into both the rates of anxiety and depression at admission and their longitudinal impacts on hospital outcomes in an extensive sample of non-psychiatric inpatients. We further probed whether an early, BS and management strategy for emotional distress could mitigate the influence of anxiety and depression on hospital outcomes. This approach encompassed a broad array of independent clinical departments, as well as the hospital as a whole. The knowledge gained from this investigation helps to underpin the need for enhanced mental health services for non-psychiatric inpatients across diverse clinical departments and general hospitals.

A key finding from this study is the considerable variation in the rates of clinically significant CSAD across the 20 clinical departments [Figure 2]. This finding has important implications for the optimal allocation of mental health resources within each non-psychiatric department in a general hospital. In this study, we have charted the sequence of CSAD rates in 20 independent clinical departments within a general hospital. Our findings align with previous studies demonstrating the strong association of emotional disorders with certain chronic physical conditions, such as stroke, multiple sclerosis, epilepsy, diabetes complications, renal failure, and rheumatic diseases.^[10–13,29] We found that the CSAD rates in departments specializing in these conditions (i.e., the departments of neurology, endocrinology and metabolism, nephrology, and rheumatology and immunology) ranked from second to fifth. However, our study also revealed new, though logical, observations. We discovered that the rehabilitation medicine department reported the highest prevalence of CSAD (20.50%) within the general hospital, a finding not previously highlighted in prior research. Additionally, the burns and plastic surgery department had the highest CSAD rate (6.83%) among the surgical departments, which is another novel yet understandable finding. In general, clinical departments with a higher prevalence of CSAD, such as the rehabilitation medicine, neurology, endocrinology and metabolism, nephrology, rheumatology and immunology, and burn and plastic surgery departments, would benefit more from prioritizing the allocation of mental health resources.

Second, this study highlights the previously unknown associations between CSAD and LOS and/or LLOS across the majority of clinical departments and at the whole hospital level. These findings strongly urge health policy-makers and non-psychiatric clinicians to place greater emphasis on the effects of emotional disorders on outcomes in non-psychiatric inpatients. While previous research has shown that patients with comorbid physical and emotional disorders tend to have negative hospital outcomes, these studies have largely focused on a limited number of specific physical conditions.^[10–13] These findings have not adequately driven most non-psychiatric medical staff and policy-makers in general hospitals to implement mental health care across broad groups of clinical departments, as indicated in our study.

Third, the discovery that BS-GPS is associated with a shorter LOS or a lower LLOS rate provides a promising avenue for providing mental health care in non-psychiatric medical wards in a cost-effective and effective manner. Only a handful of studies with small sample sizes have reported the effectiveness of specific treatment modalities (such as medications) for depression in improving hospital outcomes among inpatients with specific physical diseases. For example, Wong et al^[10] studied the use of antidepressants to reduce the impact of depressive symptoms on LOS in patients with advanced cancer. In contrast, our study evaluated a large hospital-based cohort and revealed that an early brief screening and management strategy for emotional distress (i.e., the BS-GPS) could mitigate the negative effect of CSAD on LOS and LLOS across most clinical departments and throughout a general hospital. The effect on LLOS was particularly compelling because the association between BS-GPS and LLOS remained statistically significant even after accounting for confounders, such as demographics, departments of discharge, and years of discharge, in both the multilevel analysis and propensity score-matched comparisons. The association between CSAD and LOS seemed less robust because it became non-significant after further adjusting for years of discharge in the multilevel analysis. However, it is noteworthy that controlling for the effect of the year of discharge might lead to a conservative estimate of BS-GPS effectiveness, as the shorter LOS and lower rates of LLOS in more recent years may be partly attributed to the increasing use of BS-GPS. Furthermore, the propensity score-matched comparisons reinforced the effectiveness of BS-GPS on LOS, as the shorter LOS in the BS-GPS cohort remained statistically significant after matching for demographics, department of discharge, and even the year of discharge.

In a classical liaison psychiatry model, psychiatrists or other mental health professionals are deeply integrated into diagnosis and treatment teams for patients seeking services in non-psychiatric clinical settings.^[30–32] While the liaison psychiatry model is more prevalent in developed countries, it is not common in many developing nations. The key challenges include a significant shortage of qualified mental health professionals,^[30,31] a lack of mental health awareness among non-psychiatric physicians in developing countries such as China,^[15] and a lower tendency among patients in general hospitals to proactively discuss emotional problems.^[33,34] The effective BS-GPS strategy employed in the present study was developed and utilized at WCH to address significant barriers to recognizing and treating emotional disorders in non-psychiatric clinical settings in China, thus substantially improving treatment rates.^[27] Designed for both effectiveness and practicability, the BS-GPS strategy saves human resources and is financially affordable, making it particularly promising for hospitals where a liaison psychiatry model is not commonly available. For example, the WCH has been able to implement BS-GPS in each clinical department with only half of a nurse’s manpower. The number of psychiatric consultations for inpatients in non-psychiatric departments increased from approximately 3000 person-times per year before 2015 to approximately 5000 person-times in 2018.

Furthermore, the demographic correlates of CSAD and LOS discovered in this study, which largely align with the existing knowledge, lend additional validity to our findings. For instance, higher rates of depression and anxiety among females, younger individuals, and widowed populations have been reported in previous studies.^[35,36] Associations between a longer LOS and male sex and older people in a Chinese tertiary general hospital have also been previously reported.^[37]

This study has several limitations. First, this was a single-center study with a large sample size. Second, the BS-GPS grouping should have been based on both HEI scores and other specific symptoms, such as suicidalities, psychoses, somatic symptoms, and hypochondriasis. However, these data were not available for our study, making it unclear how many patients were in each group and the differences in LOS among them. Third, while other health economics indices may also be important, we were unable to include them in this study for various reasons. For instance, the service fee per admission is unavailable, and the number of hospitalizations can only be obtained after long-term follow-up. Further research is needed to explore the direct impact of BS-GPS on patient hospitalization costs and the number of long-term hospitalizations. Furthermore, although this study provides clinical evidence with external validity from the perspective of that the subjects and conditions are more representative of real-world clinical practice,^[38,39] it may present more challenges in controlling for confounders than randomized controlled trials because of a longitudinal retrospective cohort study based on real-world clinical practice. For instance, this study involved a wide range of diseases, including medical/surgical conditions, as well as acute and chronic diseases. Although the use of both multilevel analysis and PSM comparisons substantially contributed to controlling confounders in the study,^[30] factors such as the severity of the disease, interventions, level of engagement of department staff, and the department’s workload, which may have an impact on the LOS, were not measured or controlled for in this study. Especially noteworthy, the severity of emotional distress of the non-BS-GPS cohort was not measured and was unavailable as a controllable confounder in this study. It is possible that the longer LOS and higher LLOS rate were partly attributable to a higher rate of CSAD of the non-BS-GPS cohort relative to the BS-GPS cohort, which also underscores the importance of addressing mental health care needs among non-psychiatric patients in general hospitals. Future randomized controlled studies focused on specific diseases should explore these confounding factors. Moreover, while a significantly longer LOS in the BS-GPS cohort relative to the non-BS-GPS cohort was observed in five departments, seemingly inconsistent with the overall findings, this discrepancy can be partially explained by the valuable time spent addressing, rather than ignoring, emotional disorders. This approach may have positive and sustained effects on patient recovery in the longer term.^[23]

In conclusion, the present study highlights the substantial impact of emotional distress on the length of hospital stays and the likelihood of prolonged hospitalization among non-psychiatric inpatients. The results also demonstrate the potential of the BS-GPS strategy to moderate these effects while offering resource-saving benefits. These findings emphasize the importance of integrating mental health care into general hospital settings and provide a compelling rationale for the widespread adoption of effective strategies, such as the BS-GPS approach, to promote mental health and optimize patient outcomes in non-psychiatric clinical settings.

Acknowledgments

We thank all the colleagues who participated in this study and the members of the Collaborative Group of Sunshine Hospital Project.

Funding

This work was partly funded by the National Natural Science Foundation of China Key Project (Nos. 81630030 and 81920108018); the Special Foundation for Brain Research from Science and Technology Program of Guangdong (Nos. 2018B030334001); and the National Key Research & Development Program, Ministry of Science and Technology, China (Nos. 2016YFC0904300); 1.3.5 Project for Disciplines of Excellence, West China Hospital of Sichuan University (Nos. ZY2016103, ZY2016203, and ZYGD20004).

Conflicts of interest

None.

Supplementary Material

cm9-138-2974-s001.docx ^{(554.8KB, docx)}