Evaluation of hospital length of stay and revenues as a function of admission mode,clinical pathways including observation unit stay and hospitalization characteristics

Abstract

Objectives

Hospital length of stay (days) and revenues per day (euros) could be different depending on admission mode. To determine the impact of admission mode as a function of clinical pathway, we conducted the present study.

Data sources: We included 159,206 admissions to three academic hospitals during a four-year period. Data were obtained from the electronic system of the hospital trust.

Study design

A case (through-emergency department)–control (elective (EA)) study was conducted (77,052), matched by age, stay severity and type, disease-related group, and discharge mode.

Principal findings: Through-emergency department were significantly elderly, more severe, had more intensive care stays, a higher mortality rate, longer length of stay (days) (9.5 ± 12 vs. 6.8 ± 9.5; p < 0.0001), and lower revenues per day (647 ± 451 vs. 721 ± 422; p = 0.01). In case–control study, mean differences between cases and controls were: longer length of stay −0.64 and revenues per day −75.6; for ≥75 years −1.2 and −102.1; medical −0.9 and −90.4; and discharge to facilities care centers −1.5 and −81.8. Among cases, 40% had a stay in observation unit before being admitted in hospital ward. Differences were strongly reduced for patients who did not go to observation unit before being admitted. Differences were reduced from 0.64 to 0.2 days for length of stay and from 79 to 41 euros for revenues per day when patients did not stay in observation unit before being admitted.

Conclusions

We conclude that admission mode is associated with length of stay and revenues. However, as differences are weak, elective admissions should not be prioritized on economic arguments. Otherwise, our study indicates that among through-emergency department admissions, observation unit stay is associated with longer length of stay and lower revenues.

Keywords

emergency department health economics hospital length of stay hospital revenues observation unit

Introduction

The rates of acute care hospital admissions in the European countries and the United States were quite dissimilar; however, the number of admissions are decreasing.^1,2 There are two hospital admission modes: Elective or planned admissions and through-emergency department (ED) admissions or unplanned. These two admission modes compete for insufficient or misused hospital beds.³ It has been reported that for some hospitals, length of stay (LOS) and revenues may explain the priority given to elective admissions.⁴ As most hospitals prioritize elective admissions,⁵ through-ED admissions patients stay in the ED.⁶ This situation is a major ED overcrowding cause which leads to a diminished quality of care, the increase of adverse events, the extension of in-hospital LOS and a rise in hospital cost.⁷

Patient’s complexity,⁸ diseases related-groups (DRGs) and some diagnoses⁹ have been associated with increased hospital LOS and additional costs. The question of the influence of admission mode has also been raised, but with discording results concerning the LOS¹⁰ and revenues per stay (RPS) and revenues per day (RPDs).¹¹ Elderly patients have increased LOS,¹² but few data are available concerning the revenues.¹³ As links between LOS and revenues are complex, hospital LOS and charges are correlated, but cost and revenues are not independent of LOS,^14,15 and hence their analyses are difficult. Furthermore, it is currently accepted that LOS and cost and revenues are skewed, multimodal and vary significantly with other covariates.¹⁶

Due to previous publications, we hypothesized that the admission mode may affect length of stay and revenues, but also that other demographic or hospital stay characteristics and clinical pathways including Observation unit stay could be associated with hospital length of stay and costs. Our main objective is to determine the influence of the admission mode and those covariates on length of stay and hospital revenues by using a new validated case-control study approach.^17,18 We believe that this approach could help identify populations with longer lengths of stay and higher costs, where it will be possible to target cost reduction strategies.

Methods

Study design

We conducted a retrospective multi-center observational study in three academic hospitals during a four-year study period (1 January 2010 to 31 December 2013). A case–control study was conducted, where “cases” were through-ED admissions and “controls” elective admissions.

Setting

The study was performed in an academic hospital trust, including three public academic hospitals located in urban areas. Respectively, each hospital has 961, 461, and 489 beds and an emergency department with around 75,000, 30,000, and 25,000 visits a year.

Methods and measurements

Data were collected using the hospital discharge database Programme de médicalisation des systèmes d'information (PMSI). This database collects patient’s characteristics (age, sex, domiciliation); admission mode (through-ED admission, elective admission); stay in the OU; date of entry and discharge, LOS; stay category: medical, surgical or interventional; Critical, intensive or interventional care stay; primary diagnosis which is the main diagnosis that led to hospitalization, and associated diagnoses which reflect comorbidities; destination at discharge or final disposition (home, acute care, facilities care, home care, death). Severity score (0–4) was determined from the DRG.¹⁹

French public and private hospitals funding depends on a DRG-based payment system (called T2A, Tarification à l'activité). The French equivalent of DRG is called GHM (Groupe Homogène de Malades) and depends on primary and associated diagnoses along with procedures performed during hospitalization. Cases classified as belonging to a particular GHM are characterized by a homogenous resource consumption pattern. A tariff called the GHS (Groupe Homogène de Séjour) is assigned to each GHM.²⁰ Average costs per GHM are determined using information from a sample of public and private hospitals. Corrections are then applied according to regional factors and fixes price/volume to adapt supply to national health expenditure targets. As costing tools applicable in other countries have not been validated in France, we preferred the use of the model of total revenues, which brings revenues and total hospital costs.²¹

Total RPS were calculated by adding to the T2A tariff additional incomes due to outliers LOS, and critical and intermediate care stays. RPDs were calculated by dividing total revenues by LOS.

Selection of participants

All admissions during the study period were included. We have excluded pediatric cases (age <15 years), all direct critical, intermediate, and intensive care admissions; patients admitted for obstetrical causes; patients admitted for cardiac, pulmonary or hepatic transplant, and patients admitted from ED to OU and discharged directly from OU. To increase homogeneity and robustness of our analyses, we eliminated LOS extreme values (<5th percentile and >95th percentile).

Case–control study

To each case (through-ED-admitted patient), we matched a control on several variables known to be associated with LOS: age (by category: 15–74 years and ≥75 years), type of hospital care (surgical, interventional (non-invasive procedure) or medical), severity index (0 to 4), DRG, and discharge mode (home, acute care, facilities care, home care, death). Matching process was conducted as follows: we established first a pool of eligible cases from the hospital discharge database. Controls were chosen from the same database, among patients who were admitted directly in hospital wards. We created two distinct tables with index variables for all matching criteria. Then, we used a computerized program (proc surveyselect from SAS®) to randomly draw pairs of cases and controls, similar on each matched stratum. The program used simple random sampling without replacement to choose among eligible controls. The matching design allows comparing similar patients who differ only by their admission modes. Moreover, increasing the similarity on clinical aspects and hospital pathway patterns should increase the comparability of patients on unknown factors such as associated comorbidities. Patients of our case–control sample studies were representative of all patients admitted in the three hospitals.

In a subgroup analysis, we separated through-ED patients: those with a stay in the OU before being admitted in hospital wards and those without. Then we compared those two subgroups with their controls to estimate differences in LOS and revenues.

Statistical analysis

Continuous variables are presented using means ±standard deviation (SD) and categorical variables as frequency and percentage of the total (%). To compare study groups, we used paired Student tests for continuous variables and McNemar Chi-square test for categorical data. Multifaceted analysis on both LOS and hospital revenues were built. LOS and revenues were difficult outcomes to model because of its skewness and multimodal distribution. As previously reported, LOS is a major determinant of hospital cost.^15,16 Thus, we used logarithmic transformation and built a model on log transformed LOS and dependent variables of interest. We selected the relevant variables based on the adjusted R², with a stepwise approach.

Mean differences between average cases and controls were calculated for LOS and RPDs as well as 95% confidence interval were calculated as previously reported.²² p values were two-tailed, and for each analysis, a p value of <0.05 was considered significant.

All data were completely anonymous and currently used for economic evaluation. This non-interventional study was done with the approval of Institutional Emergency ethics, informatics and research Committee.

Results

Overall, 187,456 hospital stays were registered during the study period. Of them, 28,250 were excluded and 159,206 were included. Figure 1 represents the flowchart of the study population.

Figure 1.

Study flowchart.

Through-ED represented 29.2% of total admissions. Elderly patients represented 23.2% of total admissions, but were more often admitted through-ED (30% of the through-ED admissions vs. 15% of elective admissions).

Table 1 presents main characteristics of the entire study population. Patients admitted trough-ED were mostly medical patients (68.7%), with increased severity, had more Intensive or Intermediate or Critical Care stays, and they had a higher in-hospital mortality rate (5.8% compared to 1.2% among elective patients; p<0.001). LOS was significantly longer in through-ED admissions, while their RPS and RPDs were lower than in elective admissions.

Table 1.

Main characteristics of the entire study population (159,206 admissions).

	Elective admissions	Through-ED admissions
	n=112,769 (70.8%)	n=46,437 (29.2%)
	mean ± SD or n (%)	mean ± SD or n (%)	p
Age	55.8±17.6	58±22.7	<0.001
Age groups			<0.001
<75 years	109,519 (84.7)	40,254 (70)
≥75 years	19,832 (15.3)	17,224 (30)
Stay categories			<0.001
Surgical	48,821 (37.7)	11,929 (20.8)
Interventional	8634 (6.7)	1825 (3.2)
Medical	66,519 (51.4)	39,483 (68.7)
Non-classified	5377 (4.2)	4241 (7.4)
Severity level			<0.001
0	14,900 (11.5)	4912 (8.5)
1	60,127 (46.5)	19,356 (33.7)
2	28,127 (21.7)	13,427 (23.4)
3	12,944 (10)	12,486 (21.7)
4	4396 (3.4)	4419 (7.7)
Other/non-classified	8857 (6.8)	2878 (5)
Stay in
Intensive care	2183 (1.7)	4835 (8.4)	<0.001
Intermediate care	3612 (2.8)	2198 (3.8)	<0.001
Critical care	4650 (3.6)	2784 (4.8)	<0.001
Direct payment of hospital costs			<0.001
Without social security	5320 (4.1)	2668 (4.6)
Direct payment of hospital costs	1283 (0.9)	1635 (2.8)
Social security	121,539 (94)	52,430 (91.2)
Final disposition decision			<0.001
Home	112,532 (87)	43,003 (74.8)
Residence	56 (0.1)	129 (0.3)
Transfer	4857 (3.8)	3811 (6.6)
Facilities care	9975 (8.1)	6733 (11.7)
Home service	88 (0.2)	81 (0.2)
Death	1578 (1.2)	3336 (5.8)	<0.0001
LOS (mean±SD)	6.8±9.7	9.5±12.0	<0.0001
Revenues per stay (mean±SD)	4131±4965	4070±3994	0.02
Revenues per day (mean±SD)	721±422	647±451	0.01

LOS: length of stay; ED: emergency department.

Predictors of LOS

Table 2 presents the multifaceted analysis of LOS-associated factors. As previously reported, LOS is difficult to analyze by using multivariable or multifaceted methods.^14–16 Thus, we used logarithmic transformation and built a model on log transformed LOS and dependent variables of interest. We selected the relevant variables based on the adjusted R², with a stepwise approach (Table 3). This method allowed us to define the independent variables having the strongest effect on LOS. Our data indicate that the severity of the patient’s state, as well as his destination after discharge (rehabilitation facilities or long-term care facilities) increased the LOS.

Table 2.

Final multivariate models for LOS.

Parameter	Estimate	Standard error	t	Pr > \|t\|	Estimate (original scale)
Intercept	1.12549	0.00263	427.52	<0.0001	3.08174167
Observation unit stay	0.32032	0.00594	53.91	<0.0001	1.37757152
Age category (≥75 years)	0.14347	0.00494	29.07	<0.0001	1.15426951
Severity level	0.88306	0.00529	166.95	<0.0001	2.4182844
Transfer to rehabilitation	0.50154	0.00698	71.86	<0.0001	1.65125751
Surgical stay	0.28399	0.00401	70.74	<0.0001	1.32841441

Table 3.

Selection of explicative variables for LOS based on adjusted R².

Step	Effect entered	Adjusted R²
0	Intercept	0.00001
1	Severity index	0.1892
2	Transfer to facility care	0.2231
3	Surgical stay	0.2397
4	Observation unit stay	0.2547
5	Intensive care stay	0.2612
6	Age more than 75 years	0.2703
7	Death during hospital stay	0.2750
8	Discharge at home	0.2786
9	Medical stay	0.2795
10	Through-ED admission	0.2796
11	Transfer to acute care center	0.2797
12	Absence of French health insurance	0.2798

ED: emergency department.

Case–control study

The case–control study (Table 4) included 76,512 stays: through-ED (cases=38,256 stays) and elective admissions (controls=38,256 stays). Overall, elective admissions have shorter LOS (−0.64 days) and higher RPDs (+76 euros). We found that some covariates were associated with longer LOS among through-ED admissions: age ≥75 years, interventional and medical stays, middle severity (levels 2 and 3), and final disposition to facilities care. Concerning RPDs, the difference was minimal for surgical stays (18 euros), while it was greater for medical stays (90 euros) and maximal for interventional stays (164 euros).

Table 4.

Case–control study: length of stay and revenues per day as a function of covariates (n=76,512).

	Elective		Through-ED		p	Difference	95% CI lower	95% CI upper
	Mean	SD	Mean	SD	p	Difference	95% CI lower	95% CI upper
Length of stay
Overall population	6.8	5.4	7.4	5.8	<0.001	−0.64	−0.70	−0.58
Age group
<75 years	6.2	5.2	6.7	5.5	<0.001	−0.5	−0.6	−0.4
≥75 years	8.6	5.8	9.8	6.1	<0.001	−1.2	−1.3	−1.0
Stay category
Surgical	7.1	5.3	7.0	5.9	0.500	0.0	−0.1	0.2
Interventional	5.9	5.3	6.9	5.6	<0.001	−0.9	−1.2	−0.7
Medical	6.7	5.5	7.6	5.8	<0.001	−0.9	−0.9	−0.8
Severity level
0	1.1	0.25	1.1	0.28	0.009	0.0	0.0	0.0
1	4.1	3.3	4.5	3.7	<0.001	−0.3	−0.4	−0.3
2	8.0	4.8	9.0	5.4	<0.001	−1.1	−1.2	−1.0
3	11.1	5.7	11.9	5.7	<0.001	−0.7	−0.9	−0.6
4	13.2	5.9	13.5	5.7	0.134	−0.3	−0.7	0.1
Other/non-classified	7.4	6.2	8.9	6.5	<0.001	−1.5	−2.0	−1.1
Final disposition decision
Home/residency	6.1	5.0	6.8	5.4	<0.001	−0.7	−0.7	−0.6
Facilities care	10.8	6.1	12.2	6.4	<0.001	−1.5	−1.7	−1.2
Transfer to acute care	9.1	6.4	8.6	6.5	0.003	0.5	0.2	0.8
Home care	13.7	6.1	12.7	6.3	0.509	0.9	−2.0	3.9
Death	10.2	6.5	9.7	6.6	0.06	0.5	0.0	1.0
Revenues per day
Overall population	742.7	460.3	667.1	429.2	<0.001	−75.6	−81.2	−70.1
Age group
<75 years	754.7	456.3	687.0	438.5	<0.001	−67.7	−74	−61.3
≥75 years	703.0	471.1	600.8	389.5	<0.001	−102.1	−113.6	−90.7
Stay category
Surgical	920.5	531.5	902.6	511.9	0.006	−17.9	−30.5	−5.2
Interventional	1025.8	816.3	860.2	507.7	<0.001	−165.7	−212.6	−118.8
Medical	666.4	376.3	575.9	352.5	<0.001	−90.4	−96.3	−84.6
Severity level
0	603.6	132.9	626.9	173.8	<0.001	23.3	14.9	31.8
1	761.1	487.9	693.1	448.3	<0.001	−68	−77	−58.9
2	719.1	421.4	607.7	362.1	<0.001	−111.5	−121.1	−101.8
3	732.0	422.3	625.3	382.9	<0.001	−79.7	−91.2	−68.2
4	909.7	541.6	872.1	673.2	0.028	−37.6	−71.2	−3.9
Other/non-classified	761.6	675.4	713.3	638.8	0.059	−48.2	−98.3	1.8
Final disposition decision
Home/residency	505.4	275.9	479.1	257.8	<0.001	−83.7	−89.2	−78.1
Facilities care	658.6	370.0	576.8	350.0	<0.001	−81.8	−97.3	−66.4
Transfer to acute care	745.5	620.2	781.6	545.5	0.042	36.2	1.3	71.1
Home care	479.1	257.8	505.4	275.9	0.758	−26.2	−202.8	150.3
Death	1159.3	1289.3	1158.8	1127.9	0.992	−0.5	−86.1	85.2

ED: emergency department.

Among through-ED admissions, 15,462 (40.1%) have a stay in OU. Mean LOS in OU was 0.83 ± 0.72 days. Among through-ED admissions, those without Observation unit stay have shorter length of stay (6.5±5.5 vs 8.9±6.0; p<0.001) and higher revenues per stay (3683.5±3259 vs 4011.47±2582; p<0.001) and revenues per day (727.6±462.8 vs 576.8±355; p<0.001). The results of comparisons between through-ED admissions with or without OU stay are presented in Table 5. In the subgroup analyses (through-ED with or without OU stay compared to elective admissions – Table 6), differences were markedly reduced between through-ED without an OU stay and elective admissions for LOS (−0.20) and RPDs – (41 euros). Difference in LOS was almost null in younger (−0.09) and least severe patients (−0.02 and −0.04 for levels of severity 0 and 1). For ED patients with a stay in the OU, differences in LOS are increased, especially for the elderly, patients with non-invasive (interventional) procedure, mild severity (level 2 and non-classified), and those transferred to rehabilitation and long-term facilities. In terms of revenues, through-ED patients with an OU stay had more than twice the difference in RPDs with elective patients compared to ED patients directly admitted to the ward. Being ≥75 years old, having a non-invasive procedure, and being of mild severity were again related to higher differences in revenues.

Table 5.

Patients’ characteristics as a function of observation unit stay.

	Without observation unit stay		With observation unit stay		p
	n	%	n	%	p
Age group
<75 years	18,714	81.2	10,902	70.5	<0.001
≥75 years	4347	18.8	4563	29.5
Stay category
Surgical	7537	32.7	1891	12.2	<0.001
Interventional	1094	4.7	426	2.8	<0.001
Medical	14,430	62.6	13,148	85.0	<0.001
Severity level
0	1508	6.5	390.0	2.5	<0.001
1	11,301	49.0	4833	31.3	<0.001
2	5657.0	24.5	4943	32.0	<0.001
3	3135	13.6	3483	24.9	<0.001
4	840	3.6	768	5.0	<0.001
Other/non-classified	620	2.7	688	4.5	<0.001
Final disposition decision
Home/residency	19,571	84.9	12,768	82.6	<0.001
Facilities care	1926	8.4	1426	9.2	0.003
Transfer to acute care	1093	4.7	748	4.8	0.663
Home care	13	0.1	6	0.0	0.446
Death	458	2.0	517	3.3	<0.001
Stay in
Intensive care	2644	11.5	930	6.0	<0.001
Intermediate care	855	3.7	348	2.3	<0.001
Critical care	901	3.9	304	2.0	<0.001
Length of stay (days)	6.5	5.5	8.9	6.0	<0.001
Revenues per stay (euros)	3684	3259	4012	2582	<0.001
Revenues per day (euros)	728	463	577	355	<0.001

Table 6.

Mean differences on LOS and RPD between cases (through-ED patients with or without stay in OU) and controls (elective patients).

	Elective, N = 23,061		Through-ED (without OU) N = 23,061		Difference	p	Elective, N = 15,465		Through-ED (with OU), N = 15,465		Difference	p
	Mean	SD	Mean	SD	Difference	p	Mean	SD	Mean	SD	Difference	p
Length of stay
Overall population	6.28	5.26	6.47	5.49	−0.20	<.0001	7.56	5.60	8.87	5.95	−1.31	<0.0001
Age group
<75 years	5.85	5.06	5.94	5.23	−0.09	0.04	6.90	5.31	8.07	5.73	−1.17	<0.0001
≥75 years	8.09	5.68	8.76	5.95	−0.67	<.0001	9.13	5.95	10.76	6.05	−1.63	<0.0001
Stay category
Surgical	6.88	5.22	6.63	5.74	0.25	0.00	7.78	5.63	8.57	6.43	−0.79	<0.0001
Interventional	5.67	5.28	6.28	5.18	−0.61	0.00	6.60	5.47	8.41	6.35	−1.81	<0.0001
Medical	6.01	5.25	6.40	5.37	−0.40	<.0001	7.56	5.59	8.92	5.87	−1.37	<0.0001
Severity level
0	1.06	0.23	1.08	0.27	−0.02	0.01	1.10	0.30	1.12	0.32	−0.02	0.47
1	4.15	3.28	4.19	3.53	−0.04	0.35	4.14	3.37	5.18	3.87	−1.04	<0.0001
2	8.10	4.86	8.63	5.27	−0.53	<.0001	7.78	4.71	9.50	5.42	−1.72	<0.0001
3	11.23	5.68	11.48	5.66	−0.25	0.07	11.01	5.65	12.16	5.64	−1.15	<0.0001
4	13.45	5.89	13.10	5.79	0.35	0.18	13.01	5.89	14.01	5.66	−1.00	0.00
Other/non-classified	6.34	5.68	7.17	6.07	−0.83	0.00	8.27	6.48	10.44	6.48	−2.16	<0.0001
Final disposition decision
Home/residency	5.65	4.80	5.85	5.02	−0.20	<.0001	6.88	5.16	8.24	5.59	−1.36	<0.0001
Facilities care	10.37	5.96	11.46	6.42	−1.09	<.0001	11.37	6.26	13.31	6.26	−1.94	<0.0001
Transfer acute care	8.74	6.41	7.71	6.28	1.03	<.0001	9.55	6.43	9.80	6.71	−0.25	0.39
Home service	12.46	5.08	11.23	5.69	1.23	0.39	16.33	7.63	16.00	6.99	0.33	0.93
Death	9.74	6.47	8.81	6.38	0.93	0.01	10.67	6.46	10.56	6.75	0.11	0.78
Revenues per day
Overall population	768.67	479.22	727.66	462.77	41.01	<.0001	704.05	427.52	576.76	354.94	127.30	<0.0001
Age group
<75 years	777.29	475.68	742.68	468.23	34.61	<.0001	715.91	418.03	591.47	362.83	124.40	<0.0001
≥75 years	731.57	492.53	663.03	432.65	68.55	<.0001	675.71	448.17	541.61	332.77	134.10	<0.0001
Stay category
Surgical	919.14	542.89	925.52	526.80	−6.38	0.38	925.84	483.56	811.37	436.21	114.50	<0.0001
Interventional	1018.12	765.71	902.67	514.38	115.50	<.0001	1045.68	934.38	751.09	473.59	294.60	<0.0001
Medical	671.16	377.66	611.05	375.37	60.11	<.0001	661.08	374.76	537.37	321.14	123.70	<0.0001
Severity level
0	604.95	135.43	633.14	183.42	−28.19	<.0001	598.22	122.41	602.61	127.51	−4.39	0.60
1	779.88	503.64	746.86	472.26	33.02	<.0001	717.12	446.08	567.31	356.30	149.80	<0.0001
2	744.17	428.99	663.39	384.07	80.78	<.0001	690.46	410.77	543.87	323.67	146.60	<0.0001
3	783.51	456.57	727.72	415.67	55.79	<.0001	690.02	387.06	590.80	341.80	99.22	<0.0001
4	1007.07	591.16	1019.80	820.67	−12.73	0.65	803.14	456.25	710.53	403.48	92.61	<0.0001
Other/non-classified	788.09	719.66	798.01	708.63	−9.93	0.80	737.66	632.43	637.00	558.07	100.70	0.00
Final disposition decision
Home/residency	761.52	409.15	711.83	396.89	49.69	<.0001	704.17	383.52	568.43	326.98	135.70	<0.0001
Rehabilitation and facilities care	691.25	379.87	636.87	381.73	54.38	<.0001	614.57	351.67	495.64	282.39	118.90	<0.0001
Transfer acute care	786.80	691.89	882.51	606.06	−95.71	0.00	685.04	491.54	634.23	399.26	50.81	0.02
Home service	528.66	305.31	458.11	246.66	70.55	0.52	454.93	214.23	524.67	299.30	−69.75	0.64
Death	1363.49	1510.29	1424.34	1405.81	−60.86	0.40	978.35	1023.61	923.57	730.17	54.78	0.29

ED: emergency department; OU: observation unit.

Discussion

Our study indicates that there were some differences between through-ED and Elective admissions. Through-ED admissions concern more often elderly patients, more severe and complex stays with more critical care admissions and OU stays, with longer LOS and increased mortality rate. However, their RPS and RPDs are lower than those of elective admissions. Our study confirms that usual analysis methods including multifaceted analysis are poorly adapted to length of stay. Then, case–control study method after matching admissions for most important previously reported variables associated with LOS and hospital revenues, allowed us to point out that through-ED admissions LOS are longer (0.64 days) and RPD is lower (76 euros). Our results suggest that for some categories of patients, differences of LOS and incomes between the two admissions modes are weak. In terms of care pathways, our study shows that OUs received more elderly, severe and medical patients, with more discharges to facilities care centers.

Age has been associated with longer and more complex in-hospital stays.²³ In our study, through-ED admissions were mostly elderly, and the majority of elderly hospitalized patients were admitted through-ED, indicating that ED is the main admission mode for elderly patients. In the case–control study, elderlies have longer LOS (1.2 days) and their RPDs were lower (−102 euros). As the proportion of elderly patients is increasing in ED as well as the number and proportion of elderlies among hospitalized patients,^24,25 our results suggest that pathways, costs, and revenues of aged patients need to be reevaluated. Differences in LOS and revenues have been reported as a function of stay categories, comorbidities, and severity.^26–28

It has been reported that medical stays LOSs is longer than surgical stays.²⁷ In the present study, medical and severe stays were more frequent in through-ED admissions. LOSs of surgical stays were not different between cases and controls, while medical and interventional through-ED admissions had longer LOS (0.9 days). Concerning RPD, difference is minimal for surgical stays (18 euros), while it was greater for medical stays (90 euros) and maximal for interventional stays (164 euros). Similarly, intermediate severity was associated with longer LOS and lower RPD for through-ED admissions. Disposal to facilities care was associated with longer LOS (1.5 days) and lower RPD (81.8 euros) for through-ED admissions, whereas home care was associated with a reduced differences, indicating that proposed clinical pathway is of greater importance.

In France, admission in OU is frequently used to avoid disturbing planned admissions to medical and surgical wards. As we found important differences among through-ED admissions between with- and without-OU stay, we evaluated the impact of OU stay by separating among through-ED stays those with or without an OU stay. We estimated again the differences between both categories of emergency patients with their elective controls. Our results indicated that the differences were reduced from 0.64 to 0.2 days for LOS and from 79 to 41 euros for RPD when patients did not stay in OU before being admitted. Even more, we found that differences were also narrowed in some covariates notably the age category (<75 years old) and least severe or surgical stays. After identifying the effect of the OU among cases, our study indicates that the differences in LOS and RPD were much smaller for ED patients directly admitted onto the ward compared to elective admissions (LOS: −0.20 and RPD: −41€), while ED patients with an OU stay had a difference in LOS of more than a day (1.31) and a difference of 127.3€ per day in revenues. We conclude that stay in OU had unfavorable impact on hospital LOS and revenues, that OU stay is a complex process that increases LOS and reduces RPD notably for elderlies and patients of mild severity, or those with a non-scheduled interventional procedure.

This study has several limitations. Even though we analyzed a very large number of covariates describing patients, hospital stay characteristics, clinical pathways including OU, and final disposition, a number of other measures or other factors probably exist that can influence LOS and revenues. Moreover, we did not investigate whether there were ED patients’ categories that could not be matched to controls and if they differed from our sample. However, we performed paired comparisons between through-ED and elective admissions on known criteria associated with LOS and revenues. A multivariate analysis would have allowed comparing several variables simultaneous and further exploring the impact of the admission mode. We chose in this study to describe differences in LOS and revenues and to demonstrate the impact on a prolonged stay in OU as a main factor of possible losses in revenues and organizational efficacy. Moreover, even if we carried on this study in three urban hospitals, our results are probably not generalizable. For instance, average LOS in French hospitals is 5.1 to 6.1 days, significantly lower than the average LOS we observed.

In summary, our study indicates that despite the observed differences in severity and complexity between both admissions modes explaining the increased LOS of through-ED admission, lesser revenues have been observed for through-ED admissions. We think that our study highlights an issue with our hospital costs and revenues calculation system: elderly through-ED admitted in medicine ward with complex and severe stays are less valued than elective admissions. In our opinion, this may indicate an inadequate valorization of complex in-hospital stays or a methodological bias in actual costs and revenues calculation method. Our results also indicate that easy and early available variables such as age ≥75 years, medical stay, initial stay in OU, and need of discharge to facility care centers were associated with increased LOS and reduced revenues. We consider that an LOS and costs reduction policy for both admission modes could be based on early interventions for these high-risk populations, that the missions of OU should be defined, and that it is necessary to evaluate the contribution of this stay in the fate of certain categories of patients, such as elderly patients or certain diagnostic categories. Then, our study opens new research ways for the evaluation of interventions to reduce LOS and costs and to define the benefit of the OU.

Footnotes

Authors’ note

Data are available upon request.

Author contributions

EC, AP, CC, SC, CL and RH conceived the study and designed the trial. EC, AP and RH supervised the conduct of the trial and data collection. EC, AP, CC, SC, CL and RH undertook recruitment of participating centers and patients. EC, AP and RH managed the data, including quality control. EC, AP, CC, SC, CL and RH provided statistical advice on study design, analyzed the data and chaired the data oversight committee. EC, AP, CC, SC, CL and RH drafted the manuscript. All authors had full access to the data, take responsibility for the integrity of the data, and approved the article. EC takes responsibility for the paper as a whole.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

European Hospital and Healthcare federation. Hospitals in Europe. Healthcare data, www.hope.be/wp-content/uploads/2015/10/00-hospitals_in_europe-synthesis.pdf (2012, accessed 29 September 2017).

National Center for Health Statistics. Health, United States, 2016: with chartbook on long-term. Trends in health. Hyattsville, MD, www.cdc.gov/nchs/data/hus/hus16.pdf#081 (2017, accessed 29 September 2017).

Scott

IA.

Public hospital bed crisis: too few or too misused?

Aust Health Rev 2010; 34: 317–324.

Freitas

Silva-Costa

Lopes

et al . Factors influencing hospital high length of stay outliers. BMC Health Serv Res 2012; 12: 265.

AHA Lewin Group. Emergency department overload: a growing crisis, www.aha.org/aha/content/2002/pdf/EdoCrisisSlides.pdf (2002, accessed 8 February 2015).

Lawton

Thomas

Morel

DG.

Trends in access block 2011 to 2013: the Redcliffe National Emergency Access Target experience. Emerg Med Australas 2015; 27: 11–15.

Handel

Hilton

Ward

et al . Emergency department throughput, crowding, and financial outcomes for hospitals. Acad Emerg Med 2010; 17: 840–847.

Omachonu

Suthummanon

Akcin

et al . Predicting length of stay for Medicare patients at a teaching hospital. Health Serv Manage Res 2004; 17: 1–12.

Suthummanon

Omachonu

VK.

DRG-based cost minimization models: applications in a hospital environment. Health Care Manag Sci 2004; 7: 197–205.

10.

Fay

America's debt help organization. Hospital and surgery costs, www.debt.org/medical/hospital-surgery-costs/ (2015, accessed 12 February 2016).

11.

Henneman

Lemanski

Smithline

et al . Emergency department admissions are more profitable than non-emergency department admissions. Ann Emerg Med 2009; 53: 249–255.

12.

CDC. National Hospital Discharge Survey: 2010 table, number and rate of hospital discharges. Hospital Utilization (in non-Federal short-stay hospitals), www.cdc.gov/nchs/data/nhds/1general/2010gen1_agesexalos.pdf (accessed 12 February 2016).

13.

Flood

Maclennan

McGrew

et al . Effects of an acute care for elders unit on costs and 30-day readmissions. JAMA Intern Med 2013; 173: 981–987.

14.

Polverejan

Gardiner

Bradley

et al . Estimating mean hospital cost as a function of length of stay and patient characteristics. Health Econ 2003; 12: 935–947.

15.

Ishak

Stolar

et al . Accounting for the relationship between per diem cost and LOS when estimating hospitalization costs. BMC Health Serv Res 2012; 12: 439.

16.

Shen

Melnick

Trends in hospital cost and revenue, 1994–2005: how are they related to HMO penetration, concentration, and for-profit ownership?

Health Serv Res 2010; 45: 42–61.

17.

Diehr

Yanez

Ash

et al . Methods for analyzing healthcare utilization and costs. Annu Rev Public Health 1999; 20: 125–144.

18.

Roberts

Mensah

Weinstein

RA.

A guide to interpreting economic studies in infectious diseases. Clin Microbiol Infect 2010; 16: 1713–1720.

19.

Ananda-Rajah MR, Cheng A, Morrissey CO, et al. Attributable hospital cost and antifungal treatment of invasive fungal diseases in high-risk hematology patients: an economic modeling approach. Antimicrob Agents Chemother 2011; 55: 1953–1960.

20.

Ministère de la Santé et des sports. Manuel des Groupes Homogènes de Malades. 11e version de la classification, 1re révision (11b), Version 12.11b de la fonction groupage, Mise à jour au, 1er mars 2010, Volume 1. Présentation et annexes générales, www.sante.gouv.fr/fichiers/bos/2010/ste_20100004_0001_p000.pdf (2010, accessed 12 February 2016).

21.

Direction générale de l’offre de soins. Guide méthodologique de comptabilité analytique hospitalière, www.sante.gouv.fr/IMG/pdf/GUIDE_CAH___BOS_2011-3.pdf (accessed 12 February 2016).

22.

Bras

Vieilleribiere

Lesteven

Inspection Générale des affaires sociales RM2012-024P. Evaluation de la tarification des soins hospitaliers et des actes médicaux, http://www.google.fr/url?url=http://www.ladocumentationfrancaise.fr/var/storage/rapports-publics/124000549.pdf&rct=j&frm=1&q=&esrc=s&sa=U&ved=0ahUKEwjhpYDexbrJAhWCkA8KHaaWDK4QFggaMAE&usg=AFQjCNHCROFHZKStZ3RVTb KStPb7Cvhetw (2012, accessed 12 February 2016).

23.

Friedrich

Adhikari

Beyene

Ratio of means for analyzing continuous outcomes in meta-analysis performed as well as mean difference methods. J Clin Epidemiol 2011; 64: 556–564.

24.

Manzano-Santaella

From bed-blocking to delayed discharges: precursors and interpretations of a contested concept. Health Serv Manage Res 2010; 23: 121–127.

25.

Hampton

Experts predict visits by baby boomers will soon strain emergency departments. JAMA 2008; 299: 2613–2614.

26.

HCUPnet: a tool for identifying, tracking, and analyzing national hospital statistics, http://hcupnet.ahrq.gov./HCUPnet.jsp (accessed 16 April 2016).

27.

OECD Health Statistics. Definitions, sources and methods. Hospital average length of stay by diagnostic. Categories LOS in surgical stays, www.stats.oecd.org/fileview2.aspx?IDFile=05c5f7a0 (2015, accessed 19 March 2018).

28.

Eurostat. Statistics explained. Hospital discharges and length of stay statistics, http://ec.europa.eu/eurostat/statistics-explained/index.php/Hospital_discharges_and_length_of_stay_statistics (2015, accessed 16 April 2016).