The Combined Effect of Delirium and Falls on Length of Stay and Discharge

Abstract

Introduction: 

Delirium or a fall are associated with many negative outcomes including increased length of stay (LOS) and discharge to a facility; however, this relationship is incompletely understood.

Methods: 

A cross-sectional study of all hospitalizations in a large, tertiary care hospital evaluated the effect of delirium and a fall on the outcomes of LOS and risk of being discharged to a facility.

Results: 

The study included 29,655 hospital admissions. A total of 3,707 (12.5%) patients screened positive for delirium and 286 (0.96%) had a reported fall. After adjustment for covariates, relative to patients without delirium or a fall, patients with delirium only had a 1.64-fold longer LOS; patients with fall only had a 1.96-fold longer LOS; and patients who had delirium and fall had a 2.84-fold longer LOS. The adjusted odds of discharge to a facility, relative to those without delirium or a fall, was 8.98 times higher in those with delirium and a fall.

Conclusions: 

Delirium and falls influence LOS and likelihood of being discharged to a facility. The joint impact of falls and delirium on LOS and facility discharge was more than additive. Hospitals should consider the integrated management of delirium and falls.

Introduction

Adverse events leading to patient harm are common and occur in 10% of patients hospitalized in the United States.^1,2 Among the most notable adverse events are falls and delirium.^3-5 Importantly, these events can be interrelated, having direct and indirect impacts on the patient.³ Although they each independently influence morbidity and hospital outcomes, the combined effect of delirium and falls has the potential to compound the negative consequences.^6-9

Delirium is a common and dangerous neuropsychiatric syndrome that can occur in hospitalized patients, with immediate and lasting effects.^10-12 Although most common in elderly and critically ill patients,^10,13,14 delirium can occur in patients of any age and is estimated in 11–42% of all hospital admissions.¹³ Delirium is inconsistently associated with increased length of stay (LOS), but it is a well-established risk factor for discharge to a facility, such as subacute rehabilitation or skilled nursing facility.⁹ Delirium also carries a considerable financial effect on the patient, family and health system. A review of patients with delirium compared with those without revealed up to $64,000 in additional hospital charges and paid insurance claims per delirious patient per year.¹⁵ In addition, delirium is difficult to diagnose and is often overlooked by physicians and nurses.^16-18 The routine use of validated screening tools is a potential intervention to prevent and manage delirium and is considered a guideline-driven standard of care in the intensive care setting;^19,20 however, routine screening is underused.

In-hospital falls are the most frequently reported safety events in hospitals,²¹ with an estimated 3.5 falls for every 1,000 patient days in the United States.²² They are most common in elderly patients and can have a lasting physical, mental, and financial impact on patients, and a potential financial impact on hospitals.^7,23,24 In-hospital falls have been associated with increased LOS, regardless of the degree of injury,⁶ and an increased risk of being discharged to an inpatient rehabilitation facility or skilled nursing facility.⁷ Despite increased emphasis by health systems, there is limited evidence, and no definitive consensus, on how to reduce the occurrence and impact of falls.^25,26

The patient-level risk factors for delirium and hospital falls have considerable overlap, sharing risks such as advanced age, cognitive impairment, use of psychoactive medications, and sensory impairment.^27,28 However, there is incomplete understanding of the combined impact of delirium and falls on adverse outcomes such as longer length of stay and discharge to a facility.

The interplay between these events is complex, because delirium can lead to a fall and falls can contribute to delirium.³ Building off a health-system wide initiative to identify and manage delirium, cross-sectional study leverages comprehensive electronic medical record data on delirium and falls to further understand this intersection. We examine how the relationship between delirium and falls affect length of stay and discharge to a facility and hypothesize that falls and delirium will increase both of these outcomes, and the combination of both will have greater impact than either alone.

METHODS

This study is a cross-sectional analysis of hospitalized patients at a 740-bed, United States-based tertiary-care academic hospital. Institutional review board (IRB) approval was obtained, and informed consent was waived. The reporting for this analysis follows the STROBE reporting guidelines.

Study Population and Data Collection

All data were obtained from the Data Warehouse and included all adult (18 years and older) inpatient encounters from August 2018 to January 2020. Inpatient encounter was defined as an admission to medical, surgical, neurologic, ICU and OBGYN services, and excluded radiology, laboratory, and procedural areas. Any patient who did not receive a delirium screen during hospitalization, was not admitted to these specified services, or died during the hospitalization was excluded from the analysis (Figure 1).

Figure 1.: Patient selection flow diagram.

Delirium Screening

Starting in 2017, our hospital sought to improve patient care and reduce falls by enhancing the ability to detect delirium in non-ICU patients. The initiative implemented nursing administration of the Brief Confusion Assessment Method (bCAM) once per shift on all adult non-ICU inpatients. The bCAM is a validated delirium screening tool that assesses level of arousal, attention, and presence of disorganized thinking.²⁹ Results are recorded as positive or negative in the electronic medical record (EMR). In the non-ICU setting, there was a 63% compliance with the bCAM tool on a per-opportunity-to-screen basis during the time frame of this study. Before and throughout the implementation and study period of the bCAM in the non-ICU setting, the CAM-ICU³⁰ is administered on every ICU patient at least once per shift. Across all ICUs, compliance with the screening tool was approximately 88% during the timeframe of this study. Over the study period, 87% of all admitted patients received a CAM-ICU or bCAM screen at least once during their hospitalization. Results for ICU and non-ICU were recorded in the same location in the EMR, and the results are viewed the same in data extraction. If screening was not recorded, it was assumed not to have been performed and treated as missing. If the patient never had a recorded screening during the admission, they were excluded from the analysis.

Patients were defined as positive for delirium (CAM-positive) with any positive delirium screen (CAM-ICU or b-CAM) during their hospitalization. Patients were defined as negative for delirium (CAM-negative) if all screens during the hospitalization were negative.

Fall Record

Falls were collected from the hospital's centralized documentation of adverse outcomes, maintained by the Department of Quality and Safety for the purpose of quality improvement. A fall is defined as an event in which there is uncontrolled, downward displacement of a patient's body from a standing, sitting, or lying position. Falls data were then integrated into the analytic data set by matching medical record number and hospitalization data. If a patient had more than one fall during a single admission, it was treated as a single admission with a fall. Twelve patients had more than one fall during their admission.

Outcome Definition

Our primary outcomes are LOS and discharge to facility. Length of stay is defined as the number of consecutive midnights spent in the hospital. Discharge to facility is defined as a discharge destination other than home, such as subacute or acute rehabilitation facility and skilled nursing facility and was considered a dichotomous variable.

Covariates

The following covariates were included in comparison of the delirium, falls, and the primary outcomes: demographics including age, gender, race, and marital status; distance of patient residence from the hospital based on geographic center of their zip code; socioeconomic status based 2010 census data for the patient residence zip code; history of alcohol abuse; Charlson comorbidity index (CCI) diagnoses^31,32 as documented in the medical record; medications ordered but not necessarily administered; and discharging physician speciality service. The variables were chosen as they have been shown to potentially affect the outcomes being studied in this analysis.^33,34

Statistical Analysis

Univariate analyses were performed comparing four patient groups: (1) patients negative for CAM and negative for fall; (2) patients positive for CAM and negative for fall; (3) patients negative for CAM, but positive for fall and; (4) patients positive for CAM and positive for fall. We analyzed using Pearson χ² for categorical variables and procedure of analysis of variance (PROC ANOVA) for continuous variables for Tables 1 and 2. Generalized linear regression with a gamma distribution was used to estimate the fold-change in length of hospital stay associated with independent variables of interest. Multivariable logistic regression was used to determine adjusted odds ratios for discharge to a facility. Multicollinearity was also assessed and corrected; if two variables exhibited high correlation, one was dropped from the model based on clinical relevance. Pearson coefficient and variance inflaction factor did not show evidence of multicollinearity. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc, Cary, NC) and significance was determined at the 5% level.

Table 1. - Unadjusted Comparison of the four Patient Groups (Pearson χ² for Categorical Variables and procedure of analysis of variance [PROC ANOVA] for Comparison of Means for Continuous Variable)

	CAM negative (n = 25,948)		CAM positive [n = 3,707]		p-value
	Falls neg (n = 25,767)	Falls pos (n = 181)	Falls neg (n = 3,602)	Falls pos (n = 105)
Age, mean, SD Age, median	57.3 ± 17.2 60.0	58.3 ± 16.4 61.0	63.3 ± 16.6 65.0	57.2 ± 15.5 61.0	<.0001 <.0001
Age group, %					<.0001
<50	30.6	29.8	19.1	26.7
50–64	30.5	26.5	28.4	35.2
65–79	31.2	37.0	37.6	34.3
80+	7.7	6.6	14.9	3.8
Gender, %					<.0001
Male	48.8	50.3	51.5	69.5
Female	51.2	49.7	48.5	30.5
Race, %					.0037
Black	34.9	35.9	38.3	42.9
Other	3.5	3.3	3.3	1.9
White	61.6	60.8	58.4	55.2
Marital status, %					<.0001
Married	48.8	43.1	43.1	40.0
Other	18.8	17.1	24.7	14.3
Single	32.4	39.8	32.2	45.7
Distance, mean, SD Distance, median	65.4 ± 128 39.6	70.4 ± 93 50.5	62.2 ± 151 28.6	76.6 ± 120 59.6	<.0001 .0089
Far to hospital, distance > 50 miles	46.1	51.4	43.9	55.2	.0066
Poverty, %	29.6	33.2	31.4	40.0	.0106
Score of CCI, mean, SD Score of CCI, median	4.0 ± 3.1 3.0	4.6 ± 3.2 4.0	5.0 ± 3.1 5.0	5.6 ± 3.4 5.0	<.0001 <.0001
Score of CCI, %					<.0001
0	12.8	7.2	5.4	2.9
1–2	25.6	23.8	16.2	19.1
3–4	25.3	24.3	26.0	22.9
5+	36.3	44.8	52.5	55.2
Medication exposure, %
Antipsychotic	27.4	40.3	50.6	76.2	<.0001
Anticholinergics	64.3	77.9	54.8	72.4	<.0001
Opioid	82.9	88.4	78.3	90.5	<.0001
Benzo	37.5	56.4	61.2	83.8	<.0001
ICU During hospitalization, %	15.2	20.4	62.5	73.3	<.0001
Alcohol abuse, %	6.2	6.6	13.6	27.6	<.0001
Dementia, %	3.1	3.3	15.2	6.7	<.0001
Myocardial infarction, %	11.9	14.9	18.2	16.2	<.0001
Congestive heart failure, %	18.7	19.9	26.9	31.4	<.0001
Cerebrovascular disease, %	8.6	8.3	26.3	33.3	<.0001
Chronic pulmonary disease, %	20.4	18.2	22.7	29.5	.0012
Rheumatoid arthritis, %	5.1	5.0	5.7	8.6	.2033
Peptic ulcer disease, %	1.6	4.4	2.4	4.8	<.0001
Uncomplicated diabetes, %	10.9	8.3	10.7	7.6	.4527
Complicated diabetes, %	18.2	23.2	25.5	21.0	<.0001
Hemiplegia, %	3.0	2.8	11.7	14.3	<.0001
Renal disease, %	19.5	26.5	25.7	21.9	<.0001
Cancer, %	21.1	26.5	17.8	34.3	<.0001
AIDS/HIV, %	1.3	2.8	1.6	1.9	.0586
Liver disease, %	7.6	9.9	11.9	20.0	<.0001
Discharging service, %					<.0001
ICU	0.2	0.6	0.6	1.9
Medicine	47.1	48.6	53.7	44.8
Neurology	14.4	12.2	23.7	18.1
OBGYN	2.4	5.0	0.6	3.8
Surgery	35.9	33.7	21.5	31.4

CAM = confusion assessment method; CCI = Charlson comorbidity index; ICU = intensive care unit.

Table 2. - Unadjusted Comparison of Outcomes for the four Patient Groups (procedure of analysis of variance [PROC ANOVA] for Comparison of Means and Medians for Length of Stay [LOS] and Pearson χ² for Discharge to Facility)

	CAM negative (n = 25,948)		CAM positive (n = 3,707)		p-value
	Falls neg (n = 25,767)	Falls pos (n = 181)	Falls neg (n = 3,602)	Falls pos (n = 105)
LOS, mean, SD LOS, median	5.2 ± 5.0 4.0	12.8 ± 13.9 8.0	12.9 ± 12.1 9.0	30.5 ± 26.0 22.0	<.0001 <.0001
Discharge to facility, %	9.9	24.9	44.2	62.9	<.0001

CAM = confusion assessment method.

Results

There were 29,655 unique patient admissions included in the analysis. Delirium was identified in 12.5% (3,707/29,655) of patients who were screened during their hospitalization. There were 286 admissions that included a fall, representing 0.96% of admissions. The overall mean LOS was 6.2 days. A total of 4,252 of 29,655 (14.3%) patient encounters were discharged to a facility, with 32.5% discharged to a rehabilitation facility, 50.0% to a nursing home or long-term care facility, 11.8% to another hospital, and 5.7% to an inpatient hospice or other facility.

Of the 286 admissions with falls, 105 (36.7%) patients screened positive for delirium during their admission, compared with 181 (63.3%) admissions with falls that screened negative for delirium. When considering median length of stay, there were 2.11 falls per 1,000 bed days in the delirium group and 1.34 falls per 1,000 bed days in the nondelirium group.

Demographic and Clinical Differences

Significant demographic and clinical differences among the four groups are detailed in Table 1. In general, among the four groups, there were small but significant differences by age, gender, and race. Patients experiencing falls and delirium were less likely to be married, but more likely to be male. Dementia was most common in the delirium only group (15.2%), whereas the delirium and fall group were more likely to have a history of cerebral vascular disease (33.3%) and alcohol abuse (27.6%). Admission to the ICU was highest in those who had a fall and delirium (73.3%) and lowest in those with neither (15.2%). Benzodiazepines and antipsychotics were more likely to be ordered in the delirium and falls group compared with the other three categories, being ordered in 83.8% and 76.2% of patients in the delirium and falls group, respectively. Although statistically significant because of the large sample size, Table 1 shows other statistically significant demographic and clinical variables, but less clinically relevant. There were very few discharges from the ICU and most patients with falls and delirium otherwise were discharged from Medicine and Surgery services. Thirteen percent of all admitted patients were never screened for delirium. These admissions were compared to the patients who were screened and demonstrated that the nonscreened patients were younger, had fewer Charlson comorbidities, were exposed to fewer high-risk medications, and were more likely to be admitted to OBGYN services.

Unadjusted Outcomes

The unadjusted comparison of LOS and discharge to facility among those who experienced a fall, delirium, or both compared with most patients who experienced neither are shown in Table 2. The median LOS of patients with neither a fall nor delirium was 4 days. There was a statistically significant longer unadjusted median length of stay of patients who had a fall (4 additional days) or delirium (5 additional days). Patients who had a fall and delirium had an even greater unadjusted increase in length of stay (18 additional days). The proportion of patients who were discharged to a facility was greater in those who had a fall without delirium (24.9%), those who screened positive for delirium who did not fall (44.2%), and those with delirium who fell (62.9%), when compared with patients that experienced neither delirium nor a fall (9.9%).

Length of Stay

The results from a generalized linear regression model with a gamma distribution used to estimate the fold-change in length of hospital stay associated with indicated covariates are shown in Table 3. The interaction term between falls and delirium was not statistically significant (p = .6556). However, the joint impact of falls and delirium was of interest so we modeled delirium and falls as a four category variable (i.e., neither fall nor delirium, delirium only, fall only, fall and delirium). Length of stay was 1.64-fold longer (95% CI 1.60, 1.69) in patients with a positive delirium screen without a fall compared with those without delirium or a fall. Patients experiencing a fall without delirium had a 1.96-fold (95% CI 1.78, 2.16) longer hospital stay than those without delirium or a fall. Finally, patients with delirium who fell had a 2.84-fold (95% CI 2.50, 3.23) longer hospital stay than those without delirium or a fall.

Table 3. - Generalized Linear Regression With a Gamma Distribution Was Used to Estimate the Fold-Change in Length of Hospital Stay Associated With Independent Variables of Interest

	Estimate	Std error	Rate ratio (RR)	95% lower RR	95% upper RR	p-value
Group (ref. fall−/CAM−)
Fall+ CAM−	0.671	0.049	1.957	1.776	2.155	<.0001
Fall− CAM+	0.496	0.014	1.642	1.599	1.686	<.0001
Fall+ CAM+	1.044	0.065	2.842	2.501	3.230	<.0001
Age group (ref. 18–49)
50–64	−0.004	0.011	0.996	0.975	1.017	.6978
65–79	0.003	0.011	1.003	0.981	1.026	.7836
80+	0.018	0.017	1.018	0.986	1.052	.2778
Sex: Male (ref.)	−0.046	0.008	0.955	0.940	0.970	<.0001
Race: White (ref.)
Black	0.110	0.009	1.116	1.096	1.136	<.0001
Other	0.044	0.021	1.045	1.002	1.089	.0407
Marital status: Married (Ref.)
Other	0.070	0.011	1.073	1.050	1.095	<.0001
Single	0.113	0.010	1.120	1.099	1.142	<.0001
Distance to hospital>50	0.123	0.008	1.131	1.113	1.149	<.0001
Poverty	−0.010	0.009	0.990	0.974	1.007	.2574
ICU During hospitalization	0.495	0.011	1.640	1.606	1.674	<.0001
Antipsychotic	0.186	0.009	1.204	1.183	1.225	<.0001
Anticholinergics	0.240	0.009	1.271	1.249	1.293	<.0001
Opioid	0.199	0.011	1.221	1.194	1.248	<.0001
Benzo	0.289	0.008	1.335	1.313	1.357	<.0001
Alcohol abuse	−0.047	0.016	0.954	0.924	0.985	.0042
Dementia	0.042	0.019	1.042	1.004	1.082	.0315
Myocardial infarction	−0.023	0.012	0.978	0.955	1.001	.0599
Congestive heart failure	0.161	0.011	1.174	1.149	1.200	<.0001
Cerebrovascular disease	0.095	0.014	1.099	1.069	1.130	<.0001
Chronic pulmonary disease	0.060	0.010	1.062	1.042	1.083	<.0001
Peptic ulcer disease	0.189	0.029	1.209	1.141	1.280	<.0001
Uncomplicated diabetes	−0.092	0.013	0.913	0.890	0.935	<.0001
Complicated diabetes	0.109	0.011	1.115	1.092	1.140	<.0001
Hemiplegia	0.286	0.021	1.331	1.277	1.386	<.0001
Renal disease	0.054	0.011	1.056	1.033	1.079	<.0001
Cancer	0.171	0.010	1.187	1.163	1.210	<.0001
AIDS/HIV	0.043	0.034	1.044	0.977	1.116	.2038
Liver disease	0.119	0.015	1.127	1.094	1.160	<.0001
Service groups: Medicine (ref.)
ICU	0.071	0.072	1.073	0.933	1.236	.3227
Neurology	−0.259	0.013	0.772	0.753	0.792	<.0001
OBGYN	−0.104	0.028	0.901	0.854	0.951	.0002
Surgery	−0.181	0.010	0.835	0.819	0.851	<.0001

CAM = confusion assessment method; ICU = intensive care unit.

Discharge to Facility

The details the multivariable logistic regression model for a patient being discharged to a facility are displayed in Table 4. The interaction term between falls and delirium was not statistically significant (p = .2017), but the joint impact of falls and delirium was of interest; therefore, we again modeled our main exposure delirium and falls as a four-category variable. Relative to those without delirium or a fall, the odds ratio of being discharged to a facility if experiencing delirium without a fall was 3.49 (95% CI 3.17, 3.85, p-value <.0001) and experiencing an in-hospital fall without delirium was 2.94 (95% CI 2.04, 4.24, p-value <.0001). Patients who had a fall and delirium had dramatically increased odds of discharge to a facility of 8.98 (95% CI 5.79, 13.94, p-value <.0001) compared with patients who experienced neither falls nor delirium during hospitalization.

Table 4. - Multivariable Logisitic Regression Adjusted Odds Ratio for Discharge to Facility

	Estimate	Std error	OR	95% lower OR	95% upper OR	p
Group (ref. fall−/CAM−)
Fall+ CAM−	1.080	0.186	2.944	2.043	4.242	<.0001
Fall− CAM+	1.251	0.050	3.494	3.169	3.852	<.0001
Fall+ CAM+	2.195	0.224	8.981	5.787	13.939	<.0001
Age group (Ref. 18–49)
50–64	0.631	0.063	1.879	1.661	2.124	<.0001
65–79	1.467	0.063	4.337	3.834	4.906	<.0001
80+	2.195	0.077	8.980	7.721	10.445	<.0001
Sex: Male (ref.)	−0.017	0.040	0.983	0.909	1.063	.6646
Race: White (ref.)
Black	0.039	0.045	1.039	0.952	1.135	.3901
Other	−0.025	0.113	0.976	0.781	1.219	.8291
Marital status: Married (Ref.)
Other	0.457	0.049	1.579	1.434	1.738	<.0001
Single	0.554	0.051	1.740	1.576	1.922	<.0001
Distance to Hospital>50	−0.163	0.040	0.850	0.785	0.920	<.0001
Poverty	0.059	0.042	1.061	0.977	1.152	.1621
ICU During hospitalization	0.405	0.046	1.499	1.369	1.641	<.0001
Antipsychotic	0.409	0.041	1.505	1.388	1.633	<.0001
Anticholinergics	0.199	0.042	1.221	1.124	1.326	<.0001
Opioid	0.086	0.053	1.09	0.982	1.209	.1044
Benzo	0.178	0.041	1.195	1.102	1.295	<.0001
Alcohol abuse	0.372	0.073	1.451	1.257	1.675	<.0001
Dementia	0.701	0.072	2.017	1.753	2.32	<.0001
Myocardial infarction	0.088	0.053	1.092	0.984	1.212	.0992
Congestive heart failure	0.140	0.049	1.151	1.045	1.267	.0043
Cerebrovascular disease	0.504	0.056	1.655	1.483	1.848	<.0001
Chronic pulmonary disease	0.013	0.046	1.013	0.925	1.109	.7832
Peptic ulcer disease	0.044	0.136	1.045	0.801	1.364	.7452
Uncomplicated diabetes	−0.025	0.062	0.976	0.864	1.101	.6888
Complicated diabetes	0.255	0.050	1.291	1.171	1.422	<.0001
Hemiplegia	1.490	0.077	4.436	3.812	5.161	<.0001
Renal disease	0.101	0.051	1.106	1.000	1.223	.0498
Cancer	−0.326	0.052	0.722	0.652	0.799	<.0001
AIDS/HIV	0.074	0.171	1.077	0.770	1.507	.6643
Liver disease	0.163	0.071	1.177	1.024	1.353	.0219
Service groups: Medicine (ref.)
ICU	−0.247	0.365	0.781	0.383	1.596	.4987
Neurology	0.235	0.060	1.265	1.124	1.424	<.0001
OBGYN	−0.728	0.208	0.483	0.321	0.726	.0005
Surgery	−0.092	0.050	0.912	0.828	1.006	.0653

CAM = confusion assessment method; ICU = intensive care unit; OR = odds ratio.

Limitations

There are several limitations to this study. A potential bias that was not controlled for in our study was their admission source. Patients who reside in a facility before admission are generally more frail and it can be assumed they are at greater risk of developing delirium and having a fall given the overlap of clinical features that led to their previous facility placement. In addition, it is hard to determine what impact the exclusion of those who died during hospitalization would have had on variables and outcomes, given that delirium is a clear marker of illness severity; however, the death rate is low (<1%). Although there is overlap of risk factors for these events, how they affect each other could be further elucidated and could have meaningful influence on the development of interventions.

Thirteen percent of patients were never screened during their admission and were excluded from the study. We have run a demographic and covariate analysis of screened versus not screened. This analysis revealed that the not-screened group was younger, had fewer comorbidities and were exposed to few medications associated with delirium. Although this represents a difference, we believe that this suggests that this group is less likely to become delirious, less likely to fall, less likely to have a long LOS, and less likely to be discharged to a facility. Thus, our conclusions are not applicable to patients who were not screened for delirium.

The small sample size of falls resulted in a large standard deviation in our outcomes. Despite having a limited number of events in the four categories, the confidence intervals for our two primary outcomes are relatively narrow suggesting adequate sample size for this analysis. Although this is a single-center study, the inclusion of hospital admissions of all adults and minimal exclusion criteria highlights the potential generalizability of the results.

Discussion

The consequences of falls and delirium directly affect the patient and their hospital course. Describing the intersection of the two has been challenging as falls—whereas common when compared with other hospital complications—are relatively infrequent events. With our large sample size, we were able to evaluate falls and delirium together and their separate and joint association with length of stay and discharge disposition. Our findings are consistent in that delirium and falls alone increase length of stay and the odds of discharge to a facility. The interaction between falls and delirium on LOS and discharge to facility was not statistically significant, which suggests that their impact on LOS and facility discharge are independent. In unadjusted analyses, delirium and falls independently increased the median length of stay 5 and 4 days, respectively, effectively doubling the patient's time in the hospital. The joint impact of falls and delirium was even more pronounced: when a patient experienced delirium and a fall, they had a 5-fold increase in LOS, or nearly 18 days longer, compared with patients who had neither. Even after adjusting for covariates, patients with delirium and a fall had a 2.84-fold increase in their LOS compared with patients who had neither delirium nor a fall. The combined impact of delirium and falls on LOS and discharge to facility has not been previously studied. The size of the study, the broad inclusion criteria, and the occurrence rates of delirium and falls consistent with other studies are all strengths of this study.

For those engaged in improving hospital care, this dramatic association presents an association and potential incentive for clinical innovation and investment to prevent falls and delirium. Improving these outcomes is additionally important in enhancing the quality of care of older adults, who are particularly vulnerable to falls and delirium. First, fallers need to be assessed for delirium, because 36.7% (105/286) of fallers had delirium, compared with 12.5% (3,707/29,655) of the total population. Fall risk assessments, such as the Morse assessment, which is used at our hospital, may have superficial measures of cognitive assessment,^35,36 but their use of alertness and orientation as crude surrogates for cognitive function does not capture the full cognitive features of arousal, awareness, and attention necessary for an adequate delirium assessment.⁵ In practice, it is possible to be fully oriented but inattentive to one's inability to safely plan and execute a trip to the bathroom for example. Formal cognitive testing would be ideal, but requires extensive training and time to administer, whereas bedside delirium screening can be performed quickly and effectively.²⁹ Routinely screening for delirium or the creation of integrated delirium and falls screening tools could create an opportunity to intervene with additional precautionary measures for those who screen positive for delirium, such as a best practice alert to the provider or nursing staff. Many delirium prevention and management care plans have been studied and demonstrated to be helpful in reducing falls and other important outcomes.^37-39

Second, our efforts at prevention of falls and delirium should be integrated because they are inextricably linked, and associated with significantly increased risk of adverse outcomes when they co-occur. Early supervised mobility programs are an important area of possible intervention,^40,41 because immobility is a risk factor and consequence of delirium and falls. The cognitive impact of delirium may directly contribute to patients' inability to engage in rehabilitation during and after a hospitalization.¹² Delirium or identification as a high fall risk can impede robust physical and occupational therapy while in the hospital. Staffing shortage, patient and family education, and patient willingness to participate in therapies are additional identified barriers to reduction in delirium and falls.⁴² An effective intervention that increases physical therapy to these at-risk groups could have meaningful impact on length of stay.⁴³

Third, minimizing LOS and optimizing patient throughput are priorities in all hospital systems. These data suggest that specific interventions targeting these high-risk patients could address this objective. Reducing falls and their impact has been a main concern of hospital systems; however, these results should encourage an emphasis on prevention and management of delirium as well. The combined impact of falls and delirium on LOS and facility discharge is more than additive, a phenomenon that has not been previously shown in the literature.

Conclusions

Falls and delirium are strongly associated with increased LOS and increased odds of being discharged to a facility. A patient who experiences delirium and has a fall in the same hospitalization is at high risk of being in the hospital longer and being discharged to a facility. This analysis shows that the joint impact is more than additive. This information should motivate hospitals to improve prevention and management of falls and delirium as a strategy to enhance their quality and cost of care. Future studies should focus on further understanding the impact of delirium prevention tools on the outcomes such as falls, length of stay, and discharge to a facility.

Implications

The strong association between delirium and falls and their impact on important outcomes such as LOS and facility discharge stress the importance of a proactive approach. The implementation of routine delirium screening is an important step toward identifying patients at risk as early as possible and potentially intervening before a fall occurs, hospitalization is prolonged or facility placement is required.

Authors' Biographies

Benjamin Kalivas, MD, is an assistant professor of Medicine and Psychiatry at the Medical University of South Carolina. He works clinically as a hospitalist and brain stimulation psychiatrist and serving as program director for the MedPsych residency program and as the Medical Director of GME Quality and Safety. He has ongoing quality improvement projects and research around the use of delirium screening in hospitalized patients.

Jingwen Zhang, MS, is Research Instructor within the Division of General Internal Medicine and Geriatrics at the Medical University of South Carolina. She is a statistician with research interest in longitudinal data analysis, multivariable modeling analysis in patient care and health.

Kristine Harper, MSN, RN, NE-BC, is the Patient Safety Officer for MUSC Health. Her background includes 22 years of nursing, board certifications as a Nurse Executive and Patient Safety Professional. Kristine's ongoing efforts include fostering the safety culture for the health system, harm reduction strategies, process improvement, and teamwork training.

Jennifer Dulin, MD, MPH, FAAHPM, is an assistant professor of Medicine at the Medical University of South Carolina. She works clinically as a palliative care physician and hospitalist. She has ongoing research surrounding delirium, patient outcomes, and caregiver burden.

Marc Heincelman, MD, MPH, is an associate professor at MUSC. He also serves as Division Chief of Hospital Medicine and Vice Chair of Quality for the Department of Medicine. Focus areas include interhospital transfers, delirium, and, point-of-care ultrasound.

Justin Marsden, BS, is a Program Coordinator in the Section of Health Systems Research and Policy within the Division of General Internal Medicine and Geriatrics at the Medical University of South Carolina. He manages multiple databases and performs data quality assurance to support quality improvement projects and research studies. His role also includes certifying, achieving, and maintaining NCQA Patient-Centered Medical Home (PCMH) Recognition for multiple primary care clinics.

Kelly J. Hunt, PhD, is a Professor in the Department of Public Health Sciences at the Medical University of South Carolina as well as a health service research at the Ralph H Johnson VA Medical Center. She is a diabetes and cardiovascular disease epidemiologist. Her work focuses on the complex inter-relationships between diabetes and cardiovascular disease and using geo-spatial analyses to understand the relationship between individual, neighborhood and medical center level characteristics and diabetes outcomes.

Patrick D. Mauldin, PhD, is Professor and Director of the Section of Health Systems Research and Policy within the Division of General Internal Medicine and Geriatrics at the Medical University of South Carolina. He is a health economist with methodological interests in the multivariate/multivariable modeling of patient-centered information to optimize population health and well-being.

William P. Moran, MD, MS, is a Distinguished University Professor at Medical University of South Carolina in Charleston South Carolina. His recent work has centered on local health system changes including the Patient-centered Medical Home, to support physician decision-making in clinical practice, facilitate interdisciplinary care and improve quality of care especially for older patients. This work includes analysis of existing system characteristics, systematic application of clinical evidence to care processes, and evaluation of impact on process and clinical outcome measures.

Meghan K. Thomas, MD, MS, is an assistant professor of Medicine at the Medical University of South Carolina in Charleston. She works clinically as an academic hospitalist and explores research in Point-of-Care Ultrasound education and Quality Improvement. After completing her Academic Generalist Fellowship and Masters of Public Health in May 2022, she will be expanding into various leadership roles including associate clerkship director for 3rd year medical students, associate program director for the Internal Medicine residency program and Associate Chief Quality Officer for inpatient medical services.