A Managerial Approach to Investigate Fall Risk in a Rehabilitation Hospital

Abstract

Among the issues on which health directions focus, fall risk is one of major importance since it affects patients hospitalized in both acute and rehabilitative hospitals. In this context, few publications have proposed a managerial approach aimed at (a) investigating several factors related to falls and (b) trying to acquire more knowledge and comprehension when analyzing the data collected. Consequently, this paper pursues such objectives by investigating data related to falls (and the recurrence of falls) registered in a rehabilitation hospital within the years 2020 and 2021. A multidisciplinary team (clinical staff and engineers) registered 238 first falls, and descriptive statistics were used to analyze the fall-related anamnestic and clinical data. Then, appropriate statistical analyses were used to compare the same data—this time distinguishing fallers/recurrent fallers—and, again, descriptive statistics were used to analyze the consequences of falls. The statistical analyses allowed us to gain insights into the fall mechanisms, the main places in which falls took place, the impacts of drugs, and fall consequences (e.g., the potential extra costs for the hospital). Moreover, the Morse and Stratify risk tools, state of consciousness, and fall containment measures were proven to be statistically significant features for distinguishing fallers and recurrent fallers, and they may be further investigated to define more accurate preventive measures within rehabilitation hospitals.

1. Introduction

According to the World Health Organization (WHO), falls are in the foreground as external causes of unintentional injuries, and they are usually defined as actions that consist of “inadvertently coming to rest on the ground, floor or other lower level, excluding intentional change in position to rest in furniture, wall or other objects” [1]. It is estimated that 391,000 people died in 2002 due to these events. In terms of hospital environments, falls represent a daily problem; hospitals, in fact, report from 700,000 to 1,000,000 falls every year [2], accounting for 20–30% of the adverse events reported overall [3]. Due to this high incidence, health departments must act in terms of management and prevention so that these events do not influence hospitals’ quality levels.

In addition, falls can determine an increase in the mean length of hospital stay (LOS) and/or additional non-scheduled readmissions after discharge due to fractures, anxiety, and depression that could be derived from these traumatic events [4]. Moreover, several studies have reported a mortality rate of 10.1% in older people within one year following a hip fracture [5]. Previous research in the literature reported numerous variations in fall rates, ranging from 4 to 14 falls per 1000 hospitalization days, corresponding to about 10 monthly falls in a 28-bed unit [6,7].

In the hospital context, it is necessary to identify patients who are at risk. Several evaluation systems can be found in the literature [8], with scores that allow one to quantitatively understand if patients are more or less exposed to fall risk, and these evaluation systems’ scores are based on different variables. In the Istituti Clinici Scientifici Maugeri, the most recent operative instructions prescribe the combined use of the fall risk assessment tools known as the “Morse Fall Scale” (Morse scale) and “St. Thomas Risk Assessment Tool” (Stratify scale). The Morse scale evaluates six items: (1) history of falling, (2) secondary diagnosis, (3) ambulatory aids, (4) intravenous therapy, (5) gait, and (6) mental status. The total score defines three risk levels: 0–24, low risk; 25–44, medium risk; and ≥45, high risk [9]. The Stratify scale considers other aspects, in particular, history of falls, state of agitation, visual impairment, frequent toileting, and mobility score. The achievable score ranges from 0 to 4 with a score of ≥2, which confirms the existence of fall risk [4]. The availability of accurate and high-quality surveillance and monitoring systems that are supported by integrated information systems is fundamental for maintaining a high level of attention among medical staff with the aim of defining both the features and magnitude of the issues, directing interventions, monitoring progress by using appropriate indicators, and identifying sentinel events.

Since the previous literature and evidence indicated that the monitoring and management of fall risk are complex and easily influenced by a plethora of factors (which could often lead typical fall risk screening tools to have a distorted view of the issue [10]), it has been hypothesized that investigating approaches that could allow a deeper overview of fall risk could aid in its limitation. To this aim, simulations have been proposed as an innovative solution in the healthcare field [11] due to the similarities that some healthcare systems present with other work environments [12]. In the context of simulation, discrete event systems (DES) have attracted much attention in the healthcare context due to, among other factors, the ability of this strategy to deal with complex problems, such as those encountered in healthcare settings [13]. To date, and to the best of our knowledge, DESs have been widely proposed to improve emergency departments’ issues [14,15,16], but in the last few years, DES applications have appeared for several other functional areas/hospital settings [12,13], among which rehabilitation units/hospitals can also be found [17,18,19,20]. Since a typical objective of a DES is optimizing resources, quite recently, considerable attention has even been paid to assessing the applicability of DES in the context of healthcare resource modeling [21], thus demonstrating that DES can be integrated into multidisciplinary healthcare-oriented processes such as Health Technology Assessment [22,23]. Simultaneously, even managerially inspired approaches have been proposed as a potential solution for both improving the knowledge and proposing intervention strategies in healthcare. For instance, in the healthcare field, the Stanford Biodesign Methodology (SBM) [24] has demonstrated the ability to improve several processes in the cardiovascular [25], nephrology [26], and surgical [27] fields. In the context of fall risk, Rivers et al. [28] proposed a combination of translational research principles alongside a modified version of the SBM. This methodology allowed one to think about possible lacks in current fall prevention strategies, and it suggested, among other routes, that breaking down silos between different healthcare professionals and the lack of enhanced interprofessional communications is one of the most overlooked factors. In addition to the SBM, other effective frameworks for producing systematic innovation efforts in healthcare have been proposed: among those, the Lean Six Sigma (LSS) strategy could be cited, which was developed by combining lean thinking and Six Sigma methodologies. Briefly, the former deals with the improvement of processes by eliminating the causes of possible deviation from the ideal standards of processes (e.g., wastes), while the latter focuses on the reduction in the number of errors and oscillations in processes by employing statistical and management tools. The previous literature has already proved that LSS in healthcare has helped to achieve objectives like controlling the increase in costs and improving procedure quality [29,30,31] in several medical specialties [32]. In the context of LSS, the define, measure–analyze–improve–control (DMAIC) cycle has frequently been the main one employed in Six Sigma projects as a possible problem-solving strategy [33,34]. Similarly to the SBM, even the DMAIC cycle proved successful to face different healthcare processes and/or aiming at better achieving multiple healthcare needs [35,36,37,38,39,40,41] and, as proved in a previous publication of the research group, to improve the management of healthcare associated infections in the rehabilitation hospital of Bari (southern Italy) of the Istituti Clinici Scientifici Maugeri [42]. On the wave of the promising results achieved in this last study, we decided to employ this managerial approach even to test the fall risk. Although in the last decade, the study of this issue has attracted much attention from research teams, and the previous literature has already analyzed and compared various aspects of the fall risk, very few publications have presented studies where the LSS paradigm was used to investigate the fall risk in the context of rehabilitation hospitals. Moreover, to the best of our knowledge, in the context of a managerial study focused on the risk of falls, little attention has been given to the study of the factors that support the recurrence of falls among those seeking inpatient care and the potential consequences (for both the hospital and the inpatients) on the designed rehabilitation programs. To evaluate these aspects, in this paper, we focus on studying the falls monitored in the rehabilitation hospitals of Bari of the Istituti Clinici Scientifici Maugeri. Keeping the managerial approach proposed in the previous paper of the group [42] but being interested, in this case, only in the potential contributions of the define–measure–analyze (DMA) stages of the DMAIC cycle to this issue, we aimed to conduct a more in-depth analysis of the inaccuracies/lacks in the current in-hospital methodologies, which lead to the occurrence of fall in some patients, and, in turn, we sought to determine the potential corrective measures that could limit the number of future falls from the data obtained.

2. Materials and Methods

The project was developed at the Health Direction of the Istituti Clinici Scientifici Maugeri of Bari (Italy) and is part of the Ricerca Corrente project entitled “Valutazione delle cadute in ambito ospedaliero, utilizzando approcci manageriali e Machine Learning” approved by the Scientific Technical Committee and the local Ethics Committee of the Istituti di Ricovero e Cura a Carattere Scientifico “Giovanni Paolo II” of Bari (Prot. 84/CE Maugeri). The unit has 238 beds accredited by the national health system, and 147,339 patients were admitted between 2020 and 2021. At admission, the clinical and nursing team provided an overview of the patient’s medical history, and within 24 h of admission, the Morse scale score (present in electronic medical records) was determined. Then, the patient’s data on previous falls, secondary diagnoses, intravenous therapy, and mental state were collected to decide upon their fall risk. Following this preliminary fall risk evaluation, patients were split into three risk scales: low, medium, and high. For the first two classes, standard preventive measures were applied (e.g., specific programs to improve patients’ motricity, specifical instructions regarding movement, etc.); differently, for high-risk patients, the Stratify scale was measured, and, in addition, personalized preventive measures were applied (these measures included, but were not limited to, using movement aids, assistance by a caregiver, and application of retainers on beds).

In case an accidental fall occurred, after performing first aid, clinicians and nurses who were present at the circumstance of the fall were asked to put together the circumstances of the event and to insert the clinical effect into the fall record. Once this task was completed, the fall record was sent to the health director for the purpose of evaluating compliance with Istituti Clinici Scientifici Maugeri requisites. If compliant, the fall record was signed by the Istituti Clinici Scientifici Maugeri quality referent. In case moderate/severe/serious injury (or the death of the patient) occurred during the fall, the event was reported to the Istituti Clinici Scientifici Maugeri risk manager, who was in charge of defining sentinel events and analyzing fall events, possibly opting for further clinical audits.

The whole dataset consisted of anamnestic (i.e., age) and clinical data, which included comorbidities, therapies, Morse scale scores, Stratify scale scores (for those patients whose Morse scale score resulted > 44) and all the information about the dynamic and clinical consequences (in terms of injury) of the fall. Hereafter, the main information about the tasks carried out in the DMA stages is reported.

2.1. The Define Phase

The study was conducted by a multidisciplinary team in collaboration with the health direction of the Istituti Clinici Scientifici Maugeri of Bari and the bioengineering research group of the Department of Information Technology and Electrical Engineering (DIETI) of the University of Naples “Federico II”. The team was composed of doctors, nurses, and bioengineers and was supervised by the health director. Firstly, a project charter (

Table 1. Project charter.

Project Title
A Managerial Approach to Investigating Fall Risks in a Rehabilitation Hospital
Problem Statement	Objective Statement
Excessive number of falls in a rehabilitation hospital	Introduce clinical measures that can solve and reduce the presented problem
Critical to quality	Target
Clinical effects and consequential interventions due to falls	Analyze the rehabilitative hospital context in relation to falls and eventually realize corrective measures
Fall recurrences
Extra costs due to falls
Timeline
Define → January 2022–April 2022
Measure → Μay 2022–August 2022
Analyze → September 2022–December 2022
In scope	Out of Scope
Falls	All other clinical accidents
Istituti Clinici Scientifici Maugeri, Bari, Italy	All other structures
Business need
Reducing falls and their impact on public health

) was developed to define all the details of this project: critical to quality (CTQ), problem statement, target and objective, in- and out-of-scope features, business need, and timeline.

2.2. The Measure Phase

After defining the main points of the process, including the problem to be solved, the CTQ, and the methods to adopt, measurements were carried out in order to evaluate the currently adopted process and to identify the possible causes of fall recurrences. In this phase, the data from the years 2020 and 2021 were collected, considering all patients at their first fall and excluding further eventual fall recurrences; 238 patients met these criteria. In order to have a clear overview of fall occurrence, all the clinical parameters and accidental consequences were measured referring to the first fall of each patient; specifically, the collected information and parameters were age, rehabilitation unit, mobility, state of consciousness, pharmacological therapy, the Morse scale, the Stratify scale, risk, the mechanism of falls, and consequences of falls. All these parameters were measured, and details of each parameter were collected for all 238 patients; specifically, they were estimated in terms of mean and standard deviation in the case of numerical data, and in terms of frequency in the case of categorical data.

In Figure 1, all the drugs and pharmacological therapies of the included patients are reported in terms of frequency, and in Figure 2, all the effects of falls (in terms of both injuries and interventions) are reported.

2.3. The Analyze Phase

After the parameters were obtained during the define and measure phases, further statistical analyses were performed to analyze and understand (1) all the possible causes of fall occurrence and (2) all the consequences of every fall, also considering more falls in each patient. Therefore, a comparison among all the parameters pertinent to the patients with and without fall recurrences was carried out. For this purpose, IBM SPSS Statistics (v. 29) was used to perform statistical analyses to estimate the eventual differences among clinical parameters and fall consequences. In particular, after verifying the hypothesis of normality and homoscedasticity with Kolmogorov–Smirnov’s and Levene’s tests, respectively, a t-test for independent samples (in the case of numerical data) or a chi-square test (in the case of categorical data) was performed. Finally, all the parameters related to fall consequences (including recurrences) were measured.

In addition, a basic stream map (Figure 3) was plotted to obtain a workflow of the process and, consequently, analyze each step. Finally, as shown in Figure 4, a cause-and-effect diagram, also called the Ishikawa (root cause) diagram, was mainly used to measure the current status of the issue [33,34], namely to identify the potential causes of accidental falls. In Figure 4, the four major causes along with their associated secondary causes are reported.

3. Results

Table 2. Anamnestic and clinical parameters for all the patients considering only their first fall.

Variables		Categories	Statistics
Years, mean ± St. dev.		//	72.6 ± 11.9
Mobility, n [%]		Bedridden	19 [8.0]
		Assisted with aids	111 [46.6]
		Assisted without aids	16 [6.7]
		Autonomous with aids	60 [25.2]
		Autonomous without aids	32 [13.4]
State of consciousness, n [%]		Psychomotor agitation	14 [5.9]
		Confusion	15 [6.3]
		Vigilant	209 [87.8]
Drugs acting on vital signs	Antiarrhythmics, n [%]	Yes	59 [24.8]
		No	179 [75.2]
	Diuretics, n [%]	Yes	126 [52.9]
		No	112 [47.1]
	Hypotensive, n [%]	Yes	165 [69.3]
		No	73 [30.7]
	Hypoglycemics, n [%]	Yes	13 [5.5]
		No	225 [94.5]
	Laxatives, n [%]	Yes	14 [5.9]
		No	224 [94.1]
Psychoactive drugs	Hypnotics, n [%]	Yes	21 [8.8]
		No	217 [91.2]
	Opioid, n [%]	Yes	3 [1.2]
		No	235 [98.8]
	Sedatives, n [%]	Yes	49 [20.5]
		No	189 [79.5]
Visual impairment, n [%]		Yes	24 [10.1]
		No	214 [89.98]
Antidepressants, n [%]		Yes	39 [16.4]
		No	199 [83.6]
Antiepileptics, n [%]		Yes	32 [13.4]
		No	206 [86.6]
No drug therapy, n [%]		Yes	7 [2.9]
		No	231 [97.1]
Antiparkinsonians, n [%]		Yes	3 [1.2]
		No	235 [98.8]
Systemic antihistamines, n [%]		Yes	1 [0.4]
		No	237 [99.6]
Morse, mean ± St. dev.		//	33.3 ± 18.5
Stratify, n [%]		0	9 [13.2]
		1	31 [45.6]
		2	23 [33.8]
		3	5 [7.4]
Risk, n [%]		Yes, preventive personalized measures	69 [29.0]
		Yes, preventive standard measures	169 [71.0]
Hour of fall, n [%]		Morning	97 [40.8]
		Night	48 [20.2]
		Afternoon	59 [24.8]
		Evening	34 [14.2]
Location, n [%]		Elevator	1 [0.4]
		Bathroom	32 [13.4]
		Room	183 [76.9]
		Corridor	10 [4.2]
		Outdoors	3 [1.3]
		Gym	3 [1.3]
		Waiting room	5 [2.1]
		Other	1 [0.4]
Mechanism, n [%]		Tripped	20 [8.4]
		Loss of consciousness	6 [2.5]
		Loss of balance	58 [24.4]
		Loss of strength	30 [12.6]
		Slipped	83 [34.9]
		Others	41 [17.2]
Dynamic, n [%]		Waiting in wheelchair	8 [3.3]
		Falling from bed while sleeping	5 [2.1]
		While urinating	15 [6.3]
		During assisted movements	8 [3.3]
		During personal cleaning	9 [3.8]
		While dressing up	4 [1.7]
		Not available	18 [7.6]
		While picking up items	9 [3.8]
		While getting in/out of bed	64 [26.9]
		While getting in/out of the wheelchair/chair	56 [23.5]
		While removing restrains aids	4 [1.7]
		Other	38 [16.0]
Clinical impact, n [%]		With injury	76 [31.9]
		Without apparent injury	162 [68.1]
Contusions, n [%]		Yes	28 [11.8]
		No	210 [88.2]
Distortions, n [%]		Yes	1 [0.4]
		No	237 [99.6]
Hematoma, n [%]		Yes	6 [2.5]
		No	232 [97.5]
Excoriation, n [%]		Yes	26 [10.9]
		No	212 [89.1]
Wounds, n [%]		Yes	9 [3.8]
		No	229 [96.2]
Fractures, n [%]		Yes	8 [3.4]
		No	230 [96.6]
Head trauma, n [%]		Yes	17 [7.1]
		No	221 [92.9]
No injuries, n [%]		Yes	25 [10.5]
		No	213 [89.5]
Cryotherapy, n [%]		Yes	17 [7.1]
		No	221 [92.9]
Emergency department, n [%]		Yes	4 [1.7]
		No	234 [98.3]
Medication, n [%]		Yes	35 [14.7]
		No	203 [85.3]
No interventions, n [%]		Yes	106 [44.5]
		No	132 [55.5]
Observation, n [%]		Yes	52 [21.8]
		No	186 [78.2]
RX, n [%]		Yes	37 [15.5]
		No	201 [84.5]
CT brain, n [%]		Yes	30 [12.6]
		No	208 [87.4]
Medical therapy, n [%]		Yes	8 [3.4]
		No	230 [96.6]
Prognosis, n [%]		None	169 [71.0]
		Mild ≤ 3 days	50 [21.0]
		Moderate (from 4 to 20 days)	15 [6.3]
		Severe (from 21 to 39 days)	3 [1.3]
		Serious ≥ 40 days	1 [0.4]
Sentinel event, n [%]		Yes	93 [39.1]
		No	145 [60.9]

shows all the parameters (anamnestic and clinical) considering only patients at their first fall and excluding recurrences. The collected data are referred to “alert and fall analysis reports” referring to the period January 2020–December 2021 and occurred at Istituti Clinici Scientifici Maugeri of Bari, Italy.

The data provided in Table 2 refer to the first fall of the 238 patients, who presented at the neurorehabilitation unit, the cardiac rehabilitation unit, the functional rehabilitation unit, and the pneumological unit.

To provide more detail on the level of support to patients experiencing fall events, we ascertained the following data:

• Overall, 46.6% of falls involved assisted patients with aids, i.e., those requiring help from medical staff while walking and using auxiliary aids (111 patients);
• In total, 25.2% of falls occurred among autonomous patients with aids, i.e., those not requiring help from medical staff while walking but using auxiliary aids (60 patients);
• Of those experiencing falls, 13.4% were autonomous patients without aids (32 patients);
• In total, 8% were bedridden patients (19 patients);
• Lastly, 6.7% were assisted patients without aids (16 patients).

Regarding the fall site, the results show that the site with the highest number of falls was the recovery room (76.9%), even though the toilet (13.4%) and unit corridors (4.2%) were also found to have a significant number of fall events. Furthermore, most of the reported falls occurred while the patient tried to get on and off the bed (26.9%), to get up from and sit on the chair or wheelchair (23.5%), and to pick up an object from the floor (3.8%). Furthermore, the collected data revealed that 69.3% of the fallen patients took hypotensive drugs, and 52.9% of fallers were under diuretic therapy. Other relevant drugs were antiarrhythmics (24.8%), sedatives (20.5%), antidepressants (16.4%), antiepileptics (13.5%), hypnotics (8.8%), laxatives (5.9%), and hypoglycemics (5.5%). Finally, the more common fall mechanisms were sliding (34.9%), loss of balance (24.4%), and loss of strength (12.6%).

Table 3. Comparison among anamnestic and clinical parameters between patients who fell once and those who fell twice or more. Statistical significance for p ≤ 0.01. **: 0.01 < p < 0.001. ***: p < 0.001.

Variables		Categories	without Recurrences (n = 200)	with Recurrences (n = 38)	p-Value
Years, mean ± St. dev.		-	72.6 ± 12.1	72.1 ± 10.9	0.406
Mobility, n [%]		Bedridden	15 [7.5]	4 [10.5]	0.106
		Assisted with aids	90 [45.0]	25 [65.8]
		Assisted without aids	14 [7.0]	2 [5.3]
		Autonomous with aids	53 [26.5]	5 [13.1]
		Autonomous without aids	28 [14.0]	2 [5.3]
State of consciousness, n [%]		Psychomotor agitation	6 [3.0]	8 [21.0]	<0.001 ***
		Confusion	11 [5.5]	5 [13.2]
		Vigilant	183 [91.5]	25 [65.8]
Drugs acting on vital signs	Antiarrhythmics, n [%]	Yes	51 [25.5]	5 [13.2]	0.100
		No	149 [74.5]	33 [86.8]
	Diuretics, n [%]	Yes	105 [52.5]	18 [47.4]	0.562
		No	95 [47.5]	20 [52.6]
	Hypotensive, n [%]	Yes	136 [68.0]	30 [78.9]	0.178
		No	64 [32.0]	8 [21.1]
	Hypoglycemics, n [%]	Yes	10 [5.0]	5 [13.2]	0.058
		No	190 [95.0]	33 [86.8]
	Laxatives, n [%]	Yes	14 [7.0]	3 [7.9]	0.844
		No	186 [93.0]	35 [92.1]
Psychoactive drugs	Hypnotics, n [%]	Yes	17 [8.5]	7 [18.4]	0.063
		No	183 [91.5]	31 [81.6]
	Opioid, n [%]	Yes	2 [1.0]	0 [0.0]	0.536
		No	198 [99.0]	38 [100.0]
	Sedatives, n [%]	Yes	38 [19.0]	13 [34.2]	0.036
		No	162 [81.0]	25 [65.8]
Visual impairment, n [%]		Yes	20 [10.0]	2 [5.3]	0.428
		No	180 [90.0]	36 [94.7]
Antidepressants, n [%]		Yes	32 [16.0]	13 [34.2]	0.009 **
		No	168 [84.0]	25 [65.8]
Antiepileptics, n [%]		Yes	28 [14.0]	5 [13.2]	0.890
		No	172 [86.0]	33 [86.8]
No drug therapy, n [%]		Yes	6 [3.0]	1 [2.6]	0.902
		No	194 [97.0]	37 [97.4]
Antiparkinsonians, n [%]		Yes	3 [1.5]	0 [0.0]	0.447
		No	197 [98.5]	38 [100.0]
Systemic antihistamines, n [%]		Yes	1 [0.5]	0 [0.0]	0.662
		No	199 [99.5]	38 [100.0]
Morse, mean ± St. dev.		-	32.5 ± 17.8	53.3 ± 22.1	<0.001 ***
Stratify, n [%]		0	8 [15.4]	0 [0.0]	0.004 **
		1	26 [50.0]	5 [23.8]
		2	16 [30.8]	10 [47.6]
		3	2 [3.8]	5 [23.8]
		4	0 [0.0]	1 [4.8]
Risk, n [%]		Yes, preventive personalized measures	51 [25.5]	15 [39.5]	<0.001 ***
		Yes, preventive standard measures	149 [74.5]	23 [60.5]
Hour of fall, n [%]		Morning	86 [43.0]	14 [36.8]	0.875
		Night	36 [18.0]	8 [21.1]
		Afternoon	48 [24.0]	9 [23.7]
		Evening	30 [15.0]	7 [18.4]
Location, n [%]		Elevator	1 [0.5]	1 [2.6]	0.162
		Bathroom	30 [15.0]	2 [5.3]
		Room	148 [74.0]	34 [89.5]
		Corridor	9 [4.5]	0 [0.0]
		Outdoors	3 [1.5]	0 [0.0]
		Gym	3 [1.5]	0 [0.0]
		Waiting room	5 [2.5]	0 [0.0]
		Other	1 [0.5]	1 [2.6]
Mechanism, n [%]		Tripped	17 [8.5]	2 [5.3]	0.015
		Loss of consciousness	6 [3.0]	0 [0.0]
		Loss of balance	52 [26.0]	7 [18.4]
		Loss of strength	26 [13.0]	3 [7.9]
		Slipped	70 [35.0]	11 [28.9]
		Others	29 [14.5]	15 [39.5]
Dynamic, n [%]		Waiting in wheelchair	6 [3.0]	1 [2.6]	0.350
		Falling from bed while sleeping	4 [2.0]	2 [5.3]
		While urinating	15 [7.5]	0 [0.0]
		During assisted movements	8 [4.0]	0 [0.0]
		During personal cleaning	8 [4.0]	1 [2.6]
		While dressing up	4 [2.0]	0 [0.0]
		Not available	16 [8.0]	4 [10.6]
		While picking up items	7 [3.5]	0 [0.0]
		While getting in/out of bed	52 [26.0]	8 [21.1]
		While getting in/out of the wheelchair/chair	49 [24.5]	11 [28.9]
		While removing restrains aids	2 [1.0]	1 [2.6]
		Other	29 [14.5]	10 [26.3]
Clinical impact, n [%]		With injury	69 [34.5]	13 [34.2]	0.973
		Without apparent injury	131 [65.5]	25 [65.8]
Contusions, n [%]		Yes	24 [12.0]	3 [7.9]	0.464
		No	176 [88.0]	35 [92.1]
Distortions, n [%]		Yes	1 [0.5]	0 [0.0]	0.662
		No	199 [99.5]	38 [100.0]
Hematoma, n [%]		Yes	6 [3.0]	0 [0.0]	0.280
		No	194 [97.0]	38 [100.0]
Excoriation, n [%]		Yes	22 [11.0]	3 [7.9]	0.567
		No	178 [89.0]	35 [92.1]
Wounds, n [%]		Yes	9 [4.5]	2 [5.3]	0.837
		No	191 [95.5]	36 [94.7]
Fractures, n [%]		Yes	7 [3.5]	2 [5.3]	0.601
		No	193 [96.5]	36 [94.7]
Head trauma, n [%]		Yes	17 [8.5]	2 [5.3]	0.500
		No	183 [91.5]	36 [94.7]
No injuries, n [%]		Yes	20 [10.0]	6 [15.8]	0.294
		No	180 [90.0]	32 [84.2]
Cryotherapy, n [%]		Yes	16 [8.0]	3 [7.9]	0.982
		No	184 [92.0]	35 [92.1]
Emergency department, n [%]		Yes	4 [2.0]	0 [0.0]	0.379
		No	196 [98.0]	38 [100.0]
Medication, n [%]		Yes	31 [15.5]	6 [15.8]	0.964
		No	169 [84.5]	32 [84.2]
No interventions, n [%]		Yes	82 [41.0]	17 [44.7]	0.668
		No	118 [59.0]	21 [55.3]
Observation, n [%]		Yes	44 [22.0]	7 [18.4]	0.622
		No	156 [78.0]	31 [81.6]
RX, n [%]		Yes	32 [16.0]	4 [10.5]	0.388
		No	168 [84.0]	34 [89.5]
CT brain, n [%]		Yes	27 [13.5]	5 [13.2]	0.955
		No	173 [86.5]	33 [86.8]
Medical therapy, n [%]		Yes	7 [3.5]	0 [0.0]	0.242
		No	193 [96.5]	38 [100.0]
Prognosis, n [%]		None	137 [68.5]	27 [71.0]	0.706
		Mild ≤ 3 days	45 [22.5]	7 [18.4]
		Moderate (from 4 to 20 days)	14 [7.0]	2 [5.3]
		Severe (from 21 to 39 days)	3 [1.5]	2 [5.3]
		Serious ≥ 40 days	1 [0.5]	0 [0.0]
Sentinel event, n [%]		Yes	80 [40.0]	13 [34.2]	0.851
		No	120 [60.0]	25 [65.8]

shows the results of the comparison among anamnestic, clinical, and management parameters between those patients who experienced falls only once and those who experienced fall recurrences. Notably, 200 patients fell once, while 38 patients fell twice or more.

The statistical analyses revealed that the Morse scores between the two populations were statistically significant (32.5 ± 17.0 versus 53.3 ± 22.1); moreover, the Stratify scores equal or greater to two increased from about 25% (Table 2) to more than 75% in the case of the patients who experienced multiple falls (Table 3). This result also implies that the number of preventive personalized measures increased by about 10% when considering multiple falls. Table 2 and Table 3 also show statistical differences regarding the state of consciousness; indeed, 91.0% of the patients who fell once were vigilant (about 87.8% when also considering the first fall of the remaining 38 patients, as shown in Table 2), while for those who fell more times, only 65.0% were vigilant. The results were also more significant for the patients’ psychomotor agitation (21.1%) and confusion (13.2%). Table 2 and Table 3 also reveal that the administration of antidepressants proved an important predictive value for patients who fell more than once; indeed, 34.2% of those patients took them. Moreover, the mechanism of fall was also found to be important: 39.5% of patients’ falls were due to unconventional mechanisms, such as falling off the bed or trying to overcome an obstacle. Furthermore, the chi-square test revealed that patients who took antiepileptics and hypnotics fell more frequently than those who did not take those drugs.

Finally, the main consequences of the 291 falls reported at the Maugeri Institute of Bari were analyzed (

Table 4. Statistics regarding the consequences of falls relative to the extra costs sustained by the hospital.

Variables	Categories	Statistics
No interventions, n [%]	Yes	132 [45.4]
	No	159 [54.6]
Cryotherapy, n [%]	Yes	21 [7.2]
	No	270 [92.8]
Emergency department, n [%]	Yes	4 [1.4]
	No	287 [98.6]
Medication, n [%]	Yes	42 [14.4]
	No	249 [85.6]
Observation, n [%]	Yes	61 [21.0]
	No	230 [79.0]
RX, n [%]	Yes	42 [14.4]
	No	249 [85.6]
CT brain, n [%]	Yes	37 [12.7]
	No	254 [87.3]
Medical therapy, n [%]	Yes	8 [2.7]
	No	283 [97.3]
Prognosis, n [%]	None	209 [71.8]
	Mild ≤ 3 days	59 [20.3]
	Moderate (from 4 to 20 days)	17 [5.8]
	Severe (from 21 to 39 days)	5 [1.7]
	Serious ≥ 40 days	1 [0.4]
Sentinel event, n [%]	Yes	111 [38.1]
	No	180 [61.9]

). Table S1 shows the statistics regarding the consequences of falls in terms of the health of the patients.

The prognosis and impact on the recovery, registered by doctors on fall reports, revealed significant results. Specifically, among the 291 falls, 68.4% were without any apparent injury. Contusions (11%), abrasions (10%), head trauma (6.9%), lacerations (3.8%), and fractures (3.4%) were the main consequences. Notably, 45.4% of the registered falls did not require any consequent intervention; in other cases (21%), doctors set up an observation period ranging from 2 to 48 h, while medication was required in 14.4% of the cases. Among all the diagnostic exams, RX (14.4%) and CT of the brain (12.7%) were the most performed procedures. In total, 71.5% of the falls we studied did not lead to any prognosis, which means that fall events did not influence the rehabilitative program or the clinical picture of the patient. Only one event (0.3%) had a severe prognosis, an event that led to consequences on the rehabilitative program in a temporal range greater than 40 days.

4. Discussion

This paper involves an investigation of the fall risk within a rehabilitation hospital using a managerial approach; specifically, for this study, we selected the define–measure–analyze (DMA) methodology to identify the possible risks and causes that could lead to a fall in a rehabilitative hospital. Figure 4 illustrates the main and secondary causes found by this multidisciplinary team, which, in turn, suggest the potential actions to be taken as part of the next (feasible) stages of the DMAIC cycle, i.e., improve and control. These stages, as already revealed from the previous literature on healthcare management, have proven effective to support the interventions to be introduced, as also demonstrated in the cause-and-effect diagram (i.e., a diagram equivalent to that in Figure 4) presented by Al Kuwaiti and Subbarayalu [43]. In that study, the causes to be addressed (and the consequent targeted interventions) proved effective to help the reduction in the fall rate per 1000 patients. Furthermore, our attention was focused not on the overall analysis of the falls registered within the units selected in the context of this study, but even on examining the causes that lead several of the patients to fall again (fall recurrence).

Although this methodology has been already used in several medical specialties, as indicated in Section 1, the risk of falls seemed to be scarcely investigated using managerial approaches such as LSS. In fact, previous research has documented only a couple of applications of Six Sigma—and, even less, of the define–measure–analyze–improve–control (DMAIC) strategy—in the context of fall rate monitoring and improvement. In particular, the previously cited work by Al Kuwaiti and Subbarayalu (also with the help of the multidisciplinary team involved) focused on acquiring the data of patients’ falls, which were collected throughout an entire year, and applied the DMAIC strategy to assess the (even ongoing) progress of the number of falls [43]. Following the introduction of corrective measures after data analysis, the authors observed a significant reduction in patients’ falls (from 7.18 to 1.91 per 1000 patient days), which corresponds to an improvement of 70%, higher than the most optimistic initial expectations (a reduction of 60%). Although this strategy proved to effectively limit the incidence of fall risk, the previous study did not provide an understanding of whether the presented data involved the overall patients’ falls (recurrences included); in fact, within the measure phase, similar anamnestic and clinical data (such those shown in the previous section in this study) were neither presented nor discussed and, by the authors own admission, the data acquired during the DMA steps could not help to study the impact of fall rate on the costs of patient care (by contrast, the data listed in Table 4 of this study can help to discern which type of health services that are needed to investigate patients’ status after a fall event could result in more expenses for the national health service). In this last context, Hill et al. proposed a preliminary prospect of the estimated cost saving that potentially be achieved with a reduction in fall risk. Aiming to decrease inpatient falls and falls with injury by 30%, the authors proposed the use of DMAIC, a solution that resulted in a reduction in the number of falls without consequences by 40% (from 5.00 to 3.00 per 1000 patient days) and by 70% for falls with injuries (from 0.58 to 0.16 per 1000 patient days). With these improvements, the authors estimated an annual cost saving of about USD 482,000 for the US national health system [44].

In several other studies, only statistical analyses were implemented. In this context, several researchers evaluated aspects that were also considered in this work. For instance, Castaldi et al. hypothesized that polypharmacy, and even the so-called fall-risk-increasing drugs (FRIDs), could represent a non-negligible factor of fall risk for inpatients of rehabilitation hospitals, although previous studies revealed no unequivocal results. To prove their hypotheses, a retrospective study was conducted analyzing data from inpatients hospitalized in a rehabilitation hospital in northern Italy in 2018. After collecting a conspicuous number of anamnestic and clinical parameters, a conditional logistic regression analysis was used to evaluate the impact of 13 FRIDs on the fall risk, while a logistic regression model was used to determine the associations between the number of administrated FRIDs and fall events. From the outcome of those statistical analyses, the authors found that the use of antidepressants effectively increased the risk of falls (a result we also observed when analyzing the recurrences of falls) and that polypharmacy could be another effective factor, providing evidence that suggests the idea of reconsidering the current fall risk assessment tools, which often overlook these factors [45]. In another study, the focus was instead on understanding which of the routine procedures administered at admission and discharge was the most informative in discriminating fallers and non-fallers among inpatients. According to the results of a multivariate logistic analysis, the Morse scale scores were among the factors considered a high predictor of falls in discriminating the faller and non-faller groups [46]. This finding is in line with the data shown in Table 3, where this same parameter was confirmed to also be the “distinguishing” factor for recurrences of fall events.

Despite the findings of the previous study, this seems to be the first study based on a managerial approach where, in the first stage, a large number of anamnestic and clinical data were collected and analyzed for inpatients experiencing falls, and in the second stage, similar data were compared between a group of one-time fallers and recurrent fallers. The findings presented in the previous section could nevertheless help to perform comparisons with other previous studies.

As shown in the previous section, the data listed in Table 2 provide a cross-sectional overview of the anamnestic and clinical situation of the 238 fallers in this study. Considering the results inferable from fall screening tools used in the first step, it was found that, in our institute, the mean value of the registered Morse scale was 33.3 ± 18.5. Since we ascertained that the Morse scale scores of several patients reached a total value greater than 45 (high risk of fall), following the instructions provided in the protocol, the Stratify score was also evaluated. In this study, the Stratify scale was assessed for 68 patients, and 33.8% of them had a fall risk with a score equal to 2. In the literature, to the best of our knowledge, no similar works have considered the Morse and/or Stratify scale scores as parameters to include in a managerial approach aimed at investigating (and potentially reducing) fall risk. For instance, previously, Al Kuwaiti and Subbarayalu [43] reported that the investigation of fall risks using the Morse scale was introduced later as an improvement measure. Although the introduction of Morse assessment effectively limited the registered falls, since several clinical areas of the hospital were considered, and since the previous literature reported the use of different fall risk assessment tools (alone or combined) in multiple different settings (for instance, the Hendrich Fall Risk Model II (HFRM II) in the case of an acute geriatric unit [47]; the revised Casa Collina Fall Risk Assessment Scale in three inpatient rehabilitation facilities [48]; the Conley and Stratify scales in multiple units of a university hospital [49]; the HFRM II and the Morse scale in an acute care setting [50]; the HFRM II and the Morse and Stratify scales in an acute care setting [51]), further investigations are needed to understand if the registered reduction in the number of falls in that setting (i.e., the hospital considered in the study by Al Kuwaiti and Subbarayalu) is consistent with the information the authors reported [43].

Moreover, in line with a previous study by Castellini et al. [52], who found that the use of the Stratify scale was not appropriate as the only tool for fall prevention programs, a logical and natural solution was to consider a multidimensional fall assessment followed by a multifactorial intervention. Focusing on similar aspects of the previous experts, a later investigation of the clinical risk management and patient safety of the Tuscany region (Italy) focused on, among the other objectives, designing an alternative risk assessment tool that could incorporate multiple features of the “core set” of risk factors considered by pre-existing fall risk assessment tools. The novel tool designed by these experts (ReTos) included all the core sets of the Conley, Morse, and Stratify scales (which are some of the most used scales, as indicated in this study). This strategy helped researchers to perform appropriate comparisons between these scales within the same population and to gain better insights into the information acquired (often in an inhomogeneous manner) from these three scales [53]. Overall, the evidence reported in these studies might also have inspired the experts of the Istituti Clinici Scientifici Maugeri in implementing the combined use of the Morse and Stratify scales, as reported in Section 2 and in the work of Castaldi et al. [45]. Nevertheless, the current paper seems to be the first one that aimed to consider a multidimensional fall assessment, simultaneously with more than a single assessment tool [9] for patients at high fall risk, within the context of a managerial approach.

The mean value of the Morse scale and the cataloging of patients following the Stratify cut-offs also allowed for the identification of the interventions/corrective measures to be adopted; in this study, standard preventive measures were applied to 169 patients (71.0%), while personalized preventive measures were applied to 69 patients (29.0%). Again, in the most similar previous study, Al Kuwaiti and Subbarayalu [43] did not consider these aspects in depth in their proposed managerial approach.

Considering the previously cited works, we found that the majority of first falls involved patients who used assistive aids. This finding was confirmed in a previous study by Roman de Mettelinge and Cambier [54], who ascertained that older adults (albeit with a mean age higher than that of the 238 patients enrolled for this study) using walking aids were more prone to falls than patients not using these aids in a residential aged care facility. In the most similar previous study, however, Al Kuwaiti and Subbarayalu [43] did not seem to have comprehensively considered these aspects in their proposed managerial approach.

In this study, the analysis of the 238 fallers suggested that the state of consciousness might be a very important clinical parameter since the results of our descriptive statistical data analysis highlight that 87.8% of the patients were vigilant and well oriented, in contrast to 6.3% who were confused, and 5.9% who reported a psychomotor agitation. Furthermore, 10.1% of the 238 fallen patients showed a visual impairment that can be considered one of the most predisposing factors for a fall. This finding is consistent with the systematic review of Viera et al. [55], who highlighted that, among other factors, confusion and cognitive impairment are also risk factors for geriatric patient falls in rehabilitation hospital settings. Again, in the most similar previous study, Al Kuwaiti and Subbarayalu [43] did not consider these aspects in more detail in their proposed managerial approach.

Although not analyzing the fall risk using a managerial approach, previous studies have also reported similar results regarding the time, location, and mechanisms of falls. For instance, Castaldi et al. [45] also observed that most of the falls happened during mornings and afternoons; furthermore, the same authors also ascertained that the recovery room and the toilet were the riskiest locations, and the most frequent falling mechanisms proved to be patients’ getting off the bed and wheelchairs.

Finally, the focus was also on the impact of medications on fall risk. As shown in Figure 1 and Table 2, the collected data revealed that more than half of the fallers took hypotensive drugs and were under diuretic therapy. As already discussed in this section, Castaldi et al. [45] indicated that antidepressants may increase the risk up to two times than that of a control group (non-fallers) with (among others) similar Morse scores. Antidepressants were administered to only 16.4% of the fallers in this study, suggesting that the presented result might be controversial. Conversely, Castaldi et al. [45] revealed that diuretics might be considered among the administrated FRIDs associated with fall events when matching cases and controls according to age, sex, and hospital ward; however, this association was absent when the cases and controls were matched, among others, based on their Morse scores. Nevertheless, according to the authors’ admission, since the sample size was relatively small, and there were also no homogeneous diagnoses at admission, further observation may refute or diminish the former non-statistical correlation. Finally, in other studies, different researchers established that diuretics strongly affected the risk of falls [56], another finding in line with our study.

Regarding the data shown in Table 3, particular attention was paid to analyzing the potential differences that could have led to the recurrence of falls. As previously mentioned, to the best of our knowledge, no similar managerial studies were found in which these aspects were analyzed to a similar level of detail. However, previous researchers have compared the anamnestic and clinical parameters of fallers and recurrent fallers. For instance, Fischer et al. [57] performed a retrospective study to gain insights into the potential predictors of inpatient falls in an academic hospital in the USA. The authors revealed that a total of 1082 inpatients fell during the considered period, and only about 120 patients (11% of the overall number) experienced repeated falls. The results of Pearson’s chi-square and Fisher’s exact tests revealed that the only potential predictor (among the parameters investigated) were the “elimination-related falls”—namely, falls linked to ambulating to or from the bathroom or bedside commode, reaching for toilet tissue, exiting a soiled bed, or using the toilet or bedside commode—which were more likely to occur among one-time fallers than recurrent fallers. Although, the study considered not only rehabilitation units, aspect which might suggest the result may be conflicting with such environments (in fact, for instance, in the current study we did not find a global statistical significance among the dynamics of fall of the two groups). In this context, Anderson and Lane [58] analyzed instead the characteristics of falls and recurrent falls in residents (in a timeframe of two years) of an aging-in-place community in the USA. The authors found that 60% of the subjects who fell (30/50 overall) experienced a fall only once, whereas 21/30 subjects (70%) experienced multiple falls. From the outcome of their investigation, the authors found significant differences in sitting-to-standing balance (p = 0.011) and surface-to-surface transfer balance (p = 0.011) between the participants who had recurrent falls and those with single falls, an aspect that was also highlighted in the data found in the current study (see Table 3, “Mechanism”). Furthermore, Anderson and Lane also found a significant difference in the use of antidepressant medication between those who had recurrent falls and those who had a single fall (p = 0.019), another result in line with the evidence found in the current study.

Finally, in Figure 2 and Table 4, the main consequences following both the first and recurrent falls are summarized. As indicated in previous sections, again, to the best of the authors’ knowledge, no similar managerial studies have been conducted in which these aspects were analyzed in similar detail. In this context, we found that Al Kuwaiti and Subbarayalu [43] only accounted for this issue in their study design, postponing a more accurate and deep study of the issue (possibly, using a managerial approach) to a future study. Although in the previous literature, several studies collected data regarding this aspect, only in [59] were several of the consequences (related to injuries) we considered in our work investigated. In particular, Hitcho et al. found a similar trend regarding the registered types of injuries; in fact, considering only first falls, abrasions, contusions, and lacerations were found with a similar percentage to that observed in our study, and consequently, those were found to be among the most common consequences. However, the trend of the prognosis after the fall was not so similar; in fact, although moderate, severe, and serious prognoses presented values in line with those listed in Table 4 of our study, Hitcho et al. found that only about 58% of the falls occurred without prognosis, while the percentage of the falls with mild prognosis was about 34% (which was 67% higher than the data registered in this work). Nevertheless, it should be noted that Hitcho et al.’s study was not focused on a rehabilitation hospital; in fact, several of the registered 183 falls happened in units in which rehabilitation was not the main focus (for instance, 4 falls were registered in women’s and infants services), and the extra costs incurred due to the falls was neither considered in that study nor planned for future investigations. In this context, it should be noted that interventions following falls in hospitals have, of course, consequences in terms of both patient health and the financial resources of the hospital. In this study, we estimated the extra costs sustained by the hospital due to these falls: when considering only additional interventions (cryotherapy, emergency department, medication, observation, RX, CT brain, and medical therapy), an extra cost of EUR 33,363 was estimated in two years, without taking into account the costs associated with the prolonging LOS after falls, which would also have a significant impact on this amount.

5. Conclusions

Falls are a significant clinical, legal, and regulatory issue for hospitals and pose a high risk to patient safety. Hospitalized patients are exposed to the risk of falls and their consequent injuries, which have a severe impact on both patients and their families, as well as the increase in patients’ LOS, corresponding, in turn, to higher financial costs for hospitals.

To this end, we considered a managerial approach in which the objective was to analyze (with increased clarity) the results of the data of falls that occurred in the Istituti Clinici Scientifici Maugeri during the years 2020 and 2021. The data analysis allowed us to have a cross-sectional overview of the anamnestic and clinical state of 238 patients (fallers) considering only their first fall; this overview helped the multidisciplinary team to have a more clear understanding of the current state of fall occurrence in the four wards selected for this study (the neurorehabilitation unit, the cardiac rehabilitation unit, the functional rehabilitation unit, and the pneumological unit). Such knowledge—in the “measure” phase of the DMA approach—supported our analysis, which aimed to investigate the potential statistical differences between the variables that could lead patients to repeated falls. From the outcome of this investigation, it was found that fall risk assessment tools (i.e., the Morse scale and the Stratify scale) could help in detecting potential repeated falls as well as in the use of personalized measures. It was also found that repeated fallers presented a different state of consciousness and that the administration of antidepressants could increase the risk of repeated falls. In the previous sections, the typical consequences of both first falls and recurrent falls were also highlighted, and lastly, their potential impact on the extra costs incurred by the hospital was also investigated.

From this research, it is possible to conclude that each organization should provide a systematic process for an objective evaluation of the risk of falls for each admitted patient. This process should start with the evaluation of the risk of falls of patients in order to apply all the useful interventions to invert or reduce the associated risks. However, most of the previous studies highlighted difficulties in data collection. In this study, we did not encounter this limitation since the multidisciplinary team devised a precise and punctual recording plan of the falls inside the Istituti Clinici Scientifici Maugeri of Bari with the objective of possibly obtaining a reliable report of the falls.

On the basis of the findings presented in this paper, working on the remaining issues would be of interest; firstly, the indications following the DMA phases could be further studied in the “improve” and “control” phases (which would also focus more on other relevant main causes found in the literature, which are highlighted in Figure 4, namely, medical staff and methods) so as to evaluate and validate the findings of the DMA phases. Secondly, the obtained data, even following the “control” stage, could be further analyzed to devise a model to predict the risk of fall occurrence (possibly, even to distinguish fallers from recurrent fallers). Finally, this approach might be assessed in the context of future prospective studies in order to identify and evaluate other risk factors (such as the correlation between falls and cognitive decline) or to focus on other diseases such as Parkinson’s disease, ictus, and dementia.