Association Between Characteristics of Injurious Falls and Fall Preventive Interventions in Acute Medical and Surgical Units

Abstract

Background

Hospital falls remain common and approximately 30% of falls in hospital result in injury. The aims of the study were the following: (i) to identify the association between fall interventions present at the time of the injurious fall and injurious faller characteristics and (ii) to identify the association between fall preventive interventions present at the time of the injurious fall and the injurious fall circumstances.

Methods

Secondary data analysis of deidentified case series of injurious falls across 24 acute medical/surgical units in the United States. Variables of interest were falls prevention interventions (physical therapy, bed alarm, physical restraint, room change, or a sitter) in place at the time of fall. Data were analyzed using logistic regression and hazard ratios.

Results

There were 1,033 patients with an injurious fall, occurrence peaked between Day 1 and Day 4, with 46.8% of injurious falls having occurred by Day 3 of admission. Injurious fallers with a recorded mental state change 24 hours prior to the fall were more likely to have a bed alarm provided (adjusted odds ratio [OR] 2.56, 95% confidence interval [CI] 1.61, 4.08) and receive a physical restraint as fall prevention interventions (adjusted OR 6.36, 95% CI 4.35, 9.30). Injurious fallers restrained fell later (stay Day 6) than those without a restraint (stay Day 4) (p = .007) and had significantly longer lengths of stay (13 days vs 9 days).

Conclusions

On medical/surgical units, injurious falls occur early following admission suggesting interventions should be commenced immediately. Injurious fallers who had a physical restraint as an intervention had longer lengths of stay.

Falls in hospital settings remain an ongoing problem with up to 1 million falls occurring annually in U.S. hospitals (1,2). They compromise patient safety and cause injury in approximately 23%–40% of patients who fall (3,4). Between 4% and 8% of these injuries are classified as moderate or severe including hip fractures, head injuries, and in some cases, death (4,5). Given the volume of hospital falls and the associated consequences, preventing falls and fall-related injuries is a major patient safety indicator for quality improvement initiatives.

Despite quality improvement efforts and guidelines that span decades, only a modest reduction in hospital falls have occurred (1,6). Preventing hospital falls is complex and requires understanding and targeting interventions at not only the patient level, but also at the clinician and organizational level (7,8). For patient-level strategies, national guidelines suggest conducting standardized fall risk assessments combined with clinician-determined strategies that encompass patient/family education, increased surveillance and visibility, environmental modifications, medication review, mobilization, and medical interventions for specific risk factors (eg, addressing hyponatremia or hypoglycaemia) (2,9).

Unfortunately, fall risk assessment tools provide only modest predictive validity (10) and few rigorously designed studies have been conducted to examine the effectiveness of the suggested interventions, either as single or multiple interventions (11,12). Randomized trials evaluating interventions that engage patients directly through education have shown reductions in both falls and injurious falls on subacute wards (13,14) and tailored patient risk management, including patient education and stakeholder alerts, significantly reduced fall rates, but not fall-related injuries in acute medical units (15). Trials investigating the use of individual interventions including low beds, bed alarms, and patient sitters have not demonstrated significant reductions in hospital falls or injurious falls (16–18).

Current evidence is hampered by the nature of quality improvement designs including reliance solely on incident report systems to identify and characterize fall events (19,20), retrospective nature of the designs, small sample sizes, single settings, or injurious falls reported as a smaller subset of all falls (3,21,22). Thus, little is known regarding the type or combination of interventions in place at the time of fall event or the association of faller characteristics, fall prevention interventions, and circumstances of injurious fall events. Understanding these associations will provide greater direction for hospital fall prevention strategies and quality improvement initiatives. In the present study, we hypothesized that on acute medical and surgical wards: (i) patient characteristics are not associated with the type of fall prevention interventions in place; (ii) fall circumstances are not associated with the type of fall prevention interventions in place

Therefore, the aims of the study were the following: (i) to determine the characteristics of injurious fallers and the circumstances of the injurious falls events in acute medical and surgical hospital wards; (ii) to identify the association between fall interventions present at the time of the injurious fall and injurious faller characteristics; and (iii) to identify the association between fall preventive interventions present at the time of the injurious fall and the injurious fall circumstances.

Methods

Design

We conducted a secondary data analysis of deidentified data from a multisite matched case–control study. The primary data collection was approved by the Institutional Review Board of the University of Tennessee Health Science Center and the secondary analysis was approved by the University of Florida Institutional Review Board.

Settings

The original study was undertaken to examine the effect of the Centers for Medicare and Medicaid Services (CMS) nonpayment for injurious falls as a “never event” on patient-level approaches for fall prevention (23). As part of this study, a case–control study was conducted at four hospitals in the Methodist Healthcare System, Memphis TN to gather patient-level data to determine patient-level risk factors for injurious falls, circumstances of injurious falls, and prevention strategies from January 2005 through December 2010. These four hospitals used the same electronic health record format, fall prevention protocol, and adverse incident event reporting system. The four hospitals ranged in size from 200 to 635 beds and included a university teaching hospital (Methodist University Hospital; 15 medical, surgical, or medical/surgical units) and three community hospitals (Methodist North, Methodist South, and Methodist Germantown; nine medical, surgical, or medical/surgical units). Specialty wards such as pediatric, psychiatric, maternity, and emergency departments were excluded.

Sample

Cases consisted of all patients who sustained a single injurious fall on one of the participating medical/surgical units during the study period. A fall was defined as “an unintentional change in position which resulted in coming to rest on the ground or a lower level.” Fall injury was classified as minor (resulting in pain, bruise, abrasion, laceration), moderate (resulting in muscle/joint strain, dislocation), or severe (resulting in open or closed head injury, internal injury, or fracture). Exclusion criteria were falls with no documented injury, injurious falls resulting from catastrophic clinical events, such as stroke, seizure, or arrhythmia and multiple fall events. To isolate patient-level factors from unit-level factors, for example, staffing, up to two controls were matched by (i) unit, (ii) day of the fall event, and (iii) length of stay at the time of the event. For this secondary data analysis, only cases were included since the area of interest was whether the interventions provided for these injurious fallers were associated with their risk factors or the circumstances of the fall.

Data Collection Procedure

The procedure for data collection has been described previously (5,20). Briefly, cases were identified prospectively via the incident report system or the fall reporting system within the hospital system (20). A trained research assistant, who was experienced in falls research, collected all data from the medical records, incident report, and hospital fall reports using a standard template. This training and subsequent data collection procedure was supervised by the senior nurse researcher on the study.

Variables

Demographic data included admitting hospital, age, gender, comorbidities, number of medications, receipt of anesthesia within 24 hours of the injurious fall, and length of stay (LOS). Level of fall risk (high/low) was obtained closest to the time of the injurious fall. All hospitals used the Morse risk assessment tool (24). Circumstances of the fall event were the following: witness to fall, time, hospital day, location of fall, and type and severity of injury sustained.

Fall prevention interventions in place for the patients at the time of sustaining the injurious fall had to be documented in the patient’s’ medical records within 24 hours prior to the injurious fall event. Fall prevention interventions were the following: bed alarm, physical therapy, a sitter, room change (for closer staff observation), and physical restraint.

Statistical Analysis

A deidentified data base was made available to the primary authors (J.F.C., A.M.H.) via a secured website. Descriptive summaries consisted of frequency distributions for categorical data and medians, interquartile ranges, and ranges for continuous data. Chi squared and Fisher exact tests, as appropriate, were utilized for comparisons of frequency between intervention indicator categories (physical therapy, bed alarm, restraints) within the cases.

Kaplan–Meier survival probabilities were used to estimate time to injurious fall, injurious fallers’ LOS at the time of the fall by physical therapy and injurious fallers’ LOS by restraint in place at time of fall, with Log Rank tests used for group comparisons.

Logistic regression models, adjusted for age, were used to examine the effect of patient characteristic risk factors of LOS, number of medications, comorbidities, mental state change, falls risk classification (high/low), recent anesthesia (hypothesis i), and falls circumstances of witness to fall, time, stay day, location of fall, and severity of injury sustained (hypothesis ii) on falls prevention intervention outcomes (including physical therapy, physical restraint, and bed alarm usage). Risk factors that were significant at 85% (p < .15) in the adjusted models were selected as covariates for corresponding multivariable logistic regression models. Final multivariable models included covariates that were significant to 95%. Results were summarized using odds ratios (OR) and 95% confidence intervals (CI). Statistical analysis was undertaken using IBM SPSS version 24.0 (Armonk, NY) and Stata 15 (StataCorp LP, College Station, TX). Hypothesis tests were two-sided with p values less than .05 considered statistically significant.

Results

Patient, Fall Events, and Fall Prevention Interventions Profile

There were 1,033 unique patients who sustained an injurious fall during their hospitalization; demographic characteristics of injurious fallers are presented in full in Supplementary Table S1. Age of the cohort ranged from 18 to 90 years (mean age 63.7 [SD 16] years), with 522 (50.5%) being female and 151 (14.6%) having a diagnosis of dementia. In the 24 hours prior to the injurious fall, acute mental changes were present in 27% of cases and 10% of cases had received anesthesia.

Characteristics of the injurious fall events are shown in full in Supplementary Table S2. Figure 1 presents day to injurious fall; injurious falls peaked early during admission (Median [95% CI] 4 days [3.7–4.4]); 54.8% of the injurious falls had occurred by Day 4 and 78.7% had occurred by Day 8, for all patients (see also Supplementary Table S3). The time of the injurious fall event was proportionate throughout the day and 75.1% of falls were unwitnessed. Most events (79.4%) occurred in the patient’s room and half (49.9%) were related to elimination needs. Injuries were predominantly minor in nature.

Figure 1.

Occurrence of injurious fall by stay day of patient admission

Falls prevention interventions documented as being in place at the time of the injurious fall on the units for cases are presented in full in Supplementary Table S4. Receiving physical therapy was the most frequently documented intervention with 32.4% of injurious fallers receiving physical therapy. There were n = 146 (14.1%) injurious fallers who received a physical restraint in the 24 hours prior to the injurious fall.

Falls Risk Classification

The falls risk classification of the injurious fallers was approximately equal (55% high risk, 45% low risk) when the injurious fall occurred. There was a significantly higher frequency of patients who were screened as being at high risk of falls receiving PT (61% vs 53%) and having a bed alarm (69% vs 54%;Supplementary Tables S5 and S6).

Association Between Fall Preventive Interventions in Place and Injurious Faller Characteristics

Comparisons of patient characteristics covariates and the fall prevention intervention of receiving physical therapy in injurious fallers are presented in Supplementary Table S5. Injurious fallers who had anesthesia or a reported mental state change within 24 hours of the fall, diagnosis of stroke, or hypertension, LOS 10 days of longer, were on four or more medications or at high risk of falling were significantly more likely to be receiving physical therapy as an intervention. Injurious fallers receiving physical therapy as an intervention had a significantly longer LOS compared to those who did not receive physical therapy (Supplementary Figure 3).

Comparisons of patient characteristics covariates and the fall prevention interventions of receiving a bed alarm or physical restraint in injurious fallers are presented in Supplementary Tables S6 and S7. Injurious fallers admitted through ER, with LOS 10 days or longer, diagnosis of dementia or reported mental state change 24 hours before the injurious fall were significantly more likely to receive a bed alarm or a physical restraint as an intervention. Injurious fallers with a restraint in place fell significantly later during their hospital stay compared to those without a restraint in place (p = .007) and had a significantly longer length of hospital stay (p = .011; see Figure 2 and also Supplementary Table S8 and Supplementary Figure 4).

Figure 2.

(a) Injurious faller length of hospital stay—restraint versus no restraint in place. (b) Injurious faller time to first fall—restraint versus no restraint in place.

Association Between Fall Preventive Interventions in Place and Injurious Fall Circumstances

Comparisons of fall circumstances and falls prevention intervention indicators (physical therapy, bed alarm, restraints) in injurious fallers are presented in Supplementary Tables S5–S7. There were no significant associations between the circumstances of the injurious falls (witness to fall, time, stay day, location of fall, location, and severity of injury sustained) and the fall interventions provided.

Adjusted Multivariable Analysis for Outcomes of Physical Therapy, Receiving a Bed Alarm and Physical Restraint

Adjusted multivariable analysis presented in Table 1 demonstrated that injurious fallers on four or more medications, 24 hours postanesthesia, categorized at high risk of falling, or LOS 10 days or longer, had significantly higher odds of receiving physical therapy as an intervention.

Table 1.

Adjusted Univariate and Multivariable Analysis for Outcome = Receiving Physical Therapy, Adjusted for Age

Predictor	Adjusted Univariable			Adjusted Multivariable
	OR	95% CI	p Value	OR	95% CI	p Value
4+ medications	2.24	1.44–3.49	<.001	2.20	1.38–3.51	.001
Anesthesia	1.84	1.22–2.78	.004	1.79	1.16–2.74	.008
LOS >10 days	1.93	1.48–2.51	<.001	1.83	1.40–2.40	<.001
High risk for falling	1.34	1.02–1.75	.035	1.39	1.05–1.83	.021
Comorbidity
Hypertension	1.40	1.02–1.92	.038
Stroke	1.50	1.06–2.14	.024
Mental state change	1.53	1.15–2.04	.004

Note: CI = confidence interval; LOS = length of stay; OR = odds ratio.

Multivariable analysis for the outcome of having a bed alarm is presented in Table 2. Injurious fallers with a recorded mental state change 24 hours prior to the fall had two and a half times odds of having a bed alarm provided as a fall prevention intervention. In addition, injurious fallers who had a sitter had almost four times odds of having a bed alarm and those receiving a room change had four and a half times odds of having a bed alarm provided.

Table 2.

Adjusted Univariate and Multivariable Analysis for Outcome = Having a Bed Alarm, Adjusted for Age

Predictor	Adjusted Univariable			Adjusted Multivariable
	OR	95% CI	p Value	OR	95% CI	p Value
LOS >10 days	1.58	1.01–2.48	.048
High risk for falling	1.81	1.11–2.94	.017
Dementia	1.79	1.01–3.18	.048
Intervention
Room change	3.80	1.19–12.07	.024	4.54	1.37–15.01	.013
Sitter	3.50	1.36–9.01	.009	3.97	1.52–10.40	.005
Physical therapy	1.61	1.02–2.53	.041
Mental state change	2.41	1.53–3.81	<.001	2.56	1.61–4.08	<.001
Anesthesia	0.32	0.10–1.04	.057

Note: CI = confidence interval; LOS = length of stay; OR = odds ratio.

Multivariable analysis (Table 3) also demonstrated that injurious fallers with a recorded mental state change 24 hours prior to the fall had six times odds of having a physical restraint in place as a fall prevention intervention.

Table 3.

Adjusted Univariate and Multivariable Analysis for Outcome = Physical Restraints in Place at Time of Fall, Adjusted for Age

Predictor	Adjusted Univariate			Adjusted Multivariable
	OR	95% CI	p Value	OR	95% CI	p Value
Admitted through ER	1.95	1.25–3.03	.003
Comorbidity
Dementia	2.94	1.87–4.62	<.001
Mental state change	6.80	4.65–9.96	<.001	6.36	4.35–9.30	<.001

Note: CI = confidence interval; OR = odds ratio.

Discussion

Findings from this secondary analysis of a large prospective case–control study demonstrated that injurious falls peak early in patients’ admission. Peak falls rates were previously reported as occurring during the first week of admission in subacute geriatric hospital units (25) where patients had a median length of stay of 22 days; the median stay day for injurious falls in our acute care cohort was Day 4 of admission. This peak in injurious falls may reflect patients having significant functional decline early in admission when they are acutely ill (26) and also unfamiliar with their new environment (27); this suggests that fall prevention interventions be commenced immediately following admission or transfer from another unit within the hospital.

Injurious fallers who received physical therapy were most likely to be post anesthesia, have a longer LOS, have diagnoses such as stroke and take more than four medications. This profile indicates that physical therapy was not causative of longer LOS, but suggests rather that LOS prompted orders for physical therapy, particularly targeted to frailer patients with high numbers of comorbidities and those recovering from surgery. A recent large study demonstrated that only 2% of hospital falls reported occurred during therapy-related activities (28) and in concurrence with a previous large cohort study, we found that 75% of fall events occur when patients are not supervised by staff (4). It is concerning that only 32% of the injurious fallers in the cohort were receiving physical therapy. In the acute setting, physical therapists are able to complete mobility assessments to provide guidance to staff and patients about safe mobility and concurrent falls risk, as impaired mobility is a known, strong risk factor for falls (27). Given trials have demonstrated that nurse led mobility programs and a therapy-driven program on acute units improved older patients’ mobility levels and did not increase falls (29–31), our results further pose the question, if physical therapy were ordered earlier during the hospital stay, would it result in attenuation of falls in this population?

Patients who experienced a mental state change 24 hours prior to the injurious fall were significantly more likely to have a bed alarm or restraint provided. Staff implementing falls preventive strategies in response to a documented mental state change is timely, however receiving a bed alarm as a prevention intervention has been shown in previous randomized trials to be ineffective in preventing patient falls (17,32) and other research has also suggested that physical restraint does not reduce falls (33). Our results demonstrated that these interventions were also not associated with reductions in the severity of falls or reduced falls at night. We also found that patients with a diagnosed mental state change in the 24 hours prior to their fall had six times odds of being physically restrained. An acute mental state change is characterized by increase in patients’ confusion and disorientation (34). Involuntary immobilization does not align with best practice recommendations for managing patients with acute confusion (35).

Additionally, injurious fallers who received a physical restraint as a prevention intervention stayed 4 days longer than those without restraint. The physical restraint appeared to delay the time to the first fall, but consequently these injurious fallers had longer LOS. Bed alarms and restraints may have unintended consequences of further reducing patient mobility, causing functional decline and disability (36,37), resulting in these longer LOS. Better evidence exists for keeping older patients safe, using patient-centered programs designed to educate older patients about safe mobility, prevent hospital complications, such as delirium and enhance functional recovery (14,15,35,38).

Findings also confirmed that risk scores are not useful for providing falls management with nearly half of the cohort being classified as at low risk of falls. However patients were significantly more likely to be receiving fall interventions if they were identified as being at high risk, indicating that using risk scores may divert fall prevention interventions away from a patient-centered approach. Previous research has demonstrated that risk scores do not reliably predict which patients are at risk of falls or injurious falls (10,22) and national guidelines also specifically recommend that risk scores are not used to direct fall prevention care (9). Instead it is suggested that for effective falls prevention on wards it is essential to assess and target all identified patient risk factors (2,9).

A key strength of this study was the data were collected on a large prospective cohort in both acute medical and surgical wards. Information regarding the type, frequency, and intensity of physical therapy delivered was not available for these analyses, thus limiting discussion or conclusions to be drawn regarding its true effect on fall prevention. Further documentation, detailing the physical therapy provided, would assist in understanding what components could possibly contribute to an effective multifactorial intervention for falls reduction. Additionally, all patients on medical and surgical wards potentially have fall prevention interventions in place which may not be specifically documented as “fall prevention.” These could include toileting schedules, medication review and managing individual patients’ medical conditions such as postural hypotension. Data were collected by one research assistant and intrarater reliability was not examined. However the research assistant was highly experienced in the field of hospital falls prevention and followed a standard extraction template and procedure throughout the study to collect all data on the characteristics and circumstances of the injurious falls. Data were not collected regarding staff level expertise and actions prior to the injurious fall event and organizational level fall prevention policies and programs in place when the data were collected. Staff and organizational level interventions that facilitate systematic translation of fall prevention evidence into effective implementation at patient level are known to be important (12,27).

Conclusion

Injurious falls on hospital medical and surgical wards occur early in the patient’s stay. This suggests that fall prevention interventions should be commenced on immediately on admission and re-enforced on a 24-hour basis throughout the patient’s stay. Receiving any fall intervention was not significantly associated with reductions in severity of the injuries experienced.

Injurious fallers who experienced a mental state change 24 hours prior to the fall were significantly more likely to receive a physical restraint and the use of physical restraint was associated with a longer length of hospital stay. Further research to understand which interventions are effective in reducing injurious patient falls on hospital medical and surgical units is required.

Funding

This work was supported by a merit award to A.-M.H. from the Department of Health, Western Australia, Australia. The data used in these analyses were from a primary study that was supported by a grant from NIH/R01AG033005 CMS Nonpayment for Iatrogenic Injury and the Risk of Falls in Hospitals.

Conflict of Interest

R.I.S. declares he serves as an expert witness regarding falls in hospital. The other authors have no conflicts to declare.