Falls in the inpatient setting are common, with published rates ranging from two to twenty per 1000 patient days1-4. According to the Joint Commission, the rate of falls in acute care hospitals resulting in serious injury or death has increased almost threefold since 20005. During this time period, the number of total hospital admissions only increased by 8.3% and the number of elderly patients (sixty-five years of age or older) admitted actually decreased6,7. Thus, this increase in falls cannot be attributed to a larger sample population or a greater proportion of traditionally high-risk patients.
Inpatient falls result in considerable morbidity and potential mortality2. Injury rates typically range from 25% to 45%2,4,8,9, with one study showing that the risk of serious injury from an inpatient fall was as high as 10%10. Commonly, 1% of falls result in fracture or dislocation2,8,11. However, one study of orthopaedic patients only showed a rate of 5%12. The psychological effects resulting from falls, including depression and anxiety, lead to impaired function13,14. Additionally, inpatient falls are associated with increased cost and length of hospital stay10,15. In 2008, the Centers for Medicare & Medicaid Services deemed inpatient falls as “never events” and stopped hospital reimbursement for the treatment of related sequelae from falls16,17. Many private insurance providers have followed this policy18-20, a move that has shifted substantial cost to hospitals.
Many studies have identified patient factors associated with falls2,4,9,11,15,21,22. Unfortunately, interventions that have been reported in previous fall-prevention studies, which primarily target high-risk patients, have been ineffective in the acute inpatient setting17,23-28. Studies attempting to address care environments (lighting, bed height, patient footwear, fall identifiers, and so on) have only demonstrated effects in the subacute setting3,29,30. Continuous quality improvement approaches have shown that interventions developed by means of this methodology can be effective and durable even in an acute setting31.
The present study tested the hypothesis that a systems-based fall-prevention program targeting high-risk situations would result in fewer falls with injury and fewer total falls.
Study Design
This is a prospective cohort study with consecutive historical controls adapted from a standard continuous quality improvement approach. The five key components to this approach are: (1) define area(s) for improvement, (2) identify all possible causes, (3) develop action plan and outcome metrics, (4) implement action plan, and (5) evaluate outcome metrics28.
The setting of the study was the orthopaedic ward of a large academic hospital. A prior detailed review of the hospital’s adverse event database, Patient Safety Net (University Healthsystem Consortium, Chicago, Illinois), which is used by all physicians, nurses, aides, and therapists, and into which all inpatient falls have been recorded since January 2005, allowed careful evaluation of the circumstances of each of those falls. Two common high-risk factors suspected of contributing to falls were identified. Falls were most often related to toileting activities, especially when patients attempted unassisted walking to the commode. Additionally, the analysis found that the first twenty-four hours after surgery was a particularly high-risk period for falls among patients who underwent inpatient upper-extremity surgery (mostly shoulder arthroplasty). On the basis of these findings, four systems-based interventions (Table I) were developed in partnership with the head nurse on the orthopaedic ward, professionals in the hospital’s Center for Clinical Excellence, and involved orthopaedic surgeons. The interventions were implemented on September 29, 2008, and were evaluated prospectively.
The “timed toileting” and “wake ’em, take ’em” interventions were designed to decrease the likelihood that a patient would attempt unassisted walking to the commode. When too much time elapses between voids, patients can develop a sense of urgency, making them unwilling to wait for assistance. Timed toileting was designed to limit the duration between voids, thus reducing the likelihood that a patient would act in haste. Many patients have articulated that after being awakened (for medication administration, vital assessment, wound care, and so on) and before falling back to sleep, they felt the need to use the restroom; wake ’em, take ’em seeks to minimize the chance that they will do so unassisted, particularly at night.
Often, patients require or request assistance getting to and from the commode. Even with call buttons available, some patients do not request help before or after toileting, and sometimes the delay between the call and the response results in patients trying to reach the restroom or return to bed alone. The “assist in, assist out” intervention stipulates that if a patient requires assistance into the restroom, he or she should be attended there and then assisted back to the bed or chair.
Somewhat counterintuitively, patients who underwent inpatient shoulder surgery were more likely to fall than lower-extremity surgery patients within twenty-four hours after surgery. It could not be determined whether this finding was a function of staff overestimating the ability of these patients to walk safely; patients themselves underestimating the residual effects of anesthetics, analgesics, and sedative agents; or other factors such as gait imbalance from an upper-extremity sling in an older population. In light of this, patients were informed of the finding by the physician staff and were given printed materials about fall safety and prevention during their preoperative visit in the clinic. Additionally, the floor nurses and nursing assistants emphasized the importance of the “shoulder safety” intervention by asking all shoulder surgery patients not to walk unassisted for the first twenty-four hours after surgery.
Workflow Evaluation, Employee Education, and Program Implementation
The proposed interventions were reviewed in light of existing nursing and nursing assistant workflows. Factors that were considered in the integration of the interventions included how best to encourage all staff (including registered nurses, nursing assistants, physical therapists, occupational therapists, and therapy aides) to share the responsibility of fall prevention and create a culture of safety, educate the staff and disseminate the new approaches, and document the use of the interventions in a way that would allow an audit of compliance.
Importantly, the interventions were considered performance expectations of all employees on the ward. By the nature of their jobs, nurses and nursing assistants have the most frequent contact with patients and therefore were primarily charged with executing the interventions. However, therapists and therapy aides were also expected to follow the guidelines during their interactions with patients. Given that the interventions were designed and implemented with nursing workflows in mind, they were readily incorporated into established clinical duties with minimal impact. There were no major changes in nursing organization or staffing. The nurse-to-patient ratio did not change during the entire study period.
Initial education and training targeted registered nurses and took place during periodic education days. Separate sessions were conducted for therapists and nursing assistants. Education was performed by teams composed of both physicians and nurses to stress the importance of partnership in creating the desired culture of safety. The primary data that drove the proposed interventions were shared to emphasize that the program was data-driven and evidence-based. Continuing education took place among all groups during which fall events were reviewed, audited for compliance with the interventions, and used as instructional case studies. A focus was placed on process improvement and not individual fault. Educational responsibilities were shared by the nurse manager and participating surgeons.
Population
All patients admitted to the ward were included in the study. The ward is principally an orthopaedic unit, but patients from other specialties are occasionally admitted depending on hospital flow needs. As the interventions were systems-based, the study population was not limited to orthopaedic patients. All patients admitted during the study period were eighteen years of age or older.
Definition of a Fall
The definition of an in-hospital fall has been established by the Agency for Healthcare Research and Quality, the National Quality Forum, and the Joint Commission32,33. Consistent with these organizations as well as the current body of literature, a fall was defined as a sudden, unexpected descent from a standing, sitting, or horizontal position and included patients slipping from a chair, assisted falls, patients found down, and patient self-reported falls2,4,32,33. Falls that occurred during therapy were included. Injury was defined as any acute change identified by means of physical examination or imaging study32,33. The definition of a fall and the documentation of falls remained constant from the beginning of the preintervention study period through the end of the prospective postintervention study period.
Data Tracking
During the prospective portion of the study, falls were identified by means of the hospital Patient Safety Net and verbal report by staff to management. Falls that occurred in the prospective postintervention period were reviewed in real time. Fall circumstances that included identification of injury and use of the four interventions were examined by interviewing involved staff, reading chart notes and Patient Safety Net reports, and, when necessary, speaking with patients. The compliance for the use of the four interventions among patients who fell was 100%.
All patients admitted to the orthopaedic ward from a period beginning May 1, 2007, and continuing until the implementation of interventions on September 29, 2008, were used as the historical control cohort. Falls that occurred on the ward were identified by means of the hospital Patient Safety Net. Injurious falls were determined by careful review of the designated harm score and narrative included in the Patient Safety Net event reports as well as the electronic medical records. Demographic data were pulled from the aforementioned sources as well as the central hospital discharge database.
Statistics
To achieve a power of 0.8 and to detect a reduction in falls of 20%, with use of a p value of <0.05, it was determined that two years of patient data would be required. A straightforward proportion analysis of the number of falls was performed with use of a chi-square test; significance was defined as p < 0.05. In addition, data were further analyzed with use of a Poisson multivariate regression model that controlled for age, length of stay, and sex. A relative risk <1 represented a protective effect. In this analysis, results were considered significant when the 95% confidence intervals (CI) did not include unity.
Source of Funding
This study was approved by the institution’s internal review board human subjects division. No funding was provided for this study.
Demographic Characteristics
Six thousand and sixty-six patients were admitted to the ward during the three-year study period. Roughly three-quarters (4397) of those patients were admitted to an orthopaedic service; other services included general medicine (537 patients; 8.9%) hematology/oncology (467; 7.7%), surgery (124; 2.0%), and miscellaneous (541; 8.9%). Among the orthopaedic services, the hip and knee arthroplasty service had the greatest number of admissions, followed by the oncology, shoulder, and spine services. Admission to the sports, hand, and foot services was relatively rare because of the high percentage of outpatient procedures associated with these specialties.
Baseline demographic characteristics between the two study periods were similar, with all metrics, aside from sex, statistically comparable across study periods (Table II). While a female predominance was appreciated throughout the entire study, the proportion of female patient days was significantly greater in the postintervention period. Interestingly, no significant difference was identified in the two variables contributing to patient days: number of patients (p = 0.47) and length of hospital stay (p = 0.87). Thus, the relatively greater predominance of female patient days in the postintervention period may be spurious. Regardless, results were analyzed controlling for sex in a multivariate analysis. The study intervention effect was still realized and remained significant.
Frequency and Rates of Falls
The preintervention period represented 11,082 patient days and 2811 admissions. Forty-seven total falls, of which thirteen (28%) were falls with injury, occurred during this period. The fall with injury rate and the total fall rate were 1.17 and 4.24, respectively, per 1000 patient days.
The postintervention period represented 12,267 patient days and 3255 admissions. Thirty-one total falls, of which five (16%) were falls with injury, occurred during this period. The fall with injury rate and the total fall rate were 0.41 and 2.53, respectively, per 1000 patient days. Injuries resulting from falls during the preintervention and postintervention study periods are listed in Table III.
As analyzed with use of a straightforward chi-square test, the reductions in the rates of falls with injury (p = 0.036) and of total falls (p = 0.024) were significant. These data represent a 40% reduction in the total falls rate and a 65% reduction in the falls with injury rate from the preintervention to postintervention period. The intervention effect was verified with use of a Poisson multivariate analysis that controlled for age, length of stay, and sex. When exposed to the intervention, patients’ risk of both fall with injury (relative risk, 0.32; 95% CI, 0.096 to 0.892) and total fall (with or without injury) (relative risk, 0.59; 95% CI, 0.356 to 0.957) decreased. A relative risk <1 represents a protective effect. Thus, these data indicated that exposure to the intervention decreases the likelihood of a fall with injury by 68% and a total fall by 41%.
Circumstances of In-Hospital Falls
During the preintervention period, forty-seven falls occurred involving forty-two patients, and during the postintervention period, thirty-one falls occurred involving thirty patients. Toileting was the most common activity associated with falls in both study periods. Despite implementation of interventions in all cases and several interventions targeting toileting, 61% (nineteen of thirty-one) of falls during the postintervention study period were still related to this activity. With one exception during the preintervention period, all falls occurred in patient rooms. Only two falls in each study period occurred during physical or occupational therapy. During the postintervention period, 26% (eight of thirty-one) of falls occurred while patients were being assisted by staff and none of those falls resulted in injury. During the preintervention period, only 13% (six of forty-seven) of falls occurred while patients were being assisted by staff with one fall still resulting in injury. This difference was not significant with the numbers available (p = 0.253). Two patients who underwent shoulder surgery fell during the postintervention study period, only one of whom fell on postoperative day zero. During the preintervention study period, five shoulder surgery patients fell (two on postoperative day zero). This reduction in falls among shoulder surgery patients was not significant (p = 0.372). Aside from the study interventions, there were no systems-based changes during the study period. Individualized patient care continued during both study periods on the basis of patient condition and need.
This study demonstrates the efficacy of systems-based interventions that were developed with use of a continuous quality improvement model31 in decreasing the patient risk of falling in an inpatient orthopaedic surgery unit. To our knowledge, this is the first fall-prevention program to demonstrate a beneficial effect in the acute inpatient setting; this approach reduced the rate of total falls by 40% and the rate of falls with injury by 65%. Prior intervention studies targeting high-risk patients have been largely unsuccessful23-28. Other trials of multifactorial interventions have only shown improvement in patients with long lengths of hospital stays3,29,30 and thus may not apply to the acute inpatient orthopaedic setting.
The interventions utilized in this study most likely improved patient safety by means of both fall prevention and harm reduction. We suspect that the reduction in falls with injury was a result of both the decrease in total falls and the increase in the percentage of falls occurring while patients were being assisted by staff, although the latter was not significant with the numbers available.
Consistent with previous studies2,8-10,12, toileting was commonly associated with falls in our study during both the preintervention study period and after the interventions were put into use. With few exceptions, the involved patients had been advised by nursing staff to call for assistance prior to using the restroom. In some instances, patients failed to ask for assistance; in others, they opted to walk before assistance arrived. During the study period, these events occurred, albeit less frequently than before, despite interventions of timed toileting, wake ’em, take ’em, and assist in, assist out. As such, the complete prevention of toileting-related falls may not be possible. However, the systems-based interventions addressing toileting were effective at reducing the frequency of falls and the number of falls with injury. Because toileting was identified as a high-risk activity for inpatients in prior studies8-10, the interventions in the present study may apply well to other acute-care wards.
Many fall-prevention studies have utilized interventions that target patients who were at high risk of falling1,3,23-25,27,34-37. These high-risk patients were typically identified by means of standardized assessment on admission. The interventions targeting this group of patients that have not been successful in the acute setting include low beds, bed alarms, bedside sitters, vitamin D administration, medication review, safety footwear, patient education, additional physical therapy, and fall alert bracelets1,3,24,25,27,35-38.
We postulate that these prior studies did not observe an effect because the interventions targeted high-risk patients. Individual fall risk can vary greatly over the course of a hospitalization. Postoperative patients are particularly dynamic because of factors such as anesthesia, blood loss, narcotic pain medication, and surgically altered biomechanics. Medical patients face similar challenges with infections, electrolyte imbalances, and medication changes. The identification of high-risk patients may be of limited utility as it relies on a static assessment of a changing condition. Individual patient fall risk in the acute hospital setting is potentially too dynamic a target to serve as the basis for fall prevention.
It is important to distinguish between a high-risk patient and a high-risk situation. All patients, independent of individual fall risk, face high-risk situations during their hospital stay. Even in the acute setting, these situations are constant and therefore amenable to intervention. Identification and subsequent modification of these problematic scenarios are essential components of fall prevention in the acute setting. Study interventions targeted the high-risk situations related to toileting and one group of patients (shoulder surgery) at high risk. The data in both time periods demonstrated that most falls were associated with one high-risk situation: toileting. With the numbers available, there was no intervention effect observed among shoulder surgery patients. As a result, it appears that high-risk situations, rather than high-risk patients, should be targeted through systems-based interventions.
This study had several limitations, and it should be interpreted in light of these. Patient allocation was not randomized. Because the interventions in this study were systems-based, it would have been unrealistic to expect staff to apply interventions to some but not all patients. Additionally, as there is only one orthopaedic ward at the university hospital at which the study was conducted, an attempt at cluster randomization would have required multiple sites and different patient populations. Reporting bias was a potential weakness, but the nursing staff was not aware that the nursing unit’s data would be analyzed in a study. Study investigators were not blinded; however, falls were tracked centrally by the hospital through the Patient Safety Net, and reporting was mandatory of all staff. Finally, this study used a historical control cohort, but these data were consecutive without any overlap in time between the two study periods.
Reducing the number of patients who fall will improve the safety of inpatients and reduce the overall cost of care. The cost of inpatient falls, which the Centers for Medicare & Medicaid Services has deemed as “never events,” is now fully borne by hospitals, as the Centers for Medicare & Medicaid Services and many private insurers no longer provide reimbursement for injuries incurred during these falls. This study demonstrates that the utilization of a continuous quality improvement model to develop a systems-based fall-prevention program can be effective in reducing falls with injury and total falls in an acute inpatient setting. The authors question the concept of falls being “never events.” Despite a thoughtful, multidisciplinary, intensive approach to the problem, falls did occur. We believe it is unrealistic to consider all falls to be preventable; however, injuries can be significantly reduced through utilization of systems-based interventions that target high-risk situations.