PATIENT falls in hospitals, specifically in psychiatric units, are a growing concern for health care providers. The ambulatory nature of psychiatric inpatients contributes to the higher incidence of falls. Estimates show that anywhere from 3% to 20% of inpatients will fall during their hospitalization—fall rates among medical-surgical inpatients are approximately 1.3 to 8.9 falls per 1000 inpatient days compared with 13 to 25 falls per 1000 inpatient days for psychiatric inpatients.1–4 Given the potential for injury related to falls and the associated increased cost for hospitals when inpatients fall, a psychometrically sound high-risk falls assessment for psychiatric inpatients is needed. Currently, there are 2 instruments designed to assess fall risk for this population: the Edmonson Psychiatric Fall Risk Assessment Tool (EPFRAT)5 and WilsonSims Fall Risk Assessment Tool (WSFRAT).6,7
Given a lack of assessments designed for psychiatric inpatients, many hospitals rely on fall risk instruments designed for and tested on a medical-surgical population.5,6,8 For example, the Hendrich II Fall Risk Model (Hendrich II) and Morse Falls Scale have been used with psychiatric populations.
Prospective studies designed to develop high-risk falls assessments for use with psychiatric inpatients have not been conducted. The EPFRAT and WSFRAT were designed to assess fall risk for this population but were developed retrospectively using chart reviews. In 1 study designed to develop a high-risk falls assessment for psychiatric inpatients, the resulting assessment was based on data that included only 2 patient falls.6 Another study with a similar design applied the assessment to the charts of 50 patients that had fallen to determine its predictive capabilities.5
This analysis addresses a component of data originally collected during the development of the Baptist Health High Risk Falls Assessment (BHHRFA)—a high-risk falls assessment for medical-surgical patients.9 During the original development of the BHHRFA, data were collected on medical-surgical inpatients and psychiatric inpatients. Before submitting this article for publication, the psychiatric inpatients' fall data had not been analyzed. The purpose of this analysis was to evaluate the psychometric properties of the BHHRFA when used with an inpatient psychiatric population.
The BHHRFA was developed in 2 phases. In the first phase, 118 medical-surgical patients who fell were interviewed by the supervisors of each unit immediately after the fall in an attempt to better understand why they had fallen. In the second phase, an assessment on the basis of outcomes of phase 1 and the literature was tested for 6 months on a medical-surgical unit. Data were analyzed to ascertain which of the 9 factors were most likely to predict falls. The final high-risk falls assessment contains 6 of the 9 original factors plus the addition of 1 item retrieved from the literature related to nurses' judgment about which patients were likely to fall (Supplemental Digital Content, Figure 1 available at: http://links.lww.com/JNCQ/A286). This original study was reviewed and approved by the hospital's institutional review board.
Testing of the final draft of the BHHRFA was conducted at 6 hospitals. On the basis of expert clinical judgment plus preliminary data from phases 1 and 2, a total score of 13 or greater was used in identifying patients at high risk for falls. Analysis of the data (N = 241 599 assessments) with a cutoff score of 13 provided sensitivity (0.70) specificity (0.66), and diagnostic odds ratio (DOR) of 4.73.9
Existing falls risk assessments for psychiatric inpatients
A review of the literature regarding high-risk fall assessments designed for use with psychiatric inpatients yielded minimal results. Two fall risk assessments specific to psychiatric inpatients were found (EPFRAT and WSFRAT). During their initial development, the EPFRAT and WSFRAT were each tested against a commonly used medical-surgical high-risk fall assessment. The EPFRAT was tested against the Morse Fall Scale. The WSFRAT was tested against the Hendrich II.
The creators of the EPFRAT developed this instrument by reviewing the literature and identifying 6 risk categories: age, ambulatory status, medications, psychiatric illnesses, nutrition/fluid intake, and sleep. Retrospective chart reviews were conducted to examine the prevalence of factors included on the EPFRAT. Weights were applied to items followed by an assessment of nurse usability. They then compared the Morse Falls Scale5 and the EPFRAT. Both scales were retrospectively applied to patients' charts to determine whether they accurately predicted if the patient had fallen; a total of 138 patient records were examined. Sensitivity and specificity were calculated (Morse Falls Scale 0.49 and 0.85; EPFRAT 0.63 and 0.85). The authors concluded that the EPFRAT was a better predictor of falls when used with acutely ill psychiatric patients than the Morse Falls Scale.5
The WSFRAT was developed to identify psychiatric inpatients at high risk for falls. The tool was developed in multiple phases: (1) a list of risk factors were selected by retrospectively reviewing 2 years of psychiatric inpatients' charts, from a Magnet-designated hospital, (2) a review of the literature was conducted on the existing fall risk factors for psychiatric inpatients, (3) the WSFRAT was formulated by combining the risk factors identified in phases 1 and 2, and (4) the WSFRAT was implemented into practice to test the initial validity of the tool.7 Tests of sensitivity and specificity were not conducted during the development of the assessment; however, these psychometric properties along with a cutoff score were calculated by Dyke et al.6 The WSFRAT was tested against the Hendrich II, a widely used fall risk assessment tool for medical-surgical patients. The WSFRAT and Hendrich II were applied retrospectively to a convenience sample of 50 psychiatric inpatients. Of the 50 patients examined, 2 falls occurred—this sample was used to calculate the sensitivity and specificity of the WSFRAT—the sensitivity and specificity are 1.00 and 0.63.6
Existing medical-surgical assessments used in psychiatric populations
The Morse Fall Scale and Hendrich II have been used in psychiatric units. These 2 fall risk assessments were developed before the EPFRAT and WSFRAT.
Morse Fall Scale
The Morse Falls Scale was developed by comparing 100 inpatients who fell with 100 randomly selected inpatients who did not fall (N = 2689).10 The instrument was developed and tested in multiple units of the hospital, both inpatient and outpatient. It assesses 6 risk categories: history of falling, secondary diagnosis, ambulatory aids, use of intravenous therapy, gait, and mental status. Applying the Morse Falls Scale to a psychiatric population resulted in sensitivity and specificity of 0.49 and 0.85, respectively.5
Hendrich II Fall Risk model
The Hendrich II was developed by retrospectively analyzing 1232 inpatient fall risk assessments and determining which of the 600 identified variables were most meaningful in predicting falls.11 During data analysis, 8 risk factors were found to be most prevalent: confusion, disorientation/impulsivity, systematic depression, altered elimination, dizziness/vertigo, gender (men), prescribed/administered antiepileptic (anticonvulsant) agents, and prescribed/administered benzodiazepine. When used to assess fall risk with psychiatric inpatients, data were analyzed with only 2 falls. Reported sensitivity and specificity when tested in the psychiatric population were 1.00 and 0.68, respectively.6
Defining a fall
There is an ongoing challenge in health care to accurately define a fall—different care units may experience different kinds of falls.2,12,13 For example, psychiatric inpatient units experience “intentional falls,” in which a patient intentionally falls as a means of acting out in front of hospital staff.12 Similarly, these “intentional falls” have been reported in general medical-surgical units; these occurrences can lead to confusion as to what constitutes a fall. For the purpose of the original study, a fall was defined as an unplanned (to the hospital staff) descent to the floor or other lower surface with or without an injury to the patient.14,15
During the original development of the BHHRFA, data were collected on a psychiatric unit in a 519-bed Magnet redesignated hospital. Given major differences in a population of medical-surgical and psychiatric patients, these data (N = 5910 assessments) were not included in the original analysis of sensitivity and specificity. Diagnoses of patients on the 22-bed psychiatric unit included psychoses, depressive neuroses, degenerative nervous system disorders, alcohol/drug abuse dependence, and dementia.
Data related to the use of the BHHRFA for a psychiatric population were collected at the same time as the data on medical-surgical patients used to test the BHHRFA. Assessments were collected by research assistants designated by administration at each institution. For purposes of this article, data from psychiatric inpatients (N = 5910) were analyzed to determine the psychometric properties of the BHHRFA. Data were analyzed using SPSS version 21 (IBM, Armonk, New York).
To determine the clinical utility of the BHHRFA with psychiatric inpatients, a receiver-operating characteristic (ROC) was calculated to identify the appropriate cutoff score. In conjunction with the ROC curve, Youden's index was calculated for each possible cutoff value. Youden's index identifies the maximum y-value above the chance line, which subsequentially identifies the best cutoff point. A χ2 analysis was conducted using whether the patient fell as the dependent variable and total score on the BHHRFA as the independent variable. Sensitivity, specificity, and the (DOR) were calculated. A DOR is a single indicator of test performance—higher DOR indicates greater test performance.
When analyzing data for general hospital inpatients (in the original study), a score of 13 was originally determined to be the cutoff score that best predicted the likelihood of falling.9 Results of the ROC analysis and calculation of Youden's index for psychiatric patients identified a cutoff score of 16 (Supplemental Digital Content, Figure 2 available at: http://links.lww.com/JNCQ/A287). In addition, the area under the ROC curve is an indicator of the discriminatory ability of the screen. In this case, the area under the curve (AUC) = 0.745, standard error = 0.048, 95% confidence interval = 0.65 to 0.84. A perfect diagnostic test would have an AUC of 1.16 The results of the χ2 analysis were used to calculate sensitivity (true positive), specificity (true negative), and the DOR of the BHHRFA used with psychiatric patients. The results yielded a sensitivity of 0.68, a specificity of 0.70, and a DOR of 4.964 (Table).
The purpose of this study was to evaluate the psychometric properties of the BHHRFA when used with an inpatient psychiatric population. The construction of the BHHRFA required unique initiative and rigor. Because a prospective design and a process of initial pilot studies were used in the development of the BHHRFA, it has distinct advantages over instruments of this kind. Some of these advantages when used to assess psychiatric inpatients include a higher sensitivity score (0.68) than other psychiatric specific assessments, an acceptable specificity score (0.70), and a strong DOR (4.964). Furthermore, 5910 assessments were analyzed using the BHHRFA compared with 188 total assessments analyzed during the testing of the EPFRAT (n = 138) and WSFRAT (n = 50). With lower sample sizes during the testing of the EPFRAT and WSFRAT, clinicians must cautiously weigh the risk of using these instruments in practice. The results demonstrate that using the BHHRFA with a cutoff score of 16 yields good sensitivity and specificity—when compared with other instruments designed for use with psychiatric population, the BHHRFA more accurately predicts falls.4,5
The results of this analysis, taken in conjunction with the results of the original study, demonstrate the wide applicability of the BHHRFA in psychiatric inpatients and general hospital populations. The applicability of the BHHRFA to 2 different inpatient populations may be due to a number of factors. It is possible that because the investigators who designed the BHHRFA assessed a large number of patients, a more accurate portrayal of fall risk was captured compared with other studies. Furthermore, the literature related to falls among psychiatric and medical surgical inpatients stresses the impact of medications on fall risk.1–2,4–5,7–9 Unlike other medical-surgical fall risk assessments, the BHHRFA captures medication profiles that could be considered psychiatric specific (narcotics, sedatives, hypnotics). For example, a nurse may assess an antipsychotic drug as a sedative because of the nature of the drug's side effects. It is plausible that nurses consider medications on the basis of the type and severity of their side effects rather than the pharmacologic drug classification.
Medication administration and their side effects have implications for nurses and patients alike. The identification of certain medications as additional fall risk factors provides opportunities for nurses to engage with their patients. Nurses may be able to talk to inpatients and explain how some of their medications may increase their risk for falls. Moreover, this educational timeframe serves as a great segue for nurses to implement their respective institution's fall prevention interventions to maintain patient safety.
This assessment gives providers the tools to identify and implement fall protection interventions for their patients. However, this tool is designed to identify preventable falls. Psychiatric inpatients can present with behaviors that can be difficult to classify and risk factors that may not be present on the BHHRFA. Because of this, nurses must use and apply their best clinical judgment when assessing inpatients for their fall risk.17 Psychiatric inpatients may have tendencies to act out, or engage in attention-seeking behaviors, that may increase their risk for falls—unintentional or intentional. These tendencies and behaviors are characteristics that should incline nurses to score inpatients higher in regard to the nurses' clinical judgment on the BHHRFA.
Given the lack of high risk for falls assessments for psychiatric inpatients, the BHHRFA may provide a useful tool for predicting falls. It is clinically useful in that it takes nurses approximately 38 seconds to complete.9 Further testing of the BHHRFA is warranted in psychiatric inpatient populations.
Items related to the ambulatory nature of psychiatric units as compared with general hospital inpatient units might increase the sensitivity and specificity of the BHHRFA when used with this population. Additional testing of the BHHRFA with a psychiatric population could provide information that would strengthen the ability of the assessment to predict falls.
High-risk fall assessments with strong predictive ability are needed in order for the incidence of inpatient falls to decrease. As part of The Joint Commission's Goal 9B, hospitals are required to implement a fall reduction program that includes a reliable risk assessment.18 In addition, The Joint Commission expects an organization's fall reduction program to be appropriate for each setting and population.18
To date there are limited possibilities for assessing falls within a psychiatric population. The BHHRFA has good sensitivity and specificity when used with both medical-surgical (0.70/0.66) and psychiatric patients (0.68/0.70)—the BHHRFA is an assessment that is clinically useful, has better predictive ability than other available instruments, and can be used across inpatient populations.
1. Miake-Lye I, Hempel S, Ganz D, Shekelle P. Inpatient fall prevention programs as a patient safety
strategy. Ann Intern Med. 2013;158(5):390–396.
2. Quigley P. Reducing falls and fall-related injuries in medical-surgical units: one-year multihospital falls collaborative. J Nurs Care Qual. 2016;31(2):139–145.
3. Inouye S, Brown C, Tinetti M. Perspective: Medicare, nonpayment, hospital falls and unintended consequences. N Engl J Med. 2009;360(23):2390–2393.
4. Blair E, Gruman C. Falls in an inpatient geriatric psychiatric population. J Am Psychiatr Nurses Assoc. 2005;11(6):351–354.
5. Edmonson D, Robinson S, Hughes L. Development of the Edmonson Psychiatric Fall Risk Assessment Tool
. J Psychosoc Nurs Ment Health Serv. 2010;49(2):29–36.
6. Dyke D, Singley B, Speroni K, Daniel M. Evaluation of Fall Risk Assessment
Tools for psychiatric patient fall prevention. J Psychosoc Nurs Ment Health Serv. 2013;52(12):30–35.
7. Billeen M, Kruszewski A, Fettes S, Wilson S, Sims K. Content validity of Wilson-Sims Falls Risk Assessment Tool 2011. http://eo2.commpartners.com/users/apna_kc/downloads/2034-Wilson-Handout.pdf. Accessed July 26, 2016.
8. Rutledge D, Donaldson N, Pravikoff D. Update 2003: fall risk assessment
and prevention in hospitalized patients. Am J Nurs. 2003;6(5):1–55.
9. Corley D, Brockopp D, McCowan D, et al The Baptist Health High Risk Falls Assessment. J Nurs Adm. 2014;44(5):263–269.
10. Morse J, Morse R, Tylko S. Development of a scale to identify the fall prone patient. Can J Aging. 1989;8:366–377.
11. Hendrich A, Bender P, Nyhuis A. Validation of the Hendrich II Fall Risk Model: a large concurrent case/control study of hospitalized patients. Appl Nurs Res. 2003;16(1):9–21.
12. Staggs V, Davidson J, Dunton N, Crosser B. Challenges in defining and categorizing falls on diverse unit types: lessons from expansion of the NDNQI Falls Indicator. J Nurs Care Qual. 2015;30(2):106–112.
13. Williams T, Szekendi M, Thomas S. An analysis of patient falls and fall prevention programs across academic medical centers. J Nurs Care Qual. 2014; 29(1):19–29.
14. NDNQI. Patient falls. https://members.nursingquality.org/NDNQI/learning/tutorials/modules/module_patient_falls_2014/pt_falls_data_collect_01.aspx. Published 2014. Accessed October 16, 2015.
15. Simon M, Klaus S, Gajewski BJ, Dunton N. Agreement of fall classifications among staff in U.S. hospitals. Nurs Res. 2013;62(2):74–81.
16. Sackett D, Haynes B, Tugwell P. The interpretation of Diagnostic Data. In:Sackett D, Haynes B, Tugwell P, eds. Clinical Epidemiology a Basic Science for Clinical Medicine. 2nd ed. Boston, MA: Little, Brown & Company; 1985.
17. Tzeng H, Yin C. Frequently observed risk factors for fall-related injuries and effective preventive interventions: a multihospital survey of nurses' perceptions. J Nurs Care Qual. 2013;28(3):130–138.
18. The Joint Commission. Nursing Care Center National Patient Safety
/assets/1/6/2015_NPSG_NCC.pdf. Published 2015. Accessed October 8, 2015.