Secondary Logo

Share this article on:

The Johns Hopkins Fall Risk Assessment Tool: A Study of Reliability and Validity

Poe, Stephanie, S., DNP, RN-BC; Dawson, Patricia, B., MSN, RN; Cvach, Maria, DNP, RN, FAAN; Burnett, Margaret, MS, RN, CNRN; Kumble, Sowmya, PT, NCS; Lewis, Maureen, MSN, RN, PMHNP-BC, PMHCNS-BC; Thompson, Carol, B., MS, MBA; Hill, Elizabeth, E., PhD, RN

doi: 10.1097/NCQ.0000000000000301
Articles

Patient falls and fall-related injury remain a safety concern. The Johns Hopkins Fall Risk Assessment Tool (JHFRAT) was developed to facilitate early detection of risk for anticipated physiologic falls in adult inpatients. Psychometric properties in acute care settings have not yet been fully established; this study sought to fill that gap. Results indicate that the JHFRAT is reliable, with high sensitivity and negative predictive validity. Specificity and positive predictive validity were lower than expected.

The Johns Hopkins Hospital, Baltimore, Maryland (Dr Poe and Mss Dawson, Burnett, Kumble, and Lewis); Johns Hopkins Health System, Baltimore, Maryland (Drs Poe and Cvach); The Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland (Ms Thompson); School of Nursing, The Johns Hopkins University, Baltimore, Maryland (Dr Hill); and College of Nursing, The Pennsylvania State University, University Park (Dr Hill).

Correspondence: Stephanie S. Poe, DNP, RN-BC, The Johns Hopkins Hospital, Billings Administration Bldg Room 229B, 600 N. Wolfe St, Baltimore, MD 21287 (spoe@jhmi.edu).

The authors acknowledge the contributions of Tameria Joy, BSN, RN, and Kelly Jarrett, RN, for data collection and Fall Champion services during this study.

This study was funded by the Johns Hopkins University School of Nursing Dorothy Evans Lyne (DEL) grant.

The authors declare no conflicts of interest.

Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's Web site (www.jncqjournal.com).

Accepted for publication: August 28, 2017

Published ahead of print: September 29, 2017

PATIENT FALLS remain a long-standing clinical challenge; fall-related injury is a vital patient safety concern. The Johns Hopkins Fall Risk Assessment Tool (JHFRAT) was developed to assess risk of an unanticipated physiological inpatient fall and enable early fall risk detection so that timely preventive actions could protect at-risk adults from harm. Initial tool development, testing, and revision have been published elsewhere.1,2 Although high levels of tool acceptability have been achieved in more than 100 hospitals, psychometric properties have not been fully established for use in acute care. The purpose of this article is to report results of further psychometric study.

Back to Top | Article Outline

LITERATURE REVIEW

Fall data are readily available through The Centers for Disease Control and Prevention, which publically reports mortality and morbidity rates related to unintentional falls that result in emergency department visits. Unintentional falls are the leading cause of nonfatal injuries in the United States, resulting in more than 978 000 hospitalizations and more than 8 million emergency department visits in 2013.3 Age-adjusted mortality from unintentional falls rose from 4.83 falls per 100 000 people in 2000 to 8.44 falls per 100 000 people in 2013.3 The national incidence of fall-related hospitalization in adults 65 years and older grew by 50% from 2001 through 2009.4

Current data on acute care fall rates are scant. The Agency for Healthcare Research and Quality estimates that between 700 000 and 1 000 000 hospitalized US patients fall each year.5 Fall rates vary across setting and unit type. Researchers examining unit-level data, submitted to the National Database of Nursing Quality Indicators from 2004 through 2009, reported rates ranging from 1.3 falls per 1000 patient-days (critical care units) to 7 falls per 1000 patient-days (rehabilitation units).6 A Massachusetts study of National Quality Forum nurse-sensitive hospital indicators yielded fall rates across various clinical settings, with mean falls/falls with injury per 1000 patient-days of 1.27/0.23 (critical care), 2.45/0.64 (step-down), 4.08/0.84 (medical), 2.25/0.39 (surgical), and 3.67/0.94 (medical-surgical).7

Fall-related harms are well known. Physical injuries span mild lacerations to hip fracture and subdural hematoma; psychological harms include anxiety and loss of self-confidence in mobility or activities of daily living. Inpatient fallers have been found to have longer lengths of stay, even without fall-related injury.8 Fallers stay significantly longer (6.3 days) than nonfallers.9 Inpatient fallers also have 4.8 and 4.5 times greater risk of hospital readmission within 90 and 180 days of discharge, respectively.10 Falls without serious injury increase risk of skilled nursing facility placement by 3-fold (controlling for cognitive, social, psychological, functional, and medical factors); serious fall injury increases risk 10-fold.11

Scientific data on true economic impact of inhospital falls are sparse. Incremental costs have been reported: noninjurious patient falls, $1139 to 2033; mild-to-moderate injury, $7136 to $15 444; and serious injury, $17 567 to $30 931.12 More recent reports estimate fall-related cost burden for hospitals as $17 483 per event,13 operational costs of inpatient fallers significantly higher (>$13 000 more) than costs for controls,9 and hospital costs related to serious fall-related injuries $36 897 higher than costs of non-fallers.14 Researchers report that 34% of 530 inpatient fallers incur total postfall imaging costs of $160 897, with $100 700 from head computed tomography.15 This does not include the hidden costs of litigation.

With such striking human and fiscal impact, prevention of falls and fall-related injuries is critical. Risk assessment is an essential first step per national evidence-based acute care fall prevention guidelines.16 Tools to assess risk are often used to predict anticipated physiologic falls so that preventive actions can be taken. Use of such tools allows nurses to target resource-intense fall-reduction tactics efficiently and effectively. To inspire confidence in selecting the “right” strategy for the “right” patient, risk assessment tools should be practicable and reliable, with adequate sensitivity and specificity.

The primary study objective was to test psychometrics of the JHFRAT when used in adults admitted to 1 of 4 acute care unit types. Tool creation has been described elsewhere.1 Initial testing resulted in tool revision, and the 2007 version was used for this study.2 Four hypotheses were tested: (H1) When used in adult inpatients, the JHFRAT will demonstrate acceptable interrater agreement (percent agreement for paired measurements and intraclass correlation ≥0.7 for single and paired measurements); (H2) When used in medicine, surgery, psychiatry, and medical-surgical care units, the JHFRAT will demonstrate acceptable interrater agreement; (H3) the JHFRAT will demonstrate acceptable construct validity (Spearman rank-order correlation coefficient ≥0.7) when compared with the Morse Fall Scale (MFS) in inpatient adult medicine patients; and (H4) the JHFRAT will demonstrate acceptable (70%) sensitivity, specificity, and predictive validity in accurately predicting a fall in adult inpatients.

Back to Top | Article Outline

METHODS

This research, approved by the local institutional review board with a waiver of informed consent, took place in a large mid-Atlantic academic medical center. Psychometrics studied include reliability (reproducibility of results), construct validity (tool measures what it purports to measure–-fall risk), and predictive validity (extent to which risk score predicts a fall).

Back to Top | Article Outline

Sample and setting

Reliability study

Study units included a 24-bed medical unit, a 28-bed surgical unit, a 22-bed psychiatry unit, and a 22-bed neuroscience medical-surgical unit. Units were selected on the basis of diversity of patient population. Mean fall (falls per 1000 patient-days) and fall injury (injury falls per 1000 patient-days) rates for study units for fiscal year 2008 were, respectively, medical (1.28, 0.12), surgical (1.72, 0.34), psychiatry (8.32, 1.66), and neuroscience medical-surgical (6.05, 1.48). Study unit nurse managers agreed to participate, and 4 staff nurses volunteered to serve as unit-based fall champions and study team members. Their experiences as study team members are described elsewhere.17 Other unit-based nurses completed assessments as part of standard care.

A recruitment goal of 2500 patients was set to achieve the desired accrual of 400 from each study unit. A 20% loss in subjects was assumed on the basis of past institutional studies. The intraclass correlation coefficient (ICC) was used to assess the JHFRAT rating reliability, comparing the variability of different fall ratings of the same patient with the total variation across all fall ratings and all patients. On the basis of a data set with results from nearly 20 000 assessments collected electronically over 2 months in fall 2007, a sample size of 400 patients per unit would allow estimation of an ICC (95% confidence interval [CI]) of 0.7 (0.66-0.74).

Back to Top | Article Outline

Construct validity study

A single study unit was selected to collect construct validity information, which required administration of a second instrument (MFS), to minimize burden on participating staff. The sample of patients from the medical unit was used in the construct validity study.

Back to Top | Article Outline

Predictive validity study

The original study's sample yielded insufficient numbers of falls to adequately determine sensitivity, specificity, and predictive validity. To measure predictive validity using a larger data set, the institutional review board approved an add-on retrospective substudy of all patients admitted to 18 units, including the original study units and similar units within the same departments.

Back to Top | Article Outline

Study instrument

Reliability study

The JHFRAT, a fall risk screening tool with established feasibility and content validity,1,2 is used across all adult acute care units in the study hospital. The tool is completed daily and when patient condition changes. Numeric scores are calculated as low (<6), moderate (6-13), or high fall risk (>13). The JHFRAT has several autorating criteria. If a patient meets any of the autorating criteria, tool completion is not required and the patient is automatically scored as high or low fall risk on the basis of the autorating selection. Tool completion takes an average of 5 minutes.

Back to Top | Article Outline

Construct validity study

The MFS, an established acute care fall risk evaluation scale, was employed in a comparative analysis to test JHFRAT construct validity. The MFS18 measures 6 domains: fall history, secondary diagnoses, ambulatory aid use, intravenous, gait, and mental status. The tool has well-documented psychometrics; primary developers report an interrater reliability correlation of r = 0.96, a sensitivity (true positive fall risk) of 78%, and a specificity (true negative fall risk) of 83%.18 A more recent comparative study of Korean-translated versions of the MFS, the JHFRAT, and a third fall risk scale found that the MFS had a sensitivity of 78.9%, a specificity of 55.8%, a positive predictive value (PPV) (high-risk patients who actually fall) of 30.8%; and a negative predictive value (NPV) (low-risk patients who never fall) of 91.4%.19 The same study reported sensitivity (62.0%), specificity (69.5%), PPV (33.6%), and NPV (86.0%) for the JHFRAT.19

Back to Top | Article Outline

Data collection

Reliability study

Data collection occurred from July 2009 to March 2011. To measure internal consistency and the degree to which 2 raters independently assigned the same values for the same fall attribute measured (ie, interrater agreement), independent paired JHFRAT nurse ratings were completed. Paired nurse ratings began on study day 1. The off-going night nurse rated each assigned patient's fall risk near the end of shift (6:00 AM to 7:30 AM). The nurse champion, who ensured each newly admitted patient had 2 independent paired ratings and served as a consistent rater across patients, independently rated patients' fall risk at the start of shift (8:00 AM to 9:30 AM). At the second shift change, the nurse champion and the incoming nurse rated each patient's fall risk. This paired rating occurred at change of shift, regardless of shift length (8 or 12 hours).

Patients' assigned nurse ratings were documented in the medical record. Fall champions recorded ratings on a separate paper tool that did not become part of the permanent record. To minimize confounding effects of patient status changes on fall risk scores, all paired nurse ratings occurred within a 2.5-hour time frame, allowing staff nurses with multipatient assignments to complete assessments as part of their usual care standard.

On subsequent days, each new patient admitted to the unit had 2 paired nurse fall risk assessments completed as earlier. All study patients had 4 fall ratings that included 2 sets of independent paired nurse ratings. In addition, demographic data were collected by the study coordinator and included age, race, sex, ethnicity, admitting diagnosis, and clinical service.

Back to Top | Article Outline

Construct validity study

The construct validity substudy was performed on the medical unit; the nurse champion volunteered to free up time to conduct the additional MFS assessment along with her morning JHFRAT. Administration of the first instrument may influence the rating on the second instrument; hence, the order in which these instruments were applied was randomized.

Back to Top | Article Outline

Predictive validity study

Electronic medical record data for all patients admitted to the departments of psychiatry, medicine, surgery, and neurosciences during the time frame from November 1, 2010, to January 1, 2012, were extracted and analyzed. Data extraction included all JHFRAT scores for the study sample, starting with the date of admission and ending with the date of the fall.

Back to Top | Article Outline

Data analysis

Reliability study

The ICC was used to measure interrater agreement across the quality-assured ratings for total sample and for each study unit (H1 and H2). This coefficient is used as a measure of reliability of actual ratings.20 The closer the ICC is to 1, the higher the relative consistency or reproducibility of quantitative measures of fall risk (JHFRAT ratings) made by different raters measuring the same quantity (fall risk). Variations in medical record documentation created some single ratings; however, the mixed-effects models used to assess the ICC appropriately handled both single and paired ratings to measure what proportion of all rating variability was due to patient differences compared with within-patient rater differences.

Back to Top | Article Outline

Construct validity study

The Spearman rank-order coefficient was used to determine construct validity of the JHFRAT as compared with the MFS when used in an inpatient adult medical population. The time frame between JHFRAT and MFS evaluations was 1 hour or less apart. Cut points used for the MFS were 0 to 24 (low), 25 to 54 (moderate), and 55 and over (high). Cut points were provided by the developer as part of training materials.

Back to Top | Article Outline

Predictive validity study

JHFRAT ratings are performed daily or more often as patient condition changes. The predictive validity assessment was conducted on a 4:1 frequency matching of controls to cases, where controls used for a match were randomly selected from controls having the same age category, gender, and unit type as the associated case. A control also had to have had at least 2 JHFRAT assessments in that unit no more than 28 hours apart, with the second assessment being the one compared to that of the case's JHFRAT assessment. This type of matching was selected to ensure that the 2 groups were comparable in terms of demographics and assessment timing.

Primary outcome variables to study predictive validity of screening tools are sensitivity, specificity, PPV, and NPV.21 Sensitivity measures the tool's ability to correctly identify patients with moderate or high fall risk. Sensitivity was estimated by examining JHFRAT scores within the effect period of fallers and calculating what proportion were scored as high or moderate fall risk. Specificity measures the tool's ability to correctly identify patients who are not likely to fall. To determine specificity, the proportion of persons scoring low risk (<6) was assessed among those who did not fall. PPV is the probability that the patient will actually fall if he/she screens positive (moderate or high fall risk) on the tool. NPV is the probability that the patient will not experience a fall if a he/she screens negative (low fall risk) on the tool.

Back to Top | Article Outline

RESULTS

Sample characteristics

Reliability study

All inpatient admissions to the study units during the July 2009 through March 2011 data collection period were included. A total of 2082 patients across 4 unit types were rated, where paired ratings spanned less than or equal to 2.5 hours. By unit type, enrolled patients were medical (n = 634), surgical (n = 607), medical-surgical (n = 457), and psychiatry (n = 384). Psychiatry patients tended to be younger (84% younger than 60 years), with the most patients across all units in the 41- to 59-year age range. Gender was evenly split between male and female. More than 50% of study patients were white; the surgical and medical-surgical units had the most white study patients (60% and 68%, respectively). Sample demographics (total and by study unit) are included in Supplemental Digital Content, Table 1, available at: http://links.lww.com/JNCQ/A380.

Back to Top | Article Outline

Construct validity study

MFS assessments were completed by the fall champion within 1 hour of JHFRAT scoring for a total of 352 patients. Sample demographics are noted in Supplemental Digital Content, Table 2, available at: http://links.lww.com/JNCQ/A381. Most study patients were younger than 60 years (62%), female (54%), and nonwhite (60%). This profile closely mirrored that of the medical unit reliability sample.

Back to Top | Article Outline

Predictive validity study

Historical data were used from a total of 1299 patients admitted to at least 1 of 18 units within medical, surgical, medical-surgical, or psychiatry departments. Sample demographics (total and by department type) are included in Supplemental Digital Content, Table 3, available: at http://links.lww.com/JNCQ/A382.

Back to Top | Article Outline

Findings

Reliability study

Table 1 reports percent agreement and ICC for each scale item across unit types (H1) and within unit types (H2). Staff selection of auto-rated items had 100% agreement. Paired ratings were completed on 1427 of 1615 patients at time 1 (88%) and 1025 of 1275 patients at time 2 (80%), respectively. The Overall ICC was 0.78 (95% CI, 0.76-0.80).

Table 1

Table 1

Back to Top | Article Outline

Construct validity study

The Spearman correlation coefficient was used to determine the relationship between the JHFRAT and the MFS (H3), r s = 0.66 (95% CI, 0.58-0.74). Percent agreement between the tools was 58% (low), 67% (moderate), 41% (high), and 61% (across all risk levels). Table 2 shows the score distributions in the study population (inpatient adult medical unit).

Table 2

Table 2

Back to Top | Article Outline

Predictive validity study

Table 3 reports JHFRAT sensitivity, specificity, and predictive validity (H4) used to analyze fallers (n = 261) and nonfallers (n = 1038). Among fallers, 228 were rated as moderate or high risk (sensitivity = 87.4%), and 33 patients were rated as low risk. Among nonfallers, 746 were rated as moderate or high risk, and 292 were rated as low risk (specificity = 28.1%). Among patients rated as moderate or high risk (n = 974), 228 fell and 746 did not fall (PPV = 23.4%). Among patients rated as low risk (n = 325), 33 fell and 292 did not fall (NPV = 89.8%). The odds of being a faller with a moderate or high fall risk score was 2.7 times the odds of falling with a low score (95% CI, 1.84-3.98). Area under the receiver operating characteristic curve (AUC) is 0.58 (95% CI, 0.55-0.60).

Table 3

Table 3

Back to Top | Article Outline

DISCUSSION

Reliability study

Results reveal acceptable interrater agreement for the total JHFRAT score among all units, regardless of care unit type. This finding was consistent with a published reliability analysis of a Chinese-translated version of the JHFRAT, which yielded 97% interrater agreement.21 Interrater agreement varied across units. Regardless of care unit type, percent agreement was highest for age, fall history, bowel/bladder, mobility, cognition, and JHFRAT risk categories.

Dual-rater agreement on medication and care equipment items was less than satisfactory. Changes (eg, discontinuation) in medication orders or use of tethered equipment (eg, catheters, lines, tubes, and drains) may account for differences. Variation in medication scale items could also be ascribed to how as-needed drugs are scored. Despite guidelines to score all active as-needed medication orders, even if the drug was never administered, some staff nurses only scored this item when the patient was actively taking as-needed medications. Care equipment score variations could be due to intermittent use of prescribed equipment (eg, intermittent infusions or timed on-off device application where the patient may not have been tethered at time of scoring).

The psychiatry unit had the highest levels of agreement across all JHFRAT items. The psychiatric population, with its focus on psychologic rather than physiologic risk, is more homogeneous than other study groups. High-risk drug therapies (eg, psychotropics) may be slightly more variable and may explain, in part, the medication subscale having the lowest level of agreement (89.3%).

The medical unit had the lowest levels of agreement in medication and care equipment scoring. Patient condition and treatment plans are highly variable in this setting. Of interest were the high agreement levels on cognition scale scores in this population. An education campaign on correct interpretation of this scale item may have contributed to these results.

Back to Top | Article Outline

Construct validity study

Results comparing the JHFRAT and the MFS found a moderate correlation of 0.66 (95% CI, 0.58-0.74), which is close to the expected correlation. Both tools assess fall risk but measure different constructs. The JHFRAT and the MFS both measure cognition and mental status. The MFS measures ambulatory aid, gait, and IV access; the JHFRAT assesses mobility and equipment to measure risk but broadens the equipment attribute to include all types of devices that “tether” a patient. Both tools measure fall history; the JHFRAT limits fall history to 6 months preadmission. Larger differences are inclusion of risk attributes of age, elimination, and drugs (JHFRAT), and secondary diagnoses (MFS). MFS scores are weighted differently than are JHFRAT scores. The JHFRAT uses set cut points for low, moderate, and high fall risk; the MFS suggests cut points.

Back to Top | Article Outline

Predictive validity study

Sensitivity and specificity address the percentage of patients correctly classified as people who will or will not fall. In clinical practice, sensitivity and specificity are inversely related.22 The risk of a person falling usually lies somewhere along a continuum, with some overlap. Highly sensitive tests are preferred when a false-negative result has serious outcomes. Clinicians want highly sensitive risk screening tests to avert harm. If a highly sensitive test is negative, there is greater assurance that risk is not present. Very specific tests are desired to confirm presence of risk. If highly specific test results are positive, risk is highly likely to be present.

Results revealed high sensitivity (87.4%), indicating the tool's ability to confidently screen patients with moderate to high risk. Falls can have serious consequences; avoiding false negatives increases the likelihood of identifying at-risk patients. The 28.1% specificity, much lower than hypothesized, may have been influenced by the choice of cut points. A low cut point for moderate/high risk (>6) was chosen to ensure detection of adults with risk factors amenable to prevention. This decision may have led to a highly sensitive test at the expense of specificity.

Predictive validity measures the odds that a patient with a low or moderate/high score will fall. The high NPV (89.8%) reflects how likely that a patient rated at low fall risk will not fall. Hence, nurses can be confident that basic fall preventive actions will be sufficient to protect these patients from harm. The low PPV (23.4%), reflecting the likelihood that a patient rated as moderate/high risk will fall, was lower than expected and was likely affected by preventive interventions to mitigate risk of falling (eg, bed/chair alarms).

AUC measures how good a test is at discriminating between low and moderate/high risk; the AUC of 0.58 (95% CI, 0.55-0.60) suggests that JHFRAT accuracy is sufficient to warrant preventive action. PPV, NPV, and AUC findings are similar to results of 2 published predictive validity studies. The first study tested the JHFRAT in 13 574 patients on medical units in a single large US academic medical center (PPV = 3.7%, NPV = 81.4%, AUC = 0.69).23 The second study tested a Korean-translated version in 5 acute care hospitals (PPV = 33.6%, NPV = 86.0%, AUC = 0.71).19 Sensitivity and specificity results were not consistent among the studies. The US study revealed completely opposite findings (sensitivity = 26.5%, specificity = 89.6%)23 to current findings. Korean study results revealed higher specificity (69.5%) than sensitivity (62.0%), inconsistent with current findings.19 None of the studies accounted for fall prevention actions implemented, so it is difficult to speculate as to what might account for these differences.

Back to Top | Article Outline

Limitations

A number of limitations should be noted. Limitations include: (1) single-center study design limits generalizability; (2) ethical concerns did not allow for control of preventive interventions in at-risk subjects; (3) failure to differentiate types of falls (eg, accidental, anticipated physiologic, or unanticipated physiologic) limits findings; (4) dissimilar patient knowledge between paired nurse raters, only one of whom had direct patient care responsibility, may have affected reliability analysis; (5) handwritten and electronic data collection was subject to variations in clinical documentation; (6) study did not control for preventive actions, potentially affecting predictive validity; and (7) study completion was significantly delayed because of difficulties obtaining electronic data during a systemwide change in electronic health records.

Back to Top | Article Outline

CONCLUSIONS

Effective psychometric evaluation of a fall risk screening tool is highly complex. Early studies of the JHFRAT concluded that the test appears to reflect fall risk and that aspects of fall risk screened by the tool sufficiently cover the construct being measured (face and content validity). The present study examined reliability, construct validity, and predictive validity. The JHFRAT was found to be reliable as a fall risk screening measure across 4 diverse adult acute care unit types. Construct validity, comparing JHFRAT scores with the MFS, revealed a moderate correlation. Predictive validity testing revealed high sensitivity, consistent with a desire to ensure that the tool effectively screens patients for fall risk; however, low specificity was found. The desire to avoid false negatives when assessing risk and the ethical obligation to implement preventive interventions may have served as confounders.

Back to Top | Article Outline

REFERENCES

1. Poe SS, Cvach MM, Gartrell DG, Radzik BR, Joy TL. An evidence-based approach to fall risk assessment, prevention, and management. Lessons learned. J Nurs Care Qual. 2005;20(2):107–116.
2. Poe SS, Cvach M, Dawson PB, Straus H, Hill E. The Johns Hopkins Fall Risk Assessment Tool: post-implementation evaluation. J Nurs Care Qual. 2007;22(4):293–298.
3. Centers for Disease Control and Prevention. National Center for Injury Prevention and Control Data and Statistics (WISQARS) http://http://www.cdc.gov/injury/wisqars/index.html. Published 2013. Accessed April 30, 2015.
4. Hartholt KA, Stevens JA, Polinder S, van der Cammen TJ, Patka P. Increase in fall-related hospitalizations in the United States, 2001-2008. J Trauma. 2011;1(1):255–258.
5. Agency of Health Care Research and Quality. Preventing Falls in Hospitals: A Toolkit for Improving Quality of Care. http://http://www.ahrq.gov/professionals/systems/hospital/fallpxtoolkit/index.html. Published January 2013. Accessed April 30, 2015.
6. He J, Dunton N, Staggs V. Unit-level time trends in inpatient fall rates of US hospitals. Med Care. 2012;50(9):801–807.
7. Smith DP, Jordan HS. Piloting nursing-sensitive hospital care measures in Massachusetts. J Nurs Care Qual. 2008;23(1):23–33.
8. Dunne TJ, Gaboury I, Ashe MC. Falls in hospital increase length of stay regardless of degree of harm. J Eval Clin Pract. 2014;20(4):396–400.
9. Wong CA, Recktenwald AJ, Jones ML, Waterman BM, Bollini ML, Dunagan WC. The cost of serious fall-related injuries in three Midwestern hospitals. Jt Comm J Qual Patient Saf. 2011;37(2):81–87.
10. Hong H-J, Kim N, Jin Y, Piao J, Lee S-M. Trigger factors and outcomes of falls among Korean hospitalized patients: analysis of electronic medical records. Clin Nurs Res. 2015;24(1):51–72.
11. Tinetti ME, Kumar C. The patient who falls: “it's always a trade-off” JAMA. 2010;303(3):258–266.
12. Spetz J, Brown D, Aydin C. The economics of preventing hospital falls: demonstrating ROI through a simple model. J Nurs Adm. 2015;45(1):50–57.
13. Trepanier S, Hilsenbeck J. A hospital system approach at decreasing falls with injuries and cost. Nurs Econ. 2014;32(3):135–141.
14. Zecevic AA, Chesworth BM, Zaric FS, et al Estimating the cost of serious injurious falls in a Canadian acute care hospital. Can J Aging. 2012;31(2):139–147.
15. Fields J, Alturkistani T, Kumar N, et al Prevalence and cost of imaging in inpatient falls: the rising cost of falling. Clinicoecon Outcomes Res. 2015;7:281–286.
16. Degelau J, Belz M, Bungum L, et al Prevention of Falls (Acute Care). Health Care Protocol. Bloomington, MN: Institute for Clinical Systems Improvement; 2012.
17. Burnett M, Lewis M, Joy T, Jarrett K. Participating in clinical nursing research: challenges and solutions of the bedside nurse champion. Medsurg Nurs. 2012;21(5):309–311.
18. Morse JM, Morse RM, Tylko SJ. Development of a scale to identify the fall-prone patient. Can J Aging. 1989;8:366–377.
19. Kim KS, Kim JA, Choi Y-K, et al A comparative study on the validity of fall risk assessment scales in Korean hospitals. Asian Nurs Res. 2011;5(1):1–37.
20. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86(2):420–428.
21. Zang J, Wang M, Liu Y., Glaser AN. Psychometric validation of the Chinese version of the Johns Hopkins Fall Risk Assessment Tool for older Chinese inpatients. J Clin Nurs. 2016;25(19/20):2846–2853.
22. Glaser AN. Answering clinical questions II: statistics in medical decision making. In: Glaser AN ed. High-yieldTM Biostatistics, Epidemiology And Public Health. 4th ed. Baltimore, MD: Lippincott Williams & Wilkins; 2014:68–85.
23. Klinkenberg WD, Potter P. Validity of the Johns Hopkins Fall Risk Assessment Tool for predicting falls on inpatient medicine Services. J Nurs Care Qual. 2017;32(2):108–113.
Keywords:

accidental fall; inpatients; Johns Hopkins Fall Risk Assessment Tool; psychometrics; risk assessment; validation studies

Supplemental Digital Content

Back to Top | Article Outline
Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved