Assignment: Nursing Care Quality

Multilevel Factors Associated With Injurious Falls in Acute Care Hospitals

Yunchuan (Lucy) Zhao, PhD, MPAff, RN; Marjorie Bott, PhD, RN; Jianghua He, PhD; Heejung Kim, PhD, RN; Shin Hye Park, PhD, RN; Nancy Dunton, PhD, RN

Using National Database of Nursing Quality Indicators data from July 2013 to June 2014, this cor- relational study examined the associations of injurious falls among all patient falls with multilevel factors in hospitals. The sample included all falls recorded in adult medical, surgical, combined medical-surgical, and step-down units (N = 2299) in participating hospitals (N = 488). Hierar- chical negative binominal regression analyses were performed. Results revealed hospital and unit organizational factors associated with inpatient injurious falls. Key words: fall risk assessment, falls, hospitals, injurious falls, nurse staffing, organizational structure

INPATIENT FALLS and injurious falls inacute care settings are prevalent and a se- rious concern for patient care. In the United States, the overall prevalence of falls range from 3 to 5 falls per 1000 patient-days, with about 1 million inpatient falls annually.1

Among inpatient falls, the incidence rates for fall-related injuries range from 6.8% to 72.1%, with 0.7% to 30% for severe injuries such as fractures, cranial trauma, or death.2 Injurious

Author Affiliations: Boise State University School of Nursing, Boise, Idaho (Dr Zhao); University of Kansas School of Nursing (Drs Bott, Park, and Dunton) and Department of Biostatistics, University of Kansas School of Medicine (Dr He), Kansas City; and College of Nursing, Yonsei University, Seoul, South Korea (Dr Kim).

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).

Correspondence: Yunchuan (Lucy) Zhao, PhD, MPAff, RN, Boise State University School of Nursing, 1910 University Dr, Mail Stop 1840, Boise, ID 83725 (lucyzhao@boisestate.edu).

Accepted for publication: January 15, 2017

Published ahead of print: March 20, 2017

DOI: 10.1097/NCQ.0000000000000253

falls have a negative impact on patients, fami- lies, and the health care system. With injurious falls, patients may suffer from loss of indepen- dence, depression, and decreased quality of life.1 Injurious falls can result in a prolonged length of stay (LOS) that further leads to in- creased direct patient care costs and health care resource use.1 When comparing patients with severe injurious falls with patients with- out falls, the LOS increases, on average, by 6 to 12 days, with an additional cost of $13 316 for the patient.3 In addition, indirect costs asso- ciated with injurious falls may include loss of income, placement in a skilled nursing facility or nursing home, and litigation expenses.1,3

In the last few decades, interventions aimed at fall prevention have been developed and implemented in hospitals.4 Several national initiatives also have underscored the impor- tance of preventing falls and their associated adverse consequences. Prevention of falls and fall-related serious injuries and death is one major goal in the Healthy People 2020.5 Be- ginning in October 2008, the Centers for Medicare & Medicaid Services no longer reim- burses hospitals for treatment of preventable injuries, including fall-related injuries.6 De- spite the national efforts and initiatives on

Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

20

Injurious Falls in Acute Care Hospitals 21

fall prevention, inpatients falls and fall-related injuries still are prevalent in hospitals. The results of a recent study show a total of 315 817 falls (3.56 falls per 1000 patient-days), with 26.1% (82,332) of the falls having re- lated injuries (0.93 injurious falls per 1000 patient-days) in less than a 2-year period in US hospitals.7

Inpatient falls are a complicated phe- nomenon that involves multilevel factors, including patient-specific factors, envi- ronmental factors, organizational factors, and patient-staff interaction factors in the hospital.1 Research shows that patients identified at risk for falls are more likely to experience injurious falls.8 Evidence of the associations between gender and injurious falls is controversial.8-11 Hospital character- istics such as hospital size and Magnet or teaching status are associated with injurious falls. Small hospitals (<300 beds) tend to have more injurious falls,11,12 whereas Magnet and teaching hospitals have fewer inpatient falls and injurious falls, respectively.11,13

Unit factors (eg, unit types and nurse staffing) also are found to be associated with injurious falls. Among medical, surgical, and medical-surgical units, medical units have the highest prevalence of falls and injurious falls.7,9,11 Based on a systematic review and a meta-analysis, the evidence on the association between inpatient falls and nurse staffing is inconclusive,14,15although some studies sug- gest that lower inpatient fall rates are associ- ated with higher nursing hours, higher regis- tered nurse (RN) hours, and a higher propor- tion of RN hours.13

Nursing care process factors also contribute to inpatient falls and associated injuries. Stud- ies show that fall risk assessment and fall protocol implementation help prevent or re- duce injurious falls.8,11 Falls without em- ployee assistance are more likely to result in injuries8,10,11 However, the use of phys- ical restraints is positively associated with an increased risk for falls, injuries, and even death.16,17

Given the controversial evidence on factors associated with injurious falls in the literature and the complicated nature of inpatient

falls and associated injuries, further study integrating multilevel factors contributing to injurious falls is required. Using a modified Donabedian’s Structure-Process-Outcome (SPO) model18,19 that included patient characteristics in the SPO model, this study examined multilevel factors associated with injurious falls in acute care hospitals (see Supplemental Digital Content, Figure 1, avail- able at: http://links.lww.com/JNCQ/A323). To date, this study is the first attempt at investigating the complex phenomenon of injurious falls at multiple levels. The research question was “What organizational structure, unit structure, nursing care process, and patient factors are associated with injurious falls in acute care hospitals?”

METHODS

Design and data source

The study is a cross-sectional, correlational design, using the National Database of Nursing Quality Indicators (NDNQI) data collected be- tween July 2013 and June 2014. The nursing quality indicators such as falls and injurious falls in NDNQI are National Quality Forum– endorsed measures that have demonstrated strong reliability and validity.20,21

Only adult (≥18 years) patients who had any fall event on medical, surgical, medical- surgical, and step-down units in acute care hospitals during the study period were se- lected for this study; patient-level fall data then were aggregated to the unit level. NDNQI monthly data were aggregated into annual data at the unit level or hospital level. The final sample included 2229 units (medi- cal = 587; surgical = 412; medical-surgical = 795; and step-down = 435) in 488 hospi- tals. This study was determined to be nonhu- man subject research by the Human Subjects Committee at a Midwestern academic medical center.

Measures

Injurious falls

The annual total number of injurious falls among all inpatient falls at the unit level was

Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

22 JOURNAL OF NURSING CARE QUALITY/JANUARY–MARCH 2018

the outcome variable. NDNQI defines a fall as “sudden, unintentional descent, with or with- out injury to the patient, that results in the pa- tient coming to rest on the floor, on or against some other surface (eg, a counter), on another person, or on an object (eg, a trash can).”22(p2)

Injurious falls are falls with any injuries from minor to death.22

Structure, process, and patient factor

Independent variables included structure, process, and patient factors. Structure fac- tors had 2 levels: hospital and unit. Hospi- tal characteristics were categorized into hos- pital bed size (small: <100; medium: ≥100 to <300; and large: ≥300), teaching status (teaching and nonteaching), and Magnet sta- tus (Magnet and non-Magnet). Unit structure factors included unit types (ie, medical, sur- gical, medical-surgical, and step-down) and nurse staffing factors. Nurse staffing on the unit consisted of 5 continuous variables: to- tal nursing hours per patient-day (TNHPPD), RN hours per patient-day (RNHPPD), non- RNHPPD (calculated by subtracting RNHPPD from TNHPPD), RN skill mix, and RN turnover rate.

The following process and patient factors reported by NDNQI for each inpatient fall were first coded into dichotomous variables from monthly data files and then aggregated and summed across months into annual data at the unit level that represent the propor- tions of certain characteristics among patients who fell in the units. Fall risk assessment was measured by whether a fall risk assessment was performed on the patient prior to the fall. Implementing fall prevention protocol was measured by whether a documented fall prevention protocol had been implemented prior to the fall. Falls with employee assis- tance were defined as falls in which the pa- tient was assisted by a staff member, and the patient’s descent was slowed by the staff in an effort to minimize the fall impact.22 Phys- ical restraint(s) use was measured with any physical restraints that were in use at the time of patient fall.22 Patient gender was re- ported as either male or female. Patient fall

risk status was determined on the basis of the most recent risk assessment. The estimated coefficient is related to the change of every 10 percentage points at the unit level.

Data analyses

All statistical analyses were conducted with STATA 14 (StataCorp LP, College Station, Texas). Both descriptive and regression anal- yses were performed. Prior to the regres- sion analyses, correlation and interaction tests were carried out to examine the correlations and interactions among independent vari- ables. Because of strong correlations between RNHPPD and TNHPPD (r = 0.81), RNHPPD and RN skill mix (r = 0.53), only RNHPPD, non-RNHPPD, and turnover rate among nurse staffing factors were included in the models for analyses. A quadratic term of RNHPPD also was included in the models to test the poten- tial nonlinear relationship between injurious fall rates and RNHPPD.

The hierarchical negative binominal regres- sion model was used to account for the complex sample with multiple units within hospitals.23 In the regression model, the an- nual count of total injurious falls at the unit level was the outcome measure, with the an- nual count of total falls at the unit level as the exposure variable. Falls with missing data on 1 or more variables were excluded. In addition to the fixed effects of independent variables, a random hospital intercept was included in the model to control for the correlation among units within a hospital. The STATA procedure menbreg for multilevel data was used for mod- eling, with the significance level set at .05. The incidence rate ratios (IRRs) were estimated to show the associations of independent vari- ables with the injurious fall rate.

For model selection, a teardown method was used. The initial model included all 17 predictor variables. Variables within a group (ie, structure, process, nurse staffing, and unit type) with large P values were removed (P > .05) and the resulting reduced model was tested and compared with the previous model. Using the Akaike information criterion (AIC) value as a criterion, multiple models

Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

Injurious Falls in Acute Care Hospitals 23

were tested and the final model included 6 significant predictor variables with the small- est AIC value.

RESULTS

Descriptive analysis

The frequencies and descriptive statistics are shown in Supplemental Digital Content, Table (available at: http://links.lww.com/ JNCQ/A324). Hospitals with medium bed size or teaching status accounted for about one- half of the sample, whereas Magnet hospi- tals were about 20% of the sample. Medical- surgical units (35.7%) accounted for the most units included in the sample of 2229 units. There were on average 78.9 reported falls (range, 1-864) and 5.2 injurious falls (range, 1-31) annually across the units during the study period. The average patient-days were 17.11, ranging from 0.36 to 110.24 (unit: 1000 patient-days). RNHPPD on average was 6.28, with a range of 1.15 to 14.96.

Hierarchical regression analysis

Hierarchical regression modeling was used to estimate the associations between predica- tor variables and the injurious fall rate. The Table lists IRR values, with 95% confidence intervals and P values for different variables included in the initial and final models.

Hospital characteristics

Among hospital structure characteristics, teaching status was the only significant vari- able. Compared with falls in nonteaching hos- pitals, falls in teaching hospitals were 13% less likely to be injurious falls (P = .001). There was no significant difference between Mag- net and non-Magnet hospitals and among hos- pitals with different bed sizes.

Unit characteristics

Falls on surgical units were 8% more likely to be injurious falls than those in other units (P = .021). No significant differences were found in non-RNHPPD and turnover rate on injurious fall rates. RNHPPD was the only significant variable among nurse staffing

factors that showed a significant nonlinear relationship with injurious falls. On the basis of the estimated coefficients of linear and quadratic terms of RNHPPD, the injurious fall risk was estimated to be lowest at 5.08 RNH- PPD. With all the other independent variables controlled, the injurious fall risk was ex- pected to decrease with increasing RNHPPD till RNHPPD reached 5.08. After RNHPPD reached 5.08, the expected injurious fall risk increased with increased RNHPPD. Supple- mental Digital Content, Figure 2 (available at: http://links.lww.com/JNCQ/A325), shows the relationships between injurious fall risk and RNHPPD based on models estimated for the 4 different unit types separately. This figure shows the consistent nonlinear association across different unit types ex- cept for medical units. The variable, 1000 patient-days (total annual patient-days on the unit divided by 1000), also was identified to be a significant factor: with every 1000-day increase in patient-days, there was likely a 2% decrease in injurious fall risks (P < .001).

Nursing process factors

Among nursing process factors, falls with- out employee assistance was the only signifi- cant factor. At the unit level, the risk of injuri- ous falls was expected to increase by 4% when the percentage of falls without employee as- sistance increased by 10 percentage points (P = .005). In addition, the risk of injuri- ous falls was expected to decrease by 2.5% when the percentage of patients at fall risk in- creased by 10 percentage points on the units (P = .009).

In addition to fixed effects, the random ef- fect of the model was estimated for the hos- pital level. The injurious fall rates for an indi- vidual hospital might vary from 19% lower to 24% higher than the average injurious fall rate of all hospitals, assuming all other variables are fixed.

DISCUSSION

The major contribution of the study is the examination of the associations between

Copyright © 2018 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

24 JOURNAL OF NURSING CARE QUALITY/JANUARY–MARCH 2018

Table. Estimates of Associations Between the Injurious Fall Rate and Predicator Variables

Initial Model (AICa = 9368.685) Final Model (AIC = 9351.538)

Fixed Effects IRR (95% CI) P IRR (95% CI) P

Bed size (large vs) Small 1.10 (0.95-1.26) .205 . . . . . . Medium 1.04 (0.94-1.15) .412 . . . . . .

Teaching 0.88 (0.81-0.96) .004 0.87 (0.80-0.94) .001 Magnet 1.00 (0.91-1.11) .942 . . . . . .

Unit type (medical vs) Surgical 1.07 (0.99-1.16) .066 1.08 (1.01-1.16)b .021 Medical-surgical 0.97 (0.91-1.04) .424 . . . Step-down 0.98 (0.90-1.06) .576 . . . . . .

RNHPPD 0.86 (0.78-0.95) .002 0.86 (0.78-0.95) .002 RNHPPD2 1.01 (1.01-1.02) <.001 1.01 (1.01-1.02) <.001 Non-RNHPPD 1.00 (0.97-1.04) .788 . . . . . . Turnover 0.99 (0.99-1.00) .211 . . . . . . Annual patient-days

(1000 d) 0.98 (0.99-1.00) <.001 0.98 (0.98-0.982) <.001