Skip Navigation LinksHome > July/September 2012 - Volume 27 - Issue 3 > Understanding Unassisted Falls: Effects of Nurse Staffing Le...
Journal of Nursing Care Quality:
doi: 10.1097/NCQ.0b013e318241da2d
Articles

Understanding Unassisted Falls: Effects of Nurse Staffing Level and Nursing Staff Characteristics

Staggs, Vincent S. PhD; Knight, Jeff E. MPH, BSN; Dunton, Nancy PhD

Free Access
Article Outline
Collapse Box

Author Information

Department of Biostatistics (Dr Staggs) and School of Nursing (Mr Knight and Dr Dunton), University of Kansas Medical Center, Kansas City.

Correspondence: Vincent S. Staggs, PhD, 3901 Rainbow Blvd (MS 3060), Kansas City, KS 66160 (vstaggs@kumc.edu).

Funding for this work was provided by the American Nurses Association.

The authors declare no conflict of interest.

Accepted for publication: November 14, 2011.

Published online before print: December 19, 2011.

Collapse Box

Abstract

Hierarchical Poisson modeling was used to explore hospital and nursing unit characteristics as predictors of the unassisted fall rate. Longitudinal data were collected from 1502 units in 248 US hospitals. The relation between the fall rate and total nurse staffing was positive at lower staffing levels and negative for levels around and above the median. The fall rate was negatively associated with registered nurse skill mix and average registered nurse tenure on the unit.

PREVENTION of patient falls has become a national policy issue. The National Quality Forum included patient death or serious disability due to a fall on its 2006 list of “unambiguous, serious, preventable adverse events that concern both the public and health care providers”,1 and in 2008 the Centers for Medicare & Medicaid Services stopped paying for costs associated with hospital-acquired injuries due to falls.2 While researchers have reported associations between fall rates and total nurse staffing levels,3 registered nurse (RN) staffing levels,4 proportion of nursing care hours provided by RNs,3 and RN years of experience,3 in most studies they have not distinguished between assisted and unassisted falls, and little is known about the nursing-related factors associated with each.

Back to Top | Article Outline

UNASSISTED FALL RATE AS INDICATOR OF NURSING CARE QUALITY

The American Nurses Association includes the total fall rate among its “nursing-sensitive quality indicators”,5 and the National Quality Forum6 has endorsed the total fall rate and the injury fall rate as national consensus standards for nursing-sensitive care. While the injury fall rate is important as a measure of patient safety and for its cost implications, the rate of unassisted falls is arguably a better measure of nursing care quality than the total or injury fall rate.

An assisted fall represents a partial success for the hospital staff in that someone is present to mitigate the effects of the fall. Assisted falls are less likely to result in patient injury7 and are not necessarily indicative of lack of vigilance or inadequate staffing. An unassisted fall, by contrast, represents a preventable risk of injury. Thus, unassisted falls more closely reflect the quality of nursing care provided. These important differences between assisted and unassisted falls are ignored when only the total and injury fall rates are used as measures of nursing care quality.

Back to Top | Article Outline
Purpose

The purpose of this study was to explore hospital and nursing unit characteristics, including several nurse staffing and nursing staff characteristic variables, as potential predictors of unassisted fall rates using longitudinal data from a large sample of nursing units. Predictors of monthly unassisted fall rates for nursing units were examined while accounting statistically for potential time trends and for correlations among rates for units within the same hospital. The results of this fine-grained analysis are informative both for patient safety researchers and for hospitals seeking to prevent patient injuries due to unassisted falls.

Back to Top | Article Outline

METHODS

Sample and data

The data used in the study were submitted by hospitals participating in the National Database of Nursing Quality Indicators (NDNQI), which collects data on nursing-sensitive quality measures at the nursing unit level. About one-third of all US hospitals participate in NDNQI, but because these hospitals self-select for participation, NDNQI hospitals are not necessarily representative of the population of all US hospitals.

The sample comprised 1504 nursing units from 248 US acute care hospitals. The sample was limited to 7 adult unit types and included 331 critical care units, 224 step-down units, 267 medical units, 234 surgical units, 366 medical/surgical units, and 82 rehabilitation units. Monthly data were collected from October 2009 through September 2010. Unit months with missing data were not included in the analysis. The average unit in the study submitted complete data for 8.7 of the 12 months of the study.

The results of the American Hospital Association's (AHA) annual survey8 for fiscal year 2009 were used to compare the sample with the larger set of US hospitals registered with the AHA. Hospitals with fewer than 100 beds were underrepresented in the sample (29% vs 54% of AHA hospitals), whereas those with other bed sizes, especially in the 100 to 399 range, were overrepresented. Metropolitan hospitals made up a larger share of the sample (80%) than AHA hospitals (64%), as did hospitals recognized by the American Nurses Credentialing Center as nursing Magnets (27% vs 6%).

Back to Top | Article Outline
Explanatory variables

A total of 16 explanatory variables were considered as predictors of the rate of unassisted falls. Three of these were hospital characteristic variables. Hospital Magnet status was a dichotomous variable used to differentiate American Nurses Credentialing Center–recognized Magnet hospitals from non-Magnet hospitals. Hospital teaching status had 3 levels: academic medical center (a hospital serving as the primary clinical site for a school of medicine), teaching hospital, and community (nonteaching) hospital. Each hospital's most recently reported average daily census was considered as a predictor to investigate the effect of hospital size.

There were 11 unit characteristic variables in the set of explanatory variables, including 4 time-varying variables and their squares (included to explore nonlinear effects). One of these time-varying variables—the percentage of nurses on each unit with a bachelor's or higher nursing degree—varied by quarter. The other 3 time-varying variables—total nursing hours per patient day (TNHPPD), skill mix, and total turnover—varied by month. The TNHPPD was computed by dividing the unit's total nursing care hours (provided by RNs, licensed practical nurses, or unlicensed assistive personnel) for the month by its total number of patient days for the month. Skill mix was defined as the proportion of the month's total nursing care hours provided by RNs. Total turnover was measured by dividing the number of nursing employees who separated from the unit during the month by the count of nurses employed on the last day of the month.

The remaining unit characteristic variables were not time varying. These were unit type; the average RN tenure, defined as average number of years employed on the unit as reported in the most recent response to the NDNQI's annual RN survey; and the square of the average RN tenure. In addition to the hospital and unit characteristic variables considered as predictors, the month of the study and its square were included as explanatory variables to investigate time trends in the rate of unassisted falls.

Back to Top | Article Outline
Statistical analysis

Modeling was carried out with SAS 9.2 using the GLIMMIX procedure (SAS Institute Inc, Cary, North Carolina). The associations between the unassisted fall rate and the explanatory variables were examined using a 3-level log-linear Poisson mixed model. Conditional on a random hospital intercept and random unit intercept, the monthly count of unassisted falls was assumed to follow a Poisson distribution. The monthly count of patient days was included in the model as an exposure variable.

A teardown approach was used for model selection. Beginning with the model including all 16 explanatory variables, the variable with the largest P value was removed and the resulting reduced model was fit and compared with its predecessor. At each step, the model with the smaller Akaike information criterion, corrected for sample size,9 value was retained. For quantitative variables, the linear term was removed only if the corresponding quadratic term had been removed in a previous step.

Quantitative explanatory variables were centered for use in modeling. There was no evidence of significant multicollinearity; variation inflation factors were less than 3, and the largest condition index was 13.

Back to Top | Article Outline

RESULTS

Frequencies and descriptive statistics are given in Table 1. The final model included unit type, month, linear and quadratic effects for TNHPPD, skill mix, and average RN tenure on the unit. Regression parameter estimates and P values for the quantitative variables in the final model are shown in Table 2.

Table 1
Table 1
Image Tools
Table 2
Table 2
Image Tools

The effect of unit type on the rate of unassisted falls was examined as follows: For each unit type, the expected rate of unassisted falls was computed at the average values of the other predictors, 95% confidence limits were estimated for this rate, and the difference in fall rates for each pair of unit types was tested for significance. These tests and the 95% confidence limits were adjusted using the Tukey-Kramer method10 to prevent type I error rate inflation.

Rehabilitation units had the highest expected rate of unassisted falls (4.11 falls per 1000 patient days; 95% confidence interval [3.59-4.72]), followed by medical units (2.85 [2.59-3.13]), medical/surgical units (2.60 [2.37-2.84]), step-down units (2.38 [2.17-2.61]), surgical units (1.81 [1.64-2.00]), and critical care units (1.00 [0.86-1.17]). Only the differences between step-down and medical/surgical units and between medical and medical/surgical units were not significant at α = .05.

The association between TNHPPD and the unassisted fall rate was nonlinear. According to the estimates of the linear and quadratic TNHPPD coefficients, unassisted fall rates are highest at 9.1 TNHPPD, not far below the median staffing level of 9.4 TNHPPD. With other predictors held constant, the expected rate of unassisted falls increases by 7% as TNHPPD increases from 4.0 (the minimum value in the study) to 9.1. As TNHPPD continues to increase, the fall rate begins to drop at an increasingly steep rate. For example, at the 75th percentile (12.5 TNHPPD), an increase of 1 TNHPPD is associated with a 2.1% drop in the fall rate and a full standard deviation (SD) increase of 3.8 TNHPPD is associated with a 10.3% drop. At the 90th percentile (16.7 TNHPPD), the fall rate decreases by 4.3% with an increase of 1 TNHPPD and by 17.6% with an increase of 1 SD. The rate of decrease does not begin to slow until the staffing level exceeds the 99th percentile (22.3 TNHPPD).

The effects of the 2 staff characteristic variables in the final model—skill mix and average RN tenure on the unit—were modest. An increase of 0.14 (1 SD) in the proportion of nursing care hours provided by RNs is associated with an estimated 4.0% average decrease in the rate of unassisted falls, whereas an increase of 2.8 years (1 SD) in average RN tenure is associated with a 2.3% decrease in the unassisted fall rate.

There was a slight downward trend in the rate of unassisted falls over time; with other predictors held constant, the mean rate decreased by an estimated 1% per month during the study. Given the limited number of months in the study, it is not clear whether this trend reflects a seasonal effect or a longer-term decrease in falls.

Back to Top | Article Outline

DISCUSSION

Nurse staffing level and nursing staff characteristics have been linked to the total fall rate in previous research,3 and the results of this study confirm that these variables are also important for understanding and preventing unassisted falls. Three staffing-related variables, along with unit type and month, were identified as predictors of the unassisted fall rate. None of the 3 hospital characteristic variables—Magnet status, teaching status, and average daily census—was a predictor.

The unassisted fall rate does not have a simple linear association with staffing. For staffing levels around and above the median, the unassisted fall rate tends to decrease as staffing increases; thus, adding nursing personnel may reduce the rate of unassisted falls. However, at lower staffing levels, the association between staffing and the fall rate is positive and an increase in staffing may lead to a slightly higher rate of unassisted falls. As staffing increases, it eventually reaches a level at which unassisted falls begin to become less frequent, but in this study, the expected fall rate did not drop below its value at the minimum observed staffing level until TNHPPD reached the 80th percentile. In other words, all else being equal, units at the very lowest staffing levels tended to have lower fall rates than most units with staffing levels above the median.

Understanding the positive association between the unassisted fall rate and TNHPPD at lower staffing levels will likely require further research. One possible explanation involves what social psychologists refer to as “diffusion of responsibility.”11 Nurses on units with lower staffing levels have higher levels of individual responsibility for patient safety because they have fewer coworkers to depend on, and this heightened sense of personal responsibility may result in greater focus on patient safety. At higher staffing levels responsibility is diffused, and the individual's level of focus may not be as high.

Whatever the explanation for this positive association between staffing and the unassisted fall rate, it should be noted that as TNHPPD increases from 4.1 to 9.1 (the range over which the fall rate increases with staffing), with other predictors held constant, the expected fall rate increases by 7%. This is not a dramatic change, and units with staffing levels in this range below the median should not rule out an increase in nursing personnel based solely on the results of this study.

It also should be noted that the association between staffing and the rate of unassisted falls may vary by nursing unit type. This was not accounted for in the model; the linear and quadratic TNHPPD coefficients were assumed to be the same for all unit types. On the basis of the data and model used in this study, the relation between staffing and the unassisted fall rate changes from positive to negative at a certain value of TNHPPD, which was estimated to be 9.1. Assuming this pattern holds for all unit types—and it is possible that it does not—the TNHPPD value at which the relation changes likely depends on the type of unit. This is an area for future research.

The finding that skill mix and average RN tenure are also predictors of the unassisted fall rate suggests that it is not simply the number of nursing care hours provided that is important in preventing unassisted falls but also the qualifications and experience of the personnel providing these hours. The effect of skill mix found in this study is consistent with some previous unit-level research linking skill mix to the total fall rate.3 It should be noted that the effects of TNHPPD discussed earlier hold only if skill mix remains constant. If TNHPPD is augmented by increasing only the number of support staff, the resultant decrease in skill mix will tend to mute the beneficial effect of higher TNHPPD or, in the case of lightly staffed units, exacerbate its apparent adverse effect.

The association between average RN tenure on the nursing unit and the rate of unassisted falls in this study was modest, and further research is needed to understand how and why RN tenure affects fall rates and to determine whether it also affects other patient safety outcomes. Because RNs with longer tenure on a single nursing unit may also tend to have more years of nursing experience, it is possible that nursing experience, which has been found to affect the total fall rate,3 was partly responsible for the effect of RN tenure found in this study.

Another possibility is that nursing units on which nurses have longer tenures are characterized by better teamwork, which has been linked to improvements in a number of patient safety outcomes, including falls.12 It should be noted that data on RN tenure were reported by each unit only once during the period of study, and the use of monthly tenure data reflecting changes due to turnover throughout the year might have led to different results.

Differences in fall rates among unit types were not surprising. Patients on critical care units are arguably the least likely to attempt to walk unassisted; they have the highest levels of acuity, and many are sedated or depend on a ventilator. Moreover, the presence of non–nurse health care providers (eg, respiratory therapists) on the unit reduces the likelihood that a patient attempting to get out of bed will go unnoticed by staff and fall without assistance. At the other end of the spectrum, patients on rehabilitation units have lower levels of acuity and rehabilitation often involves regaining the ability to walk without assistance—the very activity that may put patients at risk for a fall.

One of the limitations of this study is that the sample was not representative. There may be important differences between NDNQI and non-NDNQI hospitals that have not been measured, and the findings of this study might have been different had it been conducted using data from a random sample of US hospitals. Another potential limitation is that individual patient characteristics related to risk of falling were not controlled for. Differences among units in patient fall risk were partially accounted for by including unit type as a predictor, but further research in which patient characteristics are considered as predictors may shed additional light on the topic of unassisted falls.

Back to Top | Article Outline

REFERENCES

1. National Quality Forum. Serious Reportable Events in Healthcare—2006 Update. Washington, DC: National Quality Forum; 2007.

2. ECRI Institute. List of CMS hospital-acquired conditions expanded under new final rule. 2008. https://www.ecri.org/Documents/Patient_Safety_Center/CMS_New_Final_Rule.pdf. Accessed September 2, 2011.

3. Dunton N, Gajewski B, Klaus S, Pierson B. The relationship of nursing workforce characteristics to patient outcomes. Online J Issues Nurs. 2007;12(3):1–11. http://www.nursingworld.org/MainMenuCate-gories/ANAMarketplace/ANAPeriodicals/OJIN/TableofContents/Volume122007/No3Sept07/NursingWorkforceCharacteristics.aspx. Accessed September 2, 2011.

4. Lake ET, Shang J, Klaus S, Dunton NE. Patient falls: association with hospital magnet status and nursing unit staffing. Res Nurs Health. 2010;33(5):413–425.


6. National Quality Forum. National Voluntary Consensus Standards for Nursing-Sensitive Care: An Initial Performance Measure Set. Washington, DC: National Quality Forum; 2004.

7. Krauss MJ, Nguyen SL, Dunagan WC, et al. Circumstances of patient falls and injuries in 9 hospitals in a midwestern healthcare system. Infect Control Hosp Epidemiol. 2007;28(5):544–550.

8. American Hospital Association. AHA Annual Survey Database™. Chicago, IL: Health Forum; 2009.

9. Burnham KP, Anderson DR. Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach. 2nd ed. New York, NY: Springer-Verlag; 2002.

10. Kramer CY. Extension of multiple range tests to group means with unequal numbers of replications. Biometrics. 1956;12(3):307–310.

11. Darley JM, Latané B. Bystander intervention in emergencies: diffusion of responsibility. J Pers Soc Psychol. 1968;8(4):377–383.

12. Kalisch BJ, Curley M, Stefanov S. An intervention to enhance nursing staff teamwork and engagement. J Nurs Adm. 2007;37(2):77–84.

nurse staffing; nursing units; patient falls; patient safety

© 2012 Lippincott Williams & Wilkins, Inc.

Login

Search for Similar Articles
You may search for similar articles that contain these same keywords or you may modify the keyword list to augment your search.