Contract Nurses with Hospitalized Patient Pressure Injuries and Falls Paper

Association of Use of Contract Nurses With Hospitalized Patient Pressure Injuries and Falls Alice Ferguson, MSN, RN, CPHQ1, Alison Bradywood, MN, MPH, RN, NEA-BC2, Barbara Williams, PhD3, & C. Craig Blackmore, MD, MPH4,* 1 Zeta Upsilon-at-Large, Director, Quality Outcomes and Metrics, MultiCare Health System, and Fellow, Center for Health Care Improvement Science, Virginia Mason Medical Center, Seattle, WA, USA 2 Iota, Administrative Director, Clinical Quality and Clinical Administration, Virginia Mason Medical Center, Seattle, WA, USA 3 Research Scientist, Center for Health Care Improvement Science, Virginia Mason Medical Center, Seattle, WA, USA 4 Director, Center for Health Care Improvement Science, Virginia Mason Medical Center, Seattle, WA, USA Key words Falls, nurse staffing, pressure injuries Correspondence C. Craig Blackmore, 1202 Terry Ave., #324, Seattle, WA 98101. E-Mail: Craig.blackmore@virginiamason.org Accepted April 21, 2020 doi:10.1111/jnu.12572 Abstract Background: Hospital-acquired pressure injuries (HAPIs) and falls are outcomes sensitive to quality of nursing care. Use of contract (traveler) nurses varies among organizations, but there is little research on the effect of contract nurses on nurse-sensitive outcomes. Objectives: To explore the relationship between use of contract nurses and two key nurse-sensitive outcomes, HAPIs and falls. Research Design: This was a cross-sectional study of unit-level nursing, patient, and hospital factors versus HAPIs and falls from a national nursing data consortium from 2015 to 2016. We used cluster analysis to identify similar units, and compared outcomes between clusters. Subjects: 605 nursing units in 166 hospitals, 3.2 patients per nurse, and 5.3% contract nurses. Measures: Prevalence and incidence of HAPIs and number of falls, adjusted by patient days. Results: For both prevalence and incidence of HAPIs, there was a statistically significant difference between the five independent cluster groups (p = .012 and p = .001, respectively). The cluster with the highest percentage of nurse travelers (>7%) had the highest HAPI prevalence (0.84%) and incidence (0.055 per 1,000 patient days) despite higher nurse staffing, compared to HAPI prevalence of 0.32% and incidence of 0.017 per 1,000 patient days in the cluster with the lowest percentage of nurse travelers (<2%). We did not identify a consistent relationship between use of contract nurses and falls. Conclusions: Use of contract nurses was associated with higher HAPI prevalence and incidence, independent of staffing levels. Clinical Relevance: Our results suggest that institutions should either minimize the use of contract nurses, or engage in extensive training to confirm that contract nurses have understanding of the institutional practices around HAPIs. Hospitalized patients in the United States experience up to 1 million falls (Agency for Healthcare Research and Quality, 2019) and 2.5 million hospital-acquired pressure injuries (HAPIs) each year (Agency for Healthcare Research and Quality, 2014). Both are widely considered to be preventable with appropriate nursing care, and beginning in 2008, the Centers for Medicare and Medicaid Services limited reimbursement for care Journal of Nursing Scholarship, 2020; 0:0, 1–9. © 2020 Sigma Theta Tau International associated with severe (stage 3, 4, and unstageable) HAPIs and in-hospital injury falls. However, despite policy initiatives resulting in legislated nurse staffing levels in several states, the actual relationship between nurse staffing and HAPIs and falls remains unclear, and the available literature inconclusive. Previous reports and systematic reviews support an association between higher overall nurse 1 Contract Nurses, Pressure Ulcers, and Falls staffing levels and both lower in-hospital mortality (Aiken et al., 2017; Needleman et al., 2011; Stalpers, de Brouwer, Kaljouw, & Schuurmans, 2015) and lower healthcare-acquired infection (Mitchell, Gardner, Stone, Hall, & Pogorzelska-Maziarz, 2018). However, systematic reviews have identified no consistent relationship between staffing levels and HAPIs or falls outside of critical care units (Driscoll et al., 2018; Staggs & Dunton, 2014; Stone et al., 2007). Further, nursing staffing models are complex, with variability in nurse levels of experience, certification, and educational attainment, as well as differing nurse turnover, use of contract nurses, and nursing care hours per patient day. Finally, patient complexity or severity of illness is difficult to capture, and comparisons between units and healthcare organizations may represent varying levels of patient care needs. The limited published data do not support any one nurse staffing model over the others (Olley, Edwards, Avery, & Cooper, 2019). Policy efforts in California and other states have addressed the dependence of patient outcomes on nursing through mandated number of nursing hours per patient day, or ratios of nurses to patient census. These regulations do not specifically address the experience of the nurses, or the use of contract nurses. Furthermore, while studies of California’s mandated nurse staffing ratio laws have shown statistically significant increases in staffing levels, the effects on nurse-sensitive outcomes have been inconsistent (Bolton et al., 2007; Donaldson & Shapiro, 2010; Mark, Harless, Spetz, Reiter, & Pink, 2013; Serratt, 2013). Accordingly, to meet regulatory requirements or the perception that increased numbers of nurses will improve outcomes, institutions may deploy temporary contract nurses in limited numbers. Use of contract (traveler) nursing staff varies among organizations, and the literature on any relationship between use of contract nurses and nurse-sensitive outcomes is limited (Aydin, Donaldson, Aronow, Fridman, & Brown, 2015; Aydin, Donaldson, Stotts, Fridman, & Brown, 2015; Bae, Brewer, Kelly, & Spencer, 2015; Page, 2008). Though often experienced (Xue, Smith, Freund, & Aiken, 2012), contract nurses may not be familiar with the specific institutional policies, protocols, and organizational culture that may be important in preventing falls and HAPIs. Also, there is variability, and no national consensus on the appropriate deployment and site orientation of traveling nurses to an institution or unit. The objective of this study was to explore the relationship between use of contract nurses on medical and surgical units and two key nurse-sensitive outcomes, HAPIs and in-hospital falls. 2 Ferguson msn et al. Methods The overall study design was a cross-sectional evaluation of hospital characteristics, contract nurse rates (percentage of nursing care hours provided by contract nurses), nurse staffing levels, patient variables, and nurse-sensitive outcomes data aggregated for the years 2015 to 2016. The study was performed using deidentified data from organizations participating in the Collaborative Alliance for Nursing Outcomes (CALNOC) database. CALNOC is a voluntary consortium of over 284 hospitals in the United States who prospectively supply quarterly data on nurse staffing and nursingdependent outcomes (including HAPIs and falls) to a central repository (Aydin, et al., 2004; CALNOC, 2020). Data are aggregated at the unit and hospital level. We included all hospitals with data reported between 2015 and 2016, even if there was never a contract nurse in any of the units. We excluded for-profit hospitals (due to small number in the CALNOC data set), and critical care and step-down units. A small number of units (<1%) were missing data for the percentage of medical patients and they were not included in the analysis. The final number of hospitals included in the analysis was 166. The relationship between nurse staffing and patient outcomes was explored using correlation coefficients to evaluate the variables for inclusion in the model and cluster analysis to discover natural groupings within the model. Cluster analysis is a machine learning method of identifying groups of units with similar characteristics among the “independent” clustering variables included in the analysis. Inferences regarding the “dependent” outcome variables (not included in the clustering model) can then be made across the clusters (Everitt, Landau, Leese, & Stahl, 2011; Hofstetter, Dusseldorp, van Empelen, & Paulussen, 2014). Clustering variables in the model included two nursing measures (nurse staffing [patients per RN] and use of contract nurses [contract RN hours per total RN hours]); a patient measure (proportion of medical patients [vs. surgical]); and one hospital measure (total staffed beds). Patient turnover was not included in the final model due to strong correlation with the variable proportion of medical patients, which was included because it had less missing data. Patient gender and age were also not included because of their high correlation with the proportion of medical patients. Nursing turnover was not included in the final cluster model because of nonrandom missing data. HAPIs and falls were the outcomes of interest, and therefore were not included in the clustering models. Journal of Nursing Scholarship, 2020; 0:0, 1–9. © 2020 Sigma Theta Tau International Contract Nurses, Pressure Ulcers, and Falls Ferguson msn et al. The outcome variables in the final model were HAPIs (both prevalence of HAPIs of all grades, and incidence of grade 3 or higher HAPIs) and number of falls per patient day. Because HAPIs and falls were infrequent, the distribution of these outcome variables was skewed. Therefore, clustering variables in the model were first log transformed to approximate a normal distribution (with a correction of half of the minimum observed value added to values of zero), then to z scores using means and standard deviations because scales differed across variables. We used Ward’s linkage hierarchical method of cluster analysis with a Euclidean distance option because it is commonly used in other hospital studies, yielded consistent, reproducible results, had individual cluster sizes of at least 5% of the total, and produced clusters that made logical connections (face validity)(Dunn et al., 2018; Everitt et al., 2011). Hierarchical clustering starts by treating each unit as a separate cluster, then repeatedly merges similar clusters until all clusters are merged or stopped at a predetermined number of clusters. To validate the cluster assignments, we compared the results to the most frequent assignment from 100 Monte Carlo simulations of the K-means model (Cohen’s kappa = 0.71, 0.74, 0.77 for HAPI prevalence, HAPI 3+ incidence, and falls, respectively; Halpin, 2014). The optimal number of clusters was determined using Calinski/Harabasz pseudo-F scores and Duda/Hart pseudo T-squared measures, which both indicated five clusters (Everitt et al., 2011). Statistical differences between the clusters for the outcomes was determined using one-way analysis of variance (ANOVA) on unadjusted, unstandardized values for the clustering variables. Analyses were performed using Stata MP version 15.1 (StataCorp LLC, College Station, TX, USA). This study was determined by the institutional review board to be exempt from review. Results The study population consisted of 605 nursing units in 166 hospitals (Table 1). The number of staffed beds per hospital ranged from 25 to 877, with a mean of 326. The units had a mean patient age of 61.9 years, 48.4% of patients were male, and 70.6% were medical patients. There was an average of 3.2 patients per nurse, with 5.3% being contract nurses. Of the three outcomes, the mean HAPI prevalence (any stage) was 0.55%, the mean HAPI 3+ incidence was 0.032 patients with injuries per 1,000 patient days, and there were 2.48 mean falls per 1,000 patient days. Cluster analysis revealed five distinctly different groups of hospital units that were similar in the analyses for each of the three outcome measures (Table 2): Cluster A featured a low percentage of contract nurses (<1%), Cluster B featured a low proportion of medical patients, Cluster C featured small hospitals with high nurse staffing (low numbers of patients per nurse), Cluster D featured low nurse staffing, and Cluster E featured a combination of a high percentage of contract nurses, large hospitals, and a high percentage of medical patients (Figure S1). For the outcome variables HAPI prevalence and HAPI 3+ incidence, there was a statistically significant difference between the five independent cluster groups (p = .012 and p = .001, respectively, Tables 2 and 3). Cluster E, with the highest percentage of nurse travelers (>7%) also had the highest HAPI prevalence (0.84%) and HAPI 3+ incidence (0.055 per 1,000 patient days), despite having the highest nurse staffing. Cluster A, Table 1. Characteristics of Population Study Variables for 605 Hospital Units Variable Outcome variables HAPI prevalence HAPI 3+ incidence Falls Hospital variables Hospital beds Nurse variables Patients per nurse Contract nurses Patient variables Medical admission Patient gender Patient age Variable description Mean (SD) Minimum–maximum n % of patients with any HAPI Number of patients who developed grade 3 or greater HAPIs while in hospital (per 1,000 hospital days) Total number of falls per 1,000 patient days 0.55 (1.05) 0.032 (0.071) 0–8.33 0–0.554 596 601 2.48 (1.53) 0–11.36 603 Total hospital staffed beds 326 (180) 25–877 605 Number of patients per staff nurse % of total nursing hours staffed by contract nurses 3.2 (0.7) 5.3 (6.4) 1.2–6.9 0–43.2 605 605 % medical admission (vs. surgical) % male gender Mean patient age, years 70.6 (26.9) 48.4 (9.3) 61.9 (5.7) 0–100 1–100 32–74 605 605 605 Note. HAPI = hospital-acquired pressure injury. Journal of Nursing Scholarship, 2020; 0:0, 1–9. © 2020 Sigma Theta Tau International 3 Contract Nurses, Pressure Ulcers, and Falls Ferguson msn et al. Table 2. Hospital, Nurse, and Patient Characteristics in Each Cluster for HAPI Prevalence (Mean Values Unless Indicated Otherwise; N = 596 units) Number of units HAPI prevalence (%) Hospital Staffed beds Nurse Patients per nurse Contract nurses (%) Patient Medical patients (%) Male (%) Age (years) Cluster A: low % of contract nurses Cluster B: low % of medical patients Cluster C: small hospital, low number of patients per nurse Cluster D: high Cluster E: high % of contract number of patients nurses, large hospitals, high % per nurse of medical patients ANOVA 74 0.32 99 0.41 98 0.51 226 0.57 99 0.84 .012 342 394 133 293 506 <.001 3.38 0.02 3.23 5.35 2.80 6.01 3.50 5.97 2.70 7.69 <.001 <.001 73.7 49.5 61.4 26.3 46.5 59.7 78.1 48.4 64.0 79.4 48.1 62.4 86.1 50.9 61.1 <.001 .012 <.001 Note. HAPI = hospital-acquired pressure injury; ANOVA = analysis of variance. which had the lowest percentage of nurse travelers (<2%) also had the lowest HAPI prevalence (0.32%) and HAPI 3+ incidence (0.017 per 1,000 patient days). Post-hoc ANOVA confirmed the difference in HAPI prevalence (Tukey p = .012) and HAPI 3+ incidence (Tukey p = .007) between Cluster E and Cluster A. The relationship between nurse staffing levels, hospital size, and patient mix with HAPIs was less consistent. We found the highest rates of HAPIs only with the combination of large hospital size, high proportion of medical patients, and high proportion of contract nurses. Variable rates of HAPIs were observed across a range of hospital sizes and proportion of medical patients in Clusters B, C, and D. There was no consistent relationship between nurse staffing levels and HAPIs. No consistent relationship was identified between contract nurse utilization rates and falls. Fall rates did not differ greatly between clusters, with the exception of Cluster B, which had a lower proportion of medical patients and a significantly lower fall rate (1.80 falls per 1,000 patient days; Table 4). The clusters with the highest (Cluster C) and lowest (Cluster A) contract nurse rates had similar fall rates (2.68 and 2.54 falls per 1,000 patient days, respectively). There was also no consistent relationship between nurse staffing and falls. Discussion In this report, we use cluster analysis to demonstrate a significant association between higher use of contract nurses and higher rates of the nurse-sensitive patient outcomes of HAPI prevalence and incidence. The highest rates of HAPI were observed in units with a large proportion of contract nurses and a high proportion of medical patients, in larger hospitals. It may be that the combination of all three factors must be present for the Table 3. Hospital, Nurse, and Patient Characteristics in Each Cluster for HAPI 3+ (Mean Values Unless Indicated Otherwise; N = 601 units) Number of units HAPI 3+ incidence Hospital Staffed beds Nurse Patients per nurse Contract nurses (%) Patient Medical patients (%) Male (%) Patient age (years) Cluster A: low % of contract nurses Cluster B: low % of medical patients Cluster C: small hospital, low number of patients per nurse Cluster D: high number of patients per nurse Cluster E: high % of contract nurses, large hospitals, high % of medical patients ANOVA 69 0.017 74 0.040 136 0.018 241 0.035 81 0.055 .001 355 442 158 299 556 <.001 3.37 0.003 3.23 4.21 2.77 7.14 3.56 5.63 2.84 7.22 <.001 <.001 73.4 48.5 61.3 19.5 46.0 58.9 78.5 49.1 63.9 75.4 48.0 62.2 86.0 50.7 60.6 <.001 .024 <.001 Note. HAPI = hospital-acquired pressure injury; ANOVA = analysis of variance. 4 Journal of Nursing Scholarship, 2020; 0:0, 1–9. © 2020 Sigma Theta Tau International Contract Nurses, Pressure Ulcers, and Falls Ferguson msn et al. Table 4. Hospital, Nurse, and Patient Characteristics in Each Cluster for Falls (Mean Values Unless Indicated Otherwise; N = 603 Units) Number of units Falls Hospital Staffed beds Nurse Patients per nurse Contract nurses (%) Patient Medical patients (%) Male (%) Age (years) Cluster A: low % of contract nurses Cluster B: low Cluster C: small % of medical hospital, low number patients of patients per nurse Cluster D: high number of patients per nurse Cluster E: high % of contract nurses, large hospitals, high % of medical patients ANOVA 128 2.54 56 1.80 120 2.68 210 2.39 89 2.73 .0017 377 416 145 290 519 <.001 3.47 0.15 3.07 7.01 2.75 7.30 3.57 6.25 2.86 6.10 <.001 <.001 66.5 50.0 60.6 19.3 44.8 58.6 75.7 49.3 64.1 76.5 47.2 62.5 88.2 51.0 61.5 <.001 <.001 <.001 Note. HAPI = hospital-acquired pressure injury; ANOVA = analysis of variance. nurse-sensitive outcomes to be affected. In addition, units in the largest hospitals tended to have higher rates of HAPIs, while higher nurse staffing was not associated with fewer HAPIs. The cross-sectional design of this work precludes us from assigning causation. It is possible that there is some other underlying characteristic of larger hospitals with complex medical patients employing large numbers of contract nurses that is the true cause of the adverse outcomes we report (Xue, Aiken, Freund, & Noyes, 2012). However, there are theoretical reasons why high numbers of contract nurses may be associated with worse outcomes. Though often experienced, contract nurses are likely less familiar with the institutional protocols, standards, and culture (Aiken, Xue, Clarke, & Sloane, 2007). Prevention of pressure injuries is dependent on a consistent systematic approach and teamwork, rather than the skill of any one individual (Castle, 2009). Therefore, placing contract nurses in an unfamiliar setting may decrease their ability to prevent these adverse events. The prior literature on the relationship between contract nurses and nurse-sensitive patient outcomes has been inconsistent. Aiken, Shang, Xue, and Sloane (2013) found no relationship between use of contract nurses and nursing-sensitive outcomes, while Aydin, Donaldson, Stotts, et al. (2015) found an associa …