Journal of Nursing Management, 2012 Nurse turnover in New Zealand: costs and relationships with staffing practises and patient outcomes NICOLA NORTH P h D , R N 1, WILLIAM LEUNG B A ( H o n s ) M S c 2, TONI ASHTON 4 5 6 P h D , FRANCES HUGHES D N u r s , R N and MARY FINLAYSON P h D , R N PhD 3 , ERLING RASMUSSEN Associate Professor, University of Auckland, Auckland, 2Research Fellow, University of Auckland, Auckland, Professor, University of Auckland, Auckland, 4Professor, Auckland University of Technology, Auckland, 5Visiting Professor University of Sydney and Adjunct Professor, Auckland University of Technology, Auckland and 6Associate Professor, University of Auckland, Auckland, New Zealand 1 3 Correspondence Nicola North School of Population Health University of Auckland Private Bag 92019 Auckland 1142 New Zealand E-mail: n.north@auckland.ac.nz N O R T H N . , L E U N G W . , A S H T O N T . , R A S M U S S E N E . , H U G H E S F . & F I N L A Y S O N M . (2012) Journal of Nursing Management Nurse turnover in New Zealand: costs and relationships with staffing practises and patient outcomes Aims To determine the rates and costs of nurse turnover, the relationships with staffing practises, and the impacts on outcomes for nurses and patients. Background In the context of nursing shortages, information on the rates and costs of nursing turnover can improve nursing staff management and quality of care. Methods Quantitative and qualitative data were collected prospectively for 12 months. A re-analysis of these data used descriptive statistics and correlational analysis techniques. Results The cost per registered nurse turnover represents half an average salary. The highest costs were related to temporary cover, followed by productivity loss. Both are associated with adverse patient events. Flexible management of nursing resources (staffing below budgeted levels and reliance on temporary cover), and a reliance on new graduates and international recruitment to replace nurses who left, contributed to turnover and costs. Conclusions Nurse turnover is embedded in staffing levels and practises, with costs attributable to both. A culture of turnover was found that is inconsistent with nursing as a knowledge workforce. Implications for nursing management Nurse managers did not challenge flexible staffing practices and high turnover rates. Information on turnover and costs is needed to develop strategies that retain nurses as knowledge-based workers. Keywords: costs, health services research, nurse turnover, staffing practices Accepted for publication: 2 December 2011 Introduction Internationally, there is increasing recognition of the cost of nursing turnover and the impact on nurses and DOI: 10.1111/j.1365-2834.2012.01371.x ª 2012 Blackwell Publishing Ltd patients (Jones 1990a, Hayes et al. 2006, OÕBrien-Pallas et al. 2006). Studies in Canada (OÕBrien-Pallas et al. 2008, 2010), Australia (Duffield et al. 2009) and the USA (Jones 2005, 2008) reflect concerns about managing the 1 N. North et al. significant costs of nurse turnover, its impacts and the widespread use of temporary cover to address staff shortages. Research into nursing turnover in New Zealand took place in a context of long-standing nursing shortages and a tight nursing labour market, following years of deteriorating employment conditions and stagnation of pay (North 2010). A pilot study was conducted in 2001 to test a costing methodology (OÕBrien-Pallas et al. 2006); it established that turnover-related costs were seldom available (North & Hughes 2006). The first substantive study was a national survey in 2002 of directors of nursing in public hospitals, conducted to identify turnover rates and nursing workforce strategies. The survey found that despite concerns about turnover and shortages, actual turnover rates were seldom known (North et al. 2005). A national study of costs of nurse turnover in New Zealand was then conducted in 2004–2006, using a comparable research protocol to a pan-Canadian study (OÕBrien-Pallas et al. 2010), but with a more limited scope. The aims of that study were to determine turnover rates, direct and indirect costs of turnover, relationships between costs of turnover and staffing practices, and relationships between the cost of turnover with outcomes for nurses and patients. Aspects of the study have been reported previously (North & Hughes 2006, North et al. 2007). Recently, the data have been re-analysed, focusing on costs and the practise of staffing below budgeted levels, turnover and the use of temporary cover to manage nurse resources flexibly, and the impacts of staffing practices and costs of turnover on nurses and patients. While the nursing turnover study was being conducted, a national employment agreement covering nurses working in public hospitals led to an average increase in nursesÕ pay of 7%. Buchan & North (2009) found an improvement in most nursing labour market indicators in the 4 years following the employment agreement. From 2008 onwards, in recessionary conditions with high unemployment, shortages have all but disappeared, and nurse staffing practices previously employed in similar economic conditions, and believed to be efficient but that arguably contributed to shortages, have re-emerged. This article examines the role of staffing practices in turnover, and the associated impacts and costs, contributing new insights into the phenomenon of turnover as embedded in nurse management practices and culture. Previous research on costs of nurse turnover Nurse turnover has attracted increasing interest in the context of nursing shortages, an ageing workforce and 2 concerns over workforce stability. Nurse turnover contributes the highest total organizational costs because of workforce size, although cost per nurse turnover was lower than per higher-paid employee (Waldman et al. 2004). Data on rates of nurse turnover and its costs are important to inform health organization strategy and management, yet organizations do not always report such information (Jones 1990a,b, 2004, Waldman et al. 2004). A review of nurse turnover literature (Hayes et al. 2006) revealed a considerable body of research on the determinants of turnover and a growing interest in economic and systems impacts. Methodological challenges hamper comparisons, including inconsistent definitions, inclusions and exclusions, and different costing methodologies. In addition to the direct turnover costs (related to recruitment, hiring and vacancy cover), some methodologies have addressed the indirect costs that capture decreased productivity of new employees and constraints on maintaining service levels (Jones 2005, OÕBrien-Pallas et al. 2010). Some researchers attempt to balance costs against savings, for example related to a vacancy (Buchan & Seccombe 1991), and against retention strategies (Jones 2004). Different approaches to conceptualizing costs and costing methodologies have contributed to widely varying estimates of turnover costs. Two studies in the UK, cited by Gray et al. (1996), estimated the costs of replacing a nurse in 1994–1995 to be £494 and £5998, respectively. Administration costs were modest, but those costs rose twelvefold when temporary replacement costs and lost productivity of new employees were included. Studies cited by Jones (2004) reported cost per nurse turnover in the USA to be between US$22 000 and US$64 000, with the variation being attributed to differing costing methodologies. The highest costs arose from the use of temporary cover for vacancies and shortages, followed by costs of orientation and training and productivity loss (Jones 2004, 2005, 2008). Jones (1990b) costing methodology calculated costs in 1988 as US$10 000 per nurse. In 2002, when productivity loss was added, costs rose to between US$62 100 and US$67 100 (Jones 2005), and when recalculated for inflation the equivalent cost in 2007 was between US$82 032 and US$88 086 (Jones 2008). Jones (2005, 2008) did not account for the costs of impacts of nursing turnover, such as on patients, and subsequently highlighted this as an important avenue for research. OÕBrien-Pallas et al. (2010) did include those impacts in a Canadian study conducted in 2005– 2006, and found that higher turnover rates were associated with an increased likelihood of medical error. ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Nurse turnover in New Zealand The average cost per nurse turnover was CAN$25 000 (OÕBrien-Pallas et al. 2008), with highest costs related to temporary replacement and initial decreased productivity of replacements. Hierarchical linear regression (OÕBrien-Pallas et al. 2010) showed that high turnover rates were associated with a decline in both mental health and job satisfaction among nurses. Recent developments in nurse turnover research conceptualize costs of nursing turnover as human capital costs (Jones 2004), and health systems costs (OÕBrien-Pallas et al. 2006, 2010). Nurses have highvalue knowledge and skills that are related to productivity and human capital theory draws attention to the hidden costs associated with the loss of the human capital, and losing the value of investment in nurses who leave (Jones 2004). Health systems costs arise from closed beds and patient deferrals, and the impacts of turnover on outputs and outcomes. These affect patients (adverse events and satisfaction with care) and; nurses (injury, sickness, job satisfaction and further turnover) (OÕBrien-Pallas et al. 2010). Waldman et al. (2004), Jones & Gates (2007) and Jones (2008) note that as turnover generally represents a non-value added component of an organizational budget, accounting for the costs of turnover would support the case for retention. While the overall costs and impacts of turnover are becoming better understood, and research methodologies more comprehensive, research also needs to focus on the relationships between the costs of turnover and the contexts in which nurses work, including staffing practices. Materials and methods Design Using a prospective design, quantitative and qualitative data were collected. The study protocol was previously pilot tested and conceptualized using OÕBrien-PallasÕ Patient Care System and Nurse Turnover model (OÕBrien-Pallas et al. 2006). The unit of analysis was the nursing unit (ward). Quantitative data were collected over 12 consecutive months per participating unit. Qualitative data describing nursing workforce practices and working contexts were collected at the start and end of the study. Definitions and inclusion criteria Nurse turnover was defined as the process by which nurses leave or transfer from their primary employment position in health services (Jones 1990a). To limit varª 2012 Blackwell Publishing Ltd Journal of Nursing Management iation, only general medical and surgical units in public hospitals were included. Registered nurses (RNs) in staff positions were included; nurse managers and specialists, second-level nurses and unregulated assistants, were excluded. Data were collected on all RNs attached to participating units who left their primary employment position for whatever reason, including retirement, resignation and internal transfer. Costs were based on a modified cost of turnover (direct and indirect) checklist (Buchan & Seccombe 1991), piloted by OÕBrien-Pallas et al. (2006), that disaggregated turnover costs into separately identified items, grouped under four sequential processes: separation; temporary replacement; recruitment and selection; induction and training. Recruitment of units and completion rate Randomized selection from a medical and a surgical pool of units (n = 156) yielded a sample of 22 units (14% of eligible units) spread over 11 of New ZealandÕs 21 district health boards, the owners of public hospitals. The research was approved by Northern Y Regional Ethics Committee (lead committee) and six relevant regional health ethics committees (AKY/03/10/254). Data collection Data were collected in the original study between 2004 and 2006, and included: full-time equivalent (FTE) RN numbers; temporary staffing costs; productivity loss of permanent staff related to temporary staff; overtime costs; costs of preceptors and other costs related to training new RN employees; decreased productivity (estimated in three broad productivity bands) of new employees, until they met the expected productivity for their level; RN absence; work-related nurse incidents (e.g. injuries); and nurse-sensitive adverse patient incidents. Having previously established that turnoverrelated costs were seldom available (North & Hughes 2006), a series of electronic instruments were developed. Hospital employees (principally charge nurses, but also new graduate coordinators, occupational health and safety/infection control staff) were trained in the instruments and supplied data each month. All spreadsheet data sets submitted by hospital personnel were independently verified for accuracy. Qualitative data about organizational contexts and staffing practices were systematically collected by the investigators when implementing the study, and again when provisional results were returned to key informants for verification and feedback. 3 N. North et al. Analysis The purpose of the re-analysis was to identify the relationships between variables related to turnover, costs and staffing practices, and, where possible, to explain results by drawing on qualitative data. Data were extensively checked and entered into SPSS v17 (SPSS, IBM Corporation, Armonk, NY, USA). Results were analysed both separately for each unit, and nationally. To calculate the turnover rate, the FTE annual number of leavers was divided by the annual actual FTE · 100. Quantitative data were analysed using descriptive statistics and correlational analysis techniques. Throughput variables tested included: nurse staffing practices, deviation below budgeted RN levels, use of overtime and temporary staff, turnover rate and productivity loss. Output variables included patient adverse events, nurse incidents and sick leave. Variables were adjusted for resourced beds where applicable. SpearmanÕs rank order correlation was used when the assumptions required for PearsonÕs product-moment correlation were violated. All tests were two-tailed. Qualitative data were analysed thematically to augment and explain quantitative results. Results At the end of the study 19 of 22 wards had completed 12 monthsÕ data collection, representing a response rate of 86.4%, producing 228 data samples for analysis. Quantitative data are reported in Tables and numerically, and qualitative explanations reported textually. All costs are reported in 2006 New Zealand dollars. Direct and indirect turnover costs Table 1 (based on OÕBrien-Pallas et al. 2006) sets out the direct and indirect costs associated with RN turnover. Costs are reported per nurse turnover, distinguishing between costs related to new employees and leavers. The estimated average total cost per RN turnover is $23 800, comprising $3 878 per new employee and $19 922 per leaver. Qualitative data allow us to explain some costs. Termination costs, including management time and leaving rituals, are minimal because in New Zealand special payments such as long-service bonuses are not offered and unused sick leave is not paid out. Recruitment and hiring costs, including management time, uniform supply and health screening, are also low and relocation allowances seldom covered. Costs associated with orientation and training per new employee are higher and include organizational orientation, unit orientation, short courses (for credentialing) and trainer costs. The highest costs were associated with temporary cover and new employees (including loss of productivity and preceptor costs). These costs were also associated with staffing practices, explained further below. Costs associated with productivity loss Table 2 shows the national average cost per category of new employee. Costs included are those associated with new employee productivity loss and preceptor time, until expected productivity for the level of appointment is reached, and those of recruitment and hiring. These figures also indicate that 83% of replacement nurses are recruited from outside the organization (the combined total of new graduates, overseas-trained and recruits from elsewhere in New Zealand). Orientation and training of new employees make up the highest indirect turnover cost and include costs of lost productivity until 100% productivity for their level is reached, and preceptor-related productivity loss. These costs vary considerably according to the category of new employee, reflecting nursing experience and familiarity with the organization, with the highest productivity losses incurred by the two categories most frequently recruited to replace nurses who have left: new graduate nurses and overseas-trained nurses. The Table 1 Direct and indirect registered nurse (RN) turnover costs (rounded, in NZ$) Per new employee variables Direct costs Indirect costs Total costs 4 Advertising/recruitment Hiring process Subtotal Preceptor costs Decreased initial productivity of new employee Other orientation/training Subtotal Per new employee Per leaver variables 236 499 735 1518 1447 177 3142 3878 Totals per turnover Temporary cover costs 19 730 Subtotal Termination and separation 19 730 192 – Total direct costs 192 19 922 Total indirect costs Per RN turnover Subtotal Per leaver 20 465 – 3334 23 800 ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Nurse turnover in New Zealand Table 2 National average training and orientation costs for a new employee by category to reach 100% productivity and total cost to join New employee Rank category 1 2 3 4 5 * New graduate Overseas trained NZ trained – external to hospital Internal transfer Return to ward Return to nursing Totals/averages Number % 107 62 51 32 8 5 265 40.4 23.4 19.2 12.1 3.0 1.9 100.0 Costs to reach 100% Total cost per new employee to join (NZ$) productivity (NZ$) 4277 3426 1978 1059 941 4055 3142 4804 4467 3019 1711 941 4055 3878 *Numbers too low to rank. highest cost was for taking on a new graduate at $4804, with overseas-trained nurses nearly as expensive at $4467. New Zealand-trained nurses (recruited by the hospital from elsewhere) were next at $3019. The least expensive options were a nurse returning to a unit, at $941, and internal transfer, at $1711. These appointments avoided costs associated with relocation such as uniform supply, health screening and organizational orientation. However, an internal transfer may simply shift those turnover costs. Temporary cover costs and staffing practices The main contributor to turnover costs was temporary cover, incorporating costs of temporary staff, overtime, clerical time and time of experienced staff to arrange cover, and productivity loss related to permanent staff assisting temporary staff. Temporary cover is used not only to cover a vacancy until it is filled, but also to cover existing shortages, sick leave, increased demand and shortages for other reasons. It was not possible to determine the amount of temporary cover costs related only to turnover, as the purpose for temporary cover related to specific requests was not recorded. The data illustrate the practise of staffing at levels below budget, a practise widely used to maximize the flexible and efficient use of nursing resources and budget. The average number of budgeted RN FTE was close to the average number of resourced beds. There was an average of 25.57 resourced beds per unit (range 19.08– 47.33). The mean occupancy rate was 91.8%. The average budgeted RN staffing per unit was 25.84 FTE but as shown, in Table 3, the actual FTE was 24.07, a mean difference of )1.77 (range )9.7 to 4.7). Baseline data showed 497.12 RN FTE were budgeted in the 19 units and 450.91 RN FTE employed, giving an average of 2.43 RN vacancies per unit. During the 12-month data collection period, 192.6 RN FTE left their primary unit of employment, replaced by 265 new RN FTE. There were no new positions established during the study period, so some unfilled vacancies at ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Table 3 The average budgeted and actual full-time equivalent (FTE) FTE Mean Range Budgeted Actual Difference between budgeted and actual 25.84 24.07 )1.77 (13.93–46.2) (11.1–42.1) ()9.7 to 4.7) the start of data collection were filled during the period. The average annual turnover rate for the 19 participating units was 44.3% (range 13.7–90.9%). Significant relationships between variables Relationships between variables related to staffing practices and the annual cost of turnover, with outcomes for nurses and patients, were tested using SpearmanÕs rank order correlation (Table 4). Nurse outcomes were measured as turnover, sick leave in days and adverse incidents such as injury. Reported sick leave days per year ranged from 70 to 603 per unit, an average of 266 per unit. There were a total of 210 nurse injuries, an average of 11 per ward (range 1–49 per year). Patient outcomes were measured as nurse-sensitive events; the total number of patient adverse events reported was 993, an average of 52.6 per ward. The largest of these was falls with injury (319) followed by medication administration errors (222). Non-percentage variables were divided by the number of resourced beds in each unit to take account of the different size of units in the study. Table 4 shows that sick leave days are positively and significantly correlated with the cost of temporary cover (28% shared variance). The amount spent on temporary cover is correlated with the total patient adverse outcomes (27% shared variance). Adverse patient events are also correlated with new employee productivity loss (31% shared variance). Finally, the turnover rate is negatively and significantly correlated with the percentage deviation from budgeted FTE (34% shared variance). The other 12 relationships were not significant. 5 N. North et al. Per cent deviation from budgeted FTE Overtime spent* Temporary cover spent* New employee productivity loss (including preceptor costs)* Nurse incidents* Turnover as % of actual FTE Sick leave* Patient adverse events* 0.098 )0.060 )0.027 0.041 )0.586 0.096 0.033 0.323 )0.147 )0.040 0.526à 0.398 )0.233 )0.232 0.516à 0.558à Table 4 SpearmanÕs rank order correlation between measures of cost of turnover/staffing practices and nurse/ patient outcomes FTE, full-time equivalent. *Per resourced bed. Sample size n = 19. Correlation is significant at the 0.01 level (two-tailed). à Correlation is significant at the 0.05 level (two-tailed). Discussion Turnover rates Nurse turnover rates were 44.3% – more than double that of Canada (19.9%; OÕBrien-Pallas et al. 2010) and the USA (19.2%; Jones 2005). The turnover rate cannot be generalized beyond the sample (14% of general medical and surgical units in public hospitals in a 12-month period). New graduate nurses are included in unit staffing and substantially contribute to both staffing and turnover. Nurse managers displayed an indifference to turnover, conveying an acceptance and tolerance of high turnover rates. Their view of turnover as normative reflected a Ôturnover cultureÕ, reported in relation to low-skilled workers in the hospitality industry (Davidson et al. 2010). Costs The New Zealand study adds to the evidence internationally that nurse turnover represents a substantial cost to hospitals which, if reduced, would potentially free up resources for the delivery of health care. By using a similar turnover costing methodology as in the UK (Buchan & Seccombe 1991), Canada (OÕBrien-Pallas et al. 2008) and the USA (Jones 1990a), the relative contribution of types of cost to total costs can be compared. As in those studies, temporary cover costs represented the highest costs followed by orientation/ training and productivity loss. Orientation and training costs were low in New Zealand compared with a study in the USA (Waldman et al. 2004), reflecting differences between countries in investment in on-job training. Comparing the monetary value of turnover costs across countries and health systems is more problematic because contexts, such as health systems, currencies and economies, affecting salaries and costs of living, differ. Jones (2004) and OÕBrien-Pallas et al. (2008, 2010) reported a cost per nurse turnover of about $22 000 in the local currency (USA and Canadian dollars, respec6 tively). While these costs are similar to New Zealand figures ($23 800) in absolute dollar terms, it is more meaningful to compare countries by calculating the cost of turnover as a proportion of RN salaries. However, few studies report costs in this way. In five studies cited by Jones (2004) the ratio of turnover against salary was calculated as between 0.37 and 1.6. The mean cost in New Zealand of one RN turnover was about half an annual RN salary (in 2005 an average base salary was $47 000). This suggests that for every two turnovers avoided, shortages could be reduced by one RN. However, because information on costs of turnover was unknown to nurse managers, it could not be used strategically to retain nurses or reduce shortages. Because data on nurse turnover and its costs were not routinely and systematically collected by public hospitals, data had to be collected from multiple sources. This introduces the possibility that data may have been defined differently, giving rise to possible variation. Jones (1990a, 2004) and OÕBrien-Pallas et al. (2010) reported similar problems in the USA and Canada. As the research proceeded, we concluded that although a validated costing methodology was used consistently, some of the costs captured were related to staffing practices as well as turnover. Nurse staffing practices Two staffing practices contributing to costs of turnover were: the flexible use of the nursing resource (staffing units below budgeted levels and reliance on temporary cover); and the reliance on new graduates and international recruitment to fill vacancies. Flexible staffing was used strategically by the organization to manage nurse shortages, and to reduce fixed labour costs. The results showed that, while this practise may appear to be an efficient way of controlling staffing costs, there are indirect costs to the organization associated with the costs of turnover, temporary cover and adverse patient events (North et al. 2007). Employing new graduates ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Nurse turnover in New Zealand on starting salary levels was attractive to cashrestrained organizations. International recruitment highlights the severe recruitment difficulties that prevailed (Zurn & Dumont 2008, North 2010). Staffing practices intended to be efficient and offering maximum flexibility may have the opposite effect. Staffing below budgeted FTEs is attractive as a management tool, as it is easy to monitor and control the number of FTEs. However, the resulting increase in costs from increased turnover and increased adverse patient events makes it likely that this is the more costly option. The analysis found a significant association between nurse turnover and deviation of actual FTE below budgeted FTE. Turnover is one reason for the use of temporary cover, but although no association was found between rates of turnover and costs of temporary cover, there was a strong and significant relationship between the costs of temporary cover and increased sick leave. A UK study also found a positive association between the use of temporary staff and sick leave (Hurst & Smith 2010). The association between temporary cover costs and sick leave suggests possible relationships among staffing practices, work stress and sick leave. Both costs of temporary cover and costs associated with productivity loss of new employees were significantly associated with adverse patient incidents. Patient falls with injury, followed by medication error, were the highest adverse events reported in New Zealand. A study in the USA also found positive associations at unit level between temporary RN staffing (at 15% or greater than total hours) and patient falls and medication error (Bae et al. 2010). Aiken & Xue (2007) disagree that use of temporary nurses compromises quality of care; the more important variable was the level of RN involvement in direct patient care. In our study, however, costs of RN temporary cover were associated with a greater risk of adverse patient outcomes. With temporary cover shown in New Zealand, and in other studies, to be more costly and less safe than permanent cover (OÕBrien-Pallas et al. 2008, 2010, Bae et al. 2010), it would be more cost-effective and safer to employ up to the budgeted level. Budgeted FTEs already incorporate annual leave, sick leave and other leave entitlements such as bereavement and study leave. Temporary cover would then be used only to cover vacancies arising from turnover and unplanned contingencies, including parental leave, major illness and spike in demand other than expected seasonal flows. Of particular concern in the New Zealand study was that experienced nurses who left the participating units ª 2012 Blackwell Publishing Ltd Journal of Nursing Management were being replaced largely by nurses with least experience in nursing (new graduates) or who were unfamiliar with the New Zealand context of nursing (overseastrained nurses). Combined, these two categories made up nearly two-thirds of all new employees (63.8%). In addition, these two categories cost organizations the most in lost productivity. Costs of productivity loss were significantly associated with adverse patient events, an added cost to the organization, as well as to the patients themselves. Other researchers, including Waldman et al. (2004) and Jones (2005, 2008), have also attributed the highest turnover costs to lost productivity. Costs of staffing practices or costs of turnover? Although a comparable costing methodology was used as for other studies (Buchan & Seccombe 1991, Jones 2005, OÕBrien-Pallas et al. 2010), and some costs were clearly linked to turnover, other costs were also related to staffing practices. The combined costs associated with termination and separation of leavers, and recruitment and hiring of replacements, contribute to <4% of per nurse turnover costs, and are unequivocally related to turnover. The cost of lost productivity contributed another 13% to per nurse turnover costs, but in the case of new graduates it could be argued that preceptors are professional development costs related to entry to practise and not only to turnover. If new graduates were supernumerary (Cardona & Bernreuter 1996), then productivity costs related to turnover could be differentiated, but in this study new graduates filled existing vacancies and so were also related to turnover. The gap between budgeted and actual FTEs suggests that the high costs associated with temporary cover cannot be attributed solely to turnover. Temporary cover is also associated with flexible management of the nursing resource. Temporary cover was one of the few cost categories where data were both readily available and accurate, but temporary cover was commonly used to manage flexible staffing practices as well as vacancies and other absences. Because the reasons for temporary cover requests were not recorded, we could not determine the proportion to cover a vacancy, or to cover a nurse shortfall for other reasons such as sick leave or increased demand. Cost savings from vacancies theoretically offset temporary cover costs (Buchan & Seccombe 1991), but costing methodologies do not consistently allow for this. Jones (2004) defines vacancy costs as Ôcosts incurred while attempting to fill open positions that result from RN turnover and staff shortagesÕ. We also included all temporary cover costs in our analysis. Our correlational analysis supported participantsÕ feedback, showing a significant 7 N. North et al. association between sick leave and temporary cover, but not between temporary cover and turnover rates. Correlations also showed that the use of temporary cover was significantly associated with adverse patient events. The study highlighted that nursing workforces at unit level, characterized by churn (Duffield et al. 2009), are unstable. We found widespread nurse shortages, reflecting intentional staffing of nursing units below budgeted levels, and a reliance on temporary cover to meet shortfalls. There were high rates and tolerance of turnover, and a reliance on new graduates and international recruitment to fill vacancies. Our data do not allow us to claim causal relationships, but we suggest the following possible mechanisms (based on data reported and discussed above). The turnover rate is significantly and negatively correlated with the percentage deviation from budgeted FTE. This may be because shortages and managing staff levels below budgeted FTE contribute to stress, sickness, low job satisfaction and morale, and thus high turnover, or because high turnover and difficulties in recruiting staff compound shortages. Temporary cover costs, which are associated strongly with sick leave cover, also contribute to sick leave. Permanent staff are under pressure to cover staff shortages, work overtime and support temporary staff unfamiliar with the unit and its patient population. The costs of temporary cover are also significantly and positively associated with adverse patient events, which are likely to have an impact on nurse morale and so contribute to turnover. Costs of new employee productivity directly arising from turnover are also significantly and positively associated with adverse patient events. Conclusion and implications for nursing management This study supports the growing international body of evidence that nurse turnover represents a non-value added component of health organization budgets. The results also lend weight to an emergent focus of turnover research on the impacts of turnover on patients, nurses and health systems. As in earlier research, the two highest costs (temporary cover and new employee productivity losses) are both associated with adverse patient events. The present analysis goes further by highlighting the complexities of accurately distinguishing between those costs that are specific to turnover, and those that reflect accepted staffing practices, even when a robust costing methodology is used. In the NZ study, there were fewer RNs than budgeted in most units throughout the 12-month data collection period. However, whether 8 these represented unfilled vacancies and shortages was disputed. In addition to covering vacancies, temporary cover was used strategically to balance demand and supply of nurses, and included sick leave and other temporary absences that the budgeted FTE is expected to cover. This is also one of the few studies to report a positive correlation between recruitment patterns (where the largest proportion of hires are new graduates), and costs of productivity loss, and between the latter and adverse patient events. Costs of productivity loss also reflect costs of professional development. Our experiences illuminate the challenges of nurse turnover research, as the phenomenon of turnover is so intertwined with staffing practices and culture. An unsettling conclusion of our study is evidence of a Ôculture of turnoverÕ, which is found in other industries in relation to low-skilled workers (Davidson et al. 2010). Human resource specialists estimate turnover costs to be in the range of 50–150% of annual salary (De Cieri & Kramar 2004). In our study, the estimated cost of replacing an RN at about half an annual salary was at the low end and suggests that hospitals and health services are geared for high nurse turnover rates. A culture of turnover, and staffing less than budgeted levels, reflects an accepted view of a nurse as a replaceable unit of labour. Furthermore, a culture of turnover is inconsistent with nursesÕ high knowledge value and indicates a lack of recognition of the intangible costs to organizations when employees who work with knowledge and information leave (Jones 2004). There was strong evidence that nurse managers and leaders lacked information on turnover rates and costs, and so did not challenge the position of trying to deliver patient care safely by covering vacancies, shortages and sick leave with temporary cover, and continually recruiting new staff. Although turnover was regarded as normative, unit managers were surprised to learn that nearly half, and up to ninety percent of their staff nurses had left during the year. Our analysis has important implications for nursing management. For nurse managers, the results challenge beliefs that efficient budget management involves intentional understaffing to reduce fixed costs, and using temporary cover to increase flexibility. Moreover, while filling vacancies with lower-salaried new graduates may be cheaper, we have shown that the practise actually costs more. In order to challenge the assumed benefits of flexible staffing and to develop effective and sustainable staffing practices and strategies, nurse managers require information. As a minimum, regular and accurate information on nurse turnover rates, the costs of recruiting and training replacement nurses and expenditure on temporary cover should be ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Nurse turnover in New Zealand provided by hospital management. Costing methodologies, as used in this and similar studies, can be adapted for the purposes of collecting the minimum dataset. The analysis also has implications for nursing leadership, if a return to declining employment conditions, low morale and shortages in nursing (North 2010) is to be avoided. There were severe and persistent nurse shortages in New Zealand for two decades, and in tight economic times the pressure is again on the health workforce to be more cost-effective and productive. Measures such as nurse substitution and unfilled vacancies have re-emerged. It is therefore important that we understand the relationships between FTEs, nursesÕ health, adverse events, service provision and turnover. Leadership is needed to refocus strategies from cost savings to investing in staff, and to nurture cultures and working environments that retain nurses, as workers whose knowledge is an asset to the organization. Source of funding This research was funded by the Health Research Council in partnership with the Ministry of Health. Participating hospitals contributed the time of employees who collected and submitted data. Ethical approval Northern Y Regional Health Ethics Committee AKY/03/10/ 254. Acknowledgements This paper is part of an international project to examine the cost of nurse turnover and the impact of turnover on patient safety and nurse health and safety outcomes. Co-Principal investigators are Professor Linda O’BrienPallas (University of Toronto, Canada) and Dr. Judith Shamian (Victoria Nurses Association Canada). Team members include the following: Dr. Christine Duffield (University of Technology, Sydney, Australia); Dr. Heather Laschinger (University of Western Ontario, London, Ontario, Canada); Dr. Pat Griffin (Health Canada, Ottawa, Canada); Professor James Buchan (Queen Margaret University College, Edinburgh, UK); Dr. Patricia Stone (Columbia University, New York, NY, USA). References Aiken L. & Xue Y. (2007) Supplemental nurse staffing in hospitals and quality of care. Journal of Nursing Administration 37 (7/8), 335–342. ª 2012 Blackwell Publishing Ltd Journal of Nursing Management Bae S.-H., Mark B. & Fried B. (2010) Use of temporary nurses and nurse and patient safety outcomes in acute care hospital units. Health Care Management Review 35 (3), 333– 344. Buchan J. & North N. (2009) Evaluating the impact of a new pay agreement on New Zealand nursing. International Nursing Review 56 (2), 206–213. Buchan J. & Seccombe I. (1991) Nurse turnover costs. Health Manpower Management 17 (4), 23–25. Cardona S.M. & Bernreuter M. (1996) Graduate nurse overhires. A cost analysis. Journal of Nursing Administration 26 (3), 10–15. Davidson M, Timo N & Wang Y. (2010) How much does labour turnover cost? A case study of four- and five- star hotels International Journal of Contemporary Hospitality Management 22 (4), 451–466. De Cieri H. & Kramar R. (2004) Human Resource Management in Australia. McGraw-Hill, Sydney. Duffield C., Roche M., OÕBrien-Pallas L. & Catling-Paull C. (2009) Implications of staff ÔchurnÕ for nurse managers, staff, and patients. Nursing Economic$ 27 (2), 103–110. Gray A.M., Phillips V.L. & Normand C. (1996) The costs of nursing turnover: evidence from the British National Health Service. Health Policy 38, 117–128. Hayes L.J., OÕBrien-Pallas L., Duffield C. et al. (2006) Nurse turnover: a literature review. International Journal of Nursing Studies 43 (2), 237–263. Hurst K & Smith A. (2010) Temporary nursing staff-cost and quality issues. Journal of Advanced Nursing 67 (2), 287– 296. Jones C.B. (1990a) Staff nurse turnover costs: part I: a conceptual model. Journal of Nursing Administration 20 (4), 18–23. Jones C.B. (1990b) Staff nurse turnover costs: part II: measurement and results. Journal of Nursing Administration 20 (5), 27–31. Jones C.B. (2004) The costs of nursing turnover. Part 1: an economic perspective. Journal of Nursing Administration 34 (12), 562–570. Jones C.B. (2005) The costs of nurse turnover, part 2: application of the nursing turnover cost calculation methodology. Journal of Nursing Administration 35 (1), 41–49. Jones C.B. (2008) Revisiting nurse turnover costs: adjusting for inflation. Journal of Nursing Administration 38 (1), 11–18. Jones C.B. & Gates M. (2007) The costs and benefits of nursing turnover: a business case for nurse retention. Online Journal of Issues in Nursing 12 (3), Manuscript 2. Available at: http:// www.nursingworld.org/MainMenuCategories/ANAMarketplace/ ANAPeriodicals/OJIN/TableofContents/Volume122007/No3Sept 07/NurseRetention.aspx, accessed 29 June 2011. North N (2010) Can New Zealand achieve self-sufficiency in its nursing workforce?. Journal of Clinical Nursing 20 (12), 40–49. North N. & Hughes F. (2006) Methodological challenges to researching nursing turnover in New Zealand: a progress report of a national study. Asia Pacific Journal of Health Management 1 (1), 45–51. North N., Rasmussen E., Hughes F. et al. (2005) Turnover amongst nurses in New ZealandÕs district health boards: a national survey of nursing turnover and turnover costsÕ. New Zealand Journal of Employment Relations 30 (1), 49–62. North N., Rasmussen E., Hughes F. et al. (2007) Use of temporary nurse mechanisms by New ZealandÕs District Health Boards. New Zealand Journal of Employment Relations 31 (3), 36–50. 9 N. North et al. OÕBrien-Pallas L., Griffin P., Shamian J. et al. (2006) The impact of nursing turnover on patient, nurse and system outcomes: a pilot study and focus for a multi-center study. Policy, Politics and Nursing Practice 7 (3), 169–179. OÕBrien-Pallas L., Tomlin Murphy G. & Shamian J. et al. (2008) Understanding the Costs and Outcomes of NurseÕs Turnover in Canadian Hospitals (Nursing Turnover Study). Canadian Institutes of Health Research, Ottawa. Available at: http:// www.hhrchair.ca/images/CMSImages/TOS_Final%20Report. pdf, accessed 26 May 2011. 10 OÕBrien-Pallas L., Murphy G.T., Shamian J., Li X. & Hayes L.J. (2010) Impact and determinants of nurse turnover: a pan Canadian study. Journal of Nursing Management 18, 1073– 1086. Waldman J.D., Kelly F., Arora S. & Smith H.L. (2004) The shocking cost of turnover in health care. Health Care Management Review 29 (1), 2–7. Zurn P. & Dumont J.-C. (2008) Health Workforce and International Migration: Can New Zealand Compete? OECD Health Working Paper No. 33. OECD and WHO, Paris. ª 2012 Blackwell Publishing Ltd Journal of Nursing Management