Occup Environ Med 69:367-372 doi:10.1136/oemed-2011-100107
  • Methodology
  • Original article

Evaluation of direct workers' compensation costs for musculoskeletal injuries surrounding interventions to reduce patient lifting

  1. J M Dement1
  1. 1Division of Occupational and Environmental Medicine, Duke University Medical Center, Durham, North Carolina, USA
  2. 2Department of Epidemiology, University of Texas School of Public Health, Houston, Texas, USA
  1. Correspondence to Dr Hester J Lipscomb, Division of Occupational and Environmental Medicine, PO Box 3834, Duke University Medical Center, Durham, NC 27710, USA; hester.lipscomb{at}
  1. Contributors HL was the principal investigator on this evaluation effort. She contributed to study design and oversight, and conducted the analyses reported here. AS served as the project manager for this evaluation project. She contributed to conceptually framing the research approach, review of analytical findings, and writing and editing of the manuscript, as did DM, LP and JD.

  • Accepted 24 August 2011
  • Published Online First 23 December 2011


Objectives We evaluated costs for workers' compensation (WC) injuries of a musculoskeletal (MS) nature in a large tertiary care hospital and an affiliated community hospital in the 13 years surrounding an institution-wide shift to a ‘minimal manual patient-lifting environment’ supported with inpatient mechanical lift equipment.

Methods Negative binomial regression was used to model adjusted and discounted payment rates based on full-time equivalents (FTEs), and payment ratios. The risk of higher cost was assessed based on type of injury (patient-handling vs non-patient-handling), hospital, job, age, gender, institutional tenure and time since the implementation of lift equipment. Lagging was used to evaluate the latency of the intervention effect.

Results Patient-handling injuries (n=1543) were responsible for 72% of MS injuries and 53% of compensation costs among patient care staff. Mean costs per claim were 5 times higher for those over age 45 than those <25 years of age. Physical and occupational therapy aides had the highest cost rates ($578/FTE) followed by nursing aides ($347/FTE) and patient transporters ($185/FTE). There was an immediate, marked decline in mean costs per claim and costs per FTE following the policy change and delivery of lift equipment.

Conclusions The observed patterns of changes in cost likely reflect the effects of activities other than use of lift equipment, including targeted efforts to close WC claims and an almost simultaneous policy that shifted cost responsibility to the budgets of managers on individual units. Inference was facilitated through the use of longitudinal data on the workgroups and an internal injury comparison.

What this paper adds

  • Costs of occupational injury related to patient handling are high in the healthcare industry and they provide a potential measure of intervention impact.

  • These analyses provide an example of the use of longitudinal data with internal comparisons of different injury types in evaluation of efforts to reduce costs associated with work-related patient-handling injuries.

  • Cost data can provide a useful measure of safety program effectiveness by jointly incorporating incidence and severity measures to capture resource use.

  • While there was an immediate and marked decline in costs per claim and cost rates per full-time equivalent following the shift to a minimal manual lifting environment, the observed patterns likely reflect the effect of administrative activities to control costs rather than use of lift equipment.

  • These analyses demonstrate the importance of establishing a conceptual framework for evaluation of workplace safety interventions that considers context and reasonable latency before intervention effects could be expected.


The costs of occupational injury in the healthcare sector are high and vary by both occupation and industry.1 Musculoskeletal (MS) injuries, specifically patient-handling injuries, make significant contributions to the burden of occupational injury among aides and nurses on hospital units as well as some smaller work groups such as radiology technologists, physical and occupational therapists and patient transporters.1–6

The admission practices of hospitals have changed in the last 2 decades, in effect raising the bar for patient admission. Data presented in the US National Hospital Discharge Survey show a steady decline in the national rate of hospital admissions, culminating in a drop of 29% between 1980 and 1994.7 As inpatient populations in acute care facilities have become correspondingly sicker, frailer and more obese, the tasks of patient care have become more challenging.

As they reflect both injury frequency and injury severity, workers' compensation (WC) costs provide an estimate of economic burden that is not discernable in injury-rate data alone, thus providing an additional measure to use in assessing the effects of safety interventions.8 However, WC coverage for large groups of workers can be difficult to access, particularly over an extended period of time. Furthermore, WC data alone typically fail to provide information on the population at risk, making it impossible to establish cost rates based on time at work.

Using a robust occupational surveillance system for healthcare workers at a large university health system, we report on analyses of WC costs related to MS injury over a 13-year period surrounding efforts designed to specifically reduce patient-handling injuries.9 These cost analyses are part of a comprehensive evaluation of these efforts.


Setting and intervention

We evaluated the direct costs of MS injuries reported through WC among inpatient patient care staff at a large tertiary care medical centre and an affiliated community hospital. Analyses were conducted using data from 1997 through 2009, a time surrounding the initiation of a policy that emphasised minimal manual lifting of patients. Supported by the Ergonomics Division of the Occupational and Environmental Safety Office (OESO), nursing administration at the medical centre instituted a ‘minimal manual lift environment’ (MMLE) policy on inpatient units in 2004. Mechanical lift equipment was gradually placed on inpatient nursing units (paediatric units and maternity initially excluded) at the medical centre over 8 months. The community hospital adopted the policy in January 2005. A train-the-trainer approach was used, with ‘lift coaches’ on each nursing unit who were responsible for training co-workers in the safe use of the equipment and championing its use. A detailed description of the intervention program has been previously reported.10

Ancillary activities

In the years surrounding this intervention, a number of actions were initiated by the workforce, OESO, the nursing administration and the WC office that have potential relevance to this evaluation effort. In 2001, a nursing ergonomics committee was formed at the medical centre with OESO ergonomists in response to concerns about patient-handling injuries. The use of lower technology interventions, such as slip-resistant liners and gait belts, was encouraged. In 2004, the hospital administration shifted fiscal responsibility for lost time work-related injuries and temporary replacement staff from the hospital's WC office to the budgets of individual nursing unit managers. This change was implemented for the purpose of encouraging mangers to facilitate the return to work of injured employees even if restrictions were needed. The WC office focused efforts on more rapid return to work and earlier closure of claims. Employees were also required to report work injuries within 24 h using an online system.

Data sources

Data for these analyses came from the Duke Health and Safety Surveillance System; details of the system have been previously reported.9 The surveillance system uses data from multiple sources on the health system workforce; measures allow linkage of data on an individual basis without the use of personal identifiers. Data were available for the tertiary care hospital from 1997 to 2009 and for the community hospital from 2000 to 2009 when the latter became a part of the larger health system.

Human resources data were used to select patient caregivers working in patient care areas of interest at these two hospitals, as the study population. Reported work-related MS injuries were identified through nature of injury codes including sprain/strain, dislocation, twist, carpal tunnel syndrome, pain and inflammation. A subset of patient-handling injuries was identified from these WC records as described earlier by Pompeii et al, using claims that had ‘patient’ coded as an agency of injury; confirmation was made through review of limited text information on the circumstance surrounding injury.3 The primary outcome of interest was direct payment made for treatment and indemnity (paid lost time) for these WC claims. Time at risk, used in rate analyses, was defined based on estimated full-time equivalents (FTEs).

Analytical approach

Framework for evaluation

The MMLE intervention could potentially reduce costs in several ways. First, the intervention could reduce the risk of experiencing a patient-handling injury among staff with equipment access. The intervention could also reduce the severity (measured by costs for medical care and indemnity) of injuries and facilitate return to work for injured workers, even if the risk of injury did not change. These latter effects would be reflected in the reduction in costs per claim filed among the injured. Other factors, such as the administrative changes described above, could also influence costs through changes in reporting practices and lost time cost responsibilities. Lastly, there was no expectation that the MMLE policy change and equipment could have an immediate effect.

For these reasons we conducted analyses designed to evaluate and contrast cost patterns for different types of claims and for different groups of workers in different time windows before and after the policy change and equipment placement. First, costs for patient-handling MS injuries were compared to those of MS injuries that were not related to patient handling; factors that influence reporting practices and WC management practices should apply equally to both of these sets of events. Second, costs for patient-handling injuries were contrasted between personnel with access to lift equipment and groups without lift equipment (nursing units prior to phase in, radiology, transporters); effects of lift equipment should only be seen among those with access to equipment, while institutional changes in how claims were handled would affect all workers. Third, costs were stratified by hospital. The community hospital did not receive lift equipment until after the medical centre, occupational medicine services were provided through a community provider initially, rather than the health system, and the baseline WC rates and lost time patterns of the two hospitals differed.

Adjustment for inflation and discounting

Our cost data spanned 1997–2009. To compare costs incurred during different years and express them in constant dollars, we adjusted them for inflation and discounted them. We used the Consumer Price Index (CPI) for the nation, adjusting costs (medical, indemnity, impairment) for inflation to the year 2009. Medical costs were adjusted using the CPI for medical care. Those adjusted costs were then discounted by 3% per year to account for changes in the time value of money over the study period. These procedures account for differences in the values of services received or payments made at different time periods, yielding the present value of each cost stream or grouping thereof.11 These two adjustments resulted in all costs being expressed in constant dollars as of the year 2009. All WC costs were assigned to the year in which the injury occurred. In the event a WC claim was still open, projected claim reserve costs were used.


Costs were examined for medical care and indemnity (no impairment costs were awarded for any of the claims). Mean and median costs (payments) per claim were calculated by year, type of injury, hospital, job, gender and institutional tenure. Costs and FTEs were then further stratified by time before and after the lift equipment was placed on each unit. Events and time at risk were allowed to accumulate in the appropriate strata over time as age, tenure and time since intervention changed; in a few cases job and hospital also changed over the evaluation period.

The stratified data were examined using negative binomial regression to model payment rates and payment rate ratios based on FTEs (using the log of FTEs as an offset). Generalised linear models such as these have been used in analyses of medical care costs to accommodate factors that are typical of cost data. These factors include highly skewed distributions, and variability that often increases as mean costs increase rather than having a constant variance (homoscedasticity).12 All analyses were conducted using SAS V.8.2.13

All the variables of interest were included in initial models. Using a backwards elimination process, insignificant variables (based on type III likelihood ratio statistics for these class variables) which did not confound other parameter estimates were removed one at a time from the models. Under this process, no variables with a p value of <0.10 were removed. Stratified models were constructed by type of injury and time since initiation of the intervention. These analyses allowed us to assess and compare cost rates based on type of injury, age, gender, tenure and time since the intervention and in lagged windows of time afterwards in each institution. This lagging process is illustrated in figure 1 which demonstrates lagging at 6 and 12 months for a unit that received the equipment at the initial roll-out in 2004. Lagging assumes that staff need time in order to learn to use the equipment and assimilate its use into their practices of patient care.

Figure 1

Illustration of the evaluation of latency assuming different time periods for the intervention to have an effect.


Between 1997 and 2009, a total of 2156 MS injuries were filed and accepted through WC at the two hospitals, of which 71.6% (n=1543) were related to patient handling. Over 80% (80.5%; n=1736) of the injuries were among nursing staff (nurses and aides) including 82.4% of patient-handling injuries and 75.7% of non-patient-handling injuries. Over 21% (21.5%; n=462) were considered first aid only cases and 10% resulted in paid indemnity (n=215) which occurred after the 7th lost work day. The remaining 68.5% of cases (n=1472) were classified as medical only; medical costs beyond the institutional occupational medical services were incurred for a total of 805 claims. (Note: seven claims were unclassified.)

Direct WC costs for all MS injuries totalled $10 676 535. Costs per claim ranged from 0 to $2 544 803, with mean costs of $4953 per claim. Indemnity costs ($5 629 497; 52.7%) were slightly greater than medical costs ($5 047 037; 47.3%); no permanent impairment costs were awarded. Total costs for MS patient-handling injuries ($5 626 736; 52.7%) were slightly higher than those for non-patient-handling MS injuries ($5 049 799; 47.9%) and the number of patient-handling claims was considerably higher. Although there was some variability over time, non-patient-handling MS injuries tended to result in higher mean costs per claim than those resulting from patient handling. Among the injured, mean costs per claim also varied over time at both institutions (figure 2).

Figure 2

Mean costs of musculoskeletal injuries (patient-handling and non-patient-handling) over time by institution. All dollars adjusted to 2009 value and discounted 3% per year. Years 1997–1999 represent only the medical center; community hospital joined health system in 2000.

Mean costs per claim are presented by type of injury, institution, job, gender, age and tenure in table 1. Mean costs increased with increasing age, such that individuals over 45 years of age had mean costs five times greater than those under age 25; mean costs among injured workers also increased with increasing tenure. Mean costs were highest among patient transporters, physical/occupational therapists and nursing aides.

Table 1

Total, mean and median costs per claim in US dollars, for musculoskeletal injuries, 1997–2009

During this period (1997–2009) we observed 28 446 FTEs among staff involved in patient care activities. This translates to cost rates of $197/FTE for patient-handling and $178/FTE for non-patient-handling MS injury claims. Cost rates (expressed per FTE) are presented separately for patient-handling MS injuries (table 2) and non-patient-handling MS injuries (table 2) by hospital, job, gender, age and tenure. Physical/occupational therapy aides and nursing aides had the highest adjusted costs per FTE for patient-handling injuries followed by patient transporters. Adjusted cost rates per FTE for non-patient-handling MS injuries were highest among patient transporters, nurses and nursing aides, but these estimates are considerably less precise than those for patient-handling MS injuries. Not surprisingly, managers had particularly low cost rates per FTE for patient-handling MS injuries and non-patient-handling MS injuries. Individuals with 1 to <5 years of institutional tenure had the highest cost rates for patient-handling MS injuries, but the pattern was not the same for non-patient-handling MS injuries, where the most tenured workers had the highest rates. Hospital was not a significant predictor of cost rates (p>0.40) for either type of injury after accounting for job and tenure.

Table 2

Cost rates in US dollars per full-time equivalent (FTE) and cost rate ratios of musculoskeletal injuries, 1997–2009*

In figure 3, cost rate ratios (per FTE) before and after the intervention are depicted with no lagging and lagging at 6-month intervals up to 18 months following the intervention. These measures were not confounded by hospital, job, gender or tenure and the univariate ratios are presented. No lagging compares all time before to all time afterwards as if no latency of effect was anticipated. Ratio measures for patient-handling injuries were maximised with no lagging and they decreased only slightly at each point afterwards. Mean cost ratios for non-patient-handling MS injuries followed a similar pattern, although the decline in cost rate ratios per FTE was always higher for non-patient-handling MS injuries than for patient-handling MS injuries. Mean costs per claim followed the same patterns (not shown).

Figure 3

Cost rate ratios for patient-handling and non-patient-handling musculoskeletal (MS) injuries before and after introduction of the lift equipment lagged by different periods of time.


Data on direct costs associated with WC claims were used to estimate resource requirements for the provision of medical care, rehabilitation and lost productivity represented by payments for lost work time for work-related MS injuries at a major health system over time. Mean cost per claim for patient-handling MS injuries was less (∼40%) than those for non-patient-handling MS injuries; however, the cost rate per FTE was over twice as high for patient-handling events due to their prevalence. Patterns observed for MS injuries associated with patient handling were compared to other MS injuries which theoretically would not be influenced by patient lift equipment but would be affected by other institutional policies. Because the vast majority of injuries were associated with patient handling, the comparison with other MS injuries was less robust but still informative. With the anticipated latency of effect of this intervention designed to decrease costs associated specifically with patient-handling activities, ratio measures should be more extreme with increased lagging for patient-handling injuries.

Just as a number of other investigators have reported, we saw a decline in the direct institutional costs of patient-handling injuries surrounding this intervention.14–21 The decrease in costs was measured by both declining mean costs per claim and declining cost rates per FTE. The declines were immediate and largely sustained. While the magnitude of the drop in direct WC costs that we observed is not out of line with that reported by others following the introduction of lift equipment, it is of note that the initial marked decline in costs that we observed occurred prior to the time that the lift equipment could have reasonably contributed to the reduction.14 This interpretation is consistent with the decline in the costs of non-patient-handling MS injuries as well as reported findings on patterns of adoption of the equipment by staff at the institutions we were studying.10

The current report is based on the costs of WC injuries that were reported and accepted as work related, and we recognise the costs we assessed do not represent the true burden of injury but rather the direct costs absorbed by the institution. It is of note that these analyses do not include costs associated with healthcare use outside the WC system which can be important in understanding the overall health of working populations.22 23

The direct costs that we examined can be influenced by many factors. In addition to the institutional policy changes regarding lost time payments between 1997 and 2003, no more than three claims per year were denied for compensation coverage and in 3 years none were denied (average of one per year for 1997–2003); however, between 2004 and 2009, 115 claims were denied (averaging 19 per year). The range of days from injury until the work-related claim was closed for medical care and paid lost time varied from 0 to 3499 between 1997 and 2003 and ranged from 0 to 1887 in the years after. Although median time to closure between 2004 and 2009 was 61 days compared to 0 days between 1997 and 2003, mean days to closure decreased from 111 days to 93 days as there were fewer claims with extended time to closure. The proportion of claims that resulted in any paid lost time, meaning at least 8 work days were lost, decreased from 12.5% before 2004 to 7.4% afterwards.

Cost data can provide a useful measure in the assessment of program effectiveness by jointly incorporating incidence and severity measures to capture resource use. However, in using these data we must recognise that they reflect wider economic patterns (eg, inflation) which must be taken into consideration. Additionally, cost data create analytical challenges. Data such as these are typically skewed; many claims had no or very low costs with a few resulting in exorbitant costs. The analytical techniques that we used adjusted dollars for inflation and discounted to allow a comparison of constant dollar values over time; the use of a negative binomial distribution allowed the modelling of costs without transforming the skewed data.

Access to longitudinal data on this large cohort of healthcare workers drawn from an existing surveillance system that captures events of interest and time at risk is a major strength of these analyses. Institutional knowledge of competing issues that had the potential to influence costs paid through the health system's self-insured WC program was also informative and key in interpreting the findings. These data help demonstrate the importance of the broad perspective called for by others that is not restricted solely to cost data reported through compensation records but also includes reporting systems and policy changes during the study period.17 These analyses were part of a larger effort to evaluate the effectiveness of an MMLE policy supported with patient lift equipment at this large health system, and they demonstrate the utility of longitudinal observational data with an internal comparison injury group to understand workplace intervention effectiveness, or lack thereof. A randomised controlled trial was not feasible given a number of factors including the nature of the institutional change, as well as the potential for bleed over of the intervention, and marked variability by unit in both patient populations and staff injury risk.24 Although the example is drawn from the US healthcare system, the methodology has much broader application for occupational injury intervention evaluation.


These analyses provide an example of the use of longitudinal data with comparisons across different types of injuries in the same population in the evaluation of institutional efforts to reduce work-related injuries and control costs. These analyses take advantage of robust, retrospective administrative data from an existing surveillance system that can be updated on a regular basis.9

From a methodological perspective, rather than identifying what we thought was a ‘best mathematical model’, the more informative information came from the examination of multiple stratified models. Examination of cost changes among different groups of workers and different types of claims in various time windows of calendar time suggest that the primary cost control was likely driven largely by factors other than those related to the MMLE policy and certainly by factors other than the use of mechanical lift equipment; this interpretation is consistent with observed patterns of limited adoption of equipment.10

While marked declines in costs were observed following the policy change, the pattern observed likely reflects the effects of targeted efforts to close compensation claims, requirements that compensation injuries be reported within 24 h, and an almost simultaneous policy that shifted cost responsibility to the budgets of managers on individual units instead of the institution. Equipment availability, growing acceptance of the MMLE policy and use of the lift equipment may have contributed to the sustained reduction in costs, but it would be difficult to ascribe the initial marked decline to these particular interventions.

The costs of work injury among healthcare workers can be substantial and they contribute to the overall costs of medical care. With growing concerns about healthcare costs, the extent to which worker injury costs contribute should not be ignored. However, it is important to recognise that while decreasing compensation costs can signal safer working conditions or decreased risk among workers, they can also reflect procedural or institutional changes. The patterns we observed could easily have been missed in a simpler before-and-after design that did not examine effects in different calendar periods and without a comparison group of non-patient-handling injuries among the same working population.

Finally, these analyses demonstrate the importance of investigators establishing a conceptual framework for the evaluation of workplace interventions. A growing literature illustrates important barriers to the implementation of primary patient-handling preventive activities25 and individual and organisational determinants of the use of such equipment once it is introduced.26 However, evaluation designs should also consider context and latency before effects could reasonably be expected. In industrial settings it may be possible to make a production line change that influences worker exposures immediately. Such is not the case with the introduction of lift equipment, even if accompanied by a policy shift, in the acute care hospital setting. Even in situations where latency has been considered and the fidelity of the intervention has been carefully monitored, other influences should be considered before ascribing benefits and cost savings entirely to a given intervention.


  • Competing interests None.

  • Ethics approval Ethics approval was provided by Duke University Medical Center Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.


Free sample

This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of OEM.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.

You are viewing from:
Ruth Lilly Medical Library Indiana Univ-Acq Dept