While most people experience back pain at some time in their lives, many never seek care or lose time from work. Yet for some, back pain results in significant episodes of activity limitation, impairment and disability. There is evidence that both occupational and non-occupationally defined back pain tend to be recurrent,1^–4 as well as some indication that recurrent episodes may be more likely to impair activities.5 6

Methods currently used by the Bureau of Labor Statistics to count workplace injuries and illnesses do not differentiate initial events from recurrent ones, and rates of recurrent work-related back injuries are not well-defined. Recurrence is often reported as a prevalence percent7 8; information regarding return to work following the first event, when the individual actually became at risk for a recurrent injury, is often missing. In addition, the definitions of recurrence vary widely among the populations studied as do the nature of the data used in analyses, which could explain the wide variability among reports.9^–12

The rate of recurrence of back pain has been reported to decrease with time after the first event,13 suggesting that there may be periods of time when preventive measures might have greater impact. Some have reported more recurrences among those with gradual onset of symptoms, while others have reported that acute traumatic events, such as falls, are associated with a greater likelihood of recurrence.2 14 These findings have been used to suggest that different disease processes may involve different prognoses.

Intuitively, determining the severity of the injury, or disorder, would seem important in predicting outcomes and recurrence. However, the relationships between detectable tissue injury and back pain, impairment and disability are poorly defined – likely, in part, due to lack of diagnostic precision – and for back problems few case definitions have been accepted in the scientific community.15^–19 Consequently, the severity of work-related episodes is often defined in terms of another outcome such as delayed return to work or associated medical costs.

Back disorders are the most commonly reported occupational disorder among construction workers.20 21 In both the USA22 and Canada,23 being employed in construction has also been reported to be associated with delayed return to work following a back injury. Older construction workers are at greater risk of back problems than other blue collar workers24 and at greater risk of occupational disability from back problems compared with the general population.25

Despite their known exposures to recognised occupational risk factors for back pain,26 their high rates of injury and likelihood of prolonged disability, there is little literature specifically related to occupational back problems among carpenters.27^–30 We know of none that distinguish between initial and recurrent work-related injuries. Measures used to assess severity of acute trauma are not sensitive enough to grade most occupational injuries to carpenters due to a “floor effect”. The vast majority of injuries to carpenters treated in the emergency room had an abbreviated injury score (AIS) of 1, indicating minor injury30; the rating had little relationship to the length of time out of work. Hospitalisation has been used to control for severity in analyses of disability following work injury.22 Yet, in Waller’s work,30 74% of carpenters who were still impaired 6 months after injury had only received outpatient care.

We used combined administrative data sources to identify and describe claims for work-related back disorders in terms of their nature and mechanism, associated paid lost time from work and the ICD-9 (International Classification of Diseases, Ninth Revision) diagnoses assigned for treatment of these injuries among a well-defined cohort of union carpenters. In addition, analyses focused on more clearly defining the hazard functions for incident and recurrent claims and exploring risk factors for these events separately.

METHODS

Data sources

Using data from the Carpenters Trusts of Western Washington (CTWW) and the Washington State Department of Labor and Industries (L&I), we identified a cohort of union carpenters who worked in the State of Washington between 1989 and 2003, their dates of birth, gender, earliest date of union activity, their hours of union work each month and their workers’ compensation claims. No race or ethnicity information was available from these sources. Cohort membership was limited to individuals who worked at least 3 months of union hours during this 15-year period and had at least 1 month of eligibility for health insurance through CTWW.

The State of Washington has a state run worker’s compensation program which captures medical only claims as well as those which result in lost work time. Claims from companies which self-insure for compensation coverage are only coded in the L&I data if they result in paid lost time from work. The L&I data include the date of injury, American National Standards Institute (ANSI) codes describing the events in terms of body part injured, the nature of the injury, the type of event causing the injury and whether the claim resulted in medical costs or paid lost time from work, which occurs after the third lost day in the State of Washington. We also had access to the medical care claims for the treatment of these injuries including provider assigned ICD-9 diagnosis codes (excluding claims from self-insured employers) through this state run program. Data were extracted on 26 August 2004 allowing 8 months after the last claims were filed in 2003 for full development of the claims, proper identification and capture of the self-insured claims that resulted in paid lost time.

The data were provided with a blinded unique identifier allowing us to merge the records on an individual basis without the use of personal identifiers; these methods have been previously described in detail.27 28 31

Events of interest

Workers’ compensation claims for back injuries were identified by body part codes designating an injury to the back, trunk or back and neck as well as by medical claims for a back specific ICD-9 code assigned for treatment of the injury. This process captured claims in which the body part code indicated multiple injuries as well as claims that may have had some other body part code designated as the primary injury when first reported. To avoid claims that may have been identified by a single miscoded ICD-9 diagnosis, at least two claims for a back specific diagnosis were required if the body part code did not designate a back injury. Details on the process and specific codes used have been previously described.32 Up to two injuries were included per carpenter, with incidence approximated by the first back injury claim identified.

Time at risk

After the claims of interest were identified, event histories were created for each individual. Person-hours of work as a union carpenter were used as the measurement of time at risk. Person-time began accumulating for the first back injury when the individual met cohort entry criteria. After an injury, person-time at risk of a second event began accumulating in the next month in which union hours appeared. All person-hours were counted in any month in which an injury occurred. Although time at risk was counted in terms of person-hours, the person-month was effectively the unit of analysis, since it is unknown where in any given month the work hours fell.

Classification of back injury claims

Each back injury claim was classified by the nature and mechanism of injury as defined by ANSI codes, as well as by paid lost time, hierarchy of medical care received and ICD-9 diagnoses. Classifications of time out of work included no paid lost time, return within 1 month, return within 3 months and out of work for more than 3 months. Hierarchy of care included first aid only, outpatient care or inpatient hospital care. ICD-9 diagnoses included groups of codes reflecting sprains, dislocations without fracture, fractures, degenerative conditions, contusions, nerve injury and symptom descriptors; a mixed diagnosis was assigned for individuals with ICD-9 codes that reflected more than one of our case definitions. The methods used to group the diagnoses have been previously described in detail.32

Each of these characterisation schemes was used to explore whether the risk of a second back injury, based on different definitions of the incident event, differed from the risk of filing the first claim for a back disorder. It was hypothesised that individuals with degenerative diagnoses, including disc disorders, would be at greater risk of recurrence. In contrast, we felt that those who required only first aid, and those who sustained sprains, which should be self-limiting injuries, should not be at greater risk for a recurrent injury. However, we suspected that the ANSI code classifications of injury nature would prove too crude to detect meaningful differences.

Covariates of interest

Age categories were defined at 10-year intervals. Carpenters typically spend 4 years in apprenticeship; strata were constructed to assess risk in each year of this training and at 2-year intervals afterwards. The union local (branch) affiliation was the only surrogate available for characterising the work done by cohort members. The locals represented by the cohort members were grouped into categories based on the predominant type of carpentry work done by the locals. Assignments of predominant type of work from earlier work with this cohort26 27 30 were updated through interviews with business agents for each union local. These categories included light commercial, heavy commercial, drywall, millwrighting, piledriving, residential carpentry and a mixed category. Light commercial work involved construction on projects two to three stories high. Heavy commercial work involved high-rise buildings and interstate, freeway and bridge work. Millwrights are carpenters who work in industry and are often involved in the repair and maintenance of heavy machinery. Drywall carpenters in Washington State hang drywall, but they do not tape or finish, in residential or commercial settings. We were unable to identify the type of work of carpenters affiliated with a local outside the State of Washington; for this reason they were combined for the analyses.

Analyses

Descriptive statistics were generated on age, gender, time in the union, predominant type of work and hours worked by the cohort. Kaplan–Meier survival curves were constructed to compare incident and recurrent claims. Incident and recurrent back injuries were described, separately, by each of the classification schemes. Crude and stratified incidence rates were calculated per 200 000 h worked, or the equivalent of 100 carpenters working full-time for 1 year. Age and time in the union were both treated as time-varying variables, with events and time at risk accumulating in the appropriate strata over time. Poisson regression was used to calculate crude rates and rate ratios (RRs) as well as adjusted rate ratios because of its utility in the analyses of longitudinal cohort data such as these, allowing maximal use of available data.33 Claims for the first back injury identified for each carpenter were analysed separately from second events.

Next, each of the classification schemes were used in separate models to explore whether the risk of recurrence, based on different definitions of the incident back injury, differed from the risk of filing an incident claim for a back disorder. In each case those with no prior claim served as the referent group. Second injuries that resulted from acute trauma (falls, struck by) were excluded from these latter analyses since we saw no biologically plausible reason that a prior injury should contribute to subsequent acute trauma. Lastly, injury rates (95% CI) were calculated in windows of time representing 1000–2000 h increments of cumulative hours at risk following return to work. This analysis was done to explore whether periods of time could be identified when carpenters were at greater risk for a second back injury after they returned to work. All analyses were conducted with SAS v 8.2.

RESULTS

Cohort

We identified a dynamic, retrospective cohort of 18 768 carpenters who worked in the State of Washington between 1989 and 2003 who had hours worked after meeting the cohort entry criteria. The mean time the carpenters were observed over the 15 years was 45 months (median 23.0). The cohort was predominantly male (n = 17 879; 97.4%) and relatively young at first observation (range 17–76 years, mean 35 years, median 34 years). Time in the union at first entry into the cohort ranged from less than 1 year to 48 years (mean 6.1 years, median 1 year).

Identification and description of incident and recurrent back injuries

Overall, 4148 back injuries were reported by the cohort over 15 years. The number of events per person ranged from none to seven. Using our case definitions, 3037 incident and 751 recurrent back injury claims were identified. The distributions of the incident and recurrent injuries are remarkably similar by mechanism and nature of injury, medical diagnoses assigned for their treatment, type of medical care received and associated paid lost time from work (table 1). The mean difference in time lost from work was slightly higher for recurrent than for incident injuries (126 vs 110 days).

There were 151 individuals who were never observed working union hours again after their first back injury, leaving 2886 carpenters in the risk pool for a second event. Excluding 22 individuals who filed their initial claims in 2003, who may have returned to work after our observation period, 74% (n = 117) of those who never returned to work had paid lost time of over 60 days. These individuals were more likely to be assigned a degenerative diagnosis (7%; n = 9), fracture (4%; n = 5) or mixed diagnoses (40%; n = 52) than individuals who returned to work, and their injuries were more likely to have resulted from a fall from a height (n = 25; 19%). Thirty-five (27%) sustained injuries to other body parts as well. However, only 8.5% (n = 11) received hospital care for their injury. Consistent with this, the majority of injuries were coded as having resulted from overexertion or bodily motion (63%; n = 81).

Injury rates and risk factors

Incident injuries occurred at an overall rate of 5.4 per 200 000 h worked; recurrent events occurred at a rate of 9.5 per 200 000 h worked (RR 1.8). Kaplan–Meier curves demonstrating the survival functions for incident and recurrent claims are presented in fig 1. Although the survival curves for incident and recurrent injuries are significantly different from each other (log-rank test p<0.001), the patterns of relative risk are fairly similar for incident and recurrent back injuries (table 2). Because the youngest and oldest work groups are small, and consequently the least stable, age 30 to <40 years was chosen as the reference cell. Crudely, the youngest workers were at greatest risk of both incident and recurrent back injury, but this did not remain the case when adjusting for gender, time in the union and predominant type of work. Women had higher rates of incident events and slightly lower recurrent events. Overall back injury risk declined with increasing time in the union. Individuals who work predominantly in drywall were at consistently higher risk than their union counterparts doing other types of work.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Kaplan–Meier survival curves for incident and recurrent back injuries, Washington carpenters, 1989–2003.

Crude rates and rate ratios of a second back injury resulting from overexertion or bodily motion are presented in table 3 for each of the different classification schemes for incident cases, as are the adjusted rate ratios. Overall, these subsequent back injuries of a musculoskeletal nature occurred at a rate of 6.8 (95% CI 6.3 to 7.9) per 200 000 h worked; this is 30% higher than rates of initial back injuries. Minimal differences in risk of second injury were observed based on the mechanism of the first back injury. The few carpenters whose nature of injury had been coded as a fracture were at particularly high risk of a second back injury when compared with those with no prior history, but their fractures were not confirmed by appropriate ICD-9 codes. Individuals with a medical diagnosis of degenerative conditions or mixed diagnoses, those who required inpatient care for their first injury, and those who were out of work a month or longer with their initial event were at greater risk of reporting a second work-related back injury of a musculoskeletal nature. The parameter estimates for age, gender and predominant type of work were not significantly different in these models and they are not presented here.

Of note is the higher rate of injury between 1000 and 1500 h after return to work (fig 2). Other than in this period, the rate of second injuries decreased slowly over the first 6000 h after return to work (representing 3–4 years of fulltime work among this cohort) and then declined rather dramatically.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Rates of subsequent musculoskeletal back injury in windows of time after return to work, Washington carpenters, 1989–2003.

DISCUSSION

Among this large cohort of union carpenters, second back injuries occurred at a rate 80% higher than initial events. Although the mean difference in lost time from work was modestly higher for second injuries than for the first, the distributions of the first and second back injuries were remarkably similar in terms of mechanism and nature of injury, as well as medical diagnoses received for treatment. These findings are consistent with similar injuries from like work exposures being responsible for both initial and subsequent injuries among members of the cohort.

Carpenters between 30 and 40 years of age were at greatest risk of back injuries, after adjusting for gender, time in the union and predominant type of work. The fact that we observed individuals with the greatest union tenure carrying the lowest risk likely reflects a healthy worker effect or changes in direct physical exposure that occur with more seniority in the trade. Consistent with other reports,28 carpenters involved in drywall work were at relatively high risk of back injury compared with their union colleagues in other forms of work; residential carpenters were at higher risk of initial injury but not for a second event (based on a small number of observations).

The risk ratios for a subsequent musculoskeletal injury based on the different mechanisms and nature of injury were quite similar, consistent with our hypothesis that the ANSI codes assigned to the initial injury would not be sensitive enough to identify individuals at greater risk of a second injury. We observed no differences in risk of a second injury between those who had first aid only compared with outpatient medical care for their first injury. Individuals with the longest period of work disability with the first back injury were at particularly high risk (2.3 times greater than risk for first event), consistent with the fairly recent report of Wasiak et al34 based on analyses of compensation data. Carpenters whose medical diagnoses reflected a degenerative condition or multiple diagnoses were at greater risk of a second injury, as were those who required hospitalisation or were out of work for a month or more; all likely reflect more severe initial injuries or pathology at the time of injury. However, multiple diagnoses are more likely to be assigned to individuals with more medical encounters and may reflect evolving clinical impressions sometimes accompanied by more testing among individuals with delayed recovery. Consistent with Waller’s findings,29 30 it was rare for a carpenter to require hospitalisation for a back injury, even among those who never returned to carpentry work.

The risk of a back injury of a musculoskeletal nature peaked at 1500 h of time after return to work from a previous injury, and gradually decreased with increasing time back at work. This time represents close to a year of full-time work for many of these union carpenters. We have no data on accommodation or restricted work after injury. It is possible that these carpenters received less taxing duties upon return from an earlier injury or they may have looked for ways to limit stressful exposures on their own when first coming back to the job. However, the overall nature of construction work makes light duty assignments unlikely, particularly for extended periods of time.

Consistent with our findings, among Canadian workers who had lost-time back compensation cases, the risk of recurrence was greater in the first year after injury, especially for those with longer periods of absence from work.5 Of note, recurrence was not defined by the filing of another injury report (as in our study), but by exacerbation of symptoms and work absence after return. There was also a gradual increase in the duration of absence with each recurrence, potentially representing worsening symptoms or pathology, or, perhaps, the tendency of caregivers to recommend longer absence from work with recurrent episodes.

Limitations and strengths

There are a number of limitations and strengths to this work. Our approximation of an incident case does not guarantee that the person had never experienced a previous work-related back injury, and we recognise there was some misclassification in the assignment of initial and recurrent cases. We considered different definitions of incidence that would require an injury-free observation period of a year or two to be considered at risk for an incident injury. We elected not to do this because we would have lost important information about those newest to the union who are known to be at high risk of injury.27 28 31 32

The State of Washington captures medical only claims as well as those which result in lost work time, which is a distinct advantage. However, because claims from companies which self-insure are only coded if they result in paid lost time from work, we know we did not capture all injuries that occurred when carpenters worked for contractors who self-insured for workers’ compensation. We also could not identify the records of medical care with ICD-9s for the self-insured claims.

Unfortunately, because of lack of data, we were unable to look at risk differentially by race or ethnicity, which could have been enlightening and important. Through the 1990s this cohort was almost exclusively Caucasian. While this pattern largely remains, there is an increasing immigrant presence in this area of the country, even in the union workforce.

Our definitions of incident and recurrent injuries were based on reported work-related cases and should not be confused with any prior history of back problems. Additionally, risk factors for recurrence of back pain can vary depending on the definition of recurrence.6 We evaluated risk for any recurrent injury defined as the filing of a new worker’s compensation claim for a back injury, and we also explored whether information about the first injury predicted the filing of a subsequent claim for a back injury of a musculoskeletal nature. Some of these carpenters (n = 127) had their first compensation claims re-opened, including 27 of the 751 (3.6%) who filed second claims; these events were not considered a recurrent injury based on our definition.

As is always the case in analyses of workers’ compensation records, anything which influences whether a person files a compensation claim will be reflected in the results. In contrast to many construction workers,21 these union carpenters have private insurance coverage, removing some concerns that non-work-related care ended up in the workers’ compensation system.

We did not have information on psychosocial issues or work environment. However, because of the frequent overlap of significant work organisation issues and physical work exposures,35 36 we believe this information would be most enlightening with more precise knowledge of work exposures.37 We recognise that there is misclassification in the process of assigning predominant work that may well have muted the effects we observed, and beyond type of work of the local with which the carpenter affiliated, we had no real exposure information.

Despite these limitations, we were able to observe the reported, work-related back injury experiences of a large well-defined cohort of carpenters. The 15-year observation period allowed sufficient time and statistical power to explore risk factors for both incident and recurrent events and to assess periods when carpenters may be most at risk of a subsequent injury after returning to work.

The analytical methods we chose, specifically survival techniques of Kaplan–Meier curves and Poisson regression, allowed the use of all the data available for each carpenter. Access to coded compensation data and the records of actual medical care allowed us to consider the impact of the different characteristics of a prior injury on the development of a later musculoskeletal injury. The combination of work records from the union and workers’ compensation claims allowed us to clearly identify when a carpenter had returned to work regardless of whether the individual may have changed employers. This knowledge of the true population at risk of recurrence is a factor missing in many reports of recurrent back injuries.

CONCLUSIONS

Back injuries are responsible for a significant burden of injury among these union carpenters. After having an injury, individuals remain at greater risk of a second event of a musculoskeletal nature for a considerable period of time (3+ years). In large part, the ANSI coded compensation data reflecting the nature and mechanism of injury are not particularly helpful in identifying individuals who are at greater risk of a recurrent back injury. We did find some indication that more severe initial injuries put individuals at risk of a second event and they should be targeted for secondary prevention efforts. In this industry where there are considerable challenges to workplace accommodation because of the heavy nature of the work, individuals who are out of work for over a month, especially over 3 months, warrant accommodation and, perhaps, better rehabilitation efforts/attention to avoid re-injury. However, this is in fact more challenging than it might first appear.

While there is some evidence that even individuals with severe and chronic pain following work-related back injuries can reduce their risk of recurrent problems, including new work-related claims,38 there are also reports that patterns of recurrence following functional restoration efforts are consistent with the episodic nature of low back pain.39 In addition, techniques such as functional capacity testing often fail to accurately identify when individuals are ready to safely return to work following injury.40 Considering these issues in light of the lack of any particularly strong predictors of recurrence, strongly suggests a need for primary prevention of back injuries among these workers involved in strenuous tasks. Synthesis of data from other working populations41 indicates that interventions, at least for lifting and handling tasks, should focus on engineering solutions rather than changes in individual worker behaviours.