William Lohman, M.D., Medical Consultant; David Berry, Ph.D., Grant Martin and Brian Zaidman, Research Analysts
A new study by Research and Statistics examines why the denial rate among filed lost-time claims doubled between 1984 and 1998.
The study, the second phase in a larger study, looks at factors associated with claim denials, their adjudication and eventual compensation status, and their impact on future compensation claims and employment. The first phase of this project, Part one: Explaining claim denials, studied factors associated with claim denial and how rates varied by insurer.
Phase two looks at another significant variation in claim denials -- the changing rate of denials during the past two decades. All of the claims from five two-year periods were evaluated for factors associated with claim denial and changes in the rate of claim denial over time.
the denial rate among filed lost-time claims in Minnesota rose from 7.8 percent in 1984, to 16.2 percent in 1998 -- a 107.8 percent increase in a 15-year period.
the likelihood of denial associated with a delay in notice to the employer has risen since 1984.
the risk of denial for low-wage workers has climbed since 1984. By 1997 through 1998, those earning 50 percent or less of the statewide average weekly wage are almost twice as likely to be denied as the highest paid workers.
analysis indicates 62 percent of the increase in the denial rate between 1984 through 1985 and 1997 through 1998 is attributable to changes in insurer behavior and 32 percent to changes in the nature of claims and claimants.
more than 12 percent of the entire increase in denials is accounted for by an increase in the rate at which the claims of low-wage workers are denied; and another 10 percent of the increase comes from an increase in the number of claims by low-wage workers.
six percent of the overall increase is explained by an increasing intolerance with delays in notification of an injury.
the largest contribution (44 percent) to the rising denial rate has come from an increase in the number of claims with undefined injuries.
The study suggests other areas that need to be explored to elucidate the possible mechanisms underlying claim denial. In particular, future analyses should examine the effects of other aspects of the claimant's employment (occupation, job tenure, employer size, employer safety record, unionization, access to health insurance), insurer business practices (market share, changes in premiums, loss ratios, litigation experiences), claimant characteristics (other demographic factors, previous claims, previous denials, previous litigation) and the workers' compensation system (wage-replacement rates).
Under Minnesota law, an insurer may deny primary (any) liability for an injury at any time, but only if there is a legal and factual basis to do so. In practice, most denials are made when a worker first makes a claim for compensation. The insurer's denial of primary liability is based on a belief that one or more of the following is true:
the injury did not "arise out of and in the course of employment" as defined by statute or case law; 
the injury was intentionally self-inflicted; 
the injury was intentionally inflicted by a third party for personal reasons; 
the injury was caused by the injured worker's intoxication; 
the injury occurred during a voluntary recreational activity;  or
notice of the injury was not timely. 
Denials are an important measure of equity and efficiency in the workers' compensation system. If a denial of primary liability is improperly made, workers' compensation fails in its purpose of providing medical care, cash benefits and vocational rehabilitation to employees injured at work. If the insurer accepts liability for what is actually a non-work-related injury, costs are improperly shifted into the workers' compensation system, increasing losses and -- ultimately -- employer premiums. Denials are also important because they frequently generate disputes, which cause delays and increase system costs to all participants.
A Department of Labor and Industry (DLI) project is underway to study the factors associated with claim denials, their adjudication and eventual compensation status, and their impact on future compensation claims and employment. The first phase of this project was presented by David Berry in Part one: Explaining claim denials.  In an analysis of 179,000 lost-time claims  from 1996 through 2000, he found the rate at which claims were denied was higher for older claimants, lower-wage workers, workers with less tenure at the employer, workers with prior claims, claims with delays in notice to the employer or insurer, and for certain types of injuries that are not obviously connected to work. Berry (2002) also found that, even when these factors and others where statistically controlled in the analysis, there was still significant variation between insurers in the rate at which claims were denied.
The analyses presented here look at another significant variation in claim denials -- the changing rate of denials during the past two decades. The denial rate among filed lost-time claims in Minnesota rose from 7.8 percent in 1984, to 16.2 percent in 1998 -- a 107.8 percent increase in a 15-year period (Figure 1). This increase continued even when the number of claims filed started decreasing after 1989. By examining five two-year cohorts of claims, the analysis looks at how factors associated with denials have changed in relative importance over time since 1984. Underlying the analysis is a model of insurer behavior that is based on (1) intolerance of uncertainty created by missing claim information and (2) suspicion of claims originating in certain situations (Figure 2). It is hypothesized that insurers are more likely to deny claims in which important information is missing, such as the type of injury or an obvious causal relationship to work activities. In addition, insurers may learn to suspect certain types of claims or claimants that have been problematic in the past, such as occupational disease claims or claims by seasonal workers. By identifying the factors associated with claim denial in five successive cohorts with increasing rates of denials, we hope to understand how intolerance of uncertainty and suspicion have evolved as explanations of insurer behavior.
By law, all insurers in Minnesota must either begin payment or deny a worker's claim for compensation within 14 days of when the employer becomes aware of the injury. If the worker was disabled for more than three days, then the insurer's decision to accept or deny the claim must be filed with DLI along with the First Report of Injury form. The forms filed by the insurer include a variety of additional data: injury characteristics, claimant demographics, employment and vocational details, and claim and insurer information. All of this information is routinely entered into DLI's administrative database for each of the filings received, and was the sole source of the data used in this study.
Five two-year observation periods were chosen: 1984-5, 1986-7, 1991-2, 1994-5 and 1997-8. Each represents a temporary plateau in the denial rate. Claims were selected for inclusion if they represented an injury that occurred in one of these time periods and if the employee was disabled for more than three days. Claims were excluded if they were medical-only claims that had been mistakenly filed with DLI or were indemnity claims without wage loss (e.g. they had only received a permanent partial disability benefit). A claim was considered a denial if the insurer's initial decision was a denial, even if the claim was subsequently paid voluntarily, after judicial review, or by settlement.
For each cohort, all of the potential variables available in the administrative data and common to all of the time periods chosen, were examined for association with the insurer's initial claim denial. Four domains of explanatory variables were identified: demographics (age, gender, marital status, residence), injury characteristics (type of injury, cause of injury, day of the week, month of the year, source of injury), job characteristics (industry, employment status, pre-injury wages) and claim characteristics (type of insurer, timeliness of notice, onset of lost time). Unfortunately, DLI's administrative data does not include information about some factors of interest, such as race, ethnicity, union status or ICD-9 diagnostic codes. It was also not possible to use some variables that had been included in Berry's analysis, because they were not available for all of the time periods studied or their definitions were not uniform during the entire period, in particular, occupation, job tenure and prior claim status.
In lieu of ICD-9 diagnostic codes, a classification of injury types was created using the available NCCI (National Council on Compensation Insurance) part of body and nature of injury codes to assign each injury to one of 12 categories: burns, lacerations, neck or back musculoskeletal disorders (MSDs), upper extremity MSDs, lower extremity MSDs, contusions, soft tissue injuries not otherwise specified, multiple injuries, diseases or internal organ injuries, and undefined (no other type of injury could be assigned with the information available).
Given the prominent roles of uncertainty and suspicion in the hypothesized model of insurer behavior, cause and source of injury information was used to create the variable "compatibility." For each injury type, each cause of injury was identified as compatible with the injury type or not, based on principles of medical causation. For example, if the injury was a laceration and the cause of accident was listed as "overexertion" then this was "not compatible," while if the cause was "contact with sharp object" it was considered "compatible."
To control for inflation and wage growth between 1984 and 1998, pre-injury wages were expressed as an index of the statewide average weekly wage (SAWW) for the year of injury and then categorized as "greater than 150 percent SAWW," "101 through 150 percent SAWW," "51 through 100 percent SAWW" or "less than 50 percent SAWW."
For each cohort of claims, variables were assessed individually for their relationship with claim denial, next within their domains and then finally overall. All analyses were done using logistic regression. Only variables that had statistically significant relationships were kept for analysis at the next step. After the explanatory variables were identified for each cohort, final models were created by including any variable that had been significant in other cohorts.
The odds ratios for each variable were then compared across cohorts against the baseline period 1984-5. Differences in odds ratios were considered statistically significant if their 95-percent confidence intervals were mutually exclusive.
Finally, the variables of the final logistic regression models were used in linear regressions with denial rate as the dependent variable. Linear regressions were calculated for the 1984-5 and 1997-8 claims. These results were then used for an Oaxaca  decomposition to identify the relative contributions of the changes in variable means (how claims and claimants have changed) and the changes in the variable coefficients (how insurer behavior has changed) to the increased denial rate in 1997-8.
All analyses were done using SPSS for Windows version 11.5.0.
The total population of claims examined was 408,122, including 50,658 initial denials of primary liability. Figure 3 shows the population characteristics for each cohort. Several trends are immediately obvious. The proportion of claimants who are female has steadily increased (up to 35.8 percent in 1997-8), as has the percentage living in greater Minnesota (from 40.6 percent to 44.6 percent). The average age of claimants has increased from 35.7 years of age to 39.1; the percentage of claimants 40 to 49 years old has increased 63 percent. There has been no obvious trend in marital status.
There have been large changes in the types of injuries claimed. Most notable has been the dramatic increase in the number of undefined injuries, reaching 37 percent of cases in 1997-8. These are cases for which the information about the claim form was insufficient to medically identify the condition. This has been paralleled by a decline in the number of injuries that could be reliably identified as neck or back MSDs from 33.0 percent of claimants in 1984-5, to only 19.6 percent in 1997-8. The number of claims in which the cause of injury is compatible has also increased from 78.7 percent to 92.0 percent; this is due in large part to the increase in the number of undefined injuries -- any cause is potentially compatible with an unknown injury type.
While the percentage of claimants from manufacturing dropped during the observation period, the proportion from the retail trade and service industries increased. Claimants' average wage has steadily increased from $343 to $491; but the average wage index fell from 101.9 to 86.3, indicating that a larger proportion of claimants in the later years earn less than the SAWW. There has been a slight decline in the percentage of claimants who are full-time employees.
Slightly more claimants worked for self-insured employers. Though there was a large increase in claimants whose employers were in the Assigned Risk Plan from 1986 through 1995, the 1997-8 cohort was not much different from the 1984-5 baseline. There hasn't been any large change in the timeliness of notice ("delay in employer notice," "first lost-workday") or the distribution of injuries by month or day of the week (data not shown).
While a majority of the factors evaluated were found to be associated with claim denial, others were not; industry classification, day of week of the injury and month of year of the injury had no statistically significant relationship to claim denial in any of the cohorts examined. These findings are similar to those in Berry's study. Cause of injury, which had been a significant factor in Berry's analysis, was significant here only if the compatibility variable was not included.
The employee's residence at the time of injury was not a significant factor in the 1984-5 cohort, but was in all of the others. Similarly, employment status (full time, part time, seasonal, volunteer) was not related to claim denial in the 1986-7 cohort, but was in the other time periods. The remaining variables had a statistically significant association with claim denial in all five cohorts. In order to simplify comparisons, residence and employment status are included in the models for all the cohorts reported here.
Figure 4 shows the final logistic regression models for each cohort. Women, unmarried claimants, older workers and Twin Cities residents are all at higher risk of having their claim denied. Denials are much more likely for occupational illnesses, hernias and musculoskeletal disorders than for burns, lacerations and fractures. Injuries that are undefined at the time of claim, or injuries whose alleged cause is not compatible with the type of injury reported, are also denied more often.
Delays in notifying the employer about the injury also increase the likelihood of denial, and there is a clear "dose-response" relationship -- the longer the delay, the greater the chance the claim will be denied. Conversely, if lost time starts on the date of injury -- as opposed to some later date -- the claim is more often denied. Claimants in the Assigned Risk Plan (the insurer of last resort for employers unable to either purchase insurance in the voluntary market or self-insure) are also at higher risk of claim denial.
Employment status has a variable relationship to claim denial over time, but full-time workers were always more likely to be denied than part-timers. Wages have a consistent relationship with claim denials. The lowest-paid workers, those earning 50 percent or less of the SAWW for the year of their injury, are at higher risk of claim denial than those earning more than 150 percent of the SAWW. Here, again, there is a "dose-response" relationship: the risk of denial goes up as earnings fall.
In Figure 4 each cohort is compared to the 1984-5 baseline. Odds ratios where confidence intervals are mutually exclusive are considered statistically significantly different and are shown in bold. Though there are clear trends, there are few significant differences. The risk of denial associated with increasing age declines. In 1997-8 only 50- to 59-year-olds are significantly more likely to be denied than the referent group of those less than 20 years old; though only the change in odds ratio for those more than 60 years old is significantly different from the 1984-5 value by our criteria.
Odds ratios increase for all of the injury types, but none are statistically significant by our criteria. Changes in the risk of denial associated with an incompatible cause of injury are significant: incompatibility is associated with a higher risk in 1986-7, but the risk declines markedly and is significantly less in 1997-8 than at baseline.
The risk of denial to public workers and employees of self-insured employers goes down. The odds of denial associated with delay in notice rises (significantly for two- to six-day delays), but the odds associated with lost-time starting on the day of injury go down. The risk of denial to low-wage workers climbs. By 1997-8, those earning 50 percent or less of the SAWW are almost twice as likely to be denied as the highest-paid workers.
The relative contributions to changes in the population of claimants and changes in the risk of denial were estimated in an Oaxaca decomposition. Initial analysis showed 62 percent of the increase in the denial rate between 1984-5 and 1997-8 was attributable to changes in the variable coefficients (how the association of variables with claim denial had changed) and 32 percent to changes in the variable means (how the population of claims had changed). To further elucidate the nature of these changes, the relative contributions of the coefficients and means for each variable were calculated and are shown in Figure 5.
The majority of the increase in denial rates can be attributed to the increased rates of denial for low-income workers (12.7 percent of the overall increase), full-time workers (6.5 percent), neck/back MSDs (10.9 percent) and claims where notice to the employer was delayed (5.8 percent). A very large contribution is made by the increase in the number of claims with an undefined type of injury (44.1 percent of the increase) and, to a lesser extent, by an increase in the proportion of low-income workers (10.1 percent). There has also been a sizeable decrease in the number of claims for back/neck MSDs (-16.0 percent).
The results of the present analyses, in part, confirm and then extend the findings of Berry (2002). Claimant age, low wages, delays in notice, type of insurer and type of injury were significantly related to claim denial in both studies. Some important variables in the earlier study could not be included in these analyses, because of limitations in the data available for the earlier cohorts examined here. Specifically, we were not able to determine if the risk of claim denial associated with job tenure, prior claims, type of firm and insurer market share -- all factors found associated with claim denial in Berry's analysis -- has changed since 198-5. There are also a number of other potentially important factors that could not be evaluated in this study. In particular, measures of the employee's occupation, employee race and ethnicity, employer size, the employer's workers' compensation experience, the insurer's loss experience and trends in judicial decisions were not available.
There are some differences of note in the findings for comparable variables. Here gender, marital status and -- sometimes -- residence are significantly related to claim denial, with females and unmarried claimants always at higher risk of denial and Twin Cities residents at higher risk in all of the time periods, except 1984-5. Berry found gender to be significant only among unmarried claimants. In part, this may be a product of differences in analytic methods (logistic regression vs. linear regression), the specification of variables (separate variables from gender and marital status versus a combined variable) and the set of variables used (e.g., the lack of data about occupation and job tenure in this analysis).
In addition, cause of injury was not found to be related to claim denial, being replaced by a variable evaluating the compatibility of the cause with the type of injury reported -- a variable not used by Berry (2002). A striking difference in the findings is the relationship of employment status to claim denial. Berry (2002) found that part-timers had a higher rate of denial than full-timers, the opposite of the results here. It is possible the inability to control for job tenure in this analysis accounts for the disparity.
The analysis of claim denial rates over time has shown that the increase in claim denials in Minnesota is explained by changes both in the characteristics of claims and claimants on one hand, and in insurer behavior on the other, with the latter accounting for a little more than three-fifths of the increase in this model. The largest contribution (44 percent of the overall increase) to the change in the rate of denials has been the increase in the number of claims where the type of injury is undefined. More than 12 percent of the entire increase in denials is attributable to an increase in the rate at which the claims of low-wage workers are denied, when all the other factors available are controlled; and another 10 percent of the increase is attributable to an increase in the number of claims by low-wage workers. The rate at which back and neck musculoskeletal disorders have been denied has gone up almost as much. Another 6 percent of the increase is explained by an increasing intolerance with delays in notification of an injury. There has also been an increase in the kinds of claims that have always been denied, particularly those with an undefined type of injury.
Some of these results support the initial model of insurer behavior hypothesized to explain claim denials. There is some evidence for both intolerance of uncertainty and suspicion in the factors found to be associated with claim behavior. Uncertainty is created when there is not sufficient evidence about the claim to identify the type of injury, and when the cause of injury is incompatible with the type of injury. Claim denial is more likely in both of these situations, and the increase in denials in Minnesota has been strongly affected by the increase in claims with undefined injuries. Likewise, the causal link between work exposures and a disease is often subtle and controversial. Musculoskeletal disorders -- predominantly sprains, strains and cumulative trauma -- have also become the subject of a contentious national debate about their connection to work activities. Insurers are likely to be suspicious of these claims and that is supported by the high risk of denial found here for those types of injury. Delays by the employee in notifying the employer of an injury are also likely to arouse suspicion and, again, a higher risk is found with longer delays having increasing likelihood of denial.
However, other results are not so obviously explained by either intolerance of uncertainty or suspicion. The small but consistently elevated risks for women, unmarried claimants and Twin Cities residents do not appear to be explained by an uncertainty about the injury. Nor is there any obvious suspicion that attaches to these situations. It is possible that in some of these cases, the insurer is concerned about the employee's attachment to the workforce, the potential for cost shifting of uninsured non-occupational medical expenses or other factors associated with gender, marital status and residence not included in these analyses.
Another possibility is claim denial is influenced by an insurer's past experience in litigating these kinds of cases. After a denial has been made, the employee can file a claim petition and obtain a judicial hearing contesting the insurer's decision and seeking an award of benefits. Which denials are contested and which denials are ultimately upheld by the courts may have an effect on which types of claimants are denied in the future. Future studies should incorporate some measures of insurer experience with adjudication of denials as a possible explanatory variable.
Berry (2002) suggested the size of anticipated losses may influence claim denial: If a large loss is feared, the insurer may be motivated to deny the claim. This might explain the higher denial rate for older workers who may be less likely to make a full recovery from an injury and so require long-term benefits, but it does not easily explain the findings for women or unmarried claimants. It certainly does not explain the findings for low-wage workers. Since the rate at which indemnity benefits are paid is determined by the employee's wage at the time of injury, high-wage workers would receive higher benefits and so losses would be greater on these claims. But it is low-wage workers who are denied more often.
Indemnity benefits for temporary wage-loss (temporary total disability or TTD) in Minnesota are subject to maximum and minimum limits. The TTD maximum means that high-wage workers receive less than two-thirds of their gross wages in compensation. For low-wage workers, the TTD minimum means they receive more than two-thirds, and sometimes 100 percent, of their gross wages. Since TTD benefits are not subject to taxes or withholding, some low-wage workers may be getting more in wage-loss benefits than they did in earnings. This raises the possibility of an increased moral hazard  in these cases that could increase the number of claims or the duration of a claim. Therefore, it is possible that insurers are influenced by their suspicion of this moral hazard. As hypothesized by Berry (2002), the insurer then is motivated by the anticipated loss because of these concerns.
Given the increased importance of low wages as a risk for claim denial, it becomes more important to understand what leads insurers to deny these claims. Further study is needed to elucidate the possible mechanisms underlying this risk. In particular, future analyses should examine the effects of other aspects of the claimant's employment (occupation, job tenure, employer size, employer safety record, unionization, access to health insurance), insurer business practices (market share, changes in premiums, loss ratios, litigation experiences), claimant characteristics (other demographic factors, previous claims, previous denials, previous litigation) and the workers' compensation system (wage-replacement rates).
4. Berry, D. (2002). Explaining claim denials. DLI Research Reporter. May 2002
[Now at: www.dli.mn.gov/RS/ClaimDenialProj1.asp].