Instituted as part of Hiram Johnson's reforms in 1911, the Division of Workers' Compensation (DWC) is in charge of providing California workers and employers with a no-fault means of resolving claims resulting from on-the-job injuries. In exchange for the certainty of payments for medical, indemnity, and rehabilitation costs, workers waive some of their rights to sue employers. Employers, on the other hand, are provided with a means to spread risk, by having their possible exposure to injury loss capped.
1. Voluntary Payments: The insurer makes voluntary payments and issues are not in dispute.
2. Compromise and Release: The employer and worker agree on all disputed issues. The insurer makes payments and the worker, in exchange, releases the insurer from all future obligation.
3. Stipulated Award: Similar to Compromise and Release, except that the worker does not release the insurer from all obligation. Usually the stipulation is for future medical benefits.
4. Finding and Award: A judge awards the worker a particular payment from the insurer.
5. Dismissal: A judge dismisses a case, and no award is made. This is largely done in cases where the worker fails to meet a court appearance.
6. Take Nothing: A judge rules a claim without merits, and does not award a settlement to the worker.
Of these possible resolution methods, the best (in terms of DWC costs and time) is voluntary payment. The next two methods (Compromise & Release and Stipulated Award) are a second best solution, while the last three (Finding and Award, Dismissal, and Take Nothing) involve high legal and preparation costs regardless of outcome.
Medical-legal reports are conducted to determine the following issues:
* Worker's eligibility for benefits
* Ability to return to work
* Permanent and stationary status of the employee's medical condition
* Existence and extent of permanent disability
* Employee's preclusion or likely preclusion from engaging in his or her usual occupation
* Extent and scope of medical treatment
* Existence of new and further disability
These exams are conducted at the request of the worker, the employer, and on some occasions at the request of a DWC/WCAB referee/judge.
Under current law, these evaluations are carried out by the injured worker's treating physician(s) or a Qualified Medical Evaluator (QME) as approved by the Industrial Medical Council (IMC). QMEs can be acting as an Agreed Medical Evaluator (AME)/Independent Medical Evaluator (IME) agreed upon by both parties or appointed by a judge, a QME chosen by either the employer or the worker in a dispute where the worker is represented by an attorney, or a QME chosen from a panel when the worker is not represented. Prior to the 1993 reforms, the evaluator chosen by parties in a disputed case was not required to be a QME if the worker was represented.
The cost of medical-legal exams on permanent partial disability (PPD) claims has shown a steep decline since peaking for the 1991 accident year. For the insured community, the total cost of medical-legal exams performed on PPD claims by 40 months after the beginning of the accident year has declined from a high of $394 million for the 1991 accident year to an estimated $64 million for the 1994 accident year. This is a decline of 84 percent in the cost of medical-legal reports and represents a saving of $330 million.
This decline in costs reflects changes in all the components of the cost structure. Consider medical-legal exam costs on PPD claims as comprised of the following components:
As the calculation of total system costs makes use of data derived from each section of this report, a detailed description of how each element in the above equation was calculated will be given in each relevant section.
The number of PPD claims for the 1989 through 1992 accident years shown in Table 2 is based on Workers Compensation Insurance Rating Bureau (WCIRB) second level individual case reports. Projections of PPD claim levels for the 1993 and 1994 accident years at a similar level of maturity (30 months following the beginning of the policy year) were made by UC DATA for comparison. The frequency of claims is broken out by major and minor disabilities.
The decline in PPD claims is driven by a decline in the number of covered workers and the number of reported injuries. As we will discuss below, a substantial portion of the decline in PPD claim frequency may also be the result of a steep drop in the number of claims with a psychiatric component.
The number of major and minor PPD claims in the system for years 1989 through 1992 was obtained from second level individual case reports (30 months after the beginning of the policy year). The survey data was weighted to reflect the distribution of cases in the second level individual case reports. The total number and total cost of exams for the system was derived using data from the WCIRB on the total number of PPD claims for these accident years.
To make projections on the frequency of major and minor PPD claims for claims in the 1993 accident year, we first obtained 1993 data from the first level individual case reports (18 months after the beginning of the policy year). Second, we calculated the historic growth rate in PPD claim frequencies (major and minor) from first level to second level individual case reports. This growth rate was then applied to 1993 first level data to obtain projections comparable to 1989-1992 second level data.
Projections for 1994 were made by using Division of Labor Statistics and Research data on the frequency of disabling injuries for 1993 and 1994. The ratio of 1993 to 1994 reported disabling injuries was applied to the estimate of 1994 PPD claims to estimate the number of 1994 PPD claims. Since the frequency of PPD claims has been falling more rapidly than the frequency of disabling injuries, this may slightly overestimate the number of PPD claims for 1994.
The average cost per exam has declined from a high of $987 for 1990 accident year claims to the current estimate of $606 for 1994 claims. This is driven by two important changes. First, the Medical-Legal Fee Schedule underwent several revisions that substantially reduced the charges rebuttably presumed reasonable. These maximums are controlled by the date of exams rather than the date of injury. Most 1993 and all 1994 exams are concentrated in the period following Aug. 3, 1993 when the fee schedule was at its lowest level. Only a fraction of 1991 and 1992 accident year claims have exams in the lowest period. The 1994 accident year estimate of $606/exam is the best estimate of the future cost of exams under the current medical-legal fee schedule.
Second, the average cost of exams is also determined by the mix of specialties of the reporting doctors. Some provider specialties have higher than average costs. The larger the proportion of exams in costly specialties the higher the average cost. Psychiatric exams are substantially more expensive than exams by other specialties. The dramatic decline in the number of psychiatric exams is responsible for 17 percent of the decline in the average cost of all exams. To illustrate, if the proportion of exams that were psychiatric had remained steady at the 1991 accident year level, the average cost of medical-legal exams would have been $56 greater for the 1993 accident year.
For survey years 1991 and 1992, average cost per exam was obtainable from the survey, as all claims have been observed for 40 months from the beginning of the accident year.
For 1993 claims, we determined the average cost per exam at 28 months. Based on our analysis of prior years, controlling for the fee schedule in effect, the average cost of exams was not found to fluctuate over the life of the case between 28 and 40 months. Accordingly, we projected average costs per exam in 1993 to remain constant between 28 and 40 months. This could overestimate the cost/exam slightly, because a small fraction of exams in the 1993 accident year occurred prior to August 1993--in periods when more costly fee schedules were in effect. Accordingly, average cost per exam for 1993 injuries is likely to decline slightly between 28 and 40 months.
1994 accident year exams all occur in the period after the least costly fee schedule went into effect. Accordingly, we assumed the average cost per exam would remain constant at the current level.
Chart 3 illustrates the decline in the number of exams per claim in the years from 1989 through 1994. This 52 percent decline is in part a result of changes made that were meant to reduce the "dueling docs" syndrome, other changes such as presumed accuracy of treating physician reports, the number of exams allowed per specialty in litigated cases, and, as noted below, the decline in psychiatric exams due to the higher threshold required for cause.
The number of exams per claim at 40 months from the beginning of the accident year for the 1989-1992 accident years were obtained from the survey.
Data from the survey are available for the average number of exams/claim for the 1993 accident year valued at 28 months from the beginning of the accident year. Using data from the previous panels of the survey, historic growth rates were calculated between 28 and 40 months in the average number of exams per claim. The 1993 data were "grown" by this rate.
For 1994 claims, a ratio of 1993 to 1994 average number of exams/claim was estimated for comparable claims (claims incepting from January through July and valued at 16 months from the beginning of the accident year). This ratio was applied to the estimate of the 1993 exams per claim projection for 40 months to obtain an estimate of exams per claim at 40 months for the 1994 accident year claims.
The 1989 and 1993 workers' compensation reforms were intended to decrease system costs and the amount of time needed to settle cases. Many of these reforms were meant to directly or indirectly affect the medical-legal process.
Legislation has changed the amount rebuttably presumed reasonable for medical-legal reports. Prior to 4/3/93, these were determined by specialty (Orthopedic, Neurological, Internal/Cardiology, Psychiatric, and All Other). After that time, the specialty was no longer a determinant, but the reports were separated into the following four categories:
Supplemental (ML101): Includes examination of patient within one year of initial evaluation.
Basic (ML102): All evaluations, other than Supplemental, Complex, or Extraordinary.
Complex (ML103): Includes three or more of following:
a. 2 or more hours physician face to face with patient;
b. 2 or more hours of record review by the physician;
c. 2 or more hours of medical research by physician;
d. addressing the issue of medical causation;
e. addressing the issue of apportionment
Extraordinary (ML104): Used in place of ML101-103 where significantly more time is needed than the code allows.
A premium above the fee schedule amount was allowed where an interpreter was necessary or the evaluator was acting as an "Agreed Medical Evaluator" (AME).
These changes would be expected to impact the cost of medical-legal reports. The minor adjustments to the fee schedule made prior to 8/3/93 are likely to affect only the portion of tests whose cost was at or near the rebuttable maximum. Tests with costs much above or below the fee schedule targets would likely be unaffected by these small changes.
The very large change to the maximums introduced for exams conducted on or after 8/3/93 should affect a much wider range of exams, and would be expected to reduce costs more dramatically than previous reductions. Again, tests whose costs were considerably below the new maximum would presumably not be affected.
The average cost of medical-legal reports can also be affected by the mix of reports. A decrease in the relative frequency of reports by a more expensive specialty could reduce the average cost of reports even while all specialties individually were more expensive. For example, if the frequency of psychiatric reports (the most expensive reports) declined enough, that alone could reduce average cost of all reports even if all specialties, including psychiatric, had increased in cost.
Alternatively, if reports requested by employers were less expensive on the average than those requested by workers (as we will discuss below), a decrease in the frequency of reports requested by employers relative to workers would appear to drive the cost of the average report higher. This argues for an effort to disaggregate the analysis of medical-legal reports by studying them by both specialty and source of request.
The Margolin bill created the Qualified Medical Evaluator (QME) and alternative dispute resolution mechanisms. In addition, the statute limited each party to a single medical-legal report on any issue by each appropriate specialty or subspecialty. Chapter 892 also prohibits admission into evidence of any medical opinion, other than a report from the treating physician, that was obtained before the service of the assessment by the Agreed or Qualified Medical Evaluator.
AB 110 (Peace) 1993 Stat. Chap. 121, effective July 16, 1993 made numerous changes to the process of selecting medical evaluators, which could be expected to reduce the number of reports. Chapter 121 allows the treating physician, who is primarily responsible for the care of the injured worker, to serve as the worker's Qualified Medical Evaluator. Legislation in 1989 had restricted the use of the treating physician by requiring separate report(s) by Qualified Medical Evaluators.
Chapter 121 also provided that where an additional comprehensive exam is obtained, the findings of the treating physician are presumed to be correct except with a preponderance of evidence, unless both parties elect QMEs. This encourages one or both parties to avoid seeking additional evaluations.
Other legislation restricted medical-legal examinations during the first sixty days, except at the request of the employer, and limited the employer's financial responsibility to a single comprehensive evaluation covering all issues.
Several legislative changes were expected to reduce the frequency of psychiatric evaluations by restricting the range of psychiatric injuries considered compensable. The changes modified evidentiary rules, imposing a "preponderance of evidence" requirement. A series of changes increased, over time, the degree to which the cause of the injury or illness had to be related to employment.
Assuming that the proportion of psychiatric injuries/illnesses remained steady relative to other injuries in the compensation system, these restrictions should eliminate a number of these injuries from compensability, because the psychiatric component fails to meet the threshold conditions for compensability. However, it could be argued that higher standards of proof lead to the need for more evaluations to establish whether the threshold conditions have been exceeded.
The Margolin-Bill Greene Workers' Compensation Reform Act of 1989 made changes in the treatment of psychiatric injuries, effective for injuries that occur on or after January 1, 1990. Chapter 892 narrowed the scope of compensability for psychiatric conditions by imposing a requirement that the employee demonstrate industrial causation by a preponderance of the evidence. Under prior law, "substantially in the record as a whole" was the sole evidentiary standard for all workers' compensation issues. Chapter 892 further required that at least ten percent of the causation of a psychiatric condition be attributable to "actual" employment factors before the employee was eligible for benefits.
AB 971 (Peace), effective July 1991, made additional changes by applying a further restriction denying recovery for those psychiatric injuries resulting from regular and routine employment unless the employee has worked for that employer for at least six months.
SB 223 (Lockyer), effective for injuries occurring on or after July 16, 1993, changed the causation standard on psychological injuries to require that actual events of employment must be predominant as to all causes combined. However, if the psychiatric injury results from a violent act or from direct exposure to a violent act, the employee is required to prove by a preponderance of the evidence that actual events of employment were a substantial cause of the injury. Cal. Lab. Code #3208.3(b)(3) (amended by Chapter 1242) defines substantial cause as at least 35-40%. Chapter 1242 also provided that psychiatric injuries are not compensable if substantially caused by a lawful, nondiscriminatory, good-faith personnel action. Further restrictions were also placed on recovery for psychiatric injuries following notice of termination of employment or layoff. The injury must have occurred prior to the time of notice of termination and one or more of the following conditions must exist: 8 1) sudden and extraordinary events of employment caused the injury; 2) the employer had notice of the injury prior to notice of termination or layoff; 3) employment records contain evidence of previous treatment of the psychiatric injury; 4) a finding of sexual harassment; or 5) evidence that injury occurred after notice of termination or layoff, but prior to the effective date of the termination or layoff. 
Legislative changes can be expected to change outcome measures of the decision-making process by affecting the disability measurement process. The determination of permanent disability under the statutes and regulations guiding the California system is at best imprecise. This has led to numerous calls to change the way in which we determine permanent disability in our state. Different doctors can give very different evaluations to the same patient.
Doctors having "conservative" and "liberal" interpretations of the level of disability are to some extent known to informed parties. Informed parties are individuals or entities that participate in the system on a frequent basis, for example, employers, insurers, or attorneys. Workers would not be expected to know the biases of doctors since they participate once or at most infrequently. Two well-informed parties could be expected on average to reach unbiased decisions given equal levels of knowledge in choosing their experts. Workers, acting alone, might not be expected to reach a similarly unbiased outcome. Rather, their PPD awards would be expected to be biased downward if information about the evaluating doctors' interpretation is unknown to the worker and the selection is made by an informed party.
Chapter 892 required that a notice be included with an injured worker's final payment of temporary disability stating the employer's position with respect to the employee's eligibility for permanent disability compensation. Alternate procedures for resolving medical disputes were established dependent upon whether the employee was represented by an attorney.
If an employee was not represented by an attorney, Chapter 892 prohibited the employer from obtaining an agreement on the selection of a medical evaluator. Instead, the statute required that in the event of a medical dispute with an unrepresented employee, the employer must request that the Medical Director assign a panel of three Qualified Medical Evaluators from which the employee can make a selection. This eliminated the prior common practice where the injured worker was given a medical evaluation by a doctor suggested or assigned by the employer's representative.
The 1989 reform introduced new mechanisms for resolving disputes. AB 276 provided for a mandatory settlement conference conducted by a referee not less than ten days, nor more than thirty days, after the filing for adjudication. If the settlement conference does not resolve the dispute, the regular hearing will be held within seventy-five days of the filing of the application.
If an injured employee is represented by an attorney, AB 276 requires arbitration in disputes over insurance coverage and rights of contribution. Beginning 1/1/91, arbitration will be required in cases of permanent disability when there is a low percentage of disability and the case cannot be heard within 110 to 150 days of the filing of the application for adjudication.
The introduction of arbitration and mandatory settlement conferences (MSC) were meant to reduce the need for hearings and decisions and to speed the resolution of cases.
Table (4) summarizes the possible impacts of the legislative changes introduced since 1989. Here the dimensions across which the impacts are considered are:
1. Frequency of medical-legal reports
2. Cost of medical-legal reports
3. Outcomes for workers
4. Time to resolution
The following sections will examine the different impacts (listed by column above) that reform legislation had upon the California Workers' Compensation system.
Data for our analysis comes from the WCIRB Permanent Disability Survey. The survey summarizes accident claim activity, including the type and cost of specialty exams, and whether the case was settled and if so what method of settlement was employed. The survey also has data on estimated permanent disability rating, but does not include data on the geographical place of injury, any data about the worker, or the type of work.
Because accident claims are collected as part of an insurance policy, claims cannot be sampled directly. Instead, the WCIRB collects all policies incepting on or after July of the year prior to the accident year through July of the accident year. Using second level individual case reports, which value claims at 30 months after the beginning of the policy year, the WCIRB identifies accidents:
* with dates of injury in the targeted accident year
* defined as permanent disability claims
Individual claims are stratified by injury severity. Claims with estimates of permanent disability equal to or greater than 25 percent are considered "major" disabilities, while claims with ratings under 25 percent are considered "minor" disabilities.
A stratified random sample of 3,500 claims is then drawn, with 1,500 claims coming from the "major" category, and 2,000 claims coming from the "minor" category.
The sampling scheme is intentionally designed to oversample "major" disability cases. In reality, approximately one fifth of all cases are "major" when evaluated at thirty months after policy inception. The oversampling of these cases is necessary to allow enough data to be gathered so that separate inferences can be drawn about this relatively small population. A negative consequence of this oversampling, however, is that the data necessary to "re-weight" the survey sample population back to the relative distribution of major/minor injuries in the full population of injured workers has to be inferred for the 1993 and 1994 accident years. Weights were available for the 1989 through 1992 accident years. Weights for the 1993 and 1994 accident years were calculated based on estimates of the distribution of major and minor injuries at 30 months after the beginning of the policy year.
These sampled accident claims are re-examined at 52 and again at 64 months following the start of the target year to determine whether claims have been resolved or otherwise changed. At the present time, complete 64 month data in the Survey exists only for target years 1989 and 1990. The 1991 target year data has been sampled through the 52 month (2nd panel) period, and the 1992 data exists through the 40 month period (1st panel).
The data available for the 1993 and 1994 accident years comes from "special panels." These were drawn at 28 and 16 months respectively from the beginning of the accident year. Both of these special panels consist only of 350 claims. As a result variances for all estimates made for these years will be much wider than for the earlier target years.
In addition, because of the different "maturity times" by target year, our analysis is forced to limit comparisons between years by different degrees of case maturity. For example, 1992 claims can be compared at 40 months to 1991 and earlier claims. However, 1992 claims can be compared to 1993 cases only at 28 months.
The "final" word on the effects of reforms upon the workers' compensation system through target year 1994 will therefore have to wait for the third report which will be collected in April 1999 and available in early 2000.
The weights for individual claims for the major and minor strata were not available from the WCIRB. The Rating Bureau creates the weights by evaluating the claims at 30 months after policy inception months based on the Unit Statistical Report second level reports. We are required to use the PD level reported on the Permanent Disability Survey. These data are from claims when they are several months more mature than thirty months. These two data sources are likely to be very close, but not exactly the same. For future efforts, we will try to obtain the actual weights used by the WCIRB.
In most situations the analysis is done using data available on cases 28 and 40 months after the beginning of the accident year. This allows us to compare data across accident years at similar points in the average maturity of claims. It should be noted that claims occurring on Jan. 1 will have 12 months more activity than claims for accidents occurring on Dec. 31 of the same accident year. Consequently, in some situations we use the actual maturity of the claims in months since date of injury. This is done, for example, with the analysis of time to resolution and the incidence of psychiatric exams.
The 1993 reforms made substantial changes to the structure of the medical-legal fee schedule and the level of reimbursement to the physician. Because insurers are required to pay for a medical-legal examination performed within parameters set by statute, regardless of who requested the exam, fee schedule levels are intended to 1) prevent the insurer from being over billed, without 2) creating excessive legal disputes within the system over reimbursement. As such, we would expect that a lowering of the fee schedule levels would lower examination costs, all other things being equal. Table (5) presents the medical-legal fees rebuttably presumed reasonable for each type of exam between 11/1/86 through 4/3/93. Since 8/3/93 the fee schedule has been considered prima facie evidence of the reasonableness of fees charged for medical legal expenses.
Prior to 4/3/93, fee schedule amounts tended to increase each year or fluctuate in a fairly narrow range. Between 4/3/93 and 8/3/93, no fee schedule was in place. After 8/3/93, the traditional system of fee schedule levels assigned by examining doctor specialty type was replaced with a set of flat fees and the maximums were reduced substantially.
In 1993 regulations introduced a premium for exams performed by a Qualified Medical Evaluator acting as an Agreed Medical Evaluator. Setting the level of this premium presents an issue for analysis. Under the present fee schedule, AME fees are allowed to be 25 percent higher than those for either worker or employer requested examinations. It is unclear from the regulations the purpose of the premium. This difference may be intended to compensate AMEs for more complex requirements of an exam and responding to both parties. Setting this premium may be informed by analyzing prior and current premiums for AMEs. Table (8) shows the observed premium for AME first exams versus exams requested by workers and employers.
In the Table (6), we present average costs for the first specialty exam for cases opened between 1/1/89 and 12/31/92 by the source of the request (either the worker, the employer, or an AME). The first exam was used in an effort to standardize on a single type of exam as much as possible. In this case we are using the first exam as a proxy for a "comprehensive medical evaluation." For purposes of this analysis we have decided to compare the costs between the period after 8/3/93 and the four periods from 1/1/89-8/2/93.
Table (7) has the same data as above for claims with dates of injury in 1993 and 1994. These data are not strictly comparable to the above table. First, a substantial portion of the claims, especially 1994 claims, may not have had sufficient time to have had a first medical-legal evaluation. Second, the sample sizes for 1993 and 1994 are small, making the sample sizes in cells of the table below small.
* Fewer than 5 claims in sample
Average first exam costs fall for all three requesters with the exception of exams performed by AMEs in specialties other than orthopedic and psychiatric. This provides evidence that the recent changes in the fee schedule levels were effective in controlling specialty exam costs.
TABLE (8) OBERSERVED PREMIUM OF AME EXAMS OVER OTHER REQUESTS
AME Premium over Worker Requested Exams
AME Premium over Employer Requested Exams
to 8/2/93 Orthopedic exams
to latest available
to 8/2/93 Psychiatric exams
to latest available
All Other exams
to latest available
All Other Exams
Three observations are apparent from Table (8).
1. The first is that in practice, observed premiums for AME exams, prior to the adoption of the premium setting regulation had for the most part been above the 25 percent premium. However, the choice of 25 percent seems fairly consistent with the data.
2. Second, in the period after 8/3/93 the premium may have increased slightly for orthopedic exams which are by far the most prevalent type (approximately 60%) of all specialty exams. The "all other" category premium increased as well (this category has greater variability in costs and heterogeneity in exam type, so changes are harder to evaluate).
3. Third, employers are able to control the cost of medical exams far more effectively than are workers. This is not surprising as employer requested exams are more directly under the control of insurance carriers who have greater leverage to negotiate better rates. It is a debatable issue whether the premium for AMEs should be evaluated against exams requested by the employer or the employee.
The WCIRB annual reports on the reforms have indicated a decline in the number of reports requested per claim. The nature and causes of this decline have not been clear. Here we use claims that have been resolved within 40 months and 28 months after the beginning of the accident year to further analyze this trend.
The choice to base this comparison on resolved claims was done to avoid the possibility that reforms sped up or slowed down the timing of exams on claims. For example, suppose the number of exams stayed the same, but they occurred earlier in the life of a claim. If we valued all claims at similar points in time, it would appear that the number of exams was increasing. Here we avoid that problem by only looking at claims that have been resolved. On closed claims, we know the number of tests that will ever occur.
We have also split the claims into major and minor injuries. For the 1993 claims we do not have second level case report data on major and minor injuries. Therefore, it is necessary to either estimate the distribution of injuries and apply weights to the different types of cases, or to treat major and minor injuries separately. We employ both approaches in this report.
When these data are further disaggregated, it is clear that the decline in the number of exams is being driven primarily by the improvement on the represented claims. Chart (4) and Chart (5) highlight the data, broken out by represented/unrepresented and major/minor injuries.
Regardless of whether the injury is major or minor, the represented cases in this period have had more exams than the unrepresented. However, this gap is narrowing. While the frequency of exams on unrepresented claims has changed little, the change on represented cases has been substantial.
CHART (4) AVERAGE NUMBER OF SPECIALTY EXAMS AT 40 MONTHS AFTER BEGINNIG OF ACCIDENT YEAR: RESOLVED CLAIMS
CHART (5) AVERAGE NUMBER OF SPECIALTY EXAMS AT 28 MONTHS FROM BEGINNIS OF ACCIDENT YEAR
The introduction of mandatory settlement conferences (MSC) was meant to reduce the need for hearings and decisions, and to speed the resolution of cases. Have the new resolution mechanisms accomplished these goals? The answer from the WCIRB survey data is no.
Have these resolution methods had unintended impacts on the frequency of medical-legal reports? The answer is a "qualified maybe." Chart (6), Chart (7), and Chart (8) look at the resolution of cases at 40 months from the beginning of the accident year.
There has been no significant change in the rate at which cases close. Except for the decline between 1989 and 1990, the percent of cases closing within 28 months has shown no change through 1992. This issue is discussed in more detail in Section II-E, Time to Resolution.
Moreover, the demands on the DWC/WCAB for claim resolution have increased, possibly as a result of the new resolution methods. Instead of reaching voluntary settlements that do not involve active participation by the DWC/WCAB, parties are increasingly using one or more dispute resolution methods.
For this analysis, we have redefined the variables in the following manner:
Voluntary settlements (here called "True voluntary") include all cases where a compromise and release agreement (C&R), a stipulated settlement, or voluntary payment is indicated as the method of resolution (Question 2) and the only dispute mechanisms listed are "settled w/ attorney" or "none of the above" (Question 8).
Judicial settlements include all cases where the dispute resolution method was "Findings and award", "Take nothing", or "Dismissal". These cases were considered the most expensive and burdensome for the state's resources.
Other settlements are those using the several dispute resolution mechanisms, but not resulting in a formal decision.
The survey does not reveal whether a given mechanism was used more than once, so the true burden on the WCAB resources is obscured.
We see a decline in the percentage of cases where a voluntary agreement was reached without intervention by the DWC/WCAB and an increase in the percent of cases where the DWC has been involved. This is true for both major and minor injuries. Each of these subgroups exhibits the same pattern of little or no change in the rate of judicial settlements and a movement of otherwise voluntary settlements into the use of dispute mechanisms.
As will be discussed in more detail in Section II-E, this has not been associated with a more rapid settlement of cases. Consequently, costs expended by the DWC are not being balanced by quicker resolution for the parties. Apparently, claims that would otherwise have settled voluntarily are now using at least one dispute resolution mechanism. This may be simply a form of brinkmanship, where parties who would never have proceeded to a formal hearing have an intermediate option that they may feel enhances negotiation. Or it may simply be that given an option requiring a low threshold of preparation and less finality, they choose to move forward a step.
CHART (6) PERCENT OF CLAIMS RESOLVED WITHIN 40 MONTHS
CHAR (7) RESOLUTION: MAJOR PD CLAIMS RESOLVED WITHIN 40 MONTHS
CHART (8) RESOLUTION: MINOR PD CLAIMS RESOLVED WITHIN 40 MONTHS
While the frequency of reports done on claims resolved within 28 months or 40 months has declined, the apparent movement of cases from voluntary settlement to other resolution mechanisms may be attenuating some of the savings to the system.
As the following tables and Chart 9 show, claims in the "Other" category require 15-40% more tests to resolve. This gap appears to have widened even as claims have apparently moved from the voluntary processes into dispute resolution.
We cannot conclude that this migration of claims between the two categories leads to additional tests. Additional study would be required to determine this impact. Several explanations are possible. For example:
* Informal dispute mechanisms encourage parties to get additional tests to strengthen their case before a third party
* The cases migrating from "Voluntary" to "Other" categories are the ones with the most tests
* Subgroups within either category that are not migrating are driving the difference.
CHART (9) MEDICAL-LEGAL REPORTS, RESOLVED CASES AT 40 MONTHS
The legislative changes affecting psychiatric claims were expected to reduce the frequency of psychiatric evaluations by restricting the range of possible injuries considered compensable for the psychiatric component. The reforms imposed evidentiary rules requiring a "preponderance of evidence." They also increased, in several steps, the degree to which the cause of the injury had to be related to employment.
Assuming that the proportion of psychiatric injuries remained constant relative to other injuries in the compensation system, these restrictions should eliminate a number of these injuries from compensability, or at least the psychiatric component, because they fail to meet the threshold conditions for compensability. However, it could be argued that higher standards of proof lead to the need for more evaluations to establish whether the threshold conditions have been exceeded.
The period from 1990-1991 has anecdotally been considered a period when numerous stress claims with little legitimacy were put forward.
Much of the improvement in the average number and the average cost of medical-legal exams per claim is a result of reductions in the number of psychiatric claims.
As shown in Chart 10, from a high for the 1991 accident year to the 1993 accident year, the total cost of psychiatric exams dropped by $87.6 million, a savings of approximately 93 percent on the cost of psychiatric related medical-legal exams. This represents 28 percent of the overall reduction in all medical-legal costs between the 1991 and 1993 accident years.
Between 1991 and 1993, over a third (36%) of the reduction in the average number of medical-legal exams per PPD claim is a result of a drop of 87 percent in the average number of psychiatric exams per PPD claim.
The decline in the total cost of psychiatric exams is due to both a drop in the number of psychiatric claims and a decline in the cost/exam. As shown in Chart 11, the number of psychiatric claims (PPD claims with at least one psychiatric medical-legal exam) has declined by 93 percent between accident year 1991 and 1993. It should be noted that for injuries occurring after 1/1/94, a single physician (or single physician for each side) evaluates all open issues. Consequently, for 1994 and later injuries, the survey may no longer allow us to determine cases with a psychiatric component. This will often be obscured when the evaluating physician chosen is from another specialty.
CHART (11) NUMBER OF PPD CLAIMS WITH AT LEAST ONE PSYCHIATRIC EXAM
Part of this decline in number of psychiatric exams is driven by the overall decline in all types of PPD claims. In addition, as Table 5 indicates, the decline is driven by factors specific to psychiatric claims that reduced the portion of PPD claims that have a psychiatric component. Anti-fraud efforts, crackdowns on "mills" and similar efforts may be the most important factors in the declining frequency of psychiatric claims. Limits on post-termination stress claims introduced in 1991 and improvements in the California economy that reduced mass layoffs may have also played a role. The number of psychiatric exams per claim peaked in the period from 1990 through the first half of 1991. The decline in psychiatric exams per claim began in mid-1991 and accelerated in 1992. Legislation requiring that work be the "predominant" cause of a psychiatric complaint may also play a role in the decline in psychiatric claims in late 1993 and 1994.
Unlike PPD claims in general, the decline in total medical-legal costs on psychiatric claims is not due to a reduction in the average number of psychiatric exams per psychiatric claim. Although it is early to say with certainty and 1993 and 1994 samples are small, indications are that the average number of psychiatric exams on psychiatric claims has not declined sufficiently to show a clear improvement. The number of psychiatric exams per psychiatric claim, when valued at 16 months after the beginning of the accident year, has held steady at an average of approximately 1.4-1.5 exams on each claim. At 28 months, 1993 psychiatric claims had more psychiatric exams on average than psychiatric claims in the preceding years.
It is interesting to note that the rise in the proportion of all claims with psychiatric exams from 1989-91 was accompanied by only a small rise in the frequency of requests by workers. Chart (12) shows the distribution of exams by the source of request in each of four periods. For this purpose claims were evaluated at 16 months after the accident date. (Note: the 1993 sample is smaller and is subject to considerable variation with only small changes in the number of exams.)
Requests by employers for psychiatric exams rose more rapidly after 1989 than requests by workers. Moreover they declined less steeply after the reforms that went into effect in mid-1991. This is also true of claims in the 1993 sample. Consequently, as indicated in Chart (12), employers are replacing workers as the predominate source of request for these exams.
CHART (12) PSYCHIATRIC EXAMS BY SOURCE OF REQUEST
The impact of reforms upon the time to case resolution has not been significant. Based upon the hazard analysis detailed below, controlling for injury severity, the number and type of specialty exams, and presence of employee legal representation, we do not find statistically significant improvements in time to case resolution in the 1990s. Indeed, the only statistical difference attributable to year that we find is that cases following 1989 appear to be resolved more slowly than cases in the 1989 injury year.
The rest of our analysis seems to follow conventional wisdom well. Injury severity is associated with longer resolution times, as is the presence of an employee legal representative--a proxy for dispute over the initial evaluatory findings. The presence of one or more orthopedic exams was associated with more rapid time to case resolution..
As with all regression models, the results must be submitted with a set of caveats. The data set involved in this analysis did not have many explanatory variables. As a result, it was not possible to control for individual worker traits, the nature of the injury, etc. Nor was it possible to try a variety of different model forms to get a better idea of fit and robustness. This analysis, as such, represents the application of standard and suitable tools, upon a very limited data-set.
Before looking at how the reforms have changed the rate of case resolution some general discussion is needed. Whether or not a case is resolved within a given time period is a function of a number of factors including:
* degree and type of injury
* amount of dispute between workers, employers, and the DWC
* the effect of the 1989 reforms
The best measures we have of the degree of injury are the permanent disability rating and the awards for total medical and incurred indemnity. A second set of measures exists which give at least a hint at the type of injury. These measures consist of reports of specialty medical examinations. Specialty medical exams are performed by approved medical doctors at the request of either workers, insurers, or the DWC and are coded as either Orthopedic, Neurological, Internal, Psychiatric, or All Other. We have coded a variable for each of these specialty types, which counts the number of appointments of each specialty type reported for each case. A greater number of specialty exams should be similarly negatively associated with resolution time.
The amount of dispute between parties in any given case is also partly measured through the number of specialty exams. Typically, when there is dispute over the degree of injury or the level of disability both sides in the case will request their own set of specialty exams. This leads to an interesting pattern of a worker requested exam, followed by an employer exam, followed by a worker exam.... In addition to the number of specialty exams, a second measure exists to identify the level of case dispute. In cases where there is greater degree of dispute legal costs are likely to be higher. These costs are present in the data-set. The amount spent on legal services by employers seems a good indicator of case friction. We would expect employer legal costs to be negatively associated with resolution odds. We have also included a variable indicating the presence of worker legal representation.
Based on this discussion, our basic model is a Cox regression of the following form:
Results for this model are presented in the attached technical appendix, as well as a discussion of diagnostics and alternate forms that were calculated. Based on the above model, the following curves show the percent of cases not resolved by the number of months since the accident date, by the accident year of the case. This type of curve is called a "survival function."
CHART (13) SURVIVAL FUNCTION
As the above chart shows, 1989 cases (all other things being equal) have a lower survival rate at any given month than either of the three following target years. This implies that 1989 cases were resolved faster, all other things being equal. The results (presented in the appendix) also indicate that while there is not a statistically significant difference between the 1991, 1992 and 1993 time to resolution coefficients, there is a significant difference between 1989 and the following years. This suggests that whatever changes did occur in time to case resolution in the DWC system happened between 1989 and 1990, rather than later.
The data made available by the WCIRB is a sample of 3,500 cases opened in each year. These cases are sampled so that 1,500 are of workers estimated in the Unit Statistical Reports to have "major" injuries (disability equal to or greater than 25 percent), and the remaining 2,000 are drawn from the population of "minor" disability cases (disability < 25 percent). These claims are sampled in three panels, the first at 40 months, the second at 52 months, and the third at 64 months after the beginning of the accident year.
WCIRB files present several large difficulties for use in analysis. Specifically, during the analysis of the WCIRB data three major problems emerged:
* Problem 1: Sampling Methods
* Problem 2: Few Attribute Variables
* Problem 3: The Infamous "Culling" Problem
The Permanent Disability Claims Survey was designed following the 1989 reform legislation. The design was well thought out, but should be updated to satisfy issues arising from the recent legislative reforms. Several issues need to be addressed by adding:
* Two new sources of requests
* The fee schedule code (101-104) for reports
* A geographic variable
* Current WCAB office
As part of its administrative function, the WCIRB collects case information through the Unit Statistical Report (USR) on insurance policy written with each employer. In order to get a sample of workers injured in an accident year (for example 1989) the WCIRB selects policies incepted between 7/1 of the prior year (7/1/88) and 6/30 of the accident year (6/30/89). They then sample claims on these policies with dates of injury in the target year (1989). The problem with this sampling scheme is that it is not a random sample of claims and therefore provides a distorted picture of seasonal rates of injury.
CHART (14) SIMPLIFIED EXAMPLE OF WCIRB SAMPLING
For example, suppose (for the moment) that both injury occurrence and firm policy inception dates are uniformly distributed. This would give us a distribution of policies and injuries like that shown in Chart (14). The shaded lines indicate policies surveyed and the numbers (1s) above each line represent an injury. The unshaded lines show policies and injuries that would be excluded. As the chart shows, even in such a simple example, this sampling method causes over-sampling of cases incepted in summer months and under sampling of injuries in winter months.
Because of the WCIRB sampling design, it will appear that rates of injury are at a maximum in the summer and decline in the winter. Furthermore, even if in the real (but presently unknown) state of the world accident rates actually do vary between seasons, it will be difficult to identify the actual pattern of injuries. As the plot of accident dates from the survey below shows, it appears accident rates rise in the summer and fall in the winter.
CHART (15) ACCIDENT DATE DISTRIBUTION, PD SURVEY DATA
How much of this pattern is real and how much of this is due to the sampling method? No data is currently available to show the distribution of dates of injury for all injuries reported as permanent disability claims on the second level individual case reports. However, we have used data from the DWC, Rehabilitation Unit (RU) On-line System. This system reports the date of injury of all claims in the vocational rehabilitation system (approximately 160,000 for 1989-92). For this comparison, the PD Survey data and the RU data are presented as the percent of claims for each year with dates of injury in each month. This figure is included below.
While the two data sets should not match exactly, if the WCIRB data were correctly sampled from the same population as the rehabilitation claims, the graphs should generally follow one another, with deviations that show no consistent pattern. The injuries follow a cyclical pattern with regular dips in November and December and not much variation the rest of the year. However, as predicted in our earlier analysis, the deviation of the WCIRB sample from the expected distribution follow a consistent and repeating pattern. The WCIRB sample shows large underrepresentation of injuries in the earliest and latest months of the year, with large overrepresentation of injuries occurring in the middle of the year.
Does this matter? There are several reasons to think that the nature and severity of injuries varies by month. Some industries are seasonal (e.g., agriculture). These industries can be associated with more or less severe injuries, with injuries of a particular type, and with greater or lesser degrees of litigation. Also if policies incept in different industries at different times of year, under or over sampling of particular industries is increased. The average size of firms underlying the policies also varies by month of policy inception. We expect the medical-legal experience of big and small firms to differ. Finally, some important industries are also cyclical, in other words follow the business cycle (e.g., construction), and are strongly correlated with more serious injuries. Over or under sampling could capture different proportions of particular claims from year to year.
We ran several tests to compare the data between months of the year. These tests disproved the hypothesis that there was no difference between months. This was true for PD level, incurred indemnity, and time to resolution.
The impact of this sampling bias is at present unclear, however the results from this survey must be viewed with some skepticism until corrections can be made. Until these corrections are made there are two primary ways in which faulty sampling will impact the results. In the case of the Hazard modeling, the sample bias against winter months yields a distribution of start dates which may not be independent of censoring times. If this indeed turns out to be the case, analysis becomes a much more complicated matter requiring the use of a time-dependent censoring process.
The second impact this type of sampling could have is to overemphasize seasonal differences in injury characteristics. This would have the effect of biasing downward regression estimates for the impact of reforms on the type of resolution method.
Ultimately, this problem can be fixed by using weights to adjust for the relative over-sampling based on month of injury, just as they are used in the existing analysis to correct for the intentional over-sampling of severely injured workers. The real difficulty lies not in the nature of the "fix" but rather in the difficulty in obtaining data on the actual temporal distribution of cases.
The second major problem with this data-set is that it has not been integrated with other available data, for example from individual case reports. As a result the number of variables that can be used to help define traits about individual workers, as well as the ability to link other external sources of information into the file, are extremely limited. The broad effect of this problem will be low model explanatory power. More specifically, however, the limitations in the data present a challenge in being able to devise indicators related to the needs of the analysis. This is notably the case with measures of injury.
Accurate measures of the degree of injury are essential for the Hazard model presented in this paper. In this case, however, degree of injury does not mean exactly what it does in common usage. Rather than indicating a particular medical condition, in terms of the DWC's activities, the severity of an injury really means the "social cost." This cost depends upon a wide number of factors including the nature of the injury, the degree to which the injury can be repaired, the amount of time the injury will take to heal, and the degree to which the worker can be employed in his/her present occupation. In a perfect world, therefore, the determination of injury severity would be based upon a wide array of information about the worker and the exact nature of the case.
Unfortunately, the amount of information available with which to measure the severity of injury is limited to three variables. Each of these variables tells an incomplete story about the nature of the injury. The first measure is the estimated percent of permanent disability. Percent disability reflects a combined estimation by the DWC of the degree to which an injury can be repaired, and the degree to which the worker will be employable in the open labor market.
According to the DWC's guidelines, the level of permanent disability is determined by a relatively fixed set of rules. Raters use the worker's age, occupation, and impairment (as determined by a specialist) to assign an estimated percent using tables in the California Schedule for Rating Permanent Disability. In practice, however, the real practice of determining the degree of impairment also involves a fair amount of negotiation between the department, the injured worker, and the employer. It is also common for estimates of disability to be adjusted following the closure of a case, so that the final indemnity and medical schedule is consistent with the disability estimate. As a result, the WCIRB itself does not regard the rating as being very accurate, and commonly relies upon only a bivariate statistic to indicate major or minor impairment. Table (10) lists several threshold levels of injury for back and lower extremities and their associated disability ratings:
TABLE (10) DISABILITY, WORK CAPACITY GUIDELINES
precluding very heavy lifting
of 1/4 of pre-injury capacity for lifting
very heavy work
of 1/4 capacity for bending, stooping, lifting or other activities
of 1/2 pre-injury capacity for lifting
heavy lifting, repeated bending and stooping
of 1/2 capacity for bending, stooping, lifting or other activities
bending, stooping, or lifting
to light work
do work in standing or walking position, minimum demand for physical effort
to semi-sedentary work
time standing and 1/2 time seated with minimum physical effort
to sedentary work
in sitting position
Despite this pessimistic appraisal, given the subjective nature of the adjustment process about the figure determined by the Schedule, the large sample size, and no prior belief that permanent disability is systematically adjusted differently for cases of a particular type, the measure contains more information than does the bivariate major/minor classification.
Once employment trends in the state were controlled for, the distribution of permanent disability findings do not change by year. This indicates that while the measure is not perfect, it has at least been applied in a stable way over the period 1989-1992. The following chart shows the distribution of permanent disability determinations.
CHART (16) DISTRIBUTION OF PD RATINGS
Besides percent of disability, the data-set also contains information on two other measures of injury severity: total incurred indemnity and total incurred medical costs. Both of these measures give the amount of money paid to date and placed aside by the insurance companies for future indemnity or medical costs. These figures are determined by the injury, the percent of disability, and negotiations (either arbitration, legal, or voluntary) between the DWC, the worker, and the employer's insurer. Taken together, both of these measures have the advantage (with respect to percent of disability) of accurately reflecting the severity of an injury in terms of the costs involved in returning the worker to the workplace. Unlike percent of disability, however, the measures of total incurred costs do not fully reflect the severity of the injury in terms of the workers' costs, nor do they provide a good picture by themselves of the impact of the injury upon the worker's ability to seek future employment.
Given that each of these three measures captures a portion of the information on injury severity, rather than choosing one as being the most "reliable", this analysis relies on a constructed measure. Using information from all three measures, we have constructed a factor score for injury severity. This is not as ideal a solution as obtaining more data, however this tool has the advantage of combining all available information about the case.
A second solution to increasing the attributes in the data set is to link the PD Survey data to the Unit Statistical Report (USR). The USR data has numerous additional elements on each claim that could increase the usefulness of this dataset. This solution would require an effort by the WCIRB to build a more useful dataset for interested evaluators by merging data from both sources. This requires some determination of which elements should be combined from each source.
The third and final problem with the WCIRB data is so-called infamous "culling" problem. This problem stems directly from the different priorities placed on data use by administrative needs versus analysis functions. As the structure of data collection was designed primarily for administrative use, an emphasis was placed upon being able to present data in the system, rather than on being able to do anything with the data.
In the first panel of the survey sample cases were selected and data on these cases was collected. Illegible forms, incomplete entries, non-responses, or otherwise erroneous data was left blank. At each subsequent panel, however, all erroneous entries were "corrected" by bringing the appropriate data forward from past panels.
Following process of data checking this problem has now been fixed. Accordingly, this analysis is able to present results based on the latest panel data available for each year.
We are proposing a slight redesign for the PD survey to deal with several issues indicated above.
Two new sources of request are proposed for the table in Question 3. The legislative reforms created and strengthened the importance of the QME panels for unrepresented workers. The value of these changes is not clear since the current survey does not distinguish QMEs from other sources of request. While a QME from a panel acts somewhat like an AME, they are probably included under "W" for worker.
Also, the 1993 reforms increased the importance of the Treating Physician (TP) in the medical-legal process. In part, this was done by increasing the evidentiary force of the TP findings when one side obtains an alternate report by a QME. This was meant to reduce both the cost and frequency of reports. This policy cannot be evaluated using the current survey. Again, the TP is a hybrid, acting somewhat like an AME. However, a report by the TP could be reported either under "employer" or "worker."
The additions of two codes for reports by treating physicians and by QMEs from a panel would improve the value of the survey greatly. These codes could be the Medical-Legal Fee Schedule codes for Primary Treating Physician (92) and QME on uncontested claims (95).
The addition of a column for the code under which the report was billed would improve our information. Under prior law, specialty and date of service determined the applicable rebuttable maximum in effect. Under the current law, the codes ML-101 to ML-104 in combination with the date will be necessary to evaluate reports after 8/3/93 against the amounts presumed reasonable.
A geographic variable has been added at the top of the survey. This should be the current  zip code of the worker. A three digit zip code is sufficient. Often, there are issues raised about Northern v. Southern California or Los Angeles against the rest of the state. Most often, LA and Southern California are considered as having high and costly litigation rates and medical-legal report costs. Rural and urban is another distinction in analysis that is often requested. The current survey does not allow any analysis of the these data based on a geographic variable.
Several people have proposed the addition of a variable for the WCAB office (if any) where the case resides. The contention is that the WCAB office can have more to do with the dispute mechanisms used and time to resolution of cases than any other variable.
We have included a copy of a redesigned survey and new instruction sheet in Appendix 2.
The impact of reforms upon the time to case resolution is not clear. Based upon the hazard analysis detailed below, controlling for injury severity, the number and type of specialty exams, and presence of employee legal representation, we do not find a statistically significant difference for time to case resolution between pre- and post-1991. Indeed, the only statistically difference attributable to year that we find is that cases following 1989 appear to be resolved more slowly than cases in the 1989 injury year.
The rest of our analysis seems to follow conventional wisdom well. Injury severity, as measured by increased indemnity awards, is associated with longer resolution times, as is the presence of cost for employer legal representation. The presence of one or more orthopedic exams was associated with more rapid time to case resolution. The presence of an employee legal representative was associated with faster time to case resolution--presumably because representatives were able to maneuver their clients through the system more quickly.
As with all regression models, the results must be submitted with a set of caveats. The data set involved in this analysis did not have many possible explanatory variables. As a result, it was not possible to control for individual worker traits, the nature of the injury, etc. Nor was it possible to try a variety of different model forms to get a better idea of fit and robustness. This analysis, as such, represents the application of standard and suitable tools upon a very limiting data-set.
For this analysis we will look at cases with full panels (accident years 1989, 1990, 1991, and 1992). Before looking at how the reforms have changed the rate of case resolution some general discussion is needed. Whether or not a case is resolved within a given time period is a function of a number of factors including:
* degree and type of injury
* amount of dispute between workers, employers, and the DWC
* the effect of the 1989 reforms
The best measures we have of the degree of injury is the permanent disability rating, and the awards for total medical and incurred indemnities. A second set of measures exists which give at least a hint at the type of injury. These measures consist of reports of specialty medical examinations. Specialty medical exams are performed by approved medical doctors at the request of either workers, insurers, or the DWC and are coded as either Orthopedic, Neurological, Internal, Psychiatric, or All Other. We have coded a variable for each of these specialty types, which counts the number of appointments of each type reported for each case. We would expect that injury severity is negatively associated with the odds of case resolution at any given time. Likewise, the greater number of specialty exams should be similarly negatively associated with resolution time.
The amount of dispute between parties in any given case is also partly measured through the number of specialty exams. Typically, when there is dispute over the degree of injury or the level of disability both sides in the case will request their own set of specialty exams. This leads to an interesting pattern of a worker requested exam, followed by an employer exam, followed by a worker exam.... In addition to the number of specialty exams, a second measure exists to identify the level of case dispute. In cases where there is greater degree of dispute employer legal costs are likely to be higher. These costs are present in the data-set, accordingly the amount spent on legal services by employers seems a good indicator of case friction. We have also included a variable indicating the presence of worker legal representation. The presence of an attorney does not indicate the a settlement through the judiciary--the attorney may be called in by the worker to assist in arbitration, or even a voluntary settlement. We would expect employer legal costs to be negatively associated with likelihood of resolution in any time period.
Based on this discussion, our basic model is a Cox regression of the following form:
After weighting data to account for the difference between the sampling design, and the actual relative distribution of minor and major permanent disability determinations, the following regression estimates were obtained:
Variable B S.E. Wald df Z-Stat R Exp(B)
PERDISRA .0074 1.214E-03 3746.5792 1 6.1 .0140 1.0075
TOTII -3.56E-06 8.758E-07 1652.5080 1 4.1 -.0093 1.0000
TOTIMED -1.99E-05 1.247E-06 25356.168 1 16.0 -.0365 1.0000
SPECA .0434 .015 838.4955 1 2.9 .0066 1.0443
SPECI -.0106 .023 21.9165 1 .5 -.0010 .9894
SPECN .0250 .020 157.7122 1 1.3 .0029 1.0254
SPECO .0562 8.969E-03 3921.5362 1 6.3 .0143 1.0578
SPECP -.0094 .015 40.9610 1 .6 -.0014 .9906
EMLGLREP -3.03E-05 4.627E-06 4299.9890 1 6.5 -.0150 1.0000
ATTORNEY .1225 .031 1567.1919 1 4.0 .0091 1.1303
Looking at the results, our prior predictions seem to be born out--once a few interpretive complications are considered. Negative beta estimates are associated with cases which are resolved slower than the base rate, while positive estimates are resolved faster. Both Total Incurred Medical and Total Incurred Indemnity indicators move in the "right" direction--with larger awards being associated with more slowly resolved cases. Although the coefficient on permanent disability rating is positive, this is because determinations both at the low and high end of this measure are likely to be made quickly. As this measure varies only between 1 and 100, the magnitude of its impact on case time to resolution is small--and indeed composite scores for injury severity constructed using all three measures show longer times to resolution being associated with more severe injuries. Employer spending on legal representation also is associated with cases that take longer to resolve. By contrast, employee legal representation is associated with somewhat faster case resolution. As all of the coefficient estimates raised to base e are close to one, a one percent change in the explanatory variable affects resolution time by one percent.
For the most part the number of specialized exams also followed the general pattern--at least for the significant coefficients. Because weights must be integers for Cox regressions, the significance levels listed by the computer are incorrect and have not been included. Instead, we have calculated the number of corrected standard units away from zero for each coefficient (Z score). None of the negative coefficients for specialist exams are significant at the 95% level. In this first regression, we chose not to include variables for year, but instead to stratify the sample by year of injury. This makes the assumption that the process is a proportional hazard model, where only the baseline term S(t) varies by time. The following graph plots the cumulative survival functions, plotted at the mean of the covariates, obtained by this model:
FIGURE (A-1) SURVIVAL FUNCTION AT MEAN OF COVARIATES
While the coefficients on the exogenous variables followed prior predictions, the relationship between baseline survival functions do not. Rather than seeing an improvement in time to case resolution--holding our exogenous factors fixed--instead cases appear to survive longer. 1991 cases are surviving longer than either 1989 or 1990. While cases in 1992 may be slightly more prone to resolution than 1991 cases, the shorter observation time makes it very difficult to draw any final conclusions. The following table compares several values of the estimated cumulative baseline Hazard function. This is the sum of the rate of resolution prior to any period, ignoring individual case characteristics:
FIGURE (A-2) ESTIMATED CUMULATIVE BASELINE HAZARD FUNCTION
Before accepting these results a variety of diagnostics were performed. Multicolinearity is not an issue with this data, as exogenous variables are very weakly related. Point estimates were robust over a range of different forms (one of which will be discussed below) and given the limitations of the data, little more can be done to deal with misspecification. Although observations were made at identical times over the life-span of each panel, observation start times were not uniform. Cases could start anytime within the first year, and despite the sampling problems there is no prior reason to believe that censoring times are not independent of event occurrence.
In order to test this last assumption, sensitivity of the coefficient estimates to extreme deviations from censoring independence was performed. Running the regression twice, first with all cases indicated as resolved, and then with all cases indicated as being censored at 64 months (regardless of resolution) yielded very similar estimates. Accordingly, there is no reason to believe the independence assumption is violated. Similarly log-minus-log survival plots did not reveal any obvious faults in the assumption of proportional hazards between years. Accordingly, the use of time-dependent covariates does not appear necessary.
FIGURE (A-3) LML FUNCTION AT MEAN OF COVARIATES
In order to statistically test the difference in resolution between years, the above model was modified. Rather than relying upon the assumption of proportional hazards, this version assumes that base-rates between years are identical. Differences between years are accounted for by dummy variables in the covariate expression. The results of this regression are presented on the following page:
Variable B S.E. Wald df Z-Score R Exp(B)
PERDISRA .0074 1.213E-03 3743.3377 1 6.1 .0133 1.0075
TOTII -3.55E-06 8.754E-07 1648.0353 1 4.1 -.0088 1.0000
TOTIMED -1.99E-05 1.247E-06 25381.354 1 16.0 -.0346 1.0000
SPECA .0431 .015 827.7317 1 2.87 .0062 1.0440
SPECI -.0107 .023 22.1760 1 .5 -.0010 .9894
SPECN .0255 .020 164.1262 1 1.3 .0028 1.0259
SPECO .0561 8.967E-03 3919.2455 1 6.3 .0136 1.0577
SPECP -.0091 .015 38.2768 1 .6 -.0013 .9910
EMLGLREP -3.04E-05 4.627E-06 4311.8739 1 6.6 -.0142 1.0000
ATTORNEY .1214 .031 1542.4660 1 3.9 .0085 1.1291
DUM90 -.0736 .030 212.3378 1 2.5 -.0031 .9573
DUM91 -.0874 .031 795.8135 1 2.9 -.0061 .9163
DUM92 -.0967 .033 835.3219 1 2.9 -.0063 .9078
Coefficients between the two models are very similar. The coefficients for all three of the year dummy variables are negative relative to 1989, as expected from the prior model. Given that the sample sizes and standard errors are almost identical for each year coefficient, an eyeball ANOVA seems to show that the three coefficients are not significantly different from each other. This suggests that whatever changes as did occur in the DWC system happened between 1989 and 1990.
 By 40 months after the beginning of the accident year, approximately 97 percent of all medical-legal exams have been conducted.
 Data courtesy of Larry Law, Workers' Compensation Insurance Rating Bureau
 This may overestimate the number of exams/claim, especially for the 1994 accident year. Legislation in effect for injuries on or after 1/1/94 required comprehensive medical-legal evaluations to deal with all open issues, limiting this to a single exam, rather than a single exam by each appropriate specialty. This would tend to reduce the number of exams performed later in claims relative to earlier accident years.
 Some experts contend that the new fee schedule will tend to eliminate exams billed below the fee schedule levels, increasing costs. Two reasons are proposed to support this expectation. 1) Charges at the fee schedule amounts are considered "prima facie evidence" of reasonableness, as opposed to the old standard of rebuttable presumed reasonable," making it easier to defend any fee at the fee schdule level. 2) Some experts have suggested that the statute could be interpreted to require that doctors be paid at the fee schedule level, and not below.
 The impact of these alternative resolution methods has been inconsistent. While over a third of cases in the sample had had a mandatory settlement conference scheduled, arbitration, voluntary or mandatory, occurred in less than 1 percent of claims.
 See: Pacific Law Journal, vol. 215, 75-6. Also, LeVesque v. Workmen's Compensation Appeals Bd., 1 Cal. 3d 627. Note: the preponderance of evidence standard was extended to all workers' compensation issues. Leg.H 1993 ch.r, effective April 3, 1993 . L.C. 3202.5
 See: Pacific Law Journal, vol. 23, 806-7
8 See: Pacific Law Journal, vol. 25, 863-4
 Bill Molmen, General Counsel, Integrated Benefits Institute, contends that these restrictions on post-termination stress claims are easily circumvented and are unlikely to be a reason for the decline in psychiatric claims.
 see: Pacific Law Journal, Vol. 21, April 1990
 A discussion of the consequences of this sampling scheme is presented in Section VIII.
 The survey had a response rate that differed across the strata. The weighting of the survey data for this analysis takes into account the differential response rates.
 The distribution between major and minor in the original samples for 1993 and 1994 was unavailable from the WCIRB at the time of our analysis. The distribution drawn from the responses is as follows:
 see: LC sec. 5307.6
 The first exam also accounts for approximately 35% of all reports done.
 One caveat, the decline in the frequency of exams on closed cases (even when measured at similar points in time from accident) could be the result of changes in legislation on the mix of early closing claims. For example, the reforms could have worked to encourage the early resolution of long-term, low exam frequency claims. This, however, seems an unlikely explanation.
 note: This same analysis could be run with "Formal hearing" moved from the "Other" category to the "Judicial" category. This would capture cases where the WCAB was required to hold formal hearings, but 1) the issues heard did not resolve the case, and/or 2) a formal hearing was held, but before a decision was reached on case resolution, a settlement between parties was reached.
 The chart was drawn based on the survival function following from cases with the mean value for each of the factors in the model, and is useful for comparison purposes across years. The actual survival curve faced by any particular case will depend upon the exact case characteristics.
 Based on WCIRB data on California Workers' Compensation Experience by Policy Month, 1991 policy year second reports, 1992 policy year 1st reports.
 The current zip code has several advantages over the zip code at the time of injury. 1) The current zip code is more likely to be available from claims administrators who need to contact the worker and may overwrite the old address in the database if the worker moves. 2) The current zip code is useful for identifying the likely WCAB offices where the case is being handled.
 The basic problem here is that because non-integer weights cannot be used, new weights must be devised by multiplying the old weights by (in this case) 100. The SE is related to sample size by the square root of N, and therefore the computer given SE must be multiplied by 10.
 The cumulative survival function is the proportion of cases surviving at a particular moment.
 The test comes from Allison, Event History Analysis, 1984