California Commission on Health and Safety
and Workers' Compensation 

Seal of the State of California


Doctors and Courts:
Do Legal Decisions Affect Medical Treatment Practice?

An Evaluation of Treating Physician Presumption
in the California Workers’ Compensation System

Prepared for the Commission by
Frank Neuhauser
Survey Research Center, University of California at Berkeley

CHSWC Members
Tom Rankin, Chair
Allen L. Davenport
Jill A. Dulich
Leonard C. McLeod
Kristen Schwenkmeyer
Robert B. Steinberg
Darrel "Shorty" Thacker
John C. Wilson

Christine Baker
Executive Officer

November 2002


Table of Contents

Executive Summary

Introduction
PTP presumption: Statute and legal decision
Data
Methods
Results

ES-1: Average PD Award by Quarter of Injury and PTP Control
ES-2: Average medical cost in the 4th quarter of service, by control of PTP
ES-3: Percent change attributed to control and presumption

Caveats and future research:
Acknowledgements


Report

I. Introduction
II. Medical Care

(a) Workers’ Compensation Medical Care
(b) Primary Treating Physicians, Reporting Requirements, and Presumption
(c) California Workers’ Compensation Environment at the Time of Minniear

III. Model
IV. Data
V. Methodology:
VI. Results

Table 8: Mean values for dependent variables
Table 9: Percent change attributed to control and presumption
Table 10: Impact of quarter of injury > 97:1

VII. Discussion Section
Future Research

Table 4: Variable Definitions
Dependent Variables
Table 5: Impact of Minniear on Medical Treatment Cost
Table 6: Impact of Minniear on Medical Treatment Cost

    Standard errors in parentheses
          Table 7: Impact of Minniear on Medical Treatment Cost

Bibliography


Tables and Figures

Table 1: Medical Cost, Active Claims, by Controlling Party
Table 2: Natural Log of Medical Cost, Active Claims, by Controlling Party
Table 3: Average Transactions in Quarter, Active Claims, by Controlling Party
Figure 1: Probability of any services in by quarter after injury
Figure 2: Average medical across all indemnity claims by QOI and QOS
Figure 3 Average medical cost across all indemnity claims by quarter of injury and quarter from injury
Figure 4: Average PD award by quarter of injury and control of physician
Figure 5: Average medical cost for active claims in 4th quarter of service, by control of physician
Figure 6: Average medical cost, active claims, by gender and quarter of service
Figure 7: Average of natural log of medical costs, all claims, by gender and quarter of service
Figure 8: Percent of active claims under worker control


Executive Summary

Introduction

The California Legislature passed AB 110 in 1993. The bill was intended to reduce medical-legal costs by giving the opinion of the primary treating physician (PTP) the presumption of correctness in legal proceedings regarding permanent disability. In 1996 the WCAB handed down, en banc, the Minniear v. Mt. San Antonio Community College District, extending presumption to disputes over medical treatment. Subsequent to the Minniear decision the average cost of medical treatment on workers’ compensation claims began to increase at rates exceeding 15% annually. This study was conducted to determine the role that Minniear and PTP presumption played in the growth of medical treatment costs. The recent passage of AB-749, which essentially eliminates presumption, makes this analysis particularly relevant, since the impact of repeal will be factored into future insurance rates.

The study finds that Minniear had an important impact on the cost of medical treatment and the utilization of medical services. In summary:

Before the Minniear decision, when the worker controlled medical treatment, the cost in any quarter was 7.8% higher than when the insurer/employer controlled the choice of physician.

The Minniear decision had the effect of increasing this difference in average quarterly treatment costs when the worker controls the physician by an additional 11.3%.

Service utilization was 10.4% higher in any quarter when the worker controlled the physician. Minniear increased this difference by an additional 7.7%

During the period under study, the percent of medical treatment controlled by the worker increased from 27% to 40%.

The effect on total medical costs system-wide, estimated at $8.2 billion in 2003, would be at least 4.5% or $370 million.

If we attribute the change in the portion of medical treatment under worker control to Minniear, the impact would add an additional 1% or $80 million.

We did not examine the impact of Minniear on the duration of treatment. If, however, the impact on duration was similar to the impact on treatment intensity, the effect of presumption would be increased by an additional 4.5% for an upper bound estimate of the impact of $820 million.

Considering that the impact of Minniear was felt quite quickly, the impact of AB-749 should also be a quick reduction in expected medical costs for a substantial savings of $370-$820 million starting in 2003.

PTP presumption: Statute and legal decision

Workers’ compensation has generally been unsuccessful at introducing utilization control mechanisms, for example HMOs and preferred provider networks with risk sharing, that have been successful in the health benefit arena. Instead, workers’ compensation has relied on mechanism such as fee schedules that control the unit price of services and utilization review to control the units and types of services delivered. The utilization review process ultimately relies on decisions of the WCAB when there is a dispute concerning what is reasonable and necessary treatment.

In practice, only a small fraction of claims in any system actually reach a formal litigation process. However, previous research has argued that the outcomes of the small fraction of cases that are decided cast a "shadow" over the decisions of parties in the majority of disputes about their expectations and thus the value of pursuing a dispute. Consequently, rules that govern the parties’ balance of power in the legal process regarding medical treatment can have an important effect on medical costs.

The 1996 Minniear decision changed the rules governing disputes over medical treatment, casting just such a "shadow" over utilization review and treatment decisions. Minniear changed the balance by giving a legal presumption of correctness to the opinion of the PTP against all other opinions when the issue was medical treatment and, in addition, defined a higher standard of what was required to rebut the PTP’s opinion.

This decision coincided with a rapid increase in medical costs, suggesting a link between PTP presumption and higher medical treatment costs. This link and the effect of eliminating presumption as legislated in AB-749, is the focus of this study.

Data

The data that are used in this study were supplied by the California Workers’ Compensation Institute (CWCI) from its Industry Claims Information System (ICIS). These data are well suited for these analyses because we can identify individual transactions, the dates of service, and providers. Altogether we received data on 14 million transactions covering 300,000 indemnity claims with dates of injury and dates of service between 12/31/1992 and 1/1/2000.

The medical transaction files included information on the date of service, primary diagnosis (ICD9), procedure code (usually CPT), provider identification number, provider specialty, and billed and paid amounts. We are also able to discriminate between different types of transactions, such as medical treatment, medical-legal expenses, and medical cost containment expenses. For these analyses, we exclude medical-legal and cost containment expenses.

Methods

While a rapid increase in medical costs following the Minniear decisions is suggestive of a causal effect, many other characteristics of the workers’ compensation and medical treatment arenas could also be changing, making a simple before and after comparison unconvincing. Instead, we constructed a double-difference comparison that used a control group to factor out other trends in the market and isolate the effect, if any, of Minniear.

The double-difference comparison relied on the potential differences in the action of employer controlled and employee controlled doctors and the statutory provision for an initial 30-day period when employers can control treatment. In the analysis it is assumed that the employer controlled physicians have incentives aligned with those of the employer and consequently treatment patterns and the employer applied utilization review will not be substantially affected by changes to the legal process. We also identify changes of PTP after 30 days using a methodology described in the report.

We calculate medical treatment costs for each claim for each quarter of service after injury. We assign medical treatment costs prior to a change, if any, of the PTP to employer control and costs after a change to worker control. If all other trends, besides Minniear, affect employer controlled and employee controlled physicians similarly, then subtracting the change in employer controlled medical costs before and after Minniear from the change in employee controlled treatment cost, before and after, will isolate the change in cost due to Minniear.

We make these estimates for simple comparisons and after controlling for possible changes in demographic characteristics of claimants, diagnosis and severity, quarter of injury and quarter of service. We make separate estimates for the impact on average medical cost, average number of services, and a measure (average log of medical costs) which is similar to median cost.

Results

We first evaluated whether our definition of PTP and control of the physician gave us results consistent with our expectation of the timing and direction of both the statutory change (affecting permanent disability) and the legal decision (affecting medical treatment). We observed, as shown in charts ES-1 and ES-2, that in both cases costs exhibited a substantial change at the point in time the effect was expected and in the direction expected.

 

ES-1: Average PD Award by Quarter of Injury and PTP Control

 

 

 

ES-1 shows that by our definition of control, there was a significant break in the pattern of permanent disability settlements after the effective date of the legislation (1/1/1994). Prior to 1994 the worker received, on average, similar settlements regardless of who selected the physician giving treatment. After 1/1/1994, the value of the settlement was higher on average when the worker controlled relative to when the employer controlled.

ES-2 shows the impact of presumption on medical treatment for the 4th quarter of service relative to the injury date. Here, we do not see an impact on medical treatment costs coincident with the statutory change. This is consistent with the legislative intent. However, we do observe a substantial impact surrounding the Minniear decision which extended presumption to disputes over medical treatment. This effect was observed for all quarters of service studied except for the first quarter of service after injury.

ES-2: Average medical cost in the 4th quarter of service, by control of PTP

ES-2: 
  Average medical cost in the 4<sup>th</sup> quarter of service, by control of 
  PTP Chart

 

 

 

 

 

 

 

 

Combining all quarters of service and all injury dates, we used a regression model with controls for observable characteristics. The impact of control of the treating physician was estimated across all injury quarters. The impact of Minniear was estimated as the additional effect of worker control relative to employer control that resulted from Minniear. The estimates of these effects are given in ES-3 on the following page.

 

ES-3: Percent change attributed to control and presumption

Dependent Variable Independent Variable

Average Value (employer control)

Coefficient

Percent Effect

 
Average $Med in quarter (CQTR)

1551.9

   
  Worker Control (EECTRL)  

122.9*

7.9%

  Impact of Minniear (EECTRL*AFTDUM)  

175.5*

11.3%

Natural log average $Med in quarter (lnCQTR)

5.12

   
  Worker Control (EECTRL)  

.522*

68.5%a

  Impact of Minniear (EECTRL*AFTDUM)  

.160*

17.4%a

 
Average number of services in quarter (SQTR)

5.76

   
  Worker Control (EECTRL)  

.598*

10.4%

  Impact of Minniear (EECTRL*AFTDUM)  

.445*

7.7%

a For a semi-logarithmic functional form, the expression for the percentage impact of a dummy variable on the dependent variable is ecoefficient – 1. (Kennedy, 1996)

* Significant at p < .001

 

In terms of the overall impact of presumption on medical costs, the most important measure is the dependent variable for average cost (CQTR). As expected, when a worker controls the primary treating physician (PTP) the average cost in a quarter is 7.9 percent higher than when a worker with the same condition is treated by an employer selected physician. This was true even before Minniear. The impact of Minniear, applying presumption concerning medical treatment to the PTP, was to increase the differential when the worker controlled the PTP by an additional 11.3 percent. These results are all highly significant statistically (p<.01).

A characteristic of medical cost data is they are typically a highly skewed distribution with many lower and medium cost observations and a few very large observations. The average of the natural log (lnCQTR) gives an estimate that is more reflective of the central tendency of such a skewed distribution. This is much the same as using the median instead of the mean. Changes in medical care that increased very high cost services would tend to increase the average cost by a larger percent than the natural log of the average cost. The opposite would be true if the greatest change was observed on the natural log statistics. Observers who were concerned about the impact of presumption on medical costs were split on which cause might dominate.

The results for the log of medical cost are very informative. The direction of the effect of employee control and presumption are the same as for average medical cost, confirming those results. In addition, the percent changes on the natural log are larger than for the average medical cost, especially the effect of control absent presumption. Employee control increases the natural log statistic by 68.5 percent and the impact of presumption adds an additional 17.4 percent. The large impact of presumption on the average log of medical cost suggests that the impact of worker control on utilization is mostly an increase in the utilization of moderately priced services, for example physical therapy treatments. Similarly, the impact of presumption on log cost being larger than the impact on average cost was probably a product of more moderate priced services rather than increases in very costly services such as surgeries.

The results for the number of services in a quarter also confirm the impact of control and presumption. The results are most similar to the average cost variable. The effect of control is to increase the number of services by 10.4 percent while the impact of presumption is an additional 7.7 percent increase.

In general, the impact of the application of presumption to medical treatment has been to increase both the frequency and intensity of moderately priced treatments. For example, workers selecting their own doctor might have more physical therapy visits and receive more specific therapies with each visit.

Since the portion of medical treatment dollars paid for treatment under worker control in the period after Minniear was roughly 40%, we can make estimates of the impact of the Minniear decision on medical cost (and consequently, the impact of the elimination of presumption). A very conservative estimate is that presumption increases medical costs by 11.3% on the 40% of medical cost under worker control or 4.5%. With an ultimate paid medical cost of $8.2 billion for calendar year 2003 polices, this translates into $370 million.

There are several reasons to consider this estimate conservative. First, between 1996 and the end of the study period, the percent medical treatment controlled by a worker selected doctor increased from 27% to 40%. While the timing of the increase in worker control was proximal to the Minniear decision and is consistent with an increase in employee incentives to select their own treating physician, we could not definitively attribute all of this increase to Minniear. If the shift was due to Minniear, this would add an additional 1 percentage point increase to the estimate of the impact of Minniear on medical costs.

Second, the trend has been towards increasing the portion of treatment under worker control. If this trend continues, 40% would be too low as an estimate of medical dollars under worker control.

Third, we examined the impact of presumption on the cost in any given quarter. We did not examine whether Minniear caused treatment to continue for more quarters. We might expect this to be the case, but we cannot make a definitive statement without additional research. More frequent treatments may imply longer treatment and consequently more quarters of treatment, e.g., 10 PT visits over 2.5 quarters rather than 8 treatments over 2 quarters. If treatment duration was as affected by presumption as treatment intensity, this would double the estimate of 4.5 percent that we made above.

Finally, we make the assumption that employer selected physicians have incentives aligned with the employer and consequently, Minniear did not impact their treatment pattern nor the employer/insurer application of utilization review. This is a very conservative assumption. If presumption also increased in medical treatment cost when the employer selected the doctor our estimates would be low by more than twice the increase in utilization by employer doctors.

Because our estimates are conservative across the board, and particularly because we attribute no change to the employer selected doctors, it may be more appropriate to choose an estimate towards the high end of the combination of 1) observed increases in employee controlled treatment cost (4.5%), 2) the additional cost due to a greater portion of treatment under worker control (1%), and 3) likely, but untested, increases in duration (4.5%) for a total change of 10% ($820 million for calendar year 2003 injuries).

Caveats and future research:

Consistent with previous studies of pricing mechanisms, and risk sharing, we find important effects when the legal process governing utilization review is changed. Relaxing the controls on a particular set of providers does lead to substantial increases in the frequency and intensity of services delivered by those providers. However, this does not necessarily imply that the affected doctors are abusing the system and over treating patients. An important area for future research is to examine the impact of easing utilization review on outcome for injured workers. In particular, we would like to see if the increased treatment was reflected in better or worse outcomes along three dimensions 1) faster return to work, 2) improvements in self-reported health status, and 3) better earnings and labor force participation. The Commission is considering making this a focus of future efforts to evaluate medical treatment in workers compensation through the multi-state WINS survey.

The determinations of who controls initial treatment and when control of treatment switches to another party are difficult in administrative data. The method used here gives results that are consistent in timing and direction with both the statutory and legal changes. A more accurate methodology would involve examining individual files for a large number of cases and determining the status of control over the life of claims. A plan to do this is currently being evaluated. However, the impact of imperfect assignment of control on our current results is almost surely to bias down our estimates of the impact control and presumption, making our cost estimates too conservative.

Acknowledgements

The author would like to thank the Commission on Health and Safety and Workers’ Compensation (CHSWC) and particularly, Christine Baker, Executive Officer, and the CHSWC staff for the extensive support, financial as well as intellectual, that they afford throughout this project.

Special thanks should go to the California Workers’ Compensation Institute (CWCI) for making available data for the project from the Industry Claims Information System (ICIS) database and for technical assistance with use of the data, with particular thanks to Alex Swedlow, Executive Vice-President of Research & Development.

I would like to thank Jon Stiles of UC Berkeley for his insight and assistance with statistical and methodological issues.

I also acknowledge the large number of people who commented on earlier drafts and presentations making the final product more accurate and informative.


Report

I. Introduction

Medical treatment accounts for 13 percent of Gross Domestic Product, approximately $1.2 trillion. (U.S. Centers for Medicare and Medicaid Services, 2001). In workers’ compensation, medical costs represented 45 percent of total loss payments in calendar year 2000, approximately $4 billion in California (WCIRB, 2001a). On an incurred basis, the Worker’s Compensation Insurance Rating Bureau of California estimates that medical cost will rise to $8.2 billion in 2003 (WCIRB, 2001b). Consequently, for competitive and public policy reasons, employers, unions, health insurers, employer benefit plans, large providers, and the government are constantly searching for ways to control medical cost.

Most research on medical cost control has focused on pricing mechanisms and risk sharing. Pricing mechanisms involve the relative and absolute pricing of the range of services. Medicare fee schedules are an example. Virtually all payors with at least some fee for service structure will use a fee schedule to reimburse providers and influence the delivery of services. Most fee schedules are modeled on the Medicare approach developed during the 1970s and 1980s. There is a rich body of research on the impact of changes to the Medicare fee schedules on utilization (see Rice 1997 for a review). Generally, Rice finds early studies indicate decreases in Medicare payments led to decreases in billed amounts, but increases in the volume and intensity of services. There was also a pattern, within specialty, of changes in the mix of services away from the services with decreased reimbursement towards those where reimbursement was increased or unchanged.

In workers’ compensation where fee-for-service predominates, fee schedules play an important role in cost containment. A California Commission on Health and Safety and Workers’ Compensation (CHSWC) study (Neuhauser and Koehler, 1996) found that modification of the fee schedule formula for medical-legal reports reduced the average price of reports by 48 percent for an annual savings of $110 million. Two other reports by CHSWC proposed 1) a fee schedule revision for pharmaceuticals (Neuhauser, et. al., 2000), expected to save over 30 percent ($89 million in 2003) on prescription drug costs and, 2) introduction of an outpatient surgery facility fee schedule (Kominski and Gardner, 2001) estimated by CHSWC to save in 2003 between 33 percent ($70 million) and 73 percent ($160 million) (CHSWC 2002).

Risk sharing mechanisms are an approach to controlling medical cost and utilization that has been intensively studied (for reviews see, Robinson, 2001; Goosden et. al. 2001; Rice, 1997). These approaches center on shifting some of the responsibility and incentive for controlling medical treatment to the provider. Capitated rates paid health maintenance organizations (HMOs) or pooled risk arrangements like preferred provider organizations (PPOs) are examples. Again the evidence is strong that these arrangements reduce cost and service utilization. However concerns are often raised about access, under treatment, and patient satisfaction.

Risk sharing arrangements are uncommon in workers’ compensation medical delivery. Statutory limits on employer control of treatment restrict the ability of payors to contract with providers over extended periods. There is some evidence that, when used in workers’ compensation, these approaches produce results similar to those found on the non-occupational health side (Cheadle et. al, 1999; Kyes et. al, 1999; Borba and Parry, 2000). However, efforts to legislate adoption of managed care within workers’ compensation as a public policy initiative on a jurisdiction wide basis, such as in Florida managed care legislation or the California HCO Program, have not been successful.

A very common cost control approach that is similar to risk sharing is sharing some of the cost between patient and insurer. This takes two forms. First, the patient may be required to make a co-payment with each service unit (e.g., visit or prescription). There is strong evidence in the research literature that co-pays have a significant impact on the utilization of services (Harris, Stergachis, and Ried, 1990; Simon, Grothaus, Durham, VonKorff, and Pabiniak, 1996; Reuveni, et. al., 2002). On the other hand, the evidence concerning the impact of deductibles (where the patient is responsible for a portion of initial costs in a time period up to a maximum) on utilization is more mixed (Schellhorn M, 2001; Bakker and van Vliet, 1995).

Co-payments and deductibles have almost never been used in workers’ compensation where the employer is nearly always required by statute to cover the entire cost of medical treatment. The only example that the author is aware of is Florida. That state has introduced a $10 co-payment per visit for workers who receive treatment after they have reached maximum medical improvement. We are not aware of any evaluation of Florida’s program.

Utilization review (UR) and treatment guidelines represent another approach to cost control. Under these processes treatment is subject to review within an organization (medical provider network, insurer, etc.) or outside the organization through independent medical review (IMR) boards. UR typically involves the application of treatment guidelines to decisions about the appropriateness of treatment on individual cases. These guidelines may be developed internally by payor and/or provider groups or adopted from or applied by sources outside an organization. The last step in utilization review, when service denial is disputed, is almost always some form of arbitration, usually binding on the parties.

Much research has been devoted to pricing mechanisms and risk sharing, but little research has examined the impact of utilization review in general and the arbitration process in particular on treatment patterns and cost. In part, this lack of investigation may reflect that only a very small number of claims have disputes that ever reach the arbitration stage. However, Robert Mnookin and Tony Kornhauser, ("Bargaining in the Shadow of the Law," 88 Yale L.J. 950 (1979)) point out that the decision to pursue a case, whether to settle a case once pursued, and the size of the settlement agreed to by the parties are influenced by the legal decisions on previous cases which establish the expectation of the parties about future outcomes. In this paper, we examine how the process of arbitration can cast a "shadow" over treatment and utilization decisions made by parties that very rarely expect to take a dispute to formal litigation.

While little research has been devoted to analyzing the impact of the legal process on utilization review, it is a common area for public policy action through legislative intervention. These legislative actions often have their genesis in one or a small number of horror stories about critically ill patients being denied treatment. While these treatments are often expensive, the number is seen as small and actions taken to "correct" the review process may be thought to affect only a narrow range of treatments. However, if the "shadow" created by these actions has an important effect on future treatment decisions, the cost implications may be substantial.

For example, in 1984, the Social Security Administration (SSA) altered the application process for Federal disability programs. Among the changes was an increase in the weight given to evidence provided by the applicant’s own health care provider relative to the SSA’s own consultative exam when an application for disability benefits was disputed. (see Stapleton, et. al 1998). According to a recent study (Autor and Duggan, 2001) the changes to the application process had a profound effect on the number of adults awarded disability benefits.

More recently, in California the Legislature passed AB-55 (Migden) which required that all managed care health plans and disability insurers provide an enrollee, with the opportunity to seek an independent medical review whenever health care services have been denied, modified, or delayed by the plan if the decision was based on a finding that the proposed services are not medically necessary. The IMR in a substantial minority of cases approves treatment initially denied by the health care provider.

The study of Social Security Administration disability suggests that the weighting of competing medical opinions can be important, affecting even the decision to apply. The SSA’s weighting of different medical opinions is similar to the central issue of this report. However, there were several other simultaneous changes that did not allow the authors of the SSA study to separate out the effect of weighting of provider opinions. In addition, the central area in dispute involved medical-legal issues, in that case disability status, rather than medical treatment issues. The IMR process in California does affect medical treatment issues, but the legislation does not change the weight given different party’s medical evidence. Rather it added an additional opportunity for the patient to prevail, by mandating an independent review after the dispute process within a plan was exhausted. No study has yet been done to evaluate the impact of the IMR legislation on the costs faced by managed care organizations in California.

The study of changes that impact the UR process has special salience for workers’ compensation. Medical cost increases are generated through three main mechanisms, 1) cost increases on the price of specific services; 2) changes in the mix of services away from less expensive towards more expensive, often new (and sometimes better), services; and 3) increases in utilization. Of these three areas, workers’ compensation only has policy and regulatory tools that are equivalent to the non-occupational health arena for the first area, price controls on specific services (fee schedules). The mix of services and utilization are controlled in the non-occupational arena largely through risk sharing between payor and provider (e.g., HMO and PPO arrangements) and payor and patient (co-pays and deductibles). To a lesser extent non-occupational health uses UR and treatment guidelines. Workers’ compensation relies to a greater extent on the application of utilization review and resolution of disputes through the administrative law process. A formal hearing process is rarely reached in either system. For the California workers’ compensation system fewer than 1 percent of claims ever have a formal hearing to decide a medical treatment issue (Pace et. al., 2002). However, given the limited set of cost control and quality assurance tools available in workers’ compensation, if this process of adjudicating parties’ disputes casts a "shadow" over other claims for treatment, the legal process may have an important effect on medical treatment in workers’ compensation.

Additional salience is lent to this research because several recent studies have found that medical treatment in workers’ compensation is substantially more costly than non-occupational medicine (Neuhauser, et. al., 2000; Durbin, Corro, and Helvacian, 1996; Johnson, et. al., 1996; Baker and Krueger, 1995; Swedlow, et. al., 1992; Zaidman, 1990). Estimates range from 130 percent to 200 percent for the workers’ compensation relative to group health when comparing the cost of treating the same diagnosis or pricing the same service.

This report evaluates specifically the impact of changing the weight given parties’ evidence on the pattern and cost of medical treatment on all cases that fall within the "shadow" of the legal process. The next section briefly describes medical delivery in the California workers’ compensation system and the legislative changes surrounding the weight given reports by the primary treating physician. In Section III we outline the model that will be used in the analysis. Section IV describes the data. Section V describes the data analysis methodology. Section VI gives the results, estimating the impact of the changes on medical treatment costs. Section VII discusses the implications of the study. The final section focuses on the potential for future research.

II. Medical Care

(a) Workers’ Compensation Medical Care

Labor Code Section 4600.

Medical, surgical, chiropractic, acupuncture, and hospital treatment, including nursing, medicines, medical and surgical supplies, crutches, and apparatus, including orthotic and prosthetic devices and services, that is reasonably required to cure or relieve from the effects of the injury shall be provided by the employer.

Medical treatment in the workers’ compensation system is paid for by the employer / insurer. Statute establishes several conditions governing medical treatment that bear directly on the motivation for utilization controls:

Coverage is "first dollar coverage," no deductibles or co-payments by the employee.

The employer is required to pay for all treatment that is reasonably required to cure or relieve the effects of the injury, and remains liable for any medical treatment necessary for the particular work-related injury for the life of the worker.

The employer under nearly all circumstances can determine the choice of medical provider only during the first 30 days of treatment.

The employee has the option after the first 30 days to select any doctor, chiropractor, psychiatrist, acupuncturist, and the like. If the employee notifies the employer prior to an injury, the employee can predesignate a treating physician for all treatment, including the first 30 days.

Medical treatment costs are governed by the state Official Medical Fee Schedule, which establishes "reasonable maximum fees paid for medical services" under workers’ compensation (Labor Code 5307.1).

Disputes between medical providers and insurers concerning medical treatment are common. The disputes involve the length, intensity or appropriateness of treatment and the charge for treatment. Medical treatment disputes may require the intervention of the Workers’ Compensation Appeals Board (WCAB), an administrative law court.

Cost containment is a major issue because treatment is almost entirely fee-for-service and co-pay or deductible requirements are forbidden by statute. Group health models are not easily transferable to workers’ compensation. Because the models involve contracting between payer and provider, shifting cost control responsibility and sometimes risk to the provider. However, the employer controls medical care for only the first 30 days.

The duration of control of treatment is important in workers’ compensation because medical treatment costs show very skewed distributions. While a large number of claims involve no treatment after 30 days, the majority of medical costs accrue on treatment delivered after the first 30 days (Ernst & Young, 1996a, 1996b). For example, low back injuries are the most common injury category in the compensation system, between 20 and 30 percent of claims depending on the definition of low back. For 90 percent of these claims medical treatment ends in less than 30 days, which is within the period of employer medical control. However, the majority of costs are incurred on claims lasting more than 30 days. Similarly, physical therapy accounts for 30 percent of the cost of low back treatment. Seventy-five percent of low back claims have fewer than four physical therapy treatments, but ninety percent of the physical therapy dollars are spent on claims with more than four visits.

II. (b) Primary Treating Physicians, Reporting Requirements, and Presumption

In the early 1990’s rapidly rising workers’ compensation medical-legal costs had reached historic highs. The medical-legal process involves reports by doctors covering legal issues such as indemnity and causation as differentiated from reports monitoring medical treatment. In 1991, medical-legal costs for employers had reached $600 million, approximately a fourth of all medical benefit costs (treatment and medical-legal) (Neuhauser and Koehler, 1996). In response to this crisis, the California Legislature passed AB-110 (Peace) in 1993 that made significant changes to the Labor Code (Marria, 1998). The legislation included two important changes regarding the role of the primary treating physician (PTP).

First, Labor Code Section 4061.5 was added stating:

"The treating physician primarily responsible for managing the care of the injured worker or the physician designated by that treating physician shall, in accordance with rules promulgated by the administrative director, render opinions on all medical issues necessary to determine eligibility for compensation. In the event that there is more than one treating physician, a single report shall be prepared by the physician primarily responsible for managing the injured worker's care that incorporates the findings of the various treating physicians."

Second, Labor Code Section 4062.9 was added stating:

"In cases where an additional comprehensive medical evaluation is obtained under Section 4061 or 4062, the findings of the treating physician are presumed to be correct. This presumption is rebuttable and may be controverted by a preponderance of medical opinion indicating an different level of impairment. However, this presumption shall not apply where both parties select qualified medical examiners."

Prior to enactment of 4061.5, treating physicians were not required to write medical-legal evaluations even though their reports were admissible in all cases before the WCAB (Kizer, 1998). These changes had an important impact on the legal and medical-legal process, a significant effect on permanent disability awards (Neuhauser 2000), ambiguous impacts on temporary disability duration (Neuhauser 2000), and virtually no impact on the overall cost of medical-legal reports (Neuhauser, 1999).

These new Labor Code sections triggered substantial litigation over the definition of primary treating physician, who "controlled" the PTP, and related issues. Consistent with the legislative intent, the early litigation at the Workers’ Compensation Appeals Board (WCAB) and California Courts of Appeal focused on disputes surrounding presumption and medical-legal issues. (See for example, Ralphs Grocery Company v. WCAB (Lara) (1995) 38 CA4th 820, 60 CCC 840 and Vergara v. State Comp. Ins. Fund (1996) 24 CWCR 204)).

The key expansion of the reach of L.C. 4062.9 was triggered by the judiciary in late 1996 when the WCAB issued an en banc decision, Minniear v. Mt. San Antonio Community College District 61CCC 1055 24 CWCR 261. In that decision the Appeals Board held among other issues that presumption in favor of the treating physician applied to disputes over medical treatment as well as permanent disability issues. In addition, the Appeals Board made it more difficult for the side attempting to rebut the party with presumption by deciding that 1) lay testimony, in particular by the injured worker, is not substantial evidence that can rebut presumption since the statute requires medical opinion, and 2) a QME report is not more credible because it is more thorough or recent. This constituted a major expansion of the reach of L.C. Sec. 4062.9, most importantly, the inclusion of medical treatment issues.

Since the Minniear decision, subsequent cases have narrowed the impact of the decision (Swezey, 2002). For example, in Keulen v. WCAB (1998) 66 CA4th 1089, 63 CCC1125, the court of appeal relaxed the restrictions on lay testimony by giving weight to the applicant’s credible testimony in conjunction with other evidence. In Bray v. Porterville Public Schools (2000) 28 CWCR 240, the Board summed the current status of Minniear as follows:

[T]he presumption is rebutted by [the] panel QME report, considered in conjunction with applicant’s testimony and the other trial witnesses. It is proper for the WCJ and the Board to consider the injured worker’s account of symptoms and work restrictions in order to accurately assess the conclusions reached by a reporting physician.

Numerous stakeholders, particularly employers and insurers, have contended that this extension of PTP presumption to cover medical treatment and raising of the burden of proof necessary to rebut presumption are major causes for the rise in medical cost on workers’ compensation claims in the late 1990s. They contend that the statute and case law have severely restricted their ability to control inappropriate and excessive treatment by making the application of reasonable utilization review impractical for service dates after Minniear.

More specific to California’s workers’ compensation system, employers/insurers can be subject to very substantial penalties (Neuhauser and McBirnie, 2000). The extension of presumption to medical treatment, and the strengthening of the legal status given presumption may have led claims administrators to perceive a higher level of exposure to penalties, even when a dispute over the appropriateness of treatment seems reasonable. A 5814 penalty, so called because it falls under California Labor Code Section 5814, is a 10 percent penalty that applies to an entire class of benefit payments when a benefit is unreasonably delayed. For example, in the case of medical treatment, a 5814 penalty for unreasonable delay would be applied to all medical payments on the injury, past and future. For many claims and treatments, this penalty will be many times larger than the cost of a disputed treatment. A second instance within the same benefit type results in another 10 percent penalty.

A number of stakeholders claim the impact of the legal decision, or at least the interpretation of the legal decision by employers, insurers and claims administrators in the face of Labor Code 5814, is that Minniear substantially reduced the ability of employers, insurers, and claims administrators to control medical costs.

II. (c) California Workers’ Compensation Environment at the Time of Minniear

Coincident with the Minniear decision, the Workers’ Compensation Insurance Rating Bureau of California (WCIRB) began reporting rapid increases in workers’ compensation medical costs. The WCIRB reported double digit increases in the average medical cost on indemnity claims, what insurers typically refer to as "severity" (WCIRB annual reports, 1998-2001). The coincidence of the Minniear decision and this increase in average claim cost led some observers to attribute the "severity" changes to Minniear. The question for this report is, how much, if any of the increase in severity can be attributed to Minniear? In addition, stakeholders want to know if the recent changes to the statutory language (AB-749 (Calderon)) affecting presumption will lead to substantial savings in medical costs?

There are a number of alternative explanations for increasing medical costs. The most obvious explanation is changes in the pricing of services. Non-occupational medical treatment has seen consistent annual increases in per capita medical expenditures of about one-third to one-half those observed in workers compensation, 5-6 percent per year since 1995 (Centers for Medicare and Medicare Service, 2002). About half of this increase in expenditures can be attributed to medical inflation. The medical CPI during this period was about 3.1 percent annually (RAND, 2002). However, much of service unit pricing in workers’ compensation is controlled by fee schedules. The unit pricing controlled by these schedules did not change substantially during the period under study. Segments of medical treatment were outside the fee schedule mechanism during all or the majority of the period under study. The most important exceptions were in-patient hospital facility fees and out-patient surgery facility fees. However data from the WCIRB on hospital costs show these costs were stable over the period of study as a fraction of all medical costs. The data are not broken out separately for in-patient and out-patient facilities.

An important medical cost growth driver involves changes to the mix of services delivered. Medical research is constantly developing new drugs and new technologies to treat conditions. Often, these new treatments involve, more expensive services that result in quicker or more complete recovery or reduce the negative side affects experienced by patients. Even in the absence of changes in available treatments, doctors may choose a less cost-effective treatment because it offers better income potential (Rice, 1997).

Another explanation is changes to the seriousness of the average injury. The seriousness of an injury is often referred to as "severity" by observers. However, we will use the term "seriousness" to distinguish this from the way severity is used by insurers. Insurers use "severity" to refer to average cost per claim, which includes medical costs but also includes indemnity and other costs. In these discussions, severity will always refer to average cost per claim, and seriousness will refer to the acuity of the underlying injury independent of the affect on medical or indemnity costs.

Since pricing of the units of medical services was fixed by the Official Medical Fee Schedules, if Minniear had the impact of reducing the ability of employers and insurers to limit what they perceive as inappropriate treatment, then the effect would appear as either 1) more services (e.g., more physical therapy treatments or office visits, and/or 2) increases in more expensive services (e.g., MRIs, or back surgeries) as a fraction of all treatments. In the following sections we will examine both issues. Did the extension of presumption to the primary treating physician’s opinion on treatment lead to an increase in medical treatment costs? If so, has the cost effect been driven by increases in high cost treatments or has it been driven by more frequent utilization, particularly of moderate priced services?

III. Model

We define a variable, cqtrperiodqos, where "cqtr" is the medical cost on a claim during a given quarter after injury, the subscript "qos" represents the quarter of medical treatment relative to the injury quarter (injury quarter =1), and the superscript "period" indicates whether the quarter of service was pre-Minniear or post-Minniear. Then, all else equal, the effect of the change in the legal rules surrounding arbitration on the average cost of treatment in any particular quarter of service is

(1) ?cqtr (due to Minniear) = cqtrPostqos - cqtrPreqos

However, if there were systematic changes to medical treatment regimes or other system characteristics occurring at the same time as Minniear that also affected the average cost of medical treatment, then equation (1) would inappropriately attribute these effects to the Minniear decision.

As noted above, the employer/insurer can direct initial treatment by controlling the provider during the initial 30 days and thereafter until, and if, a worker changes to another provider. We would anticipate that presumption affects costs differentially depending upon who "controls" the primary treating physician. To a first approximation we expect that physicians selected by the employer/insurer will be little affected, regarding treatment decisions, by the existence of Minniear. Employer selected physicians, to a large extent will have their incentives aligned with those of the employer and insurer on whom they depend for future referrals and business. In addition, providers consistently selected by employers may "self-select" as employers’ doctors because they prefer treatment regimes that are more conservative. Discussions with claims adjusters and insurers indicate that treatment by employer selected doctors is rarely disputed through the formal legal process.

On the other hand, employee "controlled" primary treating physicians are more likely to be in conflict with the employer/insurer over what represents necessary and appropriate treatment. There are a number of reasons this is true. The incentives of doctors that draw their business through employee selection will be less consistently aligned with the employer/insurer. These doctors may select into the pool specializing in applicant treatment because they have a belief in the effectiveness of more aggressive or extensive treatment. They may be more responsive to the needs of workers and see no reason to impose limits on treatment in a trade-off between cost and effectiveness. Or doctors may simply be exhibiting income maximizing behavior in a fee for service environment [Rizzo 1996]. The bulk of litigation over medical treatment involves represented workers who selected their primary treating physician.

Consequently, if Minniear had an impact on medical treatment costs, we would expect it to show up as 1) a change in the average cost of medical treatment in the period after the Minniear decision relative to the period before, 2) this pattern should be independent of the date of injury (as long as the injury occurred after 1/1/94), and 3) the effect should be stronger for those claims were the worker controls the primary treating physician.

If we assume that when the employer selects the doctor, the incentives are aligned and the behavior for this subset of doctors does not change (as well as the insurers response to these doctors’ treatment decisions), then the impact of the change in litigation rules could be defined for the subset of doctors selected by the worker. In that case, the effect can be more precisely defined using a difference-in-difference approach that allows us to control for other unknown effects. In this case, we define a treatment that is affected by presumption as that delivered by a worker controlled provider. This treatment is affected by both presumption and all other trends and changes that are occurring in the period. We also define a comparison group covering treatment delivered by providers selected by employers. Treatment supplied by these providers is not affected by presumption, but is affected by all other trends and changes occurring in the period. We assume that all other factors, besides presumption, affecting the cost of medical treatment impact the employer controlled and worker controlled doctors equally. Then the impact of presumption is the difference between these two groups. The model for any particular quarter after injury (QOS) is:

(2) ?cqtrqos (due to Minniear on worker controlled providers) =

[ cqtrPosteecontrol - cqtrPreeecontrol ] - [ cqtrPostercontrol - cqtrPreercontrol ]

To regression-adjust this estimate for observable covariates, we define the dummy variable Controli as equal to one if during the quarter of service the primary treating physician (PTP) was selected by the worker and zero if the PTP was selected by the employer. We define the dummy variable Afteri as equal to one if the service quarter occurs during the post-Minniear period. We also define an interaction term, Afteri*Controli, that is equal to one when the worker controls the physician and the quarter of service is after the Minniear decision. The adjusted difference-in-difference estimate comes from the regression equation

(3) cqtr = 0 + 1Afteri + 2Controli + 3 Afteri*Controli + 4Xi + ei

where 0 through 4 are regression coefficients, Xi is a vector of claim and/or claimant characteristics (e.g., diagnosis, gender, age, etc.) and ei is an error term. The coefficient 3 gives the extent to which the pre-post change in average cost of medical services in a quarter when workers select the physician exceeds the change when the physician was selected by the employer, holding constant the variables in Xi.

IV. Data

The data that are used in this study were supplied by the California Workers’ Compensation Institute (CWCI) from its Industry Claims Information System (ICIS). These data are well suited for these analyses because we can identify individual transactions, the dates of service, and providers. For this study, we received a number of files covering both medical transactions, indemnity transactions, and claimant information. Altogether we received data on 14 million transactions covering 300,000 indemnity claims with dates of injury and dates of service between 12/31/1992 and 1/1/2000.

The medical transaction files included information on the date of service, primary diagnosis (ICD9), procedure code (usually CPT), provider identification number, provider specialty, and billed and paid amounts. We are also able to discriminate between different types of transactions, such as medical treatment, medical-legal expenses, and medical cost containment expenses. For these analyses, we exclude medical-legal and cost containment expenses. We are only interested in medical costs paid to medical service providers (doctors, hospitals, etc.).

Each medical transaction is linked to a specific claim through a unique identifier. Consequently we can link all medical transactions related to a claim and the claim level information regarding date of injury, demographic characteristics of the claimant and employer, and information on indemnity payments. Through the combination of date of injury and date of service, we are seeking to parse out the separate causes that are attributable to a date of injury (statutory changes such as the introduction of presumption) and causes that would be affected by the date of the transaction (e.g., the Minniear decision).

V. Methodology:

Control of the primary treating physician is a key variable in the analysis. There is no precise way to define the variable given the data. However, we can define a proxy for control and for the PTP based on the provider data and the statutory language. Later we will evaluate the success of this definition.

The employer has the ability to direct treatment during the first thirty days. The worker may switch physician any time after that period, up until the PTP finds that the worker has reached maximum medical improvement, know as "permanent and stationary" in California. The worker is allowed to change PTP once, unless an additional switch is approved by the parties or an administrative law judge. The primary treating physician is nearly always the last provider of service on a claim because of the obligation, introduced as part of the statutory language in the 1993 legislation, to write a final report indicating that the condition was "permanent and stationary." Consequently, we determine the ID of the last provider on each claim. Then we determine the first date of service of that last provider. If the first date of service of the last provider is outside the first thirty days of treatment, we assign control of all treatment on and after that date to the control of the worker. All treatment before that date, or all treatment on claims where that last provider’s first service date is within the first thirty days is assigned to employer control.

The fraction of active claims with worker control increases as a claim ages, the last quarter of treatment is weighted more heavily towards worker control, and the last quarter is more likely to be partial quarter of service. This could bias down the estimates of average quarterly cost of services under worker control. Consequently, we do not include the last quarter of observed medical treatment. Tables 1-3 (at the end of the report) give the data for each dependent variable broken down by quarter of injury and quarter of service.

Table 4 gives the definition of the variables in the regressions. Several variables require additional discussion.

Medical costs are expected to exhibit increases in cost over time. The variable "service quarter" reflects the calendar quarter in which service was delivered, and should reflect secular increases in workers compensation medical cost. As noted above, workers’ compensation medical costs are controlled to a large degree by fee schedules. While these are similar to Medicare fee schedules, during this period they were subject to very little revision. Consequently, we do not expect substantial cost increases from changes in the average cost of particular services. The coefficient on Service Quarter is likely to be driven more by changes in the basket of services delivered, changes in the intensity of service delivery, and changes in the average cost of services that are outside of fee schedules at a particular point in time rather than inflation in the cost of specific services.

There are two important dimensions that require consideration of splines in the models presented here. First, several factors affecting medical cost are driven by the date of injury (DOI). These include the statutory change that introduced primary treating physician presumption for injuries occurring on or after 1/1/1994. If this statutory change is reflected in a changed pattern of service intensity or mix, then this might affect average cost. Consequently, we introduced splines for the period 1993:1-1993:4 and 1994:1+.

Second, an examination of the data revealed a distinct pattern in the duration of claims. Figure 1 shows the probability that an indemnity claim will have any service in a given quarter after injury based on the quarter of injury. The probability that a claim would receive service during a quarter peaked in 1997:1 for virtually all quarters after injury examined (through the 12th quarter). The reason for this pattern is not clear. It is related to the quarter of injury, but no specific legislative or regulatory changes were introduced around this time that impacted workers’ compensation cases based on date of injury. To account for this unexplained pattern, we introduced a spline for the period 1997:2+ and truncate the 1994:1+ spline to the period 1994:1-1997:1.

Consequently, we introduce splines for the date of injury that break our data into 1993:1-1993:4, 1994:1-1997:1, and 1997:2+. These splines are meant to allow the rate of increase in medical cost over time to vary according to the effect of statutory changes that affect claims based on date of injury (e.g., introduction of presumption for dates of injury on or after 1/1/94) and the yet unexplained change in the pattern for claims with dates of injury after 1997:1. These splines are interval variables. They have a value of 1 in the first quarter in the range, 2 in the second quarter, etc. Outside the range they have a value of zero. This allows the level and slope to vary for each of the splines.

Quarter after injury dummies are included to control for the changes in average quarterly medical cost as claims mature. These costs decline over the early quarters after injury and increase gradually quarter-to-quarter for claims still active after the 9th quarter.

During the period under study, the WCIRB reported annual declines in the frequency of reported injuries and annual increases in both average medical and average indemnity incurred per indemnity claim. Consequently, adjusting for changes in the distribution of injury type and severity over time may be important. In most models we include characteristics of the claim in the first quarter as an independent variable. For models where the dependent variable is medical cost in a quarter, the first quarter medical cost is entered as an independent variable. When the dependent variable is the natural log of medical cost, the natural log of the first quarter medical cost is used as an independent variable. When the dependent variable is the number of services in a quarter, the number of services in the first quarter is used. The motivation for this is clear if one examines Figure 2.

Figure 2 plots the average medical cost for each quarter after injury against the quarter of injury on the x-axis. While on a percent change basis, quarters two through twelve (scaled similarly in Figure 3) show substantial trends towards higher medical cost over time, the first quarter after injury has very little trend. In addition, while regressions run with EECTRL*AFTDUM on all three dependent variables had significant results in all other quarters (2-12), there were no significant results for the first quarter after injury. This is not surprising given that with employer control for at least the first 30 days and service intensity dropping rapidly over the initial phase of treatment, and change of physician usually happens well after 30 days, we do not expect switching of control to be an important factor in the initial period.

Since the values of the dependent variables in the initial period are independent of the variable of interest, EECTRL*AFTDUM, we use them to improve the precision of the model by controlling for the seriousness of the injury. A series of dummy variables was created for diagnoses using the two digit ICD9 codes. For a small number of cases the ICD9 code was missing. In addition, a number of ICD9 2-digit categories had fewer than 0.1 percent of observations (900). Both the missing and low count 2-digit categories were grouped together as the dummy variable ICD9miss. The combination of the value of the first period variables for medical cost in quarter one (CQTRi), natural log of medical cost in quarter one (lnCQTRi), and number of services in quarter one (SQTRi), and the diagnosis dummies allows us to control for some of the heterogeneity that may result from changes in the mix and seriousness of injuries.

The remaining variables are the demographic characteristics available in the database, gender, age, and average weekly wage (AWW). Very low and very high weekly wages exhibit substantial problems that suggested that the claim data were unreliable. The sample was restricted to claims with weekly wages > $50/wk. and < $2000/wk. In practice this excluded less than 1 percent of the sample. We also limited the sample to claims with ages between 16 and 75 (again excluding less than 1 percent of the sample). A dummy variable for missing gender was included because gender was missing approximately 5 percent of the time and because, unlike age and wage, the data is unnecessary for claims handling and missing data does not infer problems with all other data on a claim.

VI. Results

First we examine the data to determine if the definition of PTP control gives results consistent with the expected impact and timing of both the statutory change legislated by AB-110 and the judicial change triggered by Minniear. Figure 4 examines the value of permanent disability payments on claims resolved within two years of the date of injury for dates surrounding the statutory change that affected dates of injury on or after 1/1/1994. We see that the statutory change triggered a discontinuity in the relative value of permanent disability awards depending on control of physician as defined above. The timing of the discontinuity is consistent with the introduction of the statutory change. The direction of the change is consistent with our expectations about control.

Figure 5 illustrates the impact of the Minniear decision on treatment cost for the 4th quarter of service after the injury. Again we see that the discontinuity in the trend lines is consistent with the timing of the decision. We also observe that the statutory change, which was not expected to affect treatment decisions, did not have an effect on treatment costs in the period surrounding 1/1/1994, even though we observed an impact on indemnity costs. The 4th quarter impact of Minniear on medical services was mirrored in every quarter except the 1st quarter of service (quarter of injury).

Both of these observations give us confidence that the definition of control and PTP are effective. To the extent that the definition is not perfect, the impact would be to miss-assign the control of treatment. Since the results are in the expected direction, miss-assignment would bias our results towards finding no effect, and when we find an effect, bias the coefficients down, giving more conservative estimates.

We evaluated a number of models using different specifications to examine the impact of different possible variables and the robustness of the results. We present the results for four different models in Tables 5-7. Each table gives the results for one of our three dependent variables. Model 1 is the simple, single differences comparison corresponding to Equation (1). The results are consistent with the expectation that medical costs were higher in the quarters of service after Minniear, in this case by approximately $90/quarter. However as noted earlier, this cannot be used to attribute this change to the rules governing presumption.

Model 2 is the preferred model. Here we use the double difference comparison defined by Equation (3) that uses services delivered under employer controlled doctors as a comparison group. We also drop the first quarter of service as observations in the regression, but include the value of the variable in the first quarter as an independent variable, which when paired with the diagnosis, gives us a control for the seriousness of the injury.

The value of the first quarter variables is dropped as an independent variable in Model 3 and the first quarter data are included as observations in the regression. Note, the control for the seriousness of the injury improves both the explanatory power of Model 2 relative to Model 3 and the precision of the coefficient estimates on the key variables (note Adjusted R2 and standard errors).

Model 4 is Model 2 with the addition of the variable (AFTCONT) to determine if Minniear led lead to a change in the rate of growth of medical cost as well as a shift in the level of medical cost. This variable is constructed as an interval variable to examine whether the trend line on the dependent variables by service quarter accelerated after Minniear, as well as allowing a discontinuous one time jump related to Minniear.

The coefficient on AFTCONT is small and not significant for the regressions for the average cost and the number of services in a quarter. Only for the log of medical costs is the coefficient significant, and even then, the coefficient is quite small (<0.3 percent).

The author also examined a number of other model specifications including ones that interacted the worker control variable with an interval variable for the service quarters after Minniear. The coefficient on this term was very small and not statistically significant. These, along with other specifications tried, confirm that the impact of Minniear was to change the level of expenditures over a short period of time, but not change the trend line for medical inflation whether or not the employee controlled the physician.

The remaining discussion will focus on the results from Model 2. This is the preferred specification for the reasons given in the above discussion.

The results can be inferred from the coefficients given in Tables 5–7. These results are most important as information to evaluate system costs. As such, we will be generally interested in percentage changes. Therefore we need a baseline against which to measure the coefficients.

Because we are most interested in the impact of presumption in today’s market, and because the impact of the spline for 97:2+ is so important, the appropriate baseline against which to measure the change demonstrated in our results is the average values for the dependent variables after Minniear when the employer controls the treatment. Against these baselines we can evaluate the impact of employee control and the additional impact of employee control that is attributable to the presumption after Minniear. The following table gives the value for the dependent variables in the period after Minniear. The data are for quarters 2-12 after injury, consistent with our preferred model.

Table 8: Mean values for dependent variables

Mean Value Dependent Variables, After Minniear

All

Employer Control

Employee Control

Average Quarterly Medical Cost,

Quarters 2-12

(CQTR)

$1648.7

(11.93)

$1551.9

(15.82)

$1828.4

(17.30)

Natural Log Average Quarterly Medical Cost,

Quarters 2-12

(lnCQTR)

5.312

(.007)

5.116

(.009)

 

5.677

(.011)

Average Number Services in Quarter,

Quarters 2-12

(SQTR)

6.088

(.016)

5.763

(.019)

6.689

(.027)

Std. errors in parentheses

The date of service coefficient picks up the overall trend in medical cost related to the date of service. We might expect this variable to have a small coefficient given the trend we observer for non-occupational medicine during the period under study. As noted above, at the national level, medical expenditures per capita grew over the period under study at a rate that varied from 1.5 percent in 1994 to 5.8 percent in 1998 (CMS 2000). However, much of this cost increase was driven by medical inflation, holding the mix of services constant. During this period, the medical CPI was approximately 3.1 percent (RAND, 2002). Since the California Official Medical Fee Schedule was virtually fixed in this period, the "CPI" portion of quarter-to-quarter medical growth trend in workers’ compensation would be very low or flat. Consequently, we might expect that the coefficient on service quarter would be quite low, representing something equivalent to national per capita medical expenditure growth (approximately 5 percent) minus Medical CPI (3.1 percent). In fact the coefficient for quarter of service is 16.57 on the base of $1648, or 1 percent/quarter, just over 4 percent/year.

That is, the long-term trend in medical cost has been about 4 percent annually given imposition of medical fee schedules and controlling for diagnosis mix, seriousness of injuries, changes in employee control, presumption, and the effect included in the 97:2 spline. While somewhat higher than estimated for the entire health arena, these long-term trends are quite close. However, it is much lower than the year-to-year changes estimated by the WCIRB for ultimate incurred medical cost per indemnity claim, about 15 percent for 1995-2001 (WCIRB 2002). We need to look to other sources to explain the remainder of the medical cost growth, specifically, changes in the mix of services and/or an increase in the number or intensity of services.

Focusing on the results for Model 2, we can evaluate the impact of presumption as defined by the coefficients on the values of the employee control variables (EECTRL and EECTRL*AFTDUM) and obtain estimates of the percent impact of the application of presumption. We get the values defined in the following table.

Table 9: Percent change attributed to control and presumption

Dependent Variable Independent Variable

Average Value (employer control)

Coefficient

Percent Effect

 
Average $Med in quarter (CQTR)

1551.9

   
  Worker Control (EECTRL)  

122.9*

7.9%

  Impact of Minniear (EECTRL*AFTDUM)  

175.5*

11.3%

Natural log average $Med in quarter (lnCQTR)

5.12

   
  Worker Control (EECTRL)  

.522*

68.5%a

  Impact of Minniear (EECTRL*AFTDUM)  

.160*

17.4%a

 
Average number of services in quarter (SQTR)

5.76

   
  Worker Control (EECTRL)  

.598*

10.4%

  Impact of Minniear (EECTRL*AFTDUM)  

.445*

7.7%

a For a semi-logarithmic functional form, the expression for the percentage impact of a dummy variable on the dependent variable is ecoefficient – 1. (Kennedy, 1996)

* Significant at p < .001

In terms of the overall impact of presumption on medical costs, the most important measure is the dependent variable for average cost (CQTR). As expected, when a worker controls the primary treating physician (PTP) the average cost in a quarter is 7.9 percent higher than when a worker with the same condition is treated by an employer selected physician. This was true even before Minniear. The impact of Minniear, applying presumption concerning medical treatment to the PTP, was to increase the differential when the worker controlled the PTP by an additional 11.3 percent. These results are all highly significant statistically (p<.01).

A characteristic of medical cost data is they are typically a highly skewed distribution with many lower and medium cost observations and a few very large observations. Using the natural log of medical cost (lnCQTR) instead of medical cost reduces the weight given to very large observations. The average of the natural log gives an estimate that is more reflective of the central tendency of such a skewed distribution. This is much the same as using the median instead of the mean. Changes in medical care that increased very high cost services would tend to increase the average cost by a larger percent than the natural log of the average cost. The opposite would be true if the greatest change was observed on the natural log statistics. Observers who were concerned about the impact of presumption on medical costs were split on which cause might dominate.

The results for our regressions using the dependent variable lnCQTR are very informative. The direction of the effect of employee control and presumption are the same as for CQTR, confirming those results. In addition, the percent changes indicated by the coefficients on the natural log are larger than for the average medical cost, especially the effect of control absent presumption. Employee control increases the natural log statistic by 68.5 percent and the impact of presumption adds an additional 17.4 percent. The large impact of control on the average log of medical cost suggests that the impact of worker control on utilization is mostly an increase in the utilization of moderately priced services, for example physical therapy treatments. Similarly, the impact of presumption on log cost being larger than the impact on average cost was probably a product of more moderate priced services rather than increases in very costly services such as surgeries.

The results from regressions using the number of services in a quarter (SQTR) as the dependent variable also confirm the impact of control and presumption. The results are most similar to the average cost variable. The effect of control is to increase the number of services by 10.4 percent while the additional impact of presumption is an additional 7.7 percent increase.

The splines for 93:1-93:4 and 94:1-97:1 have small and insignificant coefficients for all three dependent variables. This implies that after controlling for other characteristics, the date of injury did not have an effect on the cost of medical treatment or number of services during the accident quarters 1993:1 and 1997:1. On the other hand, the spline for 97:2+ is large and significant across all dependent variables and all models. This suggests an important driver of medical cost that is related to the date of injury being after the first quarter of 1997. Since the splines are interval variables, the impact of the coefficient is the quarterly change in the value of the dependent variable. Translating to percent changes the following table gives the impact of having an injury during or after the 2nd quarter of 1997.

Table 10: Impact of quarter of injury > 97:1

 

Dependent Variable

Average Value (post Minniear)

Coefficient on 97:2+ Variable

Percent Effect (quarterly)

 
Average $Med in quarter (CQTR)

1551.9

51.1*

3.3%

Natural log average $Med in quarter (lnCQTR)

5.12

.050*

5.1%a

Average number of services in quarter (SQTR)

5.76

.186*

3.2%

a For a semi-logarithmic functional form, the expression for the percentage impact of a dummy variable on the dependent variable is ecoefficient – 1. (Kennedy, 1996)

* Significant at p < .001

These are very large effects reflecting an impact consistent with the growth rates estimated by the WCIRB for ultimate incurred costs. However, these are tied to a date of injury and not a date of service and therefore cannot be attributed directly to the impact of Minniear. We will discuss possible indirect explanations for this unexplained trend in the discussion section below.

The demographic variables in general are not particularly compelling. As might be expected, age has a small, statistically significant, positive impact on medical cost and the number of services. An increase of 10 years in age increases average quarterly medical costs by approximately 3 percent. Weekly wage has a very small impact on medical cost after controlling for other variables. Male gender has a small impact on quarterly medical cost, except in the first quarter when the impact is quite large (Model 3). Further analysis of the data revealed that while 2nd-12th quarter medical costs were quite similar across gender, males had first quarter costs almost $1300 greater than females (see Figure 6). However, while average expenditures on active claims are similar for men and women after the first quarter, claims by women are more likely to receive treatment for longer periods, leading to higher costs for women, on average, in subsequent quarters. This is especially true when we observe the average of the log of medical costs. (see Figure 7).

VII. Discussion Section

Consistent with previous studies of pricing mechanisms, and risk sharing, we find important effects when the legal process governing utilization review is changed. Relaxing the controls on a particular set of providers does lead to substantial increases in the frequency and intensity of services delivered by those providers. However, this does not necessarily imply that the affected doctors are abusing the system and over treating patients. As pointed out by several observers, while we assume physicians respond to financial incentives,

"…there are counter-balancing arguments, not least the presence of a formal ethical code to which doctors are expected to adhere. Indeed, a strong system of ethics may dilute, or completely remove, the incentives in some payment systems for doctors to provide ineffective, dubious or very costly treatments merely to increase their income (Gosden, et. al. 2001)."

And,

"In a vivid illustration of the limits of payment incentives, all nations rely heavily on socialization and the inculcation of norms of behavior for physicians. While norms and cultural expectations are pervasive across all occupations, medicine seems to be subject to stronger and more explicit codes of conduct. The omnipresent and continuing reliance on socialization is testimony to the effectiveness and, indeed, the indispensable role of this quintessentially nonprice mechanism (Robinson, 2001)."

On the other hand, there is an extensive literature indicating that under certain conditions, especially where the patient is uninformed or there is uncertainty about the appropriate treatment or level of treatment, physicians may induce demand and over treat (Rizzo, 1996; Leape, 1989; Evans, 1974). Similarly, there is a body of literature that finds under certain incentive structures (e.g., capitation, salary) physicians may under treat (Robinson, 2001).

Utilization review and other non-price mechanism for influencing clinical behavior play an increasingly important role in reducing unjustified variation in treatment, improving the quality of care, and ensuring cost-effective treatment (Robinson 2001, Shortel, Bennett, and Byck 1998). Robinson argues that the impact of treatment guidelines appropriately runs opposite the direction of the prevailing payment incentives. It is against these cultural norms and the previous research on the impact of incentives that we should evaluate the effects we observe in the current data.

The differences in the percentage estimates of the impact of presumption between average cost, log of average cost, and number of services can be informative concerning the source of impact. The differences between average cost and average log cost are consistent with the impact of control and the impact of presumption driving greater utilization of moderate cost services more than they drive the most expensive services.

When this information is combined with the number of services variable, we can infer an additional insight. The "services" variable is more accurately interpreted as the number of transactions, that is, payments to providers. Consequently, a number of services can be bundled into a single payment. This can represent multiple visits to the provider and/or multiple services during the same visit. In the vast majority of records, the payment from and through dates reflect a single date, indicating a single visit. In general, the impact of the application of presumption to medical treatment has been to increase both the frequency and intensity of moderately priced treatments. For example, workers selecting their own doctor might have more physical therapy visits and receive more specific therapies with each visit.

Since the average cost 1) is less affected by control than the average number of transactions, 2) much less affected than the average log cost, 3) the average cost is more affected by presumption than the number of transactions, but 4) still much less affected by presumption than log cost, we might infer that the impact of both control and presumption is mostly to increase the intensity of service delivery within visits and less importantly to drive increases in the number of visits.

Assembly Bill 749 (Calderon) eliminated the presumption given the PTP for nearly all claims. An important question is what is the impact of our estimates presented here on total medical cost? And, how can this be expected to affect future employer cost and insurance ratemaking? We have defined the effect of presumption, an 11.3 percent increase in average quarterly medical cost when the worker controls the physician. There are two other issues to consider. First, what portion of the total medical treatment dollar is delivered under worker control? Second, was the portion of medical treatment under worker control affected by the introduction of presumption?

The first question is reasonably easy to answer. Analysis of these data for a different effort arrived at an estimate that approximately 27 percent of medical treatment dollars during the first five years after injury were delivered under the direction of a worker controlled PTP (Neuhauser 2001).

That analysis was based on a sample of claims with dates of injury from 1993 to 1995, straddling the statutory change, but with the majority of treatment delivered before Minniear. The statutory change increased the incentive to control the treating physician. However, since the focus was on medical-legal issues, particularly permanent disability, the incentive to change became most important near the conclusion of treatment when these medical-legal evaluations were made. Once the medical treatment came under the umbrella of presumption, the incentive to switch increased, and the incentive was more important early in the treatment regime.

We examined the percent of active claims in each quarter after injury where treatment is controlled by the worker. Figure 8 graphs the pattern over the observation period for selected quarters after injury. The X-axis is the calendar quarter of service and the graphed lines show quarter of service relative to the injury quarter. It is clear that there has been an increasing movement by workers to switch control, more often and more quickly. While this effect appears to accelerate after the Minniear decision, that pattern is not as clear as that observed for the medical treatment cost. The trend lines do not clearly break at the time of the Minniear decision. Consequently, we cannot conclude that all of the effect observed is the result of Minniear, even though the change is proximal in time and in the direction of the incentives.

Even though the increased portion of switching cannot entirely be attributed to the presumption change, the higher portion of claims under worker control and the greater amount of time under worker control give us a projection of 40 percent of treatment under worker control for more recent claims.

A lower bound estimate of the impact of presumption on medical treatment costs is simply the product of the percent change in medical treatment due to presumption times the percent of medical treatment under worker control.

Percent change in system medical costs =

% change from presumption when worker controls * % of treatment controlled by the work =

11.3% * 40% = 4.5%.

Given the WCIRB’s most current estimate that total ultimate medical cost on 1993 claims will be approximately $8.2 billion, the lower bound estimate of the impact of presumption is $370 million.

There are several reasons to suspect that this estimate is low. First, as mentioned above, around the time of Minniear, workers began to switch to a doctor of their choice more frequently and earlier in the claim. Since, even absent Minniear and presumption, medical treatment was on average 7.9 percent higher in a quarter when the worker controlled the doctor, if the greater tendency to switch control is a product of Minniear, then an additional impact of 7.9% * 13% = 1% of medical treatment cost could be attributed to presumption and Minniear. That is, the greater incentive to switch generated by Minniear means that employers face the historical premium of worker control (7.9 percent) on an additional 13 percent of medical treatment. Moderating this effect, we did not find convincing evidence that would attribute all additional tendency to switch physicians to presumption.

Second, the trend to earlier and more frequent switching of physician was still increasing at the end of our observation period. Consequently, our estimate of 40 percent of treatment delivered under employee control may be conservative for injury dates in 2003 and beyond.

Third, we examined the impact of presumption on the cost in any given quarter. We did not examine whether Minniear caused treatment to continue for more quarters. We might expect this to be the case, but we cannot make a definitive statement without additional research. More frequent treatments may imply longer treatment and consequently more quarters of treatment, e.g., 10 PT visits over 2.5 quarters rather than 8 treatments over 2 quarters. If treatment duration was as affected by presumption as treatment intensity, this would double the estimate of 4.5 percent that we made above.

The combination of the observed impact on medical costs within service quarter, the increase in the portion of medical services delivered under worker control, and the potential for longer duration would yield an upper bound estimate on the impact of presumption on system wide medical cost of 10.0 percent, or approximately $820 million for the 2003 accident year ultimate total medical cost.

For the purposes of estimating the impact of AB-749 on future employer costs, it is appropriate to be conservative. We do not know for example whether the impact of eliminating presumption, except when the worker predesignates, will have an effect that is symmetrical with the impact of introducing presumption. Some stakeholders have argued that a ‘cultural bias’ towards application of presumption will remain in place at the WCAB. Also, while we consider it more likely that the treatment under worker control lasted longer as well as being more intense, we do not have firm evidence. More intense treatment may resolve the medical condition sooner. For these reasons, we would go with the more conservative estimate of 4.5 percent for the likely impact of repealing presumption.

There are several caveats and concerns related to these analyses. First, we observed a large and significant effect related to dates of injury after the first quarter of 1997. This effect is more important than the effect of presumption. And, we do not have a completely satisfactory explanation for the change. It appears that for at least some portion of claims, medical duration was shorter and that this effect increased quarter to quarter over the period. At the same time, the remaining active claims experienced rapid cost increases across all of our measures. The table 10 above gave the estimates.

One possible explanation is that the claims that closed early were the less serious injuries, and the remaining pool of active claims reflects increasing average cost simply due to selecting claims further along the tail of the injury distribution.

Why an increasing fraction of claims would begin to close early, reversing a trend observed over the previous 3 years and why this is related to date of injury is an interesting question. Two explanations have been suggested. First, the realization of the impact of Minniear may have focused claims administrators on getting claims closed more quickly while medical control was held by the employer. Second, the "sports medicine" model of treatment was increasingly popular during this period. Proponents of the sports medicine model argue that early intense treatment speeds recovery and return to work. While there may be no savings on medical cost, treatment will be shorter, return to work quicker, and therefore the employer may save on indemnity costs.

A second caveat concerns the way in which we defined change in control of the PTP. Since medical treatment is not concluded on all claims, the final provider that we use is actually the last observed provider. As the injury date is closer to the last observation date, we are censoring a greater fraction of the observations. Therefore our "last provider" is less accurately defined for more recent claims. The bias, if any, introduced by this censoring is not clear. To the extent that we are miss-assigning a change of control, we will likely bias the coefficient on the key control variable (EECTRL*AFTDUM) down, making it more likely that we will not find an effect and that if found, the estimate of the effect will be conservative. On the other hand, if we are picking up a change to a specialist still under employer control, it is possible that the treatment would be more costly and, some fraction of the time, miss-assigned to worker control. We will still assign control accurately to the worker for an unknown fraction of these cases.

The assigning of control could be made in a way that avoided the censoring issue. One could define control for each quarter based on the last observed provider in the subsequent quarter. Since we do not include the last quarter of treatment in the observations, this would give us a consistent definition for all quarters, even the last injury quarter used (1999:1). However, control could change from quarter to quarter and miss-assignment of control would become even a bigger problem.

Another concern raised during review was that the impact observed on the important control variable (EECTRL*AFTDUM) may be driven by the changing portion of cases under employee control. For example, if the cases most likely to shift control at the margin are those most in need of treatment, the worker controlled pool will increase in medical treatment intensity relative to the employer controlled pool. There are two reasons to think that this is an unlikely explanation for the effect we observe, and that if affecting the outcome, the effect will be small. First, the changing portion of claims that are worker controlled has been a trend, and one not convincingly linked to the timing of the Minniear decision. On the other hand, the impact of Minniear on medical treatment was a one-time discontinuity in the pattern and was coincidental with the timing of Minniear. Second, if claims at the margin of intensity of medical treatment are switching, this would be true of earlier switches. Consequently, we would expect the marginal claims to bring down the averages of those that switched under the previous incentives at the same time that they reduced the averages for the employer controlled group. This action would mitigate the impact on the effect we are measuring.

One might also be concerned that the differences observed on the "control" variable (EECTRL) reflect unobserved differences in the types of injuries and workers that switch from employer to worker control. However, even if that is true, the impact of presumption (EECTRL*AFTDUM) is still observed even after controlling for the worker selection into the worker control PTP group using the before and after comparison.

Finally, it should be noted that groups of observations are systematically related. Each quarter of service on the same claim is not independent of the preceding or succeeding quarters. The most appropriate approach for these kinds of data is to use a statistical package that adjusts the standard errors, taking into account the grouping of observations. This statistical adjustment was not available on the software used. However, these adjustments are typically quite small, and all of the statistically significant results are well below standard thresholds. In no cases would we expect the results to change.


Future Research

The most important focus of future research should be examining the impact of presumption on worker health outcomes. We have a very strong finding that the relaxation of the ability of insurers and employers to apply utilization controls had a significant impact on the frequency, intensity, and cost of treatment. However, it cannot be assumed that this represents "over treatment." Absent empirical evidence it is not possible to determine if the additional treatment is excessive, potentially causing poorer outcomes (both medically and financially); or beneficial, improving health and/or financial outcomes.

The optimal level of treatment requires measuring the cost of that treatment against the workers’ health outcomes. We have a unique opportunity to test whether relaxing utilization controls results in a more or less optimal treatment regime. Assembly Bill 749 removes the presumption regarding medical treatment effective for injuries occurring on or after 1/1/2003. Workers injured before and after that date will be subject to different legal standards concerning treatment, essentially the reverse of the natural experiment analyzed here.

There are two potential measures of outcomes. First, self reported health outcomes are one measure of workers’ perception of the ‘success’ of their health care. There is a fairly extensive literature supporting the reliability of self-reported health status. The Commission on Health and Safety and Workers’ Compensation (CHSWC) will be conducting a survey of workers perceptions of the quality of their health care and their health status. This survey, the WINS survey, was developed and pilot tested in several states under a grant from the Robert Wood Johnson Foundation. The Commission has planned the timing of the survey to draw samples from workers injured before and after the statutory change at 1/1/2003. This will allow the Commission to make an initial evaluation of the impact at an early point in a claim, approximately six months after injury. A second wave of the survey would allow follow-up on health status at a later period to gather data on the ultimate health outcomes, possibly 18 months after injury. This second wave is not currently funded.

A second outcome measure would reflect labor market outcomes for injured workers. Following the design developed at RAND and several co-researchers (Reville 1999, Boden and Galezzi, 1999, Biddle, Boden, and Reville, 2001) we can measure workers’ outcomes as changes in earnings and labor force participation using carefully constructed controls. This approach offers a unique opportunity to measure the impact of additional medical treatment cost in terms of an economic outcome with similar dimensions.

As noted in the discussion section, incentive structures in medicine have been hypothesized to lead to both under and over utilization of medical treatment. Using the labor market outcome variables, linked to self-reported health outcomes, we would have a unique opportunity to test which of these effects predominate under the various incentive structures that are part of workers’ compensation medical delivery. This is a powerful opportunity to cost-effectively improve medical care for injured workers.

Another important area of future research is to examine the change in the services delivered that led to the sharp increases in medical cost observed in this study. The CWCI/ICIS data is an unusually rich set of data for research related to medical treatment. While this study looked at data at the transaction level, these transactions are also broken down into various services. These services are coded according to standard methods (generally CPT codes) and linked to diagnoses. It would be possible to isolate the impact of changing medical cost drivers by specific groups of services (e.g., physical therapy, radiology, and evaluation and management).

 

Table 4: Variable Definitions

Dependent Variables

CQTR Dependent Variable = medical cost in quarter of observation
lnCQTR Dependent Variable = natural log of medical cost in quarter of observation
SQTR Dependent Variable = number of services in quarter of observation

Independent Variables

Constant

EECTRL

AFTDUM

EECTRL*AFTDUM

AFTCONT

Service Quarter

93:1-93:4 spline

94:1-97:1 spline

97:2+ spline

CQTR1

LnCQTR1

SQTR1

QTR1DUM

QTR2DUM

QTR3DUM to

QTR12DUM

ICD9 Dummies

ICD9MISS

Age

Age Square

Weekly Wage

Male Gender

Gender missing

Intercept

Dummy for control of PTP (1=worker control, 0=employer control)

Dummy for quarter of service after Minniear (1= after Minniear, 0=before)

Interaction term indicating worker control and after Minniear

Interval variable indicating number of quarters after Minniear for service quarter (1=97:1, 2=97:2, etc.)

Interval variable for quarter of service (93:1=1, 93:2=2,…97:1=17, etc.)

Interval variable for quarter of injury prior to statutory change(93:1=1 to 93:4=4, otherwise = 0)

Interval variable for quarter of injury after statutory change but before 97:2 (94:1=1 to 97:1=13, otherwise = 0)

Interval variable for quarter of injury after 97:1, (97:2=1 to 99:1=7, otherwise = 0)

Medical treatment cost in quarter of injury

Natural log of medical treatment cost in quarter of injury

Number of services in quarter of injury

Dummy for service quarter equal to quarter of injury

Dummy for service quarter in second quarter (excluded variable in regressions including dummies)

Series of dummy variables for quarter of service

 

Series of dummy variables for ICD-9 (diagnosis) at 2-digit level (84 is excluded diagnosis)

Dummy variable when diagnosis missing or number of observations at 2-dgit level < 900.

Age of worker at injury

Age squared

Weekly wage at injury

Dummy for gender (1 = male, otherwise = 0)

Dummy for missing gender (<1% of cases)

 

Table 5: Impact of Minniear on Medical Treatment Cost

Dependent Variable

CQTR=medical cost in quarter

 

Model 1

 

Model 2

 

Model 3

 

Model 4

Constant

EECTRL

AFTDUM

EECTRL*AFTDUM

AFTCONT

Service Quarter

93:1-93:4 spline

94:1-97:1 spline

97:2+ spline

CQTR1

QTR1DUM

QTR2DUM

QTR3DUM

QTR4DUM

QTR5DUM

QTR6DUM

QTR7DUM

QTR8DUM

QTR9DUM

QTR10DUM

QTR11DUM

QTR12DUM

ICD9 Dummies

Age

Age Square

Weekly Wage

Male Gender

Gender missing

916.84 (28.13)

90.38 (16.82)

17.1 (2.14)

3.77 (5.30)

1.35 (1.78)

53.67 (7.18)

.054 (.000)

Excluded Variable

-344.14 (14.49)

-546.06 (15.73)

-683.48 (17.23)

-759.07 (19.00)

-817.75 (20.81)

-784.45 (22.82)

-811.83 (25.09)

-777.40 (28.19)

-632.44 (32.72)

-455.86 (34.29)

Yes

4.64 (.438)

-.005 (.000)

-.019 (.006)

58.45 (9.29)

-181.05 (36.05)

903.65 (28.20)

122.88 (12.51)

32.183 (17.73)

175.51 (18.48)

16.57 (2.14)

4.50 (5.30)

1.13 (1.78)

51.13 (7.18)

.054 (.000)

Excluded

-355.37 (14.50)

-566.96 (15.76)

-712.79 (17.35)

-762.04 (19.36)

-860.73 (20.89)

-833.52 (22.92)

-866.86 (25.20)

-838.21 (28.30)

-701.59 (32.85)

-541.53 (34.48)

Yes

4.72 (.438)

-.005 (.000)

-.018 (.006)

57.76 (9.29)

-193.60 (36.04)

618.08 (52.09)

93.73 (24.34)

54.09 (32.44)

227.67 (35.82)

12.61 (3.82)

-9.19 (9.89)

-2.30 (3.13)

37.77 (11.28)

2115.56 (27.24)

Excluded

-332.73 (29.71)

-526.38 (31.94)

-652.44 (34.77)

-673.97 (38.46)

-770.41 (41.27)

-728.31 (45.20)

-741.50 (49.78)

-696.66 (55.96)

-538.40 (65.22)

-381.73 (68.25)

Yes

3.57 (.802)

-.004 (.001)

-.006 (.011)

501.35 (17.18)

24.23 (67.83)

907.71 (39.67)

123.08 (12.52)

30.52 (18.13)

174.57 (18.60)

1.69 (3.84)

16.20 (2.30)

4.14 (5.37)

1.22 (1.80)

50.16 (7.51)

.054 (.000)

Excluded

-355.50 (14.50)

-567.21 (15.77)

-713.16 (17.37)

-762.69 (19.42)

-861.30 (20.93)

-834.13 (22.97)

-867.52 (25.25)

-838.80 (28.34)

-702.14 (32.87)

-542.00 (34.50)

Yes

4.72 (.438)

-.005 (.000)

-.018 (.006)

57.77 (9.29)

-193.41 (36.04)

R2

N

.083

873,363

.082

873,363

.024

1,088,090

.082

873,363

Standard errors are in parentheses.

 

Table 6: Impact of Minniear on Medical Treatment Cost

Dependent Variable lnCQTR = natural log of medical cost in quarter

 

Model 1

 

Model 2

 

Model 3

 

Model 4

Constant

EECTRL

AFTDUM

EECTRL*AFTDUM

AFTCONT

Service Quarter

93:1-93:4 spline

94:1-97:1 spline

97:2+ spline

lnCQTR1

QTR1DUM

QTR2DUM

QTR3DUM

QTR4DUM

QTR5DUM

QTR6DUM

QTR7DUM

QTR8DUM

QTR9DUM

QTR10DUM

QTR11DUM

QTR12DUM

ICD9 Dummies

Age

Age Square

Weekly Wage

Male Gender

Gender missing

3.009 (.022)

.092 (.012)

.038 (.002)

.005 (.004)

-.002 (.001)

.055 (.005)

.266 (.001)

Excluded

-.596 (.010)

-.986 (.011)

-1.232 (.013)

-1.443 (.014)

-1.590 (.015)

-1.574 (.017)

-1.530 (.018)

-1.410 (.021)

-1.018 (.024)

.176 (.025)

Yes

.007 (.000)

-.000 (.000)

.000 (.000)

-.257 (.007)

-.704 (.026)

2.912 (.022)

.522 (.009)

.028 (.013)

.160 (.013)

.0375 (.002)

.0006 (.004)

-.0013 (.001)

.0500 (.005)

.270 (.001)

Excluded

-0.630 (.011)

-1.047 (.011)

-1.318 (.013)

-1.518 (.014)

-1.713 (.015)

-1.712 (.017)

-1.682 (.018)

-1.574 (.021)

-1.201 (.024)

-0.047 (.025)

Yes

.008 (.000)

-.000 (.000)

.000 (.000)

-.261 (.007)

-.737 (.026)

5.085 (.018)

.422 (.008)

.084 (.011)

.207 (.012)

.026 (.001)

-.003 (.003)

-.006 (.001)

.027 (.004)

1.435 (.009)

Excluded

-0.621 (.010)

-1.034 (.012)

-1.298 (.012)

-1.483 (.013)

-1.678 (.014)

-1.668 (.016)

-1.624 (.017)

-1.500 (.019)

-1.095 (.023)

0.062 (.024)

Yes

.001 (.000)

-.000 (.000)

-.000 (.000)

-.061 (.006)

-.625 (.024)

2.95 (.023)

.524 (.009)

.013 (.013)

.152 (.014)

.015 (.003)

.032 (.002)

.003 (.004)

-.001 (.001)

.041 (.005)

.270 (.001)

Excluded

-.631 (.011)

-1.050 (.001)

-1.321 (.013)

-1.524 (0.14)

-1.718 (.015)

-1.717 (.017)

-1.687 (.018)

-1.580 (.021)

-1.206 (.024)

-.051 (.025)

Yes

.001 (.000)

-.000 (.000)

.000 (.000)

-261 (.007)

-736 (.026)

 

 

 

R2

N

.106

873,363

.113

873,363

.128

1,088,090

.113

873,363

Standard errors in parentheses

 

Table 7: Impact of Minniear on Medical Treatment Cost

Dependent Variable

SQTR = Number of Services in Quarter

 

Model 1

 

Model 2

 

Model 3

 

Model 4

Constant

EECTRL

AFTDUM

EECTRL*AFTDUM

AFTCONT

Service Quarter

93:1-93:4 spline

94:1-97:1 spline

97:2+ spline

SQTR1

QTR1DUM

QTR2DUM

QTR3DUM

QTR4DUM

QTR5DUM

QTR6DUM

QTR7DUM

QTR8DUM

QTR9DUM

QTR10DUM

QTR11DUM

QTR12DUM

ICD9 Dummies

Age

Age Square

Weekly Wage

Male Gender

Gender missing

2.918 (.041)

.075 (.024)

.057 (.003)

.012 (.008)

.003 (.003)

.194 (.010)

.206 (.001)

Excluded

-1.225 (.021)

-1.916 (.022)

-2.374 (.025)

-2.640 (.027)

-2.879 (.030)

-2.772 (.033)

-2.818 (.036)

-2.684 (.040)

-2.175 (.047)

-1.984 (.049)

Yes

.014 (.001)

-.000 (.000)

-.000 (.000)

-.399 (.013)

-1.135 (.051)

 

 

 

2.845 (.041)

.598 (.018)

-.078 (.025)

.445 (.026)

.0545 (.002)

.0140 (.008)

.0024 (.003)

.186 (.010)

.206 (.001)

Excluded

-1.270 (.021)

-1.998 (.022)

-2.489 (.025)

-2.698 (.028)

-3.046 (.030)

-2.961 (.033)

-3.028 (.033)

-2.913 (.040)

-2.433 (.047)

-2.302 (.049)

Yes

.014 (.001)

-.000 (.000)

.000 (.000)

-.402 (.013)

-1.181 (.051)

5.572 (.042)

.658 (.020)

-.079 (.026)

.478 (.029)

.029 (.003)

-.022 (.008)

.023 (.003)

.293 (.009)

4.391 (.022)

Excluded

-1.209 (.024)

-1.891 (.026)

-2.337 (.028)

-2.498 (.031)

-2.809 (.033)

-2.685 (.036)

-2.706 (.040)

-2.525 (.045)

-1.942 (.053)

-1.796 (.055)

Yes

.004 (.001)

-.000 (.000)

-.000 (.000)

.075 (.014)

-1.067 (.055)

2.86 (.043)

.598 (.018)

-.085 (.026)

.442 (.026)

.005 (.005)

.053 (.003)

.013 (.008)

.003 (.003)

.183 (.010)

.206 (.001)

Excluded

-1.270 (.021)

-1.999 (.022)

-2.491 (.025)

-2.700 (.028)

-3.048 (.030)

-2.963 (.033)

-3.030 (.036)

-2.915 (.040)

-2.435 (.047)

-2.304 (.049)

Yes

.014 (.001)

-.000 (.000)

-.000 (.000)

-4.02 (.013)

-1.181 (.051)

 

 

 

R2

N

.143

873,363

.146

873,363

.143

1,088,090

.146

873,363

Standard errors in parentheses.

 

Bibliography

Author, David H. and Mark G. Duggan. "The Rise in Disability Recipiency and the Decline in Unemployment," Working Paper 8336, National Bureau of Economic Research. Cambridge, MA. June 2001.

Baker, Laurence C., and Krueger, Alan B., " Medical costs in workers’ compensation insurance." Journal of Health Economics, 14 (1995) 531-549.

Bakker F., R. van Vliet. The Introduction of Deductibles for Prescription Drugs in a National Health Insurance: Compulsory or Voluntary?" Health Policy 1995 Jan;31(1):53-65.

Biddle, Jeffrey, Leslie I. Boden, and Robert T. Reville, "Permanent Partial Disability from Occupational Injuries: Earnings Losses and Replacement in Three States." In Budetti, Burkhauser, Gregory and Hunt (eds.), Ensuring Health and Income

Security for an Aging Workforce. W. E. Upjohn Institute for Employment Research, Kalamazoo, MI 2001.

Boden, Leslie I. and Monica Galizzi. Economic Consequences of Workplace Injuries: Lost Earnings And Benefit Adequacy. American Journal of Industrial Medicine. 36(5) 1999.

Borba, Phillip S. and Thomas Parry (2000). An Evaluation of the Comprehensive and Organized Managed Care Program: Final Report. Robert Wood Johnson Foundation and The Joint Industry Board of the Electrical Industry.

Centers for Medicaid and Medicare Services (CMS), Health Care Finance Administration Web Site, "National Health Care Expenditures, 2000." http://www.hcfa.gov/stats/nhe-oact/tables/t1.htm Accessed, 4/23/02.

CHSWC, "CHSWC Summary Findings and Recommendations to the Inpatient Surgery Schedule and Outpatient Surgery Study," Commission on Health and Safety and Workers’ Compensation Staff Report, February, 2002.

CWCI, "CMS, #98: Medical Treatment." Claims Monitoring Survey Report, California Workers’ Compensation Institute. March 1998.

Durbin, David L., Dan Corro, and Nurhan Helvacian, "Workers’ Compensation Medical Expenditures: Price vs. Quantity," Journal of Risk and Insurance 63(1), 1996

Ernst & Young LLP, "Treatment Guidelines for Low Back Injuries: A Qualitative and Quantitative Analysis of the Industrial Medical Council and the Agency for Health Care Policy and Research Low Back Guidelines." Prepared by for the California Workers’ Compensation Institute, June 1996a

Ernst & Young LLP, "Supporting Documentation for Treatment Guidelines for Low Back Injuries. Prepared for the California Workers’ Compensation Institute, June 1996b.

Evans, R., "Supplier Induced Demand: Some Empirical Evidence and Implications." Perlman, M. ed., "The Economics of Health and Medical Care." Stockton Press, New York, 1974.

Gosden, Toby, Frode Forland, Ivar Sonbo Kristiansen, Matthew Sutton, Brenda Leese, Antonio Giuffrida, Michelle Sergison, and Lone Pedersen, "Impact of Payment Method on Behaviour of Primary Care Physicians: A Systematic Review," Journal of Health Services Research and Policy, Vol6:1, January 2001.

Harris B., A. Stergachis, L. Ried, "The effect of drug co-payments on utilization and cost of pharmaceuticals in a health maintenance organization." Med Care 1990 Oct;28(10):907-17

Johnson, William G., Marjorie L. Baldwin and John F. Burton Jr., "Why is the Treatment of Work-Related Injuries So Costly? New Evidence from California." Inquiry. (Spring 1996)

Kennedy, Peter, A Guide to Econometrics, 3rd Edition. The MIT Press, Cambridge, 1996.

Kizer, David. "So Minniear, and Yet So Far." Workers’ Compensation Quarterly, Vol. 11:2, Fall/Winter 1998.

Kominski, Gerald and Laura Gardner, "Inpatient Hospital Fee Schedule and Outpatient Surgery Study," Final Report to the Commission on Health and Safety and Workers’ Compensation, December 2001.

Leape, L., "Unnecessary Surgery," Journal of Health Services Research, Vol. 24, 1989.

Marria, Suzanne P., "Keys for Primary Treating Physicians for Writing ratable Reports." Presentation to the 9th Annual Conference of the Division of Workers Compensation, February, 1998.

Mnookin, Robert and Tony Kornhauser, "Bargaining in the Shadow of the Law," 88 Yale L.J. 950 (1979).

Neuhauser, Frank and Andrew Koehler, "Evaluating the Reforms of the Medical-Legal Process Using the WCIRB Permanent Disability Survey," Report to the Commission on Health and Safety and Workers Compensation. University of California at Berkeley, 1996.

Neuhauser, Frank, "Report on the Quality of Treating Physician Reports and Cost-Benefit of Presumption in Favor of the Treating Physician." Report to the Commission on Health and Safety and Workers’ Compensation. August, 1999.

Neuhauser, Frank, "Preliminary Findings: Primary Treating Physician, Impact of Presumption on Claim Costs." Presentation to the Commission on Health and Safety and Workers’ Compensation, December, 2000.

Neuhauser, Frank and Tom McBirnie, "Issue Paper on Labor Code 5814", Report to the Commission on Health and Safety and Workers’ Compensation, April 2000

Neuhauser, Frank, Alex Swedlow, Laura Gardner, and Ed Edelstein, "Study of the Cost of Pharmaceuticals in Workers’ Compensation," Report to the Commission on Health and Safety and Workers’ Compensation. June 2000.

Neuhauser, Frank, "Medical cost, change of physician, and length of employer control," memorandum to CHSWC and DIR, dated June 26, 2001.

Pace, Nicholas, Robert Reville, Lionel Galway, Amanda Geller, Orla Hayden, Laural Hill, Christopher Mardesich, Frank Neuhauser, Suzanne Polich, and Jane Yeom. "Improving the Peoples’ Court: Issues Facing the Adjudication of Claims before the California Workers’ Compensation Appeals Board," RAND, DRU-2739-ICJ, January, 2002.

RAND, California Inflation Statistics Summary, http://ca.rand.org/cgi-bin/monthly.cgi, 2002.

Reuveni, Haim, Boaz Sheizafa, Asher Elhayanya, Michael Sherfa, Yehuda Limonia, S. Scharffa, and Ronit Peleda, "The effect of Drug Co-payment Policy on the Purchase of Prescription Drugs for Children with Infections in the Community." Forthcoming in Health Policy, 2002.

Reville, Robert T, "The Impact of a Disabling Workplace Injury on Earnings and Labor Force Participation," In Haltiwanger, John and Julia Lane, ed. "The Creation and Analysis of Linked Employer-Employee Data," Contributions to Economic Analysis, Amsterdam; London and New York: Elsevier Science, North-Holland 1999.

Rice, Thomas, "Physician Payment Policies: Impacts and Implications," Annual Review of Public Health, Vol. 18:549-65, 1997.

Rizzo, J. and D. Blumenthal, "Is the Target Income Hypothesis an Economic Heresy?" Medical Care Research and Review, Vol. 53, 1996.

Robinson, James C., "Theory and Practice in the Design of Physician Payment Incentives," The Milbank Quarterly, Vol. 79:2. 2001.

Schellhorn M., "The effect of Variable Health Insurance Deductibles on the Demand for Physician Visits." Health Economics 2001 Jul;10(5):441-56

Shortell, Stephen, .M., Charles L. Bennett, and Gayle R. Byck, "Assessing the Impact of Continuous Quality Improvement on Clinical Progress," Milbank Quarterly 76(4) 1998.

Simon G., L. Grothaus, M. Durham, M.VonKorff, and C. Pabiniak. "Impact of Visit Copayments on Outpatient Mental Health Utilization by Members of a Health Maintenance Organization." American Journal of Psychiatry, 1996 Mar;153(3):331-8

Stapleton, David, Devin Coleman, Kimberly Dietrich, and Gina Livermore. "Empirical Analyses of DI and SSI Application and Award Growth." In Dalman Rupp and David Stapleton (eds.) Growth in disability Benefits: Explanations and Policy Implications. Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 31-92.

Swedlow, A., Johnson, G., Smithline, N., and Milstein, A., "Increased costs and rates of use in the California workers’ compensation system as a result of self-referral by physicians." New England Journal of Medicine, 1992 321(21), 1502-06.

Swezey, Charles L. "Cases on Treating Physician Presumption," Memorandum to the Commission on Health and Safety and Workers’ Compensation. January 2002.

Zaidman, B., "Industrial strength medicine: a comparison of workers’ compensation and Blue Cross health care in Minnesota." Minnesota Department of Labor and Industry. 1990.

U. S. Centers for Medicare and Medicaid Services, Health Care Finance Review, Summer 2001.

WCIRB, "2000 California Workers’’ compensation Losses and Expenses," Report by the Workers’ Compensation Insurance Rating Bureau of California, June 2001a.

WCIRB, letter from the Workers’ Compensation Insurance Rating Bureau to Suzanne Marria, November 7, 2001b.

WCIRB, "Preliminary Summary of December 31, 2001 Experience," WCIRB Bulletin No. 2002-07. Workers’ Compensation Insurance Rating Bureau of California. April 12, 2002.


Table 1: Medical Cost, Active Claims, by Controlling Party

Table 
  1:</a> Medical Cost, Active Claims, by Controlling Party

Table 2: Natural Log of Medical Cost, Active Claims, by Controlling Party

Table 
  2: Natural Log of Medical Cost, Active Claims, by Controlling Party

Table 3: Average Transactions in Quarter, Active Claims, by Controlling Party

Table 3: Average Transactions in Quarter, Active Claims, by Controlling Party

Figure 1: Probability of any services in by quarter after injury

Figure 
  1: Probability of any services in by quarter after injury

 

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

Figure 2: Average medical across all indemnity claims by QOI and QOS

Figure 
  2: Average medical across all indemnity claims by QOI and QOS

 

Figure 3 Average medical cost across all indemnity claims by quarter of injury and quarter from injury

Figure 
  3 </a>Average medical cost across all indemnity claims by quarter of injury 
  and quarter from injury

Figure 4: Average PD award by quarter of injury and control of physician

Figure 
  4: Average PD award by quarter of injury and control of physician

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 5: Average medical cost for active claims in 4th quarter of service, by control of physician

Figure 
  5: Average medical cost for active claims in 4

 

 

 

 

 

 

 

 

 

 

 

 

Figure 6: Average medical cost, active claims, by gender and quarter of service

Figure 
  6: Average medical cost, active claims, by gender and quarter of service

Figure 7: Average of natural log of medical costs, all claims, by gender and quarter of service

Figure 
  7: Average of natural log of medical costs, all claims, by gender and quarter 
  of service'> 
  </font></p>
</font> 
<p align=  

Figure 8: Percent of active claims under worker control

Figure 
  8: Percent of active claims under worker control