Author: WorkCompAcademy

In January 2019, Diana Bronshteyn, whom a doctor had diagnosed with fibromyalgia, sued the California Department of Consumer Affairs under the California Fair Employment and Housing Act, section 12900 et seq. She sued for failure to accommodate, failure to engage in an interactive process, disability discrimination, and failure to prevent discrimination.

The case was a long shot. Jean Hyams, one of Bronshteyn’s lead attorneys, admitted to a “significant risk that the jury might be swayed by the fact that the [Department] permitted [Bronshteyn] to take a long leave of absence and, on paper at least, offered her all of the accommodations she had previously requested.” From her work in disability rights, Hyams was aware that “women with fibromyalgia are often stigmatized and stereotyped as exaggerating or making up their symptoms.” Hyams also conceded the doctor’s notes Bronshteyn relied on in requesting leave were “ambiguous,” which compounded the risk in taking the case.

The Department fought the case hard from the start. The Department refused to discuss settlement. Many of plaintiffs motions were unsuccessfully opposed. The Department moved for summary adjudication. The court denied the motion in July 2021. Discovery was contentious. Both sides propounded multiple sets of written discovery requests, and both sides filed multiple ex parte applications and motions to compel further responses and documents.

In June 2022, Bronshteyn made a Code of Civil Procedure section 998 offer to compromise for $600,000 and for reasonable attorneys’ fees and costs. The Department rejected the section 998 offer and did not counter.

Trial by jury went for six weeks in the summer of 2022. It too was contentious. The jury found for Bronshteyn on all counts. Its verdict was $3,324,262 in damages. This sum is more than five times greater than Bronshteyn’s section 998 offer. The Department filed a motion for judgment notwithstanding the verdict. The court denied the motion. The Department moved for a new trial. The court denied the motion. The Department appealed the verdict. It lost. (See Bronshteyn v. Dept. of Consumer Affairs (May 12, 2025, B325678) [nonpub. opn.] (Bronshteyn I).) This was in 2025. By now the case was more than six years old.

Bronshteyn moved for statutory attorney fees and costs as the prevailing party under the Act, pursuant to section 12965, subdivision (c)(6). Her counsel requested a lodestar amount of $2,987,583.11, reflecting the number of hours they worked on the case, multiplied by the requested hourly rates: $1,000 for Wendy Musell; $1,100 for Jean Hyams; $1,200 for Leslie Levy; $1,100 for Sharon Vinick; $1,050 for Darci Burrell; $900 for Maraka Willits; $425 for Brittany Wightman; $350 for bar-certified law students; and $225 for paralegal and legal assistants.

Bronshteyn’s attorneys noted they had put more than 3000 hours into the case. They submitted contemporaneous and precise timesheets recorded in six-minute intervals, with breakdowns by day and person. They also filed detailed declarations describing their litigation backgrounds, past fee awards, the work on Bronshteyn’s case, the division of labor, and their efforts to limit duplication. Counsel voluntarily applied a five percent reduction to account for clerical work and travel time.

There was a contentious battle between the parties over the amount attorney fees. At the end of the day, the total fee award was $4,889,786.03. In justifying its decision to award rates at the higher end of Los Angeles rates, the court noted: “[t]he quality of lawyering was high; far beyond what the court would expect of an average lawyer – even with the years of experience the lawyers here exhibited.”

The Department appealed (again). The court of appeal affirmed the trial court award of attorney fees in the published case of Bronshteyn v. Dept. of Consumer Affairs -B329890 (Sept 2025).

In affirming the trial court, the opinion noted “In this second appeal, the Department argues the trial court abused its discretion in awarding $4,889,786.03 in attorney fees to Bronshteyn’s counsel. As the trial court aptly observed: ‘[t]he fact that [the Department] did not settle the case early might or might not be good litigation strategy or bad litigation strategy . . . But the biggest thing it does is it makes it hard for [the Department] to claim that [Bronshteyn] shouldn’t have spent money litigating to try the case, which is reflected in the number of hours that [counsel] billed.’ ”

“We affirm the trial court’s careful and well-reasoned ruling.”

Although large language models (LLMs) have performed well on the United States Medical Licensing Examination (USMLE) and at answering medical-related questions in studies, there was currently no benchmark testing how well LLMs can function as agents by performing tasks that a doctor would normally do, such as ordering medications, inside a real-world clinical system where data input can be messy.

Thus a team of Stanford University researchers recently developed benchmarks for measuring the accuracy and effectiveness of AI agents to assist physicians and published their findings in the New England Journal of Medicine AI.

While the researchers note the enormous potential of this new technology to transform medicine, the tech ethos of moving fast and breaking things doesn’t work in healthcare. Ensuring that these tools are capable of doing these tasks is vital, and then they can be used as tools that augment the care clinicians provide every day.

“Working on this project convinced me that AI won’t replace doctors anytime soon,” said Kameron Black, co-author on the new benchmark paper and a Clinical Informatics Fellow at Stanford Health Care. “It’s more likely to augment our clinical workforce.”

Black is one of a multidisciplinary team of physicians, computer scientists, and researchers from across Stanford University who worked on the new study, MedAgentBench: A Virtual EHR Environment to Benchmark Medical LLM Agents. Unlike chatbots or LLMs, AI agents can work autonomously, performing complex, multistep tasks with minimal supervision. AI agents integrate multimodal data inputs, process information, and then utilize external tools to accomplish tasks, Black explained.

While previous tests only assessed AI’s medical knowledge through curated clinical vignettes, this research evaluates how well AI agents can perform actual clinical tasks such as retrieving patient data, ordering tests, and prescribing medications.

“Chatbots say things. AI agents can do things,” said Jonathan Chen, associate professor of medicine and biomedical data science and the paper’s senior author. “This means they could theoretically directly retrieve patient information from the electronic medical record, reason about that information, and take action by directly entering in orders for tests and medications. This is a much higher bar for autonomy in the high-stakes world of medical care. We need a benchmark to establish the current state of AI capability on reproducible tasks that we can optimize toward.”

The study tested this by evaluating whether AI agents could utilize FHIR (Fast Healthcare Interoperability Resources) API endpoints to navigate electronic health records.

The team created a virtual electronic health record environment that contained 100 realistic patient profiles (containing 785,000 records, including labs, vitals, medications, diagnoses, procedures) to test about a dozen large language models on 300 clinical tasks developed by physicians. In initial testing, the best model, in this case, Claude 3.5 Sonnet v2, achieved a 70% success rate.

“We hope this benchmark can help model developers track progress and further advance agent capabilities,” said Yixing Jiang, a Stanford PhD student and co-author of the paper.

Many of the models struggled with scenarios that required nuanced reasoning, involved complex workflows, or necessitated interoperability between different healthcare systems, all issues a clinician might face regularly.

“Before these agents are used, we need to know how often and what type of errors are made so we can account for these things and help prevent them in real-world deployments,” Black said.

What does this mean for clinical care? Co-author James Zou and Dr. Eric Topol claim that AI is shifting from a tool to a teammate in care delivery. With MedAgentBench, the Stanford team has shown this is a much more near-term reality by showcasing several frontier LLMs in their ability to carry out many day-to-day clinical tasks that a physician would perform.

A new California Workers’ Compensation Institute (CWCI) study offers some of the first comprehensive data on the use of Functional Restoration Programs (FRPs) to treat California injured workers. FRPs are multi-disciplinary programs used to treat injuries that involve chronic pain and improve patient function when the injuries do not respond adequately to traditional therapies.

For its study, CWCI analyzed 635 indemnity claims that involved FRPs compiled from the utilization review (UR) systems of 6 California workers’ comp insurers. Using demographic, bill review, and medical management data, the authors compared the FRP claim sample to 270,165 non-FRP indemnity claims to determine how they differ in terms of claim characteristics, geographic region, and clinical conditions. In addition, to gain a more precise assessment of FRP vs. non-FRP treatment costs, temporary disability (TD) duration, and claim durations, the FRP claims were compared to a matched sample of 2,361 non-FRP indemnity claims that had similar clinical, demographic, and claim characteristics to those of the FRP claims. The analysis found that unlike the broader indemnity claim population, which is primarily based in Southern California, FRP claims are heavily concentrated in the Bay Area, which represented nearly half of the FRP claims, and the Central Valley, which accounted for more than a quarter of the claims. In addition, the FRP claims were far more likely to involve attorney representation (94.0%) than other indemnity claims (50.8%), and despite the MTUS guideline criteria that FRPs be used for chronic pain patients, only 42.8% involved a chronic pain diagnosis, though it is unclear whether this was due to clinical inattentiveness in coding, or because some FRP participants did not have chronic pain.    

FRPs typically began after extended periods of conventional treatment, with an average of 792 days between the first medical service date on the claim to the first FRP service date, during which time the injured workers averaged 37 conventional physical medicine visits. Though the Medical Treatment Utilization Schedule (MTUS) treatment guidelines recommend that FRPs run for 4 to 6 weeks, the study found that they tend to run longer, with a median duration of 8 weeks. While that exceeds the MTUS recommended level, it was consistent with the intent of typical UR approvals for FRPs. In addition, while the MTUS does not recommend a set number of days for an FRP, most injured workers in these programs received 20 to 40 days of FRP services, which was also consistent with the number of days approved by UR. On the other hand, injured workers in FRPs averaged 3.8 days of treatment per week, which was less than the MTUS recommendation of at least 5 days per week for tertiary rehab programs, but they averaged a total of 29.1 FRP treatment days. The study’s review of 2,896 FRP-related UR decisions across 1,059 claims (2.8 decisions per claim) found that 25.3% were denied and 4.8% were modified, significantly higher than the 7.7% denial/modification rate for non-FRP medical services noted in prior Institute research.

The cost data showed that FRP claims averaged $234,003, or 59.3% more than the similar non-FRP claims, as they were more expensive across all cost categories, with medical costs averaging twice as high, and indemnity costs and expenses each averaging 28% higher. Total inpatient and outpatient medical costs on FRP claims averaged $127,816, with FRP medical service costs alone averaging $59,106, or 72.6% of total outpatient payments on those claims. In contrast, inpatient and outpatient medical costs on the matched non-FRP claims averaged $64,062, with outpatient costs averaging $38,613.

A key reason for the higher outpatient costs on the FRP claims was that a third of the procedure codes used to bill FRP services were not listed in the Official Medical Fee Schedule, and payments associated with unlisted codes accounted for 84.3% of total treatment costs on the FRP claim. The payment data also show that unlisted procedure codes used for bundled FRP services led to much higher payments than conventional services, averaging $1,751 per code or an estimated $350 per hour of treatment.

FRP claim durations were also longer as these claims averaged 520 temporary disability (TD) days, 25.2% more than the 415-day average for the matched non-FRP claims, which helped push average indemnity costs on the FRP claims to $87,855 vs. $68,533 for the matched claims. Average claim durations measured from the first medical service date to the claim closure date were considerably longer than the average number of TD days for both the FRP claims (1,287 days) and the matched non-FRP claims (1,013 days) as TD for most injuries in California is capped at 104 weeks (728 days) within a 5-year period from the date of injury, though the average duration of the FRP claims was still 27.0% longer than for the matched non-FRP claims.

CWCI has published its FRP analysis in a Report to the Industry which CWCI members and research subscribers can access under the Research tab at www.cwci.org. Others may purchase a copy from CWCI’s online store.