Medical Severity Has a Story Arc: What Raji Chadarevian’s AIS 2026 Session Made Plain

Raji Chadarevian opened his session in Orlando with three fictional injured workers. A 23-year-old bartender with low back pain. A 41-year-old hospitality worker with a shoulder strain. And a 67-year-old hotel manager with a slip-and-fall shoulder injury, pre-existing shoulder pain, and hypertension. By the time he closed, he had used those three lives to make a point most severity dashboards still miss.

Medical severity is not a single number. It is a sequence of demographic, treatment, and timing decisions that compound over the life of a claim, and the levers that move it are not where most carriers and administrators are looking. Chadarevian is NCCI’s Executive Director of Actuarial Research, and his AIS 2026 session was the most actionable severity session NCCI has presented in years, if you have the data infrastructure to apply it to your own book. We covered the pre-conference outlook on medical severity earlier in this series, and the rest of the pre-conference coverage lives on our AIS 2026 Resource Hub.

Why Severity Is Better Understood as a Trajectory

Chadarevian framed the entire session around a sequence: Demographics, Injury, Condition, Treatment, Outcome. Each step adds inputs that shape the next. The worker’s age, wage, and pre-existing conditions affect the type and severity of injury. The injury and the medical condition that develops shape the treatment path. The timing and setting of treatment shape the outcome. And the outcome (claim closure, time to MMI, medical duration) shapes the final cost.

Chadarevian put the principle directly. Severity is not a single concept. It is the cumulative impact of the injury, the injured worker’s medical conditions, and their treatment over time.

Reading severity as a trajectory rather than a number changes how you manage it. The lever is not in the final cost. It is in the upstream decisions that compound into that cost. That is what the rest of the session made visible.

Demographics Drive More Severity Than Most Reserving Models Assume

The first part of the trajectory is the one most carriers underweight in their reserving models. Demographics do not just describe who is injured. They predict how much medical care the claim will absorb.

Start with wages. Higher-wage workers generate more medical utilization. Frequency of major surgery increases by wage group, physician payments show the same divergence, and the majority of low-wage workers’ indemnity claims come in under $5,000. That pattern persists across age groups and across industry sectors, which means it is structural rather than an artifact of any single class.

Age compounds the effect. Utilization for workers 65 and over is almost double that of the youngest workers and roughly 36% greater than the 25-44 age group. That ties directly to the aging-workforce dynamics Paul Hendrick laid out in the demographics session. As the 65+ cohort grows as a share of the labor force and as claims, the share of total medical utilization concentrated in that cohort grows with it.

Gender also matters in ways that are not always reflected in standard reserving assumptions. Chadarevian showed that women have approximately 11 more days of medical treatment beyond their temporary disability period, but take 7 fewer days to close after treatment ends. For rotator cuff injuries, women have 23 more days of treatment beyond TD and 10 fewer days to close. For low back pain, women have 12 more days of treatment and 1 fewer day to close. These are not large gaps individually. They are large in aggregate as women continue to gain share of claim counts and loss dollars.

Comorbidities Are the Multiplier, Not the Side Note

If there is one finding from the session that should change how reserving models work, it is the scale of the comorbidity effect on utilization. Most carriers know comorbidities matter. The magnitude is bigger than most reserving assumptions reflect.

Chadarevian’s data showed relative utilization intensity for claims with treated comorbidities. Degenerative conditions drive more than 2x utilization. Hypertension drives more than 3x. Diabetes drives about 4x. Blood vessel diseases drive more than 5x. The duration effect is just as real. Claims with comorbidities run materially longer in both the temporary disability and medical treatment phases. A claim without comorbidities might involve roughly 9,300 utilization units. A claim with degenerative conditions can run multiples higher, with substance abuse, obesity, and hypertension comorbidities each adding measurable utilization on top of the base.

The aging-workforce story we covered in the pre-conference demographics piece collides directly with severity math here. Workers 65 and over are 10% more likely than the average worker to have a comorbidity. As that cohort grows as a share of claims, the share of claims carrying comorbidity multipliers grows with it. Reserving models that treat comorbidity prevalence as flat are quietly understating the trajectory of severity in claim populations that are aging.

Treatment Has Shifted: Less Surgery, More Physical Therapy, More Delay

Chadarevian spent meaningful time on what he called the composition effect: the way the mix of treatment is changing in ways that affect utilization even when individual prices are held in check. The headline is that treatment looks different than it did a decade ago. The implication is that severity calculations built on older treatment patterns may understate where the line is actually going.

From accident year 2016 to 2023, major surgery declined approximately 8% and physical therapy increased approximately 12%. That is a meaningful structural shift, and within physical therapy the mix is also changing. The PT codes growing fastest are the higher-utilization codes, particularly 97530 (therapeutic activities at approximately 51 units per session) and 97112 (neuromuscular reeducation at approximately 46 units), rather than the lower-utilization 97110 and 97140 codes.

The major surgery story is more nuanced than a simple decline. Frequency per 1,000 claims fell from 367 to 331 over that period, driven mostly by reductions in hospital inpatient and hospital outpatient settings. Overall utilization for major surgery in hospital inpatient settings dropped 24%. But on a per-episode basis, major surgery utilization actually rose: hospital inpatient +9%, hospital outpatient +8%, ambulatory surgical center +4%. Fewer surgeries, but each surgery using more resources. That is the kind of pattern that aggregate severity numbers easily hide.

When and Where Treatment Starts Now Affects Severity Materially

The timing story is just as important as the treatment-mix story. The first 30 to 60 days of a claim have always mattered. The 2026 data shows they matter more than they used to.

Times to first treatment have grown longer across the board between 2016 and 2023. First treatment is delayed by approximately 6 days. Physical therapy is delayed by approximately 9 days. Major surgery is delayed by approximately 6 days. Same-day treatment dropped from 48% of claims to 43% over that period.

Delay matters because Chadarevian’s data showed that delays in physical therapy or major surgery put upward pressure on medical utilization. The treatment that comes later often comes more intensively, and the longer disability period in the interim adds its own cost.

Setting matters as much as timing. The share of claims treated at a nonfacility on the day of injury rose from 79% to 82%. Hospital inpatient first-visit share dropped 13%. Emergency room first-visit share dropped 17%. Urgent care facility first-visit share rose 20%. Those shifts have direct severity implications. Claims with first visits at a hospital outpatient setting show 30% greater overall utilization than claims with first visits at a nonfacility. Claims with first visits at a hospital inpatient setting show 6x greater utilization.

Chadarevian’s own framing was succinct. Starting care at a nonhospital provider typically means far lower medical utilization. That is one of the most actionable utilization levers in the entire severity story.

Outcomes Are State-Specific in Ways Industry Averages Hide

The closing chapter of Chadarevian’s session focused on outcomes: temporary disability duration, medical treatment duration, and time to claim closure. This is where the state-by-state reality becomes obvious.

For claims closed within two years, average time to close varies by jurisdiction between 30 and 50 weeks. That range is large enough to materially change the financial profile of a multi-state book. The drivers behind the variation include carrier practices, administrative system features, attorney involvement, comorbidity prevalence, and the injured worker’s own engagement with the claim.

The state-level pattern is also non-intuitive. Chadarevian showed that states grouped as high in medical utilization per lost-time claim do not necessarily fall into the same group on temporary disability or time to closure. A state that runs high on utilization may run low on duration, and vice versa. The implication for carriers is that single-metric state benchmarking misses the picture. The right question is not whether a state is high or low on severity. It is which combination of utilization, duration, and closure patterns the state produces, and what each of those means for reserves, claims staffing, and pricing.

Injury type matters too. Rotator cuff injuries take roughly 62 weeks total to close, with extended medical treatment periods. Low back pain claims close in roughly 30 weeks. Permanent disability injuries take almost double the time to close that temporary disability injuries do. None of this is new conceptually. The value of seeing it laid out in NCCI’s data is that it puts numbers on intuitions claims professionals have always carried, which makes them actionable in operational decisions.