LRLossReserves.com
Back to The WireWorkers Comp

Colorado EDI Switch Tests WC Development Triangles

Colorado's workers compensation EDI Claims Release 3.1 cutover starts July 9, 2026, after a hard July 8 deadline for EDI 1.0 files. The reserve issue is whether reporting lag, rejected transactions, and changed claim-status coding distort paid and incurred development triangles.

Colorado’s Division of Workers’ Compensation says Electronic Data Interchange (EDI) stakeholders must submit all EDI 1.0 files by 5:00 p.m. on July 8, 2026, and that Verisk will accept only EDI 3.1 files starting July 9. The Division also posted Claims Release 3.1, Version 1.2 requirements with an effective date of July 9.

That is a filing-system event, not a benefit expansion. But for self-insured workers compensation programs, a filing-system event can still move the number on the balance sheet. The risk is that First Report of Injury (FROI), Subsequent Report of Injury (SROI), denial, closure, payment, and reopening transactions enter the valuation extract at a different speed or with different coding than they did before the cutover.

From comparing third-party administrator (TPA) claim extracts before and after system migrations, the misleading signal is often not the new claim count. It is the temporary change in how quickly transactions are accepted, rejected, corrected, and coded.

Who it affects

The direct audience is Colorado self-insured employers, including public entities, universities, hospitals, construction firms, manufacturers, ski and hospitality employers, and large multi-state companies with meaningful Colorado payroll. Captives and large-deductible programs that receive Colorado workers compensation feeds from a TPA or fronting carrier should ask the same questions.

The Colorado EDI website says the state has used Release 1 standards for FROI and SROI reporting since 2004 and will continue receiving Release 1 claims until the move to Release 3.1 on July 9, 2026. That creates a clean break for compliance. It also creates a dirty quarter for actuarial data unless the employer can isolate the Colorado transition period inside its workers compensation IBNR review.

The reserve mechanism

The lever is development pattern, with a secondary case-reserve adequacy issue. A chain ladder projection assumes historical reporting and payment patterns remain useful. If Colorado FROI or SROI transactions sit in a rejection queue in July, arrive in a correction batch in August, and appear in the September 30 extract all at once, the triangle may read that as a real frequency jump or a paid-development acceleration.

The same problem can work in reverse. If closure, denial, or payment transactions are accepted faster under the new mapping, early paid and incurred development can look cleaner than the underlying claims really are. Case reserves may also be misread if the state EDI status no longer reconciles cleanly to the internal open, closed, reopened, denied, paid, and outstanding-reserve fields used in the actuarial file.

This is exactly the kind of data-quality issue Actuarial Standard of Practice No. 23 is meant to surface: whether data are appropriate for the analysis, what review was performed, and what limitations remain. For a finance leader, the practical version is simpler. Do not let a system-transition quarter become an unexamined selected development factor. The reserve diagnostic guide is the right frame for separating operational noise from true loss emergence.

What this means for your next review

Put Colorado’s July through September 2026 data on the agenda before year-end reserve work starts. Ask whether rejected or delayed FROI and SROI transactions are excluded from the valuation extract, whether they enter later as new activity, and whether the actuary can flag Colorado accident-year and report-year data around the cutover before adjusting selected age-to-age factors. The analytical call: treat third-quarter Colorado data as a system-transition quarter until rejection reports, TPA correction queues, and the September 30 extract prove otherwise.

Sources