Lumbar MRI · Same case · Agent v2 · Convergent deny

DENYmanual_concept_demo_v2

borderline deny

Lumbar MRI · Same case · Agent v2 · Convergent deny

What this proves: The agent improved between v1 and v2 — the verdict is the same (DENY), but the evaluator convergence is tighter and the cited alternative pathway is enriched. The v1 record is preserved exactly as it was at decision time. Improvement platforms cannot show v1 after v1 is gone; Audicta does, by design.

record_id: rec_2026-04-26T09-14-22Z_a4f7c92e
decided_at: 2026-04-26T09:14:22.341Z

Case

Patient ref: Patient/synthea-9b2f-771
Imaging request: MRI Lumbar Spine without contrast (CPT 72148)
Chief complaint: Low back pain, 3 weeks duration, post-lifting injury
Red flags documented: (none)
Exam findings: no_radicular_symptoms, slr_negative, no_motor_deficit, no_bladder_bowel_dysfunction

Agent chain

Clinical reviewer

confidence: 0.84

What: Identified absence of red flags and inadequately documented conservative trial; case fits LCD standard pathway, prerequisites not met. Compliance gap (PT 2 of 6 sessions) is the dominant adequacy factor — duration alone is two weeks short of LCD §B.1 minimum.
Why: Patient presents with 3-week mechanical low back pain following a lifting injury — no radicular symptoms (no leg pain, SLR negative), no motor deficit, no bladder/bowel involvement, no constitutional symptoms. This is the prototypical uncomplicated acute LBP that ACR AC variant 1 covers. Conservative trial documentation is incomplete however — chart shows 1 week PT with 2 of 6 prescribed sessions attended, plus NSAIDs with subjective 'some improvement' but no quantified outcome metric. Of the two adequacy gaps, the compliance gap is the dominant factor: even at full duration, 2-of-6 PT sessions does not meet LCD §B.1.a 'documented adherence' language, while NSAIDs trajectory is at least directionally positive. The combination of (a) no red flags and (b) inadequate conservative trial places this case in LCD L34220's standard pathway, where the 4-week-wait + adequate-management prerequisites must both be met before imaging is covered.
Tradeoff: Recommend complete trial: defensible per LCD / patient endures 3–4 more weeks symptomatic. Approve now: faster diagnostic certainty / contradicts coverage policy and incentivizes premature imaging. Naming compliance as the dominant gap aids downstream patient communication (the actionable factor is adherence, not waiting alone).
Scope check: I am assessing red-flag presence and conservative-trial adequacy against documented criteria. I am NOT making treatment recommendations or judging the patient's pain experience.
Alternatives considered (4): Trigger red-flag-bypass pathway despite no documented red flags
Pattern is uncomplicated mechanical LBP; no neuro deficit, no constitutional symptoms — bypass criteria definitionally not met
Treat 'some improvement' on NSAIDs as adequate response
LCD §B.1.a defines adequacy as 'no clinically meaningful improvement' — patient self-report of 'some' does not reach that bar, and PT compliance gap remains separately unaddressed
Defer all conclusions, request clarifying clinical note
Chart documentation is sufficient for adequacy assessment; further clarification does not change the core gap (compliance + duration)
Surface 'compliance vs duration' as separately gradable factors for downstream chains
Out of scope for clinical_reviewer; criteria_mapper handles factor-level mapping
Flagged uncertainties (1): PT compliance at 2/6 sessions — chart doesn't document patient-reported reason for missed sessions; assumption that this represents inadequate adherence rather than clinically-justified avoidance
Citations (2): CMS_LCD_L34220 · Coverage Indications.B.1
sha256:kb-lcd-l34220-sec-b1-c3a82e
ACR_AC_Low_Back_Pain · variant_1 · MRI lumbar without contrast · rating 1/9
sha256:kb-acr-lbp-v1-row-mri-noctrt-7d2e91

Criteria mapper

confidence: 0.81

What: Mapped to LCD L34220 §B.1 standard pathway: prerequisites not met. Compliance gap is the binding factor; duration gap is secondary. Cited covered alternative service per LCD §D — supervised PT extension to 4–6 weeks with documented compliance and a structured re-evaluation visit (telehealth or in-person both covered under §D).
Why: LCD L34220 §Coverage Indications.B.1 (standard pathway) requires both (1) >=4 weeks of documented conservative management AND (2) inadequate response to that management. Per the upstream clinical_reviewer's factor analysis, the compliance gap is the binding factor: 2-of-6 PT sessions fails LCD §B.1.a regardless of duration. The red-flag-bypass pathway (§B.2) is not applicable per the clinical_reviewer chain. The covered alternative service per LCD §Coverage Indications.D is supervised PT extension to 4–6 weeks with documented compliance, then re-evaluation — the LCD permits both telehealth and in-person re-evaluation under §D, broadening the patient-facing access option.
Tradeoff: Cite LCD §D alternative explicitly with both modalities: defensible deny + maximally constructive next step / requires patient to engage a re-evaluation visit. Bare deny without alternative: faster decision / fails LCD §D citation requirement and undermines audit defensibility.
Scope check: I am mapping documented findings to LCD criteria rows and citing the LCD-specified alternative pathway. I am NOT recommending specific PT protocols or making the final denial decision.
Alternatives considered (4): Map to LCD §B.2 red-flag bypass
Clinical_reviewer chain confirms no red flags — pathway prerequisite definitionally not met
Treat 5-week duration as meeting 4-week prerequisite standalone
LCD §B.1 requires duration AND adequate trial conjunctively, not disjunctively; partial fulfillment does not satisfy the conjunction
Map to LCD §B.4 EMG-correlation pathway
No neurologic findings documented; pathway prerequisites (positive EMG or persistent radicular sx) not present
Cite §D alternative without specifying telehealth eligibility
LCD §D explicitly enumerates both modalities — omitting the telehealth option understates the access pathway and reduces the constructive value of the alternative for the patient
Flagged uncertainties (1): LCD §B.1.a 'adequate trial' definition uses 'no clinically meaningful improvement' — applying this to 'some improvement' chart text is a textual-pattern judgment, not a quantitative match
Citations (2): CMS_LCD_L34220 · Coverage Indications.B.1
sha256:kb-lcd-l34220-sec-b1-c3a82e
CMS_LCD_L34220 · Coverage Indications.D
sha256:kb-lcd-l34220-sec-d-f8e2c4

Evidence retriever

confidence: 0.86

What: Retrieved ACR Appropriateness Criteria evidence (2026-04 KB snapshot): LBP variant 1 (uncomplicated, <6 weeks), MRI lumbar without contrast rated 1/9 (Usually Not Appropriate). Evidence base unchanged from prior KB snapshot.
Why: ACR AC LBP variant 1 — 'acute or subacute, uncomplicated, less than 6 weeks duration, no red flags' — is the direct fit for this case. Variant 1 rates MRI lumbar without contrast at 1/9, the lowest 'Usually Not Appropriate' band. This rating is consistent across the 2018 and 2026-04 snapshots; the underlying evidence base — that early imaging in uncomplicated LBP increases utilization without changing clinical outcomes — has not changed in the interval. ACR AC also publishes variant 1's preferred pathway: structured conservative management with re-evaluation at 4–6 weeks, which is what the decision record recommends as the alternative covered service.
Tradeoff: Variant 1 evidence: directly applicable / negative recommendation may feel adversarial to patient. Multi-variant analysis: more robust / overstates evidence depth needed for this clear classification.
Scope check: I am retrieving and citing published evidence rows. I am NOT interpreting evidence ratings or recommending the alternative pathway specifics.
Alternatives considered (3): ACR AC LBP variant 2 (subacute >6 weeks)
Case is at 3 weeks, well under 6-week threshold defining variant 2
ACR AC LBP variant 4 (with red flags)
Clinical_reviewer chain confirms no red flags — variant prerequisite not met
ACR AC LBP variant 6 (post-trauma)
Lifting injury is mechanical strain, not trauma per ACR variant 6 definition (which requires >=6/10 force injury or fall)
Citations (1): ACR_AC_Low_Back_Pain · variant_1 · MRI lumbar without contrast · rating 1/9
sha256:kb-acr-lbp-v1-row-mri-noctrt-7d2e91

Decision

DENYby clinical_reviewer → criteria_mapper → evidence_retriever

LCD L34220 §B.1 prerequisites unmet; compliance gap is the binding factor (PT 2-of-6) with duration as secondary (3 of 4 weeks). ACR AC LBP variant 1 rating 1/9 unchanged in 2026-04 KB. Covered alternative cited: LCD §D supervised PT extension to 4–6 weeks with documented compliance and a structured re-evaluation visit (telehealth or in-person).

Covered alternative pathway

Supervised PT extension to 4–6 weeks with documented compliance, then re-evaluation (telehealth or in-person under LCD §D)

CMS_LCD_L34220 §Coverage Indications.D

Evaluation

Evaluator function received only this record; case data, agent code, and KB content are not in the evaluator's input scope (compile-time enforced — see /architecture).

Dimension	Local	Cloud	Final	Divergence
Scope awareness	3.5	3.5	3.5	0.0
Alternative evaluation	4.0	4.0	4.0	0.0
Tradeoff clarity	3.5	4.0	4.0	0.5
Goal alignment	4.0	4.0	4.0	0.0
Uncertainty flagging	4.0	4.0	4.0	0.0
Decision chain	3.5	3.5	3.5	0.0

Max divergence

0.50

Avg divergence

0.08

Convergent (strict)

All dimensions converge

Tightly convergent — the maximum cross-evaluator divergence (0.5 on tradeoff_clarity) sits well within the strict ≤0.7 threshold; no human review required. Both evaluators score the alternative_evaluation identically at 4.0, the dimension that diverged most in the v0.4 snapshot of this case.

Local evaluator: qwen2.5-coder:7b · Cloud evaluator: claude-sonnet-4-20250514

Integrity

Content integrity

Compute the SHA-256 of this record (excluding content_hash itself) in your browser and compare to the stored hash.

Stored: sha256:cd1a22443b2387c1a456454eae888bb380694712848675858fd3a4b11001611d

Provenance

kb_snapshot_hash: sha256:kb-snap-2026-04-26-d2e3f1
agent_genome_hash: sha256:genome-healthcare-imaging-v0.5-c7f4ab
audicta_version: 0.5.0-drydraft
reproducible_with: docker pull audicta/healthcare-imaging:0.5.0-drydraft