Prior authorization (PA) is a review process used by healthcare payers to determine if a prescribed medication, procedure, or treatment is eligible for coverage before the patient receives it. The purpose is to confirm that the request aligns with medical necessity standards, formulary guidelines, and plan-specific rules. A PA case begins when a clinician or pharmacist submits a request that includes relevant clinical details and documentation. The payer then evaluates the submission against established criteria (an example of a PA Criteria can be found in the Appendix), often guided by structured logic or clinical rules. Based on that review, the payer issues a determination: approval, denial, or a request for additional information.
Many companies use decision trees, created manually, to execute the prior authorization process. Decision trees are rule-based logic flows that guide reviewers (or automated systems) through a series of yes/no questions (for example: “Does the patient meet the diagnosis code?” → “Has the patient tried first-line therapy?” → “Is the requested drug on formulary?”) until the tree reaches a final verdict. These decision-trees require continuous maintenance because plan criteria, clinical guidelines, and formularies evolve frequently.
When executed manually, the PA process increasingly acts as a bottleneck, delaying treatment, frustrating clinicians, and consuming staff resources, despite payers’ intent to control costs. New federal standards under the CMS Interoperability and Prior Authorization Final Rule now require most Medicare Advantage, Medicaid, CHIP, and QHP plans on the federal exchange to decide on prior authorization requests for medical services within 7 calendar days for standard cases and 72 hours for expedited cases, starting in 2026.
Due to these pressures, high volume, tight timelines, evolving regulations, and administrative burdens, there is a strong business case for automation and AI-driven solutions that can reduce friction while maintaining clinical quality.
This article introduces our internal benchmark, PABench, which evaluates the performance of AI models in prior authorization workflows. Our findings demonstrate that AI can reduce review cycles from days to minutes and achieve high accuracy of approximately 83% with minimal engineering up front, resulting in faster care, lower administrative overhead, and more auditable decisions.
AI Assisted Decision Making
We structured the benchmark in three scenarios:

Question-by-Question: The first scenario leverages decision trees and utilizes AI to navigate them, question by question, at machine speed. Staff retain a familiar workflow, and auditability improves because every step is captured consistently. In our tests, this approach delivered a five-to-tenfold speedup with an average accuracy of nearly 83 percent.

All Questions at Once: The second scenario asks a different question. If the decision tree already encodes the clinical logic, do we really need to traverse every branch for every case? Here, AI can view the entire decision tree and have that context to answer the questions, rather than looping through one question at a time. This delivered the best balance in our benchmark, achieving approximately 84 percent accuracy with an average time of two and a half minutes, and providing strong comfort for clinical staff who prefer established frameworks.

Text Only: The third scenario steps beyond trees altogether. Instead of translating criteria into a navigable flow, AI reads the raw policy documents, reasons over them, and decides whether an experienced reviewer is needed. There is nothing to build or maintain, which removes an entire class of work and makes adaptation to new drugs and guidelines far easier. In our benchmark, it is two to three times faster than giving the model one question at a time, with average accuracy around 64 percent. In addition to lower accuracy, the tradeoff is change management, since teams must adapt to a workflow that no longer revolves around tree navigation.
Datasets
We started with three real prior authorization criteria for Medicaid Puerto Rico (link to an example) and converted those documents into structured decision trees with clear yes/no questions and approval pathways. To create patient test cases, we relied on large language models to generate initial clinical profiles, which we then reviewed and refined. Furthermore, we enhanced the cases to add realistic complexity, ambiguity, and missing information. The final dataset comprises 24 cases of varying complexity, including bilingual scenarios that mix English and Spanish documentation to reflect real-world healthcare settings.
The goal was to test the clinical reasoning capabilities of existing models, not just pattern matching against simple criteria. We evaluated three AI scenarios against this set and ran head-to-head comparisons of the latest models from OpenAI, Anthropic, xAI, and Google.
Even though the final decision is what matters in a prior authorization workflow, we also investigated the auditability and traceability of using AI-enhanced workflows. We analyzed how models navigated decision paths, the completeness of their clinical rationale, and whether they cited appropriate evidence for each determination.
Results

From the results shown above, both approaches with a decision tree design, all questions at once and question-by-question, show stronger performance when evaluating PA criteria for a given case. The gap between these and the performance of the text only task is quite significant. This improvement is because the decision tree provides the models with a logical flow of questions to follow, derived directly from the criteria. In contrast, the text only task requires models to read the entire document and determine which criteria to apply and evaluate solely based on the patient profile, without structured guidance.
To illustrate these differences, we’ll walk through two representative cases that show how reasoning behavior shifts across tasks.
Example #1: Question-by-Question and All Questions at Once vs Text Only
Case Details:
- Diagnosis: Ankylosing Spondylitis
- Drug: Adalimumab
- Approved Dosage per criteria: 40 mg every other week (Q2W)
- Requested Dosage: 50 mg every other week
- Outcome: Deny
- Model: GPT-5
Question-by-Question and All Questions at Once:
Includes a question in their decision trees stating:
“Is the requested adalimumab dose/frequency within approved dosing for AS (SC 40 mg every other week)?”
GPT-5 answers this correctly for both tasks, reasoning that:
“The requested frequency is 40 mg once weekly, while the approved dosing for AS is 40 mg every other week. Therefore, the requested dose is not within the approved dosing schedule.”
This leads to the deny outcome, consistent with the decision tree logic.
Text Only Task:
GPT-5 correctly identifies the dosage amount criteria and that the requested dosage exceeds the approved amount, but outputs “approved” as the outcome of the case.
Its reasoning states:
“Requested weekly dose exceeds plan approved AS dosing; will approve at Q2W per criteria”.
In other words, the model interprets the discrepancy as an opportunity to adjust and approve the request at the plan-approved dosage rather than requesting clarification.
This example illustrates how the decision tree framework constrains model reasoning to a structured, criteria-based flow, while the criteria-only setup allows for interpretive shortcuts that can lead to inconsistent or incorrect outcomes.
Example #2: Question-by-Question vs All Questions at Once
Case Details:
- Diagnosis: Ankylosing Spondylitis
- Drug: Adalimumab
- Correct ICD-10 code: M45.9
- ICD-10 Code in request: J44.9
- Outcome: deny
- Model: Gemini
Question-by-Question:
Gemini followed a step-by-step path, beginning with the question:
“Is there a documented diagnosis of Ankylosing Spondylitis with ICD-10 (e.g., M45.9) in the submission?”
The model correctly determined that the specific diagnosis was not documented, citing that the only ICD-10 code listed in the patient profile was J44.9, which is unrelated to AS, and the absence of an explicitly stated diagnosis. Based on the evidence, the model correctly provided a "deny" outcome for the requests because it aligns with the criteria requirement for confirmed diagnosis documentation.
All Questions at Once:
With this task, the model inferred that sufficient evidence was provided to determine the diagnosis, resulting in an approval decision.
Its reasoning stated:
“Although the prescription form lists an incorrect ICD-10 code (J44.9), the comprehensive clinical documentation provides an unequivocal diagnosis of Ankylosing Spondylitis. Evidence includes classic imaging findings, positive HLA-B27, and a characteristic clinical presentation, which a clinical expert would recognize as sufficient documentation of the disease being treated.”
This example illustrates how the entire decision tree structure enables inferential reasoning, allowing models to make unverified assumptions. In contrast, the question-by-question format constrains reasoning narrowly, which reduces the risk of overinterpretation.
However, this behavior can also reverse in other cases, as reflected by the alternating accuracies between the two decision tree formats across models. The all questions at once structure can be advantageous when relevant information is scattered or partially implied, but the question-by-question approach may overlook broader connections. These findings suggest that the effectiveness of the tree design depends on the balance between the interpretation of flexibility and the evidence of strictness.

When we compare model accuracy to task completion, a clear tradeoff emerges between performance and efficiency across tasks. As shown in the previous visualization, both decision tree approaches consistently outperform the text only task in accuracy. However, the graph above shows that this improvement comes at the cost of longer processing times, especially for the traversal task. However, every agentic approach remains far cheaper and faster than human review. A typical prior authorization case takes around 50 minutes to review, compared to under five minutes for any of the AI configurations we tested. If a payer typically processes 1,000,000 cases per year, that gap represents roughly 750,000 hours of manual labor avoided, or $20M+.

*Costs may vary based upon length of case an example of one input for a case can be found in the Appendix.
Viewed in terms of cost per case, the time savings from AI review translate directly into lower operational expense. Among the configurations, the question-by-question method incurs the highest AI processing cost due to its step-by-step flow but offers the most substantial reasoning depth. The all questions a once approach achieves nearly the same accuracy while cutting review time, providing the best balance between precision and efficiency. The text only setup remains the most cost-effective and fastest, though its reduced accuracy limits use in high-stakes cases.
ROI Rubrics
Each approach balances speed, accuracy, and the potential for changes to the operating model.
- Efficiency versus infrastructure: Traditional enhancement provides immediate gains with no rebuilds, optimization delivers the best speed-to-familiarity ratio, and criteria-driven offers maximum long-term flexibility.
- Cost profile: Enhancing existing trees reduces costs by approximately 70% from the speed alone, optimization approaches 85% by cutting both labor and maintenance, and criteria-directed nears 90% by eliminating tree development and upkeep.
- Timeline and complexity: Enhancement typically lands in weeks with minimal training, while optimization often requires three to six months to pilot and scale, and criteria-driven changes usually require six to twelve months with a deeper change-management plan.
Regardless of the workflow you choose, every step is logged. Reviewers can see the evidence, verify the reasoning, and either confirm the decision or change it. We hope this benchmark will serve as a reference for decision-makers as they choose the right workflow, investment level, and timeline to modernize prior authorization without compromising clinical quality.
Appendix
PA Criteria Example:
PA Description
Dupilumab (Dupixent)
Managed by Managed Care Organizations (MCOs) contracted by the Puerto Rico Health Insurance Administration (known in Spanish as Administración de Seguros de Salud or ASES) to provide pharmacy services to the insured of the Government Health Plan.
Covered Uses
(a) Treatment of patients aged 6 months and older with moderate-to-severe atopic dermatitis whose disease is not adequately controlled with topical prescription therapies or when those therapies are not advisable. (ICD-10-CM L20.81, ICD-10-CM L20.9)
(b) As an add-on maintenance treatment in patients with moderate-to-severe asthma aged 6 years and older with an eosinophilic phenotype or with oral corticosteroid dependent asthma. (ICD-10-CM J45.5, ICD-10-CM J82)
(c) As an add-on maintenance treatment in adult patients with inadequately controlled chronic rhinosinusitis with nasal polyposis (CRSwNP). (ICD-10-CM J32.9, ICD-10-CM J33.9)
(d) Treatment of adult and pediatric patients aged 12 years and older, weighing at least 40 kg, with eosinophilic esophagitis (EoE). (ICD-10-CM K20.0)
Exclusion Criteria
(a) None
Required Medical Information
For Atopic Dermatitis:
For first prescription only:
(a) For moderate atopic dermatitis: Physician documents that:
(i) Patient has previous use of topical corticosteroids and topical calcineurin inhibitors, and crisaborole,
OR
(ii) Contraindicated use of topical corticosteroids and topical calcineurin inhibitors and crisaborole.
(b) For severe atopic dermatitis: Physician documents that:
(i) Patient has previous use of topical corticosteroids and topical calcineurin inhibitors.
OR
(ii) Contraindicated use of topical corticosteroids and topical calcineurin inhibitors.
For Asthma
For the first prescription:
(a) Physician must document all of the following:
(i) Classification of asthma as uncontrolled or inadequately controlled as defined by at least one of the following:
1. Poor symptom control (e.g., Asthma Control Questionnaire [ACQ] score consistently > 1.5 or Asthma Control Test [ACT] score consistently < 20)
2. Two or more bursts of systemic corticosteroids for at least 3 days each in the previous 12 months
3. Asthma-related emergency treatment (e.g., emergency room visit, hospital admission, or unscheduled physician’s office visit for nebulizer or other urgent treatment)
4. Airflow limitation (e.g., after appropriate bronchodilator withhold, FEV1 < 80% predicted, with reduced FEV1/FVC below lower limit of normal)
5. Patient is currently dependent on oral corticosteroids for the treatment of asthma.
– AND –
(ii) Dupixent will be used in combination with one of the following:
1. One high-dose (appropriately adjusted for age) combination inhaled corticosteroid (ICS)/long-acting beta2 agonist (LABA)
OR
2. Combination therapy including both of the following:
a. One high-dose (appropriately adjusted for age) ICS product
– AND –
b. One additional asthma controller medication
– AND –
(iii) One of the following:
1. Submission of medical records (e.g., chart notes, laboratory values, etc.) documenting that asthma is an eosinophilic phenotype as defined by a baseline (pre-dupilumab treatment) peripheral blood eosinophil level ≥ 150 cells/μL within the past 6 weeks.
OR
2. Patient is currently dependent on oral corticosteroids for the treatment of asthma.
For Renewal:
(a) Documentation of positive clinical response to Dupixent therapy as demonstrated by at least one of the following:
(i) Reduction in the frequency of exacerbations
(ii) Decreased utilization of rescue medications
(iii) Increase in percent predicted FEV1 from pretreatment baseline
(iv) Reduction in severity or frequency of asthma-related symptoms (e.g., wheezing, shortness of breath, coughing, etc.)
(v) Reduction in oral corticosteroid requirements
– AND –
(b) Dupixent is being used in combination with an ICS-containing controller medication.
For Chronic Rhinosinusitis with Nasal Polyposis:
(a) For the first prescription only: Documentation evidencing that chronic rhinosinusitis remains uncontrolled despite recommended doses of oral (e.g., prednisone) or intranasal glucocorticoids (e.g., fluticasone propionate, mometasone furoate, or beclomethasone).
(b) Document use in combination with other systemic glucocorticoids as maintenance treatment.
For Eosinophilic Esophagitis:
(a) Documented diagnosis of EoE by endoscopic biopsy
(b) For the first prescription only: Documentation evidencing 2 or more episodes of dysphagia per week.
Age Restriction
(a) For atopic dermatitis: Patients aged 6 months or older
(b) For asthma: Patients aged 6 years or older
(c) For chronic rhinosinusitis with nasal polyposis: 18 years of age or older
(d) For eosinophilic esophagitis: Patients aged 12 years or older weighing at least 40 kg
Prescriber Restriction
(a) For Atopic Dermatitis: Allergist/Immunologist, Dermatologist
(b) For Asthma and Chronic Rhinosinusitis with Nasal Polyposis: Pulmonologist/Pneumologist, Allergist/Immunologist, ENT Physician
(c) For Eosinophilic Esophagitis: Allergist/Immunologist or Gastroenterologist
Coverage Duration
(a) For Atopic Dermatitis, Asthma, Chronic Rhinosinusitis with Nasal Polyposis, and Eosinophilic Esophagitis: Six (6) months
(b) Renewal: Twelve (12) months
Other Criteria
(a) Refer to the prescribing information for dosage and administration.
Source: Dupilumab PA Criteria
Case Input Example for the Text Only Task:
PRIOR AUTHORIZATION REVIEW
You are a clinical expert evaluating a Prior Authorization (PA) request. You will be given:
- PA criteria document
- A patient's clinical profile
YOUR TASK
Review the payer's PA criteria and the patient's clinical profile below. Determine whether the request should be approve, deny, or pend (more information needed).
PAYER CRITERIA
Government Health Plan (GHP) of Puerto Rico
Authorization Criteria – Tumor Necrosis Factor Alpha (TNFα): Adalimumab (Humira®)
Managed by MCO
Section I. Prior Authorization Criteria
A. Physician must submit evidence of NEGATIVE intradermal tuberculin (PPD) test result, or Negative results of a chest X-ray, or a certification of negative tuberculosis risk.
B. Patient must meet all criteria / requisites according to diagnosis (see Section II for specific criteria per diagnosis).
- Rheumatoid Arthritis
- Juvenile Rheumatoid Arthritis
- Psoriatic Arthritis
- Ankylosing Spondylitis
- Plaque Psoriasis
- Hydradenitis Suppurativa — not covered
- Uveitis
- Adult Crohn’s Disease
- Pediatric Crohn’s Disease
- Ulcerative Colitis
C. Assess clinical response after the first three months of treatment.
D. Treatment should be discontinued if failure to therapy or toxicity is documented.
E. Follow Package Insert instructions for dose administration.
Section II. Specific Criteria per Diagnosis
a. Rheumatoid Arthritis
- Prescriber Restriction: Rheumatologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
- Rheumatoid Arthritis (ICD-10: M06.9, M05.0, M05.30, M05.60)
- Physician certifies that patient was treated with one or more disease-modifying anti-rheumatic drugs (see Table I) and failed therapy after 3 months of treatment or presented toxicity.
Approved Doses for Rheumatoid Arthritis:
- Adalimumab — Route: SC — Dose: 40 mg — Frequency: Every other week
- Note: Some patients not using MTX could benefit from an increase in dose frequency to 40 mg once a week
b. Juvenile Rheumatoid Arthritis
- Prescriber Restriction: Rheumatologist or Pediatric Rheumatologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
- Juvenile Rheumatoid Arthritis (ICD-10: M08.0, M08.2, M08.3, M08.40, M08.9)
- Patient is ≥ 2 years of age.
- Physician certifies prior treatment with one or more disease modifying anti-rheumatic drugs (see Table I) and failed therapy after three months of treatment or presented toxicity
- Systemic Juvenile Idiopathic Arthritis and various degrees of synovitis:
- Physician certifies that patient was treated with anakinra and failed therapy after one month of treatment, or presented toxicity.
Approved Doses for Juvenile Rheumatoid Arthritis:
- Adalimumab — Route: SC — Dose: 20 mg — Frequency: Body weight > 15 kg - <30 kg – every other week
- Adalimumab — Route: SC — Dose: 40 mg — Frequency: Body weight > 30 kg - every week
c. Psoriatic Arthritis
- Prescriber Restriction: Rheumatologist or Dermatologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
- Psoriatic Arthritis (ICD-10: L40.54, L40.59)
- Physician certifies that patient was treated with one or more disease modifying anti-rheumatic drugs (see Table I) and failed therapy after six months of treatment or presented toxicity.
d. Ankylosing Spondylitis
- Prescriber Restriction: Rheumatologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
- Ankylosing Spondylitis (ICD-10: M45.9)
- Patient must meet the following criteria according to the type of ankylosing spondylitis presented:
Symptomatic Axial disease
- Physician documents that patient failed treatment with at least two NSAIDs for at least three months, except if NSAIDs are contraindicated or if patient has presented toxicity or intolerance.
Symptomatic Enthesitis
- Physician documents that patient failed treatment with at least two NSAIDs for at least three months, except if NSAIDs are contraindicated or if patient has presented toxicity or intolerance, and
- Physician certifies that patient failed treatment with at least two intra-articular steroid injections, except if these are contraindicated or patient presents intolerance.
Symptomatic Periferal Arthritis
- Physician documents that patient failed treatment with at least two NSAIDs for at least three months, except if NSAIDs are contraindicated or if patient has presented toxicity or intolerance, and
- Physician certifies that patient failed treatment with at least two intra-articular steroid injections, except if these are contraindicated or patient presents intolerance, and
- Physician certifies that patient presented intolerance to treatment with sulfasalazine for at least four months, or that its use is contraindicated.
Approved Doses for Ankylosing Spondylitis:
- Adalimumab — Route: SC — Dose: 40 mg — Frequency: Every other week
e. Plaque Psoriasis
- Prescriber Restriction: Dermatologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
- Plaque Psoriasis (ICD-10: L40.8)
- Physician documents that:
- a. Patient has failed treatment with one or more topical agents used for the management of plaque psoriasis, andb. Patient has failed treatment with one of the following systemic agents: methotrexate, cyclosporine, acitretin, oral corticosteroids (or use is contraindicated), andc. Patient has failed phototherapy, or use is contraindicated, or patient does not have access to phototherapy.
f. Uveitis
- Prescriber Restriction: Ophthalmologist.
- Physician must document the diagnosis on the prescription:
- Uveitis – non-infectious intermediate, posterior, or panuveitis (ICD-10: H20.0, H30.9)
- Physician documents that:
- The patient has failed corticosteroid therapy, or that its use is contraindicated, or certifies that adalimumab should be the first line of treatment because of a high risk of visual loss.
g. Adult and Pediatric Crohn’s Disease
- Prescriber Restriction: Gastroenterologist or Pediatric Gastroenterologist (applies to 1st prescription and every 12 months).
- Physician must document the diagnosis on the prescription:
Adult Crohn’s Disease (CD):
Pediatric Crohn’s Disease: - Reducing signs and symptoms and inducing and maintaining clinical remission in adult patients with moderately to severely active Crohn’s disease who have had an inadequate response to conventional therapy. Reducing signs and symptoms and inducing clinical remission in these patients if they have also lost response to or are intolerant to infliximab. (ICD-10-CM K50.00 o ICD-10-CM K50.10 o ICD-10-CM K50.90).
- Reducing signs and symptoms and inducing and maintaining clinical remission in patients 6 years of age and older with moderately to severely active Crohn’s disease who have had an inadequate response to corticosteroids or immunomodulators such as azathioprine, 6-mercaptopurine, or methotrexate. (ICD-10-CM K50.00 o ICD-10-CM K50.10 o ICD-10-CM K50.90).
- Patient is ≥ 6 years of age
- Induction Treatment: Physician must certify one of the following:
- a. Patient has documented intolerance or failure to therapy with:i. Corticosteroids (e.g., Budesonide, Hydrocortisone)ii. 6-mercaptopurine (6-MP, Purinethol) and/or azathioprine (Imuran) and/or Methotrexate (MTX)iii. TNF-α antagonist: Infliximab (applies only to adult patients).b. Patient has been diagnosed with Crohn’s disease in severe stage or presents one or more of the following:i. Intestinal obstructionii. Severe malnutritioniii. Abscess formationiv. Perianal fistulas
- Maintenance Treatment for patients using corticosteroids or TNF-α antagonist:
- a. Physician certifies that patient has achieved remission, andb. Physician documents previous use of corticosteroids.
h. Ulcerative Colitis
- Prescriber Restriction: Gastroenterologist (applies to 1st prescription and every 12 months).
- Physician must document diagnosis on the prescription:
a. Inducing and sustaining clinical remission in adult patients with moderately to severely active ulcerative colitis who have had an inadequate response to immunosuppressants such as corticosteroids, azathioprine or 6-mercaptopurine (6-MP). The effectiveness of HUMIRA has not been established in patients who have lost response to or were intolerant to TNF blockers. (ICD-10-CM K51.90, K51.911, K51.912, K51.913, K51.914, K51.918, K51.919) - Physician documents that patient has been treated with of the following drugs and there is evidence of intolerance or failure to therapy:
- 5-aminosalicylic acid on formulary (5-ASA: mesalamine, olsalazine, sulfasalazine), andCorticosteroids (e.g. Budesonide, Hydrocortisone).
Table I. Agents classified as disease-modifying anti-rheumatic drugs (DMARDs)
- Hydroxychloroquine
- Leflunomide
- Methotrexate (MTX)
- Minocycline
- Sulfasalazine
Section III. References
- Enbrel [package insert], Amgen, November 2013
- Humira [package insert], AbbVie Inc., June 2016
- Cimzia [package insert], UCB Inc., October 2013
- Singh J, Furst D, Bharat A, et al. 2012 Update of the 2008 ACR Recommendations for DMARDs and Biologic Agents in RA. Arthritis Care & Research. 2012;64(5):625–639.
- Ringold S, Weiss P, et al. 2013 Update of 2011 ACR Recommendations for Juvenile Idiopathic Arthritis. Arthritis Care & Rheumatism. 2013;65(10):2499–2512.
- Gottlieb A, Korman N, et al. Guidelines of Care for Psoriasis and Psoriatic Arthritis. J Am Acad Dermatol 2008;58:851–864.
- Braun J, Pham T, et al.
PATIENT CLINICAL PROFILE
Demographics
- Patient: José A. Rivera (DOB: 1991-09-13), 34-year-old male
- Height: 178 cm; Weight: 92 kg; BMI: 29.0 kg/m²
- Insurance: Government Health Plan (GHP) of Puerto Rico – Medicaid, Member ID: GHP-PR-000123456
- Prescriber: Karla Morales, MD — Rheumatologist, NPI: 1487692604
- Facility: Hato Rey Rheumatology Associates, San Juan, PR
- Request date: 2025-10-20
- Requested drug: Adalimumab (Humira) 40 mg/0.4 mL pen; dose 40 mg SC every other week; quantity 2 pens per 28 days; indication: Ankylosing Spondylitis with symptomatic peripheral arthritis
Diagnoses (ICD-10)
- Primary: Ankylosing Spondylitis, unspecified site (M45.9); active disease with symptomatic peripheral arthritis (knees/ankles) and axial involvement
- Comorbidities:
- Type 2 diabetes mellitus without complications (E11.9)Essential hypertension (I10)Nonalcoholic fatty liver disease (K76.0)History of upper gastrointestinal hemorrhage related to NSAID use (K92.2) — event on 2025-06-18
Medication History (Step Therapy)
- Naproxen (NSAID) 500 mg orally twice daily
- Start: 2025-01-02; Stop: 2025-04-05Duration: ~13 weeksOutcome: Ineffective (inadequate response)Reason for discontinuation: Persistent back pain and peripheral joint swelling; BASDAI remained >6
- Indomethacin (NSAID) 50 mg orally three times daily
- Start: 2025-04-10; Stop: 2025-06-20Duration: ~10 weeksOutcome: Intolerant (toxicity)Reason: Developed melena and drop in Hgb; ED evaluation on 2025-06-18; indomethacin discontinued; GI advised avoidance of high-dose NSAIDs due to bleed risk
- Celecoxib (NSAID, COX-2) 200 mg orally twice daily
- Start: 2025-07-10; Stop: 2025-08-21Duration: ~6 weeks (borderline short trial)Outcome: IneffectiveReason: Minimal improvement in joint pain; stopped when moving to injections; note: does not meet full 3-month threshold
- Intra-articular steroid injection #1: Triamcinolone acetonide 40 mg, left knee
- Date: 2025-03-15Outcome: Ineffective (transient relief ~2–3 weeks; symptoms recurred)Reason for subsequent change: Continued knee swelling and pain
- Intra-articular steroid injection #2: Triamcinolone acetonide 30 mg, right ankle
- Date: 2025-07-01Outcome: Intolerant (adverse effects)Reason: Marked hyperglycemia (BG 340 mg/dL same day, ED visit), minimal symptom relief; additional steroid injections considered risky due to diabetes and prior hyperglycemia
- Sulfasalazine (DMARD)
- Dose: Initiated 500 mg orally twice daily; titrated to 1000 mg twice dailyStart: 2024-11-10; Stop: 2025-04-25Duration: ~5.5 monthsOutcome: Intolerant (adverse effects)Reason for discontinuation: Pruritic rash and ALT elevation to 75 U/L; symptoms resolved after discontinuation and ALT normalized
- Current non-biologic meds: Metformin 1000 mg BID; Lisinopril 20 mg daily; Pantoprazole 40 mg daily; Acetaminophen PRN; No current NSAID therapy due to GI bleed risk
Laboratory and Imaging Results
- PPD (intradermal tuberculin): 0 mm induration — Negative (Date: 2025-05-31)
- Chest X-ray: No acute cardiopulmonary process; no apical scarring (Date: 2024-12-01)
- ESR: 33 mm/hr (Date: 2025-10-08) [Normal: 0–20]
- CRP (high sensitivity): 10.2 mg/L (Date: 2025-10-08) [Normal: 0–8]
- CBC: WBC 7.2 x10^3/µL; Hgb 13.4 g/dL; Plt 265 x10^3/µL (Date: 2025-07-15)
- CMP: Creatinine 1.05 mg/dL; eGFR 92 mL/min/1.73 m² (Date: 2025-07-15)
- LFTs: ALT 75 U/L (Date: 2025-04-22) [Normal: 10–40]; ALT 36 U/L (Date: 2025-05-05) [Normal: 10–40]
- HbA1c: 8.2% (Date: 2025-07-15)
- Fasting Glucose: 142 mg/dL (Date: 2025-07-15)
- Hepatitis B surface antigen: Non-reactive (Date: 2025-05-06)
- HLA-B27: Positive (Date: 2024-12-20)
- MRI Sacroiliac joints: Bilateral sacroiliitis with bone marrow edema (Date: 2025-02-18)
Clinical Notes
- 2025-02-20 — Rheumatology new patient consult: 34-year-old man with chronic inflammatory back pain, morning stiffness >90 minutes, limited lumbar flexion; MRI SI joints confirms sacroiliitis. BASDAI 6.6. Peripheral arthritis of knees and ankles active with synovitis; Achilles enthesitis noted. Plan: begin naproxen; consider injections for swollen knee; TB screening ordered.
- 2025-06-25 — Follow-up after ED visit: Indomethacin stopped due to upper GI bleed on 2025-06-18. Patient continues to have peripheral swelling of left knee and right ankle; axial pain persists. GI advises avoiding high-dose NSAIDs; started pantoprazole. One prior left knee steroid injection (3/15/25) had only brief benefit.
- 2025-10-10 — Pre-authorization visit: Persistent symptomatic peripheral arthritis (knees/ankles) with swelling and warmth; BASDAI 5.9; CRP 10.2 mg/L; ESR 33 mm/hr. Two intra-articular steroid injections completed (left knee 3/15/25 with transient benefit; right ankle 7/1/25 complicated by severe hyperglycemia). Sulfasalazine trial for >5 months resulted in rash and LFT elevation; discontinued 4/25/25. PPD negative 5/31/25; CXR 12/1/24 clear. Request adalimumab for active ankylosing spondylitis with symptomatic peripheral arthritis refractory to NSAIDs, injections, and sulfasalazine.
Plan and Prescriber Justification
- Diagnosis: Ankylosing spondylitis with predominant symptomatic peripheral arthritis; axial disease also present.
- Step therapy met: failed naproxen 3 months (ineffective); indomethacin stopped early due to GI bleed (toxicity/contraindication); completed two intra-articular steroid injections with inadequate response and intolerance; sulfasalazine >4 months with intolerance.
- Safety: TB screening negative; hepatitis B negative; LFTs normalized after discontinuation of sulfasalazine; will monitor for infection and glycemic control.
- Requested therapy: Adalimumab 40 mg SC every other week per package insert for AS.
- Monitoring: Assess clinical response at 3 months; discontinue for failure/toxicity; coordinate GI follow-up for NSAID contraindication and diabetes management.
INSTRUCTIONS
- Read the PA criteria carefully — Understand all requirements
- Review the patient profile — Extract relevant clinical information
- Apply criteria to patient — Check if requirements are met
- Make a decision — approve, deny, or pend with detailed reasoning
- Document key findings — List the critical factors that influenced your decision
OUTPUT FORMAT (JSON)
{ "case_id": "pa-criteria-ankylosing-spondylitis-approve", "decision": "approve" | "deny" | "pend", "rationale": "Comprehensive explanation of your decision (2-4 paragraphs). Explain which criteria were met or not met, cite specific evidence from the patient profile, and justify your final decision.", "key_findings": [ { "criterion": "Brief description of the criterion being evaluated", "finding": "What you found in the patient profile", "met": true | false, "evidence": "Specific quote or reference from patient profile" } ]}
IMPORTANT NOTES
- Be thorough — consider all aspects of the PA criteria
- Be specific — cite exact evidence from the patient profile (dates, values, medications, etc.)
- Be conservative — if critical information is missing or unclear, consider pending
- Document your reasoning — explain your thought process clearly
- Decision should always be approve, pend and deny
Begin your review now.
