Problem List Review

Overview

Perform a structured audit of the active problem list by pulling Condition resources and cross-referencing with MedicationRequests, recent Encounters, and Observations. Identify clinical discrepancies: conditions without active treatment, resolved conditions still listed as active, medications implying undocumented conditions, and ICD-10 coding specificity opportunities. Output actionable findings for clinical review.

FHIR Resources Used

Instructions

Step 1: Pull All Conditions

Tool: fhir_search
resourceType: "Condition"
queryParams: "patient=[patient-id]"

Retrieve all conditions regardless of status. Categorize each by:

• •clinicalStatus: active, recurrence, relapse, inactive, remission, resolved
• •verificationStatus: confirmed, unconfirmed, provisional, differential, refuted, entered-in-error
• •category: problem-list-item, encounter-diagnosis, health-concern
• •code.coding: Extract SNOMED CT code and ICD-10-CM code if present

Build two lists:

1. •Active conditions (clinicalStatus = active, recurrence, or relapse)
2. •Inactive/resolved conditions (clinicalStatus = inactive, remission, or resolved)

Step 2: Pull Active Medications

Tool: fhir_search
resourceType: "MedicationRequest"
queryParams: "patient=[patient-id]&status=active"

For each medication, extract:

• •Drug name and RxNorm code from medicationCodeableConcept
• •Drug class (infer from name suffix or RxNorm hierarchy)
• •Dosage from dosageInstruction

Step 3: Pull Recent Encounters (Last 12 Months)

Tool: fhir_search
resourceType: "Encounter"
queryParams: "patient=[patient-id]&date=ge[12-months-ago]&_sort=-date"

Extract encounter diagnoses from reasonCode or associated Condition references with category = encounter-diagnosis.

Step 4: Cross-Reference Medications Against Problems

For each active medication, check if a corresponding condition exists on the active problem list. Use the medication-condition mapping reference.

Flag Type A -- Medication Without Matching Condition (Implied Diagnosis Missing):

If medication is present but no matching condition exists on the active problem list, flag as: "Patient is on [medication] but [condition] is not on the active problem list."

Step 5: Check for Conditions Without Active Treatment

For each active condition, check if there is at least one active medication or recent encounter addressing it.

Flag Type B -- Untreated Active Condition:

• •Active condition with no medication from the expected drug class
• •No encounter in the last 12 months with the condition as a reason code
• •Exception: conditions that may not require pharmacotherapy (e.g., obesity managed with lifestyle, resolved fracture under monitoring)

Step 6: Identify Resolved Conditions Still Listed as Active

Flag Type C -- Potentially Resolved Conditions:

• •Condition has clinicalStatus = active but abatementDateTime is in the past
• •Condition is an acute diagnosis (e.g., pneumonia, UTI, fracture) with onset > 6 months ago and no recent encounter referencing it
• •Condition was treated with a medication that has been stopped

Verify with lab data where applicable:

Tool: fhir_search
resourceType: "Observation"
queryParams: "patient=[patient-id]&code=[relevant-LOINC]&_sort=-date&_count=1"

Examples:

• •"Diabetes" active but latest HbA1c (LOINC 4548-4) is consistently < 5.7% for 2+ years and no diabetes medications active
• •"Hypothyroidism" active but latest TSH (LOINC 3016-3) normal and no levothyroxine prescribed
• •"Iron deficiency anemia" active but latest ferritin (LOINC 2276-4) and hemoglobin (LOINC 718-7) both normal

Step 7: Assess ICD-10 Coding Specificity

For each active condition, check if the ICD-10 code can be made more specific:

Flag Type D -- Coding Specificity Opportunities:

Step 8: Format Output

PROBLEM LIST REVIEW
===================
Patient: [name] ([patient-id])
Reviewed: [timestamp]
Active Conditions: [count]
Inactive/Resolved: [count]
Active Medications: [count]

ACTIVE PROBLEM LIST
-------------------
1. [Condition] - SNOMED: [code] / ICD-10: [code] - Onset: [date]
   Treatment: [matching medication(s)]
2. ...

FINDINGS
--------

[A] MISSING DIAGNOSES (Medications imply undocumented conditions)
  1. On Metformin 1000mg BID but Type 2 Diabetes not on problem list
     -> Recommend adding: E11.9 Type 2 diabetes mellitus without complications
  2. ...

[B] UNTREATED CONDITIONS
  1. Essential hypertension (I10) - Active since 2020 - No antihypertensive medication found
     -> Verify: Is patient on lifestyle management only? If pharmacotherapy indicated, address.
  2. ...

[C] POTENTIALLY RESOLVED CONDITIONS
  1. Acute bronchitis (J20.9) - Onset: 2025-03-15 - Still listed as active
     -> Recommend: Update clinicalStatus to "resolved"
  2. ...

[D] CODING SPECIFICITY OPPORTUNITIES
  1. E11.9 (DM2 without complications) - Patient has documented diabetic neuropathy
     -> Recommend: E11.40 (Type 2 DM with diabetic neuropathy)
  2. ...

SUMMARY
-------
- [N] conditions reviewed
- [N] missing diagnoses identified
- [N] conditions potentially untreated
- [N] conditions potentially resolved
- [N] coding improvements suggested

Examples

Example 1: Routine Problem List Audit

User says: "Review the problem list for patient 44321"

Actions:

1. •fhir_search Condition?patient=44321 -- returns 8 conditions (6 active, 2 resolved)
2. •fhir_search MedicationRequest?patient=44321&status=active -- returns 7 active medications
3. •fhir_search Encounter?patient=44321&date=ge2025-02-07&_sort=-date -- returns 4 encounters

Cross-reference findings:

• •Metformin 1000mg and Glipizide 5mg on med list, "Type 2 Diabetes" IS on problem list. OK.
• •Lisinopril 20mg on med list, "Hypertension" IS on problem list. OK.
• •Omeprazole 20mg on med list, but NO GERD or peptic ulcer on problem list. FLAG A.
• •Sertraline 100mg on med list, but NO depression or anxiety on problem list. FLAG A.
• •"Acute sinusitis" (J01.90) onset 2025-01-10, still active, no recent encounter. FLAG C.
• •Diabetes coded as E11.9 but patient also has documented CKD Stage 3. FLAG D: suggest E11.22.

Result: Formatted report with 2 missing diagnoses, 1 potentially resolved condition, 1 coding improvement.

Example 2: Post-Discharge Problem List Reconciliation

User says: "Audit the problem list for patient 77890 after their hospital stay"

Actions:

1. •fhir_search Condition?patient=77890 -- returns 12 conditions
2. •fhir_search MedicationRequest?patient=77890&status=active -- returns 10 medications
3. •fhir_search Encounter?patient=77890&date=ge2025-02-07&_sort=-date -- returns 6 encounters including inpatient stay

Findings:

• •New medications from discharge (apixaban, metoprolol) without corresponding new diagnoses added (atrial fibrillation). FLAG A.
• •Hospital-acquired pneumonia on problem list from admission, patient discharged 3 weeks ago with completed antibiotic course. FLAG C.
• •"Heart failure" (I50.9, unspecified) could be I50.22 (chronic systolic, compensated) based on echo results. FLAG D.

Troubleshooting

Condition Resources Use Only ICD-10 Codes Without SNOMED

• •Cross-reference is still possible. Use the ICD-10 code directly for matching.
• •Note that some medication-condition mappings are more reliable with SNOMED codes. Use the display text as a fallback for matching.

MedicationRequest Uses medicationReference Instead of medicationCodeableConcept

• •The medication details are in a separate Medication resource. Resolve the reference:
  code

  Tool: fhir_read
  resourceType: "Medication"
  id: "[medication-reference-id]"
  

• •Extract the drug name from Medication.code.coding.

Too Many Conditions Returned (Including Historical)

• •Filter client-side by category = "problem-list-item" to exclude encounter-only diagnoses.
• •Sort by clinicalStatus to prioritize active conditions.
• •If the list exceeds 50 items, paginate with _count=50 and follow Bundle.link where relation = "next".

Related Skills

• •clinical-summary-generator -- for full chart summary including problems in context
• •allergy-adverse-reaction-summary -- for allergy-specific review
• •patient-demographics-summary -- when problem list review reveals missing demographic data