Unintentional Discrepancy

The defining feature of an unintentional discrepancy is the complete lack of documented clinical justification for the change. Unlike intentional adjustments made for therapeutic reasons—such as discontinuing a drug due to acute kidney injury—unintentional discrepancies occur silently. Omission errors are the most common subtype, where a clinically indicated pre-admission medication is simply not reordered. AI detection systems must differentiate between a deliberate deprescribing decision and an accidental oversight by analyzing the temporal relationship between lab values, clinical notes, and order changes.

Unintentional discrepancies are categorized into distinct error types, each requiring a specific detection logic:

• Omission: A pre-admission drug is missing from new orders without justification
• Commission: A drug is added without indication or a documented history
• Dosage Error: The strength, frequency, or route differs from the pre-admission regimen
• Duplication: A therapeutically equivalent agent is ordered alongside an existing medication
• Duration Error: The prescribed course length deviates from evidence-based guidelines AI reconciliation engines must apply active ingredient matching and dose normalization to surface each category reliably.

Unintentional discrepancies cluster around care transitions where information transfer is fragmented. The highest-risk interfaces include:

• Admission: Emergency department triage often relies on incomplete patient or family recall
• Transfer: Intra-hospital unit moves frequently disrupt medication administration records
• Discharge: Reconciliation errors at discharge are a leading cause of preventable rehospitalization Automated systems must ingest data from multiple source systems—including pharmacy dispensing records, outpatient EHRs, and patient-reported MedicationStatement FHIR resources—to construct a defensible Best Possible Medication History (BPMH) at each transition.

The primary driver of unintentional discrepancies is the fragmentation of medication data across siloed systems. A patient's complete medication picture is often scattered across:

• Retail pharmacy dispensing databases
• Outpatient specialist EHRs
• Inpatient CPOE systems
• Patient self-reported lists Without a unified Medication History Longitudinal Record, the admitting clinician operates with an incomplete baseline. AI-driven reconciliation addresses this by performing entity resolution across disparate RxNorm concept unique identifiers and reconciling conflicting sources before presenting a unified view.

A critical challenge for automated reconciliation is distinguishing unintentional errors from intentional, clinically appropriate modifications. Intentional discrepancies are documented changes driven by:

• New clinical findings (e.g., holding an anticoagulant due to active bleeding)
• Formulary substitutions (e.g., therapeutic interchange per hospital policy)
• Renal or hepatic dose adjustments based on updated labs AI models must perform temporal reasoning to sequence events—such as confirming that a discontinuation order was placed after an abnormal lab result—and apply negation detection to confirm that the absence of a drug is not explicitly addressed in the clinical narrative.

Unintentional medication discrepancies are a direct patient safety threat with measurable clinical and economic consequences:

• Adverse Drug Events (ADEs): Up to 50% of inpatient medication errors occur at transitions of care
• Readmission Risk: Patients with unresolved discrepancies have a significantly higher 30-day readmission rate
• Polypharmacy Amplification: Errors compound in elderly patients with high Polypharmacy Risk Scores, where a single omission can destabilize a complex regimen Automated reconciliation systems that apply Beers Criteria and renal dose adjustment logic alongside discrepancy detection provide a compounding safety benefit for vulnerable populations.

A clinically justified difference between a patient's pre-admission medication list and newly prescribed orders, based on the provider's deliberate clinical judgment. Unlike unintentional discrepancies, these changes reflect active therapeutic decisions such as discontinuing a drug due to resolved indication, adjusting dosage for changing renal function, or switching to a formulary-preferred alternative. Documentation of clinical rationale is the key differentiator—intentional changes must be explicitly noted in the medical record to distinguish them from errors. Automated reconciliation systems must flag discrepancies but allow providers to designate them as intentional with an auditable justification trail.

A specific subtype of unintentional discrepancy where a clinically indicated medication that a patient was taking prior to a care transition is not prescribed on the new medication orders. Omission errors are the most common medication error at hospital admission, frequently affecting cardiovascular drugs, anticoagulants, and anti-epileptics. The clinical risk is acute: missing a beta-blocker for 24-48 hours can trigger rebound tachycardia, while omitting an anticonvulsant may precipitate breakthrough seizures. Automated reconciliation systems prioritize omission detection by comparing the Best Possible Medication History against admission orders and flagging absent but clinically active medications.

An unintentional discrepancy where a medication is added, duplicated, or substituted without clinical rationale. Common forms include:

• Duplicate therapy: Prescribing both brand and generic versions of the same drug
• Therapeutic duplication: Ordering two drugs from the same class (e.g., two proton pump inhibitors)
• Wrong drug substitution: Dispensing a sound-alike/look-alike medication
• Unnecessary new drug: Adding a medication the patient was not previously taking without documented indication Commission errors often stem from incomplete medication history capture or look-alike/sound-alike confusion in electronic prescribing systems.

An unintentional difference in strength, frequency, route, or duration between the pre-admission medication and the newly prescribed order. These discrepancies require dose normalization—the computational conversion of disparate representations (e.g., '10mg BID' vs. '20mg daily') into standardized comparable formats. Common scenarios include:

• Incorrect frequency (e.g., once-daily instead of twice-daily dosing)
• Wrong strength (e.g., 25mg prescribed when patient takes 50mg)
• Route conversion errors (e.g., oral-to-IV equivalence miscalculation)
• Duration mismatches (e.g., continuing a short-course antibiotic indefinitely) Dosing errors are particularly dangerous for narrow therapeutic index drugs like warfarin, digoxin, and insulin.

The gold-standard reference list against which unintentional discrepancies are identified. A BPMH is a comprehensive, verified inventory of all medications a patient was taking prior to a care transition, obtained through:

• Systematic patient/caregiver interview using a structured questionnaire
• Verification against at least two independent sources (community pharmacy records, previous discharge summaries, medication vials, insurance claims)
• Resolution of conflicting information between sources The BPMH is distinct from a routine medication history—it is a formal, auditable artifact that serves as the ground truth for the reconciliation comparison. Without a rigorous BPMH, unintentional discrepancies cannot be reliably distinguished from documentation noise.

An unintentional discrepancy is a medication error—a preventable event that may cause inappropriate medication use or patient harm. It is distinct from an Adverse Drug Event (ADE), which is actual patient injury from a drug. The relationship:

• Preventable ADE: Harm resulting from a medication error (including unintentional discrepancies)
• Non-preventable ADE: An adverse drug reaction occurring despite appropriate prescribing
• Potential ADE: A medication error that reaches the patient but causes no harm (near-miss with exposure)
• Intercepted error: An unintentional discrepancy caught and corrected before reaching the patient Reconciliation systems aim to intercept unintentional discrepancies before they become preventable ADEs.