Case study 01
The prescribing hazard
A case study from the 433-record AI indexing review
What the human summary covered
The human summary for this record covered the patient’s chest pain, paroxysmal atrial fibrillation, myocardial infarction (2014), echocardiogram findings (left ventricular hypertrophy, regurgitations), atelectasis, metoprolol, and lab results. It was a competent, clinically focused summary.
What the AI found that the human missed
| Finding | Type · clinical risk |
|---|---|
Sulfa allergy Drug allergy | Prescribing a sulfonamide antibiotic (e.g., Bactrim for a UTI or skin infection) to this patient risks anaphylaxis. Sulfonamides are commonly prescribed empirically, the prescriber would have no signal to avoid them. |
Tramadol Opioid analgesic (Schedule IV) | An active opioid prescription absent from the summary means the next clinician may prescribe another opioid (duplicate prescribing) or underestimate the patient’s current pain regimen. |
Nitroglycerin (Nitrostat) Cardiac emergency medication | Nitroglycerin use indicates active or recent angina episodes. Missing this means downstream clinicians don’t know the patient has acute cardiac symptom management beyond their documented MI history. |
Ondansetron (Zofran) Anti-nausea medication | Relevant to the patient’s overall symptom burden and medication interaction profile. |
Abdominal pain Clinical finding | A documented symptom that could indicate GI pathology, relevant to differential diagnosis. |
Frozen right shoulder Clinical finding | A musculoskeletal condition requiring its own treatment track (physical therapy, possible injection, surgical consideration). Invisible in the summary means it’s invisible to the next treating provider. |
How this record compares
Clinician-documented findings the human summary missed, against the 433-record study
Why this case matters
The sulfa allergy is the headline. It is documented in the clinical notes. It is not in the human summary. If the next provider prescribes a sulfonamide antibiotic, which is a routine, common prescription for urinary tract infections and skin infections, there is nothing in the summary to stop them. The AI index caught it. The human summary did not. This is not a quality-of-summary problem. It is a bandwidth problem: the human reviewer read the chart, wrote a thorough summary of the cardiac and pain issues, and the allergy on page N simply didn’t make it in.
“A missed sulfa allergy before prescribing a sulfonamide is not a theoretical risk.”
See how ChartInsight indexes your records