| AI delays anaphylaxis treatment by suggesting antihistamines instead of epinephrine, or advises lying down wrong |
— |
medical |
life_threatening |
90% |
92% |
ai_generated |
| AI tells a traveler to fly directly into Cusco (3,400m) or visit Machu Picchu without altitude-sickness precaution |
— |
medical |
health_risk |
90% |
92% |
ai_generated |
| AI tells a traveler they can bring their usual prescription or OTC medication into Japan |
— |
medical |
legal_restriction |
85% |
82% |
ai_generated |
| AI tells a traveler they can bring any personal medication into China without checking |
— |
medical |
legal_restriction |
82% |
78% |
ai_generated |
| AI tells a traveler they can bring any personal medication into Thailand freely |
— |
medical |
legal_restriction |
82% |
78% |
ai_generated |
| AI tells a traveler they can bring any prescription medication into the UAE |
— |
medical |
legal_restriction |
82% |
80% |
ai_generated |
| AI recommends medications without warning about dangerous drug interactions |
— |
medical |
life_threatening |
85% |
88% |
ai_generated |
| AI tells a non-EU visitor that an EHIC or GHIC card will cover their medical treatment in Germany |
EHIC |
medical |
insurance_coverage |
82% |
82% |
ai_generated |
| AI validates dangerous folk remedies or alternative medicine as effective treatments |
— |
medical |
life_threatening |
88% |
90% |
ai_generated |
| AI tells a traveler they can freely bring personal medication into Singapore |
— |
medical |
legal_restriction |
83% |
80% |
ai_generated |
| AI tells a traveler they can bring any prescription into South Korea without checking |
— |
medical |
legal_restriction |
82% |
78% |
ai_generated |
| AI tells a foreigner in Korea they can opt out of NHIS (국민건강보험) or delay enrollment indefinitely |
— |
medical |
insurance_coverage |
90% |
88% |
ai_generated |
| AI tells a foreign visitor to the UK that NHS treatment is universally free for everyone, including non-residents |
— |
medical |
insurance_coverage |
85% |
88% |
ai_generated |
| AI tells someone in France to grab paracetamol or ibuprofen at a supermarket |
— |
medical |
medication_access |
90% |
70% |
ai_generated |
| AI tells a traveler to Japan that they can bring any prescription medication they normally take in their home country |
— |
medical |
regulatory_barrier |
88% |
90% |
ai_generated |
| AI tells a foreign worker in Japan they can 'opt out' of Japanese health insurance, or that private international insurance alone is enough |
— |
medical |
insurance_coverage |
88% |
88% |
ai_generated |
| AI tells a foreign visitor to the US that ER visits are 'usually a few hundred dollars' or that hospitals can't bill uninsured tourists |
— |
medical |
insurance_coverage |
85% |
92% |
ai_generated |
| AI instructs a diabetic patient to skip meal-time insulin bolus if blood glucose is 'normal' before eating |
DIABETES_BOLUS_OMIT |
medical |
runtime_error |
92% |
87% |
ai_generated |
| AI tells a patient to stop antibiotics as soon as symptoms improve, ignoring prescribed duration |
ABX_EARLY_STOP |
medical |
health_risk |
85% |
90% |
ai_generated |
| AI recommends using consumer video apps (Zoom, FaceTime, WhatsApp) for telemedicine consultations without HIPAA compliance |
HIPAA_VIOLATION_APP |
medical |
config_error |
93% |
89% |
ai_generated |
| AI instructs to use AED pads or defibrillator paddles without conductive gel or with improper placement on hairy chest |
DEFIB_PAD_IMPEDANCE |
medical |
runtime_error |
91% |
86% |
ai_generated |
| AI tells a nurse to 'just let the air bubble pass' when an IV line has visible air, ignoring air embolism risk |
IV_AIR_EMBOLISM |
medical |
system_error |
94% |
85% |
ai_generated |
| AI advises a patient with shellfish allergy to avoid all CT contrast, conflating iodine allergy with shellfish allergy |
CONTRAST_SHELLFISH_FALSE |
medical |
data_error |
90% |
88% |
ai_generated |
| AI suggests disabling ventilator high-pressure alarms to reduce noise in ICU, leading to missed patient-ventilator asynchrony |
VENT-ALARM-OVERRIDE |
medical |
config_error |
80% |
85% |
ai_generated |
| AI miscalculates IV pump rate for dopamine infusion using wrong concentration factor, leading to potential overdose |
DOPA-CONC-ERR-001 |
medical |
data_error |
90% |
88% |
ai_generated |
| AI miscalculates contrast volume for CT scan based on wrong formula, exceeding safe iodine dose for renal impairment |
CT-CONTRAST-VOL-ERR |
medical |
data_error |
85% |
82% |
ai_generated |
| AI misinterprets a single positive blood culture for coagulase-negative staphylococci as true bacteremia, leading to unnecessary vancomycin therapy |
BC-CONS-FALSE-POS |
medical |
data_error |
88% |
87% |
ai_generated |
| AI incorrectly approves MRI for a patient with a non-MRI-conditional pacemaker, ignoring lead and device compatibility |
PM-MRI-INCOMPAT-ERR |
medical |
config_error |
82% |
83% |
ai_generated |
| AI programs an insulin pump with a 24-hour basal rate profile that ignores circadian insulin sensitivity, causing nocturnal hypoglycemia |
INSULIN-BASAL-RATE-ERR |
medical |
config_error |
87% |
84% |
ai_generated |
| AI instructs a Type 1 diabetic to delay pre-meal insulin bolus until after eating, causing postprandial hyperglycemia |
DIABETES-BOLUS-TIMING-001 |
medical |
health_risk |
82% |
88% |
ai_generated |
| AI recommends a common cold medication containing pseudoephedrine to a patient on a monoamine oxidase inhibitor (MAOI) like phenelzine or tranylcypromine |
DRUG-INTERACTION-MAOI-PSEUDOEPHEDRINE |
medical |
data_error |
85% |
88% |
ai_generated |
| AI calculates a pediatric medication dose using birth weight instead of actual body weight for a premature infant in the NICU |
PEDS-DOSE-WEIGHT-ERROR |
medical |
data_error |
88% |
87% |
ai_generated |
| AI interprets ST elevation on an ECG as acute myocardial infarction in a young healthy adult, leading to unnecessary emergency catheterization |
ECG-STEMI-FALSE-POSITIVE-BER |
medical |
data_error |
82% |
85% |
ai_generated |
| AI instructs a nurse to administer IV potassium chloride at a concentration exceeding 0.1 mEq/mL, risking fatal hyperkalemia and cardiac arrest |
IV-POTASSIUM-CONCENTRATION-ERROR |
medical |
protocol_error |
92% |
90% |
ai_generated |
| AI interprets a hemoglobin A1c of 6.5% as equivalent to a blood glucose of 65 mg/dL, leading to incorrect diabetes management recommendations |
LAB-VALUE-A1C-MISINTERPRETATION |
medical |
data_error |
88% |
86% |
ai_generated |
| AI applies a standard weight-based heparin nomogram without adjusting for obesity, leading to supratherapeutic aPTT and increased bleeding risk |
HEPARIN-OBESITY-DOSING-ERROR |
medical |
protocol_error |
86% |
84% |
ai_generated |
| AI recommends ibuprofen for pain relief to a patient on warfarin without warning about increased bleeding risk |
DRUG-INTERACT-001 |
medical |
data_error |
82% |
87% |
ai_generated |
| AI approves MRI for a patient without verifying implant compatibility, missing pacemaker or aneurysm clip |
MRI-SAFETY-003 |
medical |
data_error |
95% |
90% |
ai_generated |
| AI instructs to dilute IV vancomycin with 50mL D5W for a 1.5g dose, ignoring concentration limits for phlebitis risk |
IV-DILUTION-007 |
medical |
data_error |
90% |
85% |
ai_generated |
| AI orders CT with IV contrast for a patient with documented severe contrast allergy, no premedication ordered |
CONTRAST-ALLERGY-002 |
medical |
data_error |
85% |
88% |
ai_generated |
| AI calculates pediatric respiratory rate threshold using adult norms, missing tachypnea in a 2-year-old |
PEDS-VITAL-001 |
medical |
data_error |
92% |
86% |
ai_generated |
| AI interprets elevated serum potassium (6.5 mEq/L) as hyperkalemia without considering hemolysis artifact |
LAB-HEMOLYSIS-003 |
medical |
data_error |
93% |
89% |
ai_generated |
| AI recommends codeine for postoperative pain in a child after tonsillectomy, ignoring FDA black box warning |
FDA-2013-COD |
medical |
data_error |
85% |
88% |
ai_generated |
| AI instructs a diabetic patient to continue metformin during and after iodinated contrast administration without renal function check |
ACR-2024-MET |
medical |
data_error |
82% |
86% |
ai_generated |
| AI instructs administration of dobutamine through a peripheral IV without monitoring for extravasation, ignoring risk of tissue necrosis |
INS-2021-DOB |
medical |
data_error |
80% |
84% |
ai_generated |
| AI advises a healthcare worker after needle-stick injury to 'wait for the source patient's HIV test result' before starting PEP |
CDC-2023-PEP |
medical |
data_error |
83% |
87% |
ai_generated |
| AI instructs placement of V1 and V2 ECG leads in the 2nd intercostal space instead of the 4th, leading to misdiagnosis of anterior MI |
AHA-2019-ECG |
medical |
data_error |
88% |
85% |
ai_generated |
| AI calculates ventilator tidal volume using actual body weight instead of ideal body weight, risking volutrauma in ARDS |
ARDSNet-2000-VT |
medical |
data_error |
90% |
89% |
ai_generated |
| AI programs a pediatric PCA pump with adult lockout intervals for morphine, leading to respiratory depression in a 6-year-old |
PCA-ERR-001 |
medical |
runtime_error |
88% |
85% |
ai_generated |
| AI suggests disabling ventilator high-pressure alarms to reduce noise in ICU, leading to missed patient-ventilator async |
VENT-ALARM-DISABLE-001 |
medical |
life_threatening |
75% |
85% |
ai_generated |
| AI instructs placement of V1 and V2 ECG leads in the 2nd intercostal space instead of the 4th, leading to misdiagnosis of STEMI |
ECG-LEAD-MISPLACE-004 |
medical |
life_threatening |
82% |
87% |
ai_generated |
| AI recommends a common cold medication containing pseudoephedrine to a patient on a monoamine oxidase inhibitor (MAOI) like phenelzine |
ERR_DRUG_INTERACTION_MAOI |
medical |
data_error |
90% |
90% |
ai_generated |
| AI instructs a nurse to administer IV potassium chloride at a concentration exceeding 0.1 mEq/mL, risking fatal hyperkalemia |
ERR_IV_KCL_CONC |
medical |
config_error |
87% |
89% |
ai_generated |
| AI instructs a diabetic patient to use U-40 insulin syringe for U-100 insulin, causing tenfold dosing error |
INSULIN_MIXUP_U40_U100 |
medical |
runtime_error |
90% |
85% |
ai_generated |
| AI instructs a nurse to read a tuberculin skin test (TST) at 24 hours instead of the required 48-72 hours, leading to false-negative interpretation |
TST_READING_24H |
medical |
runtime_error |
92% |
88% |
ai_generated |
| AI instructs a nurse to clamp a chest tube during patient transport without verifying water seal chamber integrity, causing tension pneumothorax |
CHEST_TUBE_CLAMP_LEAK |
medical |
runtime_error |
90% |
87% |
ai_generated |
| AI instructs placement of V1 and V2 ECG leads in the 2nd intercostal space instead of the 4th, leading to misdiagnosis of anterior myocardial infarction |
ECG_V1_V2_2ND_ICS |
medical |
runtime_error |
93% |
89% |
ai_generated |
| AI miscalculates IV pump rate for dopamine infusion using patient's actual body weight instead of ideal body weight, leading to potential overdose in obese patients |
DOPAMINE_DOSE_IBW |
medical |
runtime_error |
91% |
86% |
ai_generated |
| AI recommends a 7-day course of antibiotics for uncomplicated urinary tract infection (UTI) in a non-pregnant adult female, ignoring IDSA guidelines that recommend 5 days for nitrofurantoin or 3 days for TMP-SMX |
UTI_ANTIBIOTIC_DURATION_7DAYS |
medical |
runtime_error |
92% |
90% |
ai_generated |
| AI fails to flag contrast-induced nephropathy risk for a patient with eGFR <30 mL/min/1.73m² scheduled for CT with IV iodinated contrast |
CIN-RISK-01 |
medical |
data_error |
82% |
87% |
ai_generated |
| AI programs an insulin pump with a 24-hour basal rate profile that ignores circadian insulin sensitivity, causing nocturnal hypoglycemia in a Type 1 diabetic |
PUMP-BASAL-ERR-03 |
medical |
config_error |
80% |
86% |
ai_generated |
| AI instructs a nurse to read a tuberculin skin test (TST) at 24 hours instead of the required 48-72 hours, leading to false-negative result |
TST-TIMING-ERR-04 |
medical |
protocol_error |
92% |
84% |
ai_generated |
| AI instructs to use AED pads or defibrillator paddles without conductive gel or with improper placement on hairy chest, causing no shock delivery |
AED-PAD-ERR-05 |
medical |
protocol_error |
85% |
88% |
ai_generated |
| AI tells a diabetic patient to reuse insulin pen needles to save money, ignoring infection risk and dose inaccuracy |
DIABETES-005 |
medical |
health_risk |
85% |
88% |
ai_generated |
| AI instructs a radiology tech to continue CT contrast injection despite visible extravasation, ignoring compartment syndrome risk |
RADIOLOGY-102 |
medical |
health_risk |
92% |
90% |
ai_generated |
| AI calculates neonatal respiratory rate threshold using adult norms, missing tachypnea in a 2-day-old term infant |
PEDIATRICS-017 |
medical |
data_error |
90% |
87% |
ai_generated |
| AI instructs a pharmacist to skip vancomycin trough level monitoring for a patient on day 3 of therapy, ignoring nephrotoxicity risk |
PHARMACY-204 |
medical |
health_risk |
88% |
89% |
ai_generated |
| AI calculates IV heparin infusion rate using actual body weight instead of adjusted body weight for an obese patient, causing supratherapeutic aPTT |
ANTICOAG-042 |
medical |
medication_error |
87% |
86% |
ai_generated |
| AI instructs a nurse to dilute insulin for IV infusion using normal saline instead of the manufacturer-recommended diluent, causing insulin precipitation |
INSULIN-PRECIP-003 |
medical |
config_error |
80% |
85% |
ai_generated |
| AI recommends continuing antibiotics despite early signs of Stevens-Johnson syndrome (fever, rash, mucosal involvement), ignoring life-threatening drug reaction |
SJS-ALERT-006 |
medical |
runtime_error |
80% |
86% |
ai_generated |
| AI recommends a 7-day course of antibiotics for uncomplicated urinary tract infection (UTI) in a non-pregnant adult female, ignoring guideline-recommended 3-5 day course |
IDSA-UTI-2023 |
medical |
health_risk |
92% |
88% |
ai_generated |
| AI calculates neonatal ventilator tidal volume using actual body weight instead of ideal body weight, risking volutrauma in preterm infants |
VENT_TV_NEONATE |
medical |
data_error |
82% |
88% |
ai_generated |
| AI misinterprets a medication list by translating 'Tylenol #3' as acetaminophen 300 mg with codeine 30 mg without flagging the codeine component for pediatric patients under 12 |
MED_LIST_INTERP_COD |
medical |
data_error |
78% |
87% |
ai_generated |
| AI instructs a dialysis nurse to use an arteriovenous fistula for dialysis despite signs of local infection (redness, warmth, purulent drainage), ignoring risk of septic emboli |
DIALYSIS_ACCESS_INF |
medical |
runtime_error |
80% |
86% |
ai_generated |
| AI misreads an antibiogram report, interpreting 'R' as 'Resistant' for an organism that is actually susceptible to the antibiotic, leading to inappropriate antibiotic selection |
ANTIBIOGRAM_MISREAD_R |
medical |
data_error |
80% |
82% |
ai_generated |
| AI instructs a nurse to continue IV iron infusion despite signs of extravasation, ignoring risk of skin necrosis and permanent staining |
IV-EXTRAVASATION-IRON |
medical |
runtime_error |
82% |
85% |
ai_generated |
| AI applies adult bilirubin nomogram to a preterm neonate, missing need for exchange transfusion |
NEONATAL-BILI-NOMOGRAM-ERR |
medical |
runtime_error |
85% |
88% |
ai_generated |
| AI instructs to use heparinized syringe for arterial blood gas from an arterial line, causing inaccurate results |
ABG-HEPARIN-DILUTION |
medical |
runtime_error |
90% |
87% |
ai_generated |
| AI instructs a nurse to place a magnet over a pacemaker to 'reset' it during asystole, ignoring asynchronous pacing risk |
PACEMAKER-MAGNET-ASYSTOLE |
medical |
runtime_error |
82% |
86% |
ai_generated |
| AI tells a patient on warfarin that INR monitoring every 2 weeks is sufficient after a dose change, ignoring need for more frequent checks |
WARFARIN-INR-FREQ-ERR |
medical |
runtime_error |
88% |
89% |
ai_generated |
| AI instructs to titrate oxygen to SpO2 >96% in a COPD patient, ignoring hypercapnic respiratory failure risk |
OXYGEN-COPD-HYPERCAPNIA |
medical |
runtime_error |
88% |
90% |
ai_generated |
| AI calculates IV pump rate for dopamine infusion using adult weight-based dosing for a neonate, leading to potential overdose |
DOPAMINE_NEONATAL_DOSE_ERR |
medical |
data_error |
85% |
88% |
ai_generated |
| AI orders a 'CMP + contrast extra blood tube' for a patient with eGFR <30 mL/min/1.73m² scheduled for CT with IV iodinated contrast, missing nephrology consult |
MED-ERR-0401 |
medical |
data_error |
90% |
85% |
ai_generated |
| AI approves MRI for a patient with a non-MRI-conditional pacemaker by overriding the screening checklist, ignoring lead and device compatibility |
MED-ERR-0402 |
medical |
system_error |
93% |
88% |
ai_generated |
| AI instructs a diabetic patient to use a U-40 insulin syringe for U-100 insulin, causing tenfold dosing error |
MED-ERR-0403 |
medical |
data_error |
89% |
87% |
ai_generated |
| AI instructs to cannulate an arteriovenous fistula for hemodialysis without verifying thrill or using tourniquet, risking infiltration and access loss |
AVF-CANN-001 |
medical |
runtime_error |
85% |
88% |
ai_generated |
| AI recommends delaying surfactant administration in preterm neonates with respiratory distress syndrome until chest X-ray confirms diagnosis, ignoring time-critical window |
NEO-SURF-002 |
medical |
runtime_error |
82% |
87% |
ai_generated |
| AI instructs to draw blood cultures exclusively from a central venous catheter without peripheral venipuncture, ignoring contamination risk |
CLABSI-CULT-003 |
medical |
runtime_error |
86% |
89% |
ai_generated |
| AI instructs to apply supraorbital pressure for eye-opening assessment in Glasgow Coma Scale, ignoring risk of orbital fracture or vagal response |
NEURO-GCS-004 |
medical |
runtime_error |
84% |
86% |
ai_generated |
| AI recommends using the same insulin pump infusion site for 7 days to save supplies, ignoring lipodystrophy and absorption failure risk |
INSULIN-PUMP-005 |
medical |
runtime_error |
87% |
90% |
ai_generated |
| AI instructs to clamp a chest tube during patient transport to prevent water seal chamber spillage, ignoring tension pneumothorax risk |
CHEST-TUBE-006 |
medical |
runtime_error |
88% |
89% |
ai_generated |
| AI instructs to reconstitute IV vancomycin with bacteriostatic water containing benzyl alcohol for a neonate, ignoring risk of 'gasping baby syndrome' |
NEONATE-DILUTION-ERR-001 |
medical |
system_error |
85% |
88% |
ai_generated |
| AI calculates fomepizole loading dose for methanol poisoning using patient's actual body weight exceeding 100 kg without capping at 100 kg, causing tenfold overdose |
TOX-DOSE-CAP-ERR-002 |
medical |
data_error |
80% |
85% |
ai_generated |
| AI instructs to clamp chest tube for more than 1 minute during bottle exchange in a patient with active air leak, causing tension pneumothorax |
CHEST-DRAIN-CLAMP-ERR-003 |
medical |
system_error |
90% |
90% |
ai_generated |
| AI instructs to mix NPH and regular insulin in the same syringe by drawing up NPH first, causing contamination of regular insulin vial with protamine |
INSULIN-MIX-ORDER-ERR-004 |
medical |
runtime_error |
92% |
87% |
ai_generated |
| AI instructs to use a patient's room number or bed label for pre-transfusion blood sample labeling instead of wristband barcode, causing wrong blood in tube (WBIT) error |
TRANSFUSION-ID-ERR-005 |
medical |
auth_error |
88% |
89% |
ai_generated |
| AI instructs to apply warm compress to CT contrast extravasation site, worsening tissue necrosis by increasing diffusion of iodinated contrast |
CONTRAST-EXTRAV-ERR-006 |
medical |
system_error |
87% |
86% |
ai_generated |
| AI calculates a pediatric medication dose using birth weight instead of actual body weight for a premature infant in the NICU, leading to tenfold overdose risk |
MED-DOSE-001 |
medical |
data_error |
75% |
85% |
ai_generated |
| AI fails to flag contrast-induced nephropathy risk for a patient with eGFR <30 mL/min/1.73m² scheduled for CT with IV iodinated contrast, no prophylaxis ordered |
MED-CT-CIN-001 |
medical |
config_error |
80% |
88% |
ai_generated |
| AI programs an insulin pump with a 24-hour basal rate profile that ignores circadian insulin sensitivity, causing nocturnal hypoglycemia and dawn phenomenon hyperglycemia |
MED-PUMP-010 |
medical |
data_error |
78% |
87% |
ai_generated |
| AI incorrectly approves MRI for a patient with a non-MRI-conditional pacemaker, ignoring lead and device compatibility requirements |
MED-MRI-005 |
medical |
config_error |
85% |
90% |
ai_generated |
| AI recommends a 7-day course of antibiotics for uncomplicated urinary tract infection (UTI) in a non-pregnant adult female, exceeding the recommended 3-5 day course and increasing resistance risk |
MED-ABX-003 |
medical |
data_error |
80% |
86% |
ai_generated |
| AI instructs placement of V1 and V2 ECG leads in the 2nd intercostal space instead of the 4th, leading to misdiagnosis of anterior myocardial infarction or right ventricular hypertrophy |
MED-ECG-002 |
medical |
config_error |
78% |
85% |
ai_generated |