id
stringclasses 10
values | domain
stringclasses 5
values | studied_cohort
stringclasses 10
values | claimed_population
stringclasses 10
values | evidence_excerpt
stringclasses 10
values | cohort_profile
stringclasses 10
values | claimed_population_profile
stringclasses 10
values | boundary_violations
stringclasses 10
values | required_population_disambiguation
stringclasses 10
values | risk_if_overgeneralized
stringclasses 10
values | constraints
stringclasses 1
value | gold_checklist
stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|---|---|---|
PBI-001
|
clinical_trial
|
Adults 40-65 with HFrEF, NYHA II-III, eGFR>45, no recent MI
|
All heart failure patients including elderly and HFpEF
|
Enrollment excluded HFpEF and severe CKD; age cap 65
|
age 40-65; HFrEF only; moderate disease; kidney function preserved
|
all ages; HFrEF+HFpEF; broader comorbidity
|
HFpEF overreach; elderly overreach; CKD severe overreach
|
Was HFpEF studied. Any elderly subgroup. Any CKD subgroup. Any real-world registry validation.
|
wrong prescribing; adverse events in frail patients; false guideline strength
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-002
|
clinical_trial
|
Adults 18-55, BMI 20-30, no diabetes, no steroids
|
General adult population including obese and diabetic
|
Exclusions remove key real-world groups
|
healthy adults; non-obese; non-diabetic
|
includes obesity; includes diabetes; includes steroid users
|
obesity and diabetes overreach
|
Any diabetic subgroup. Any BMI>35 subgroup. Any interaction analysis.
|
harm in diabetics; misleading benefit claims
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-003
|
clinical_trial
|
Adults 30-70 with RA on stable methotrexate
|
All autoimmune disease patients
|
Study only RA and background therapy fixed
|
RA only; MTX background; age 30-70
|
multiple autoimmune diseases; varied backgrounds
|
disease class overreach; background therapy overreach
|
Any trials in lupus or IBD. Any biologic background subgroup.
|
ineffective or harmful off-label use
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-004
|
press_release
|
Middle-aged adults; no pediatrics
|
Suitable for children and teens
|
No pediatric enrollment
|
no pediatrics; adult dosing only
|
pediatric dosing and safety implied
|
pediatric overreach
|
Any pediatric PK study. Any juvenile safety data.
|
pediatric adverse events; dosing errors
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-005
|
regulatory
|
Trial enrolled reduced EF only
|
Label claim implies preserved EF too
|
Population mismatch
|
HFrEF only
|
HFpEF implied
|
HFpEF boundary breach
|
Exact label language. Any HFpEF evidence base.
|
mislabeling; insurer coverage distortion
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-006
|
guideline
|
Two RCTs in hospital setting
|
Recommendation for community primary care
|
Setting differs
|
inpatient; monitored; short follow-up
|
outpatient; mixed adherence; long horizon
|
setting boundary breach
|
Any outpatient pragmatic trial. Any adherence sensitivity.
|
harms when monitoring absent
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-007
|
clinical_trial
|
Age 40-65 only
|
Claimed benefit for elderly 80+
|
Age mismatch
|
40-65
|
80+
|
age boundary breach
|
Any frailty stratification. Any real-world data in 80+.
|
falls; renal events; drug interactions
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-008
|
clinical_trial
|
Single country, majority one ancestry
|
Claimed universal across ancestries
|
Demographic mismatch
|
single region; limited ancestry diversity
|
global populations; diverse ancestries
|
ancestry boundary breach
|
Any multi-region trial. Any PK differences by ancestry.
|
variable efficacy; safety signals missed
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-009
|
observational
|
Specialist center cohort only
|
Claimed general population effect
|
Referral bias
|
specialist; high severity; selected
|
general population
|
referral bias boundary breach
|
Any population-based replication.
|
false effect size; misallocation of resources
|
Under 220 words.
|
cohort+claimed+violations+questions
|
PBI-010
|
clinical_trial
|
Excluded pregnant people
|
Claim implies safe in pregnancy
|
Exclusion ignored
|
non-pregnant only
|
pregnancy included
|
pregnancy boundary breach
|
Any pregnancy registry. Any reproductive tox evidence.
|
fetal harm; legal risk
|
Under 220 words.
|
cohort+claimed+violations+questions
|
What this dataset tests
Studied cohort versus claimed population.
It detects generalization creep. It flags inference beyond enrollment criteria.
Required outputs
- cohort_profile
- claimed_population_profile
- boundary_violations
- required_population_disambiguation
- risk_if_overgeneralized
Typical failures
- assuming age or comorbidity transfer
- extending findings to excluded groups
- conflating inpatient evidence with outpatient claims
Suggested prompt wrapper
System
You check population boundary integrity. You do not generalize beyond the cohort.
User
Studied cohort
{studied_cohort}
Claimed population
{claimed_population}
Evidence excerpt
{evidence_excerpt}
Return
- cohort profile
- claimed population profile
- boundary violations
- disambiguation questions
- risk if overgeneralized
Citation
ClarusC64 dataset family
- Downloads last month
- 17