Medical history-taking is among the most powerful diagnostic tools for healthcare professionals. However, its accuracy and reliability are underexplored areas. The present post-hoc study compares medical histories from 463 people in a rural part of Côte d'Ivoire. The medical histories of the same individuals were taken by physicians and experienced field enumerators who were blinded to the results of the others. Kappa (κ) statistics for 14 symptoms revealed only poor-to-moderate agreement between physicians and field enumerators (κ = 0.01-0.54). Participants reported consistently more symptoms to field enumerators than physicians. Only 33 (7.1%) participants gave no discordant statement at all. The average number of discordant statements per participant was 3.7. Poisson regression revealed no significant association between the number of discordant statements and participants' age, sex, educational attainment, occupation, or socioeconomic status. Operational research should further explore best practices to obtain reliable medical histories in resource-constrained settings.