Aims Medication-related harm (MRH) is common in older adults following hospital discharge. In resource-limited health systems, interventions to reduce this risk can be targeted at high-risk patients. This study aims to determine whether (1) doctors can predict which older patients will experience MRH requiring healthcare following hospital discharge, (2) clinical experience and confidence in prediction influence the accuracy of the prediction. Methods This was a multicentre observational prospective study involving five teaching hospitals in England between September 2013 and November 2015. Doctors discharging patients (aged =65 years) from medical wards predicted the likelihood of their patient experiencing MRH requiring healthcare (hospital readmission or community healthcare) in the initial 8-week period post-discharge. Patients were followed up by senior pharmacists to determine MRH occurrence. Results Data of 1066 patients (83%) with completed predictions and follow-up, out of 1280 recruited patients, were analysed. Patients had a median age of 82 years (65–103 years), and 58% were female. Most predictions (85%) were made by junior doctors with less than 5 years' clinical experience. There was no relationship between doctors' predictions and patient MRH (OR 1.10, 95% CI 0.82–1.46, P = 0.53), irrespective of years of clinical experience. Doctors' predictions were more likely to be accurate when they reported higher confidence in their prediction, especially in predicting MRH-associated hospital readmissions (OR 1.58, 95% CI 1.42–1.76, P < 0.001). Conclusions Clinical judgement of doctors is not a reliable tool to predict MRH in older adults post-discharge.