Patients with pre-surgery cognitive impairment cannot currently be assessed for cognitive recovery after surgery using the Postoperative Quality of Recovery Scale (PostopQRS), as they would mathematically be scored as recovered. We aimed to validate a novel method to score cognitive recovery in patients with low-baseline cognition, using the number of low-score tests rather than their numerical values. Face validity was demonstrated in 86 participants in whom both the Postoperative Quality of Recovery Scale and an 11-item neuropsychological battery were performed. The Postoperative Quality of Recovery Scale agreed with neuropsychological categorisation of low vs. normal cognition 74% of the time, with all but five incorrectly coded participants deviating by only one neurocognitive test. Cognitive recovery over time was comparable for groups with differing baseline cognitive function, irrespective of whether the Postoperative Quality of Recovery Scale or neuropsychological methods were used. Discriminant validation was demonstrated in a post-hoc analysis of the steroids in cardiac surgery substudy by allocating groups to normal (n = 246) or low-baseline cognition (n = 231) stratified by cognitive recovery on day 1. Recovery was similar for participants with low and normal baseline cognition. Postoperative length of stay was longer in patients with failed cognitive recovery whether they had normal mean (SD) (10.4 (10.0) vs. 8.0 (5.9) days, p = 0.02) or low-baseline cognition (12.0 (11.1) vs. 8.2 (4.7) days, p < 0.01). Overall quality, as well as cognitive, emotive and physiological recovery was independent of baseline cognition. The modified scoring method for the Postoperative Quality of Recovery Scale cognitive domain demonstrates acceptable face and discriminant validity.