Researchers have developed a robust method for imputing dementia status and onset time using data from the Cardiovascular Health Study (CHS). Utilizing a linear mixed-effects model, the study estimated individual cognitive trajectories and integrated these estimates into an accelerated failure time model to predict incident dementia. This approach was calibrated on a 60% random sample of participants with known dementia classifications and validated on the remaining 40%, ultimately allowing for the imputation of dementia onset time and status across the entire CHS cohort.

The findings reveal high specificity (98.5%), positive predictive value (81.9%), and overall accuracy (91.3%) in the validation sample when compared to the reference standard dementia classifications. However, the sensitivity was lower at 43.8%, indicating that while the method is effective at confirming dementia status in many cases, it may miss some instances, particularly among certain participant demographics. Notably, the method classified 16.0% of participants without prior dementia classifications as having dementia, showcasing its potential to identify previously unrecognized cases.

This advancement in dementia ascertainment has significant implications for clinical research and therapeutic strategies. By improving the accuracy of dementia status identification in large-scale studies, this method can enhance the understanding of cognitive decline and its risk factors. Moreover, it may streamline the development of targeted interventions and therapies by providing more reliable data on dementia prevalence and onset, ultimately influencing drug development timelines and research paradigms in aging biology.

Source: academic.oup.com