Earlier hippocampal loss gave an uncertain 0.8 percent error reduction in 87 DLBS test adults with known follow-up time

We tested whether earlier hippocampal loss adds prediction of later delayed word recall in selected DLBS adults. The comparison used the recorded follow-up time. A fixed split used 86 adults for training and 87 for testing. The baseline used two earlier recall scores, age, sex, education, brain volume, head size, and visit intervals. Adding earlier loss reduced test mean squared error from 3.276 to 3.249 words squared, or 0.83 percent. In 5,000 bootstrap repeats, the 95 percent interval for the reduction was -15.02 to 9.10 percent. It included higher and lower error. Mean reduction across repeated cross-validation was -0.24 percent. The evidence does not establish a reliable added benefit under this model. We used a recovered release 1.2.0 memory file after an audit found that the release 1.3.0 file duplicates working memory. New downloads reproduced all numerical results. The package preserves source data, a fixed plan, selection records, code, predictions, uncertainty estimates, and reproduction instructions.

hippocampus · delayed recall · episodic memory · longitudinal study · MRI · prediction · brain aging · uncertainty