Department of Psychiatry, University of Connecticut Health Center, Farmington, CT, USA.
Biostatistics Center, University of Connecticut Health Center, Farmington, CT, USA.
Arch Clin Neuropsychol. 2023 Feb 18;38(2):247-257. doi: 10.1093/arclin/acac083.
Major depression in older adults increases the statistical likelihood of dementia. It is challenging to translate statistical evidence of cognitive decline at the group level into knowledge of individual cognitive outcomes. The objective of the current study is to investigate 2-year reliable cognitive change in late-life depression (LLD), which will enhance understanding of cognitive changes in LLD and provide a means to assess individual change.
In a sample of non-depressed cognitively normal older adults or NDCN (n = 113), we used linear regression to predict tests of global cognition, processing speed-executive functioning, and memory administered 1 and 2 years later. Stepwise regression was used to select covariates among demographics and raw test scores (either baseline or year 1) and we cross-validated the final models using the predicted residual error sum of squares (PRESS). We then derived a z-change score from the difference between actual and predicted follow-up scores and investigated the proportion of LLD patients (n = 199) and NDCN adults who experienced reliable "decline" (a z-score < -1.645), "stability" (z-scores between + - 1.645), and "improvement" (z scores > +1.645).
A greater proportion LLD compared with NDCN experienced cognitive decline in processing speed/executive functioning and global cognition over 2 years. When compared to NDCN, a greater proportion of LLD also significantly improved on one test of processing speed over 2 years.
Older adults with LLD are at risk of meaningful cognitive decline over a relatively short period, particularly in the domain of executive functioning and processing speed. This study provides a series of reliable change equations for common neuropsychological tests that can be applied clinically.
老年人的重度抑郁症会增加痴呆的统计学可能性。将群体水平的认知衰退的统计证据转化为个体认知结果的知识具有挑战性。本研究的目的是调查晚年抑郁症（LLD）的 2 年可靠认知变化，这将增强对 LLD 认知变化的理解，并提供评估个体变化的方法。
在一组非抑郁认知正常的老年人或 NDCN（n=113）中，我们使用线性回归来预测 1 年后和 2 年后进行的全球认知、加工速度-执行功能和记忆测试。逐步回归用于在人口统计学和原始测试分数（基线或第 1 年）中选择协变量，我们使用预测残差平方和（PRESS）交叉验证最终模型。然后，我们从实际和预测随访分数之间的差值中得出 z 变化分数，并研究了 LLD 患者（n=199）和 NDCN 成人中经历可靠“下降”（z 分数<−1.645）、“稳定”（z 分数在+−1.645 之间）和“改善”（z 分数> +1.645）的比例。
与 NDCN 相比，在 2 年内，更多的 LLD 患者在加工速度/执行功能和整体认知方面经历了认知下降。与 NDCN 相比，在 2 年内，更多的 LLD 患者在一项加工速度测试中显著改善。
患有 LLD 的老年人在相对较短的时间内面临着有意义的认知下降的风险，尤其是在执行功能和加工速度方面。本研究为常见神经心理学测试提供了一系列可靠的变化方程，可用于临床。
