Spouse-side heterogeneity in widowhood: Preloss cognitive performance and survivors' depressive-symptom change

Widowhood research has documented substantial heterogeneity in survivors' psychological adjustment, but explanations have focused mainly on survivor-side characteristics. Using Health and Retirement Study waves 3-15, we examined whether variation in the spouse who is lost constitutes another dimension of this heterogeneity, focusing on preloss spouse cognitive performance.

We identified 4983 first observed married-to-widowed transitions. Fully adjusted comparisons included 3190 widowhood intervals; the risk-set design comprised 99,335 intervals from 20,617 respondents, and 2967 respondents formed the common within-widowhood sample. OLS interaction models compared spouse-cognition slopes across first-widowhood and married-to-married intervals, followed by functional-role and attribution analyses.

In the fully adjusted risk-set model, the spouse-cognition slope for CES-D8 change was negative during married intervals (b = -0.039) and positive during first-widowhood intervals (b = 0.150), yielding a slope difference of b = 0.188 (p < .001). The restrictive comparison yielded a similar slope difference of b = 0.171 (p < .001). Within widowhood, receipt of spouse help was associated with greater symptom increases, and adjustment for that help modestly reduced the spouse-cognition coefficient; adjustment for care provided to the spouse produced no comparable reduction. These directional patterns were compatible with a resource-removal interpretation, although the role measures came from different samples and time windows. Spouse education and measured marital quality changed the cognition estimate little, whereas demographic adjustment including both spouses' ages attenuated it.

Preloss spouse cognition was differently associated with survivors' depressive-symptom change across first-widowhood and continuing-marriage intervals. This spouse-side contrast broadens accounts of widowhood heterogeneity, while age sensitivity limits attribution to cognition specifically.