Could differential under‐reporting of loneliness between men and women bias the gender‐specific association between loneliness and memory aging? A quantitative probabilistic bias analysis

Abstract

AbstractBackgroundLoneliness is associated with later‐life dementia risk, with conflicting findings on gender as an effect modifier. The causality of this effect modification is difficult to assess, as men may have a lower propensity than women to admit loneliness potentially due to self‐perceived “masculinity”, leading to distorted gender‐specific estimates. We aimed to estimate the potential magnitude of information bias due to differential loneliness under‐reporting by gender in the gender‐specific associations between loneliness duration and memory aging.MethodData were from respondents aged ≥50 in the US Health and Retirement Study from 1996‐2016 (N = 9,032). We conducted quantitative probabilistic bias analyses to simulate various bias‐parameter selections and reclassify loneliness status (yes; no) at the record level as measured biennially from 1996‐2004. We used sensitivity values of loneliness self‐reports of 0.20‐0.45 among men and 0.30‐0.55 among women, and specificity values of 0.90‐1.00, as informed by a validation analysis. We ran mixed‐effects linear regression models on the corrected data with a three‐way interaction between loneliness duration (0, 1, 2, or ≥3 time points), years of follow‐up (0‐12), and gender (men; women) to test gender differences in the association between loneliness duration and memory decline (100 replications).ResultCompared to the uncorrected estimate (‐0.008), mean estimates of the three‐way interaction attenuated in magnitude, ranging from ‐0.006 (sensitivity 0.40‐0.45 for men and 0.50‐0.55 for women) to ‐0.001 (sensitivity 0.20‐0.25 for men and 0.30‐0.35 for women; see Figure 1). Approximately 64%, 84%, 96%, and 100% of simulation estimates remained negative while holding sensitivity distribution at 0.20‐0.25, 0.25‐0.30, 0.30‐0.35, and >0.35 among men, respectively.ConclusionThe likelihood of observing gender‐specific loneliness‐memory relationship reduces as sensitivity values of loneliness self‐reports declines.