Modifiable dementia risk factors associated with objective and subjective cognition

Abstract

Introduction: Early detection of both objective and subjective cognitive impairment is important. Subjective complaints in healthy individuals can precede objective deficits. However, the differential associations of objective and subjective cognition with modifiable dementia risk factors are unclear.

Methods: We gathered a large cross-sectional sample (N = 3327, age 18 to 84) via a smartphone app and quantified the associations of 13 risk factors with subjective memory problems and three objective measures of executive function (visual working memory, cognitive flexibility, model-based planning).

Results: Depression, socioeconomic status, hearing handicap, loneliness, education, smoking, tinnitus, little exercise, small social network, stroke, diabetes, and hypertension were all associated with impairments in at least one cognitive measure. Subjective memory had the strongest link to most factors; these associations persisted after controlling for depression. Age mostly did not moderate these associations.

Discussion: Subjective cognition was more sensitive to self-report risk factors than objective cognition. Smartphones could facilitate detecting the earliest cognitive impairments.

Highlights: Smartphone assessments of cognition were sensitive to dementia risk factors. Subjective cognition had stronger links to most factors than did objective cognition. These associations were not fully explained by depression. These associations were largely consistent across the lifespan.

Keywords: age interactions; cognitive flexibility; cognitive impairment; cross‐sectional study; depression; executive function; model‐based planning; modifiable risk factors for dementia; smartphone assessment; smartphone data collection; subjective cognitive complaints; subjective memory problems; visual working memory.

FIGURE 1

Screenshots from gamified cognitive assessments used in current study. (A) The game “Memory Match” assesses visual working memory. Participants must recognize a set of previously presented shapes or letters and selected them from a grid. (B) The game “Star Racer” assesses cognitive flexibility. Participants must tap stars in ascending order, alternating between numbers (1 to 13) and letters (a to l). (C) The game “Cannon Blast” assesses model‐based planning. Participants must pick one of two ball containers, aim a cannon, and shoot balls to collect diamonds, but their success depends on their ability to adjust to changing probabilities and picking the right container.

FIGURE 2

Associations of cognitive measures with age and gender. Points correspond to mean raw scores on (A) visual working memory, (B) cognitive flexibility, (C) model‐based planning, and (D) mean proportion of participants with memory problems. The means were calculated for 5‐year bins, split by gender. Error bars represent standard errors (A–C) or standard errors of proportion (D). Only cisgender participants are displayed in this plot (N = 3259); for plots including non‐cisgender participants please see Supplement S7.

FIGURE 3

Associations between cognitive measures and risk factors, controlling for age and gender. Filled circles represent values significant at p < .0038, that is, after applying Bonferroni correction per dependent variable. Higher estimates indicate worse cognitive outcomes. Circles and lines represent (A) standardized beta estimates with 95% confidence intervals (CIs) or (B and C) odds ratios with 95% CIs. In (C), objective cognitive scores were binarized to enable direct comparison of odds ratios with subjective memory problems.

FIGURE 4

Associations between cognitive measures and risk factors, expressed as standardized beta estimates (A) or odds ratios (B) with 95% confidence intervals. Lighter shades represent depression‐adjusted analyses; darker shades represent original analyses. All analyses additionally controlled for gender and age. Filled circles represent values significant at p < .0038, that is, after applying Bonferroni correction per dependent variable. Higher estimates indicate worse cognitive outcomes.