Cognitive and affective networks in patients referred for cognitive screening: latent profiles, bridge nodes, and...

Lower cognitive performance and anxiety and depression frequently co-occur in patients referred for cognitive screening, but how the two domains are connected, and whether this differs across patient subgroups, is unclear. Prior network studies used abbreviated cognitive tools, self-report affective measures, and undirected models.

In this cross-sectional study of 548 patients referred for cognitive screening at a Chinese general hospital, 16 network nodes were derived from the seven Montreal Cognitive Assessment domains, the two Hamilton Anxiety Rating Scale factors, and the seven factors of the 24-item Hamilton Depression Rating Scale. Gaussian graphical models with bridge expected influence were estimated for the overall sample and for each profile identified by latent profile analysis. Bayesian directed acyclic graphs with bootstrap resampling explored the direction of cross-domain associations in the overall sample and within each profile. The network comparison test compared profiles.

Cognitive and affective nodes formed two nearly completely separated communities (normalized mutual information = 1.000), connected mainly through language and anxiety/somatization, the two primary bridge nodes. Latent profile analysis identified three profiles: low cognitive performance (n = 114), high affective burden (n = 167), and low affective burden (n = 267). Few cross-domain edges were stable under bootstrap (five in the overall sample; six, three, and none in the profiles), and direction certainty was close to 0.50 for most. The two edges with the highest direction certainty, orientation to retardation and language to anxiety/somatization, ran from cognitive to affective nodes where their direction was determined, before and after adjustment for age and education. The profiles differed in which cross-domain edges appeared, not in the direction of any edge. Exploratory network comparison indicated differences between the high and low affective burden profiles in global strength and structure, with no edge-level difference surviving correction for multiple comparisons.

Cognitive and affective domains formed largely separate communities linked by a narrow bridge, and the sample comprised three distinct profiles. The directional structures were exploratory. The findings support stratified assessment in cognitive screening settings; they do not establish a causal direction or a treatment sequence, which require longitudinal and interventional studies.