The intervention targets of family functioning and cognitive insight in community-dwelling individuals with...

This study aimed to explore core and bridge elements in the network structure linking family functioning and cognitive insight among community-dwelling individuals with schizophrenia. The NodeIdentifyR (NIRA) algorithm was applied to simulate node-level perturbations within this cross-sectional network. Findings were interpreted as hypothesis-generating evidence for potential targets rather than as proof of preventive intervention effects.

Individuals with schizophrenia receiving community follow-up care in Pengzhou City, China, were recruited using convenience sampling. Data were collected through an online questionnaire platform and included demographic and clinical characteristics, the Family Intimacy and Adaptability Scale, and the Beck Cognitive Insight Scale. Because all scale items were ordinal Likert-type responses, ordinal item handling and node coding were specified before network estimation. Gaussian graphical models were estimated using polychoric correlations, while Ising/NIRA simulations were conducted on dichotomized, consistently risk-coded nodes. Sensitivity analyses were performed to examine the robustness of the main network findings after adjustment for available demographic and clinical covariates.

A total of 785 participants, including 414 males and 371 females, were analyzed. In the Gaussian graphical model, F13, representing lower family democratic decision-making, and F26, representing a lower atmosphere of free expression, showed the highest expected influence and were identified as core elements in the risk-coded network. B4, representing cognitive impulsivity, had the highest bridge expected influence, indicating its key role in linking family functioning and cognitive insight. Ising-model simulations showed that mitigating activation of F6, representing reduced intergenerational equal expression, produced the greatest reduction in overall network activation, suggesting this node as a promising hypothesis-generating prevention target. In contrast, exacerbation simulations showed no significant differences in cognitive insight scores across intervention targets after multiple-comparison correction. The covariate-adjusted sensitivity network yielded the same leading expected-influence and bridge-expected-influence nodes.

Network centrality indices, including core and bridge elements, should be interpreted cautiously when identifying intervention targets, because they do not necessarily correspond to the most effective simulated intervention nodes. Reduced intergenerational equal expression within families may represent an important hypothesis-generating focus for future early intervention and prevention research aimed at improving cognitive insight among community-dwelling individuals with schizophrenia.