Network Control Theory in Personalized Intervention Selection: A Proof-of-Concept Approach to Simulating Therapeutic Intervention Effects in Clinical Time-Series Data

AI Summary

Proof-of-concept application of network control theory to EMA clinical time-series using linear dynamic system modelling and ridge regression to simulate therapeutic interventions.
Two metrics, average controllability and cumulative impulse response, index theoretical intervention reach across the system and magnitude and direction of variable change.
Approach revealed variable and intervention-specific controllability, sometimes differing from delivered treatments; method rests on idealised assumptions and requires empirical validation.

Psychother Res. 2026 May 15:1-17. doi: 10.1080/10503307.2026.2666624. Online ahead of print.

ABSTRACT

Network Control Theory (NCT) is concerned with the control of dynamic complex systems, allowing the modeling and assessing of (theoretical) intervention effects.

This study presents a proof-of-concept approach for applying NCT in a process-oriented manner to ecological momentary assessment (EMA) data from patients.

We applied a linear dynamic system model (LDS) using ridge regression to pretreatment time-series data of 20 patients with depression or anxiety (Δdata points/person = 96). Effects of 31 therapeutic interventions were simulated based on nomothetic knowledge and expert group consensus. Two NCT metrics were applied: average controllability (AC), indexing an intervention’s theoretical effect over the system, and cumulative impulse response (CIR), indexing theoretical magnitude and direction of change per variable.

Certain nodes exhibited elevated controllability. In exemplar cases, NCT-derived recommendations seemed to differ from the interventions actually received. Theoretical controllability varied across interventions, with some showing consistently higher controllability across patients.

This proof-of-concept approach demonstrates how NCT can, in principle, be applied via process-oriented intervention simulations. However, its current application rests on idealized assumptions. Whether identified interventions are clinically meaningful needs empirical testing. We highlight opportunities for NCT research and discuss critical next steps toward empirically refining estimations of intervention effects.

PMID:42137962 | DOI:10.1080/10503307.2026.2666624

Document this CPD