Continuous-Time System Identification of a Smoking Cessation Intervention

Abstract

Cigarette smoking is a major global public health issue and the leading cause of preventable death in the United States. Toward a goal of designing better smoking cessation treatments, system identification techniques are applied to intervention data to describe smoking cessation as a process of behavior change. System identification problems that draw from two modeling paradigms in quantitative psychology (statistical mediation and self-regulation) are considered, consisting of a series of continuous-time estimation problems. A continuous-time dynamic modeling approach is employed to describe the response of craving and smoking rates during a quit attempt, as captured in data from a smoking cessation clinical trial. The use of continuous-time models provide benefits of parsimony, ease of interpretation, and the opportunity to work with uneven or missing data.

Keywords: behavioral science; continuous-time identification; self-regulation; smoking cessation; statistical mediation.

Figure 1

Plots of two group average (solid blue, AC; dashed green, PNc) and two single subject (dash-dot magenta, AC; dotted red, PNc) data sets.

Figure 2

(a) Path diagram, used in structural equation modeling (SEM) to describe the relationship between variables, for a classic mediational model. (b) Fluid analogy for mediated behavior change developed from production inventory management models in supply chains.

Figure 3

(a) Generalized fluid analogy for a mediated behavioral intervention developed from production inventory management models in supply chains. (b) Block diagram of statistical mediation.

Figure 4

Craving and Cigsmked data and models for AC and PNc group averages (solid blue, AC data; dashed light green, AC mediation model; dash-dot magenta, AC self-regulation model; dash-dot red, PNc data; dotted brown, PNc mediation model; dashed dark green, PNc self-regulation model).

Figure 5

Craving and Cigsmked data and models for AC and PNc single subject examples (solid blue, AC data; dashed light green, AC mediation model; dash-dot magenta, AC self-regulation model; dash-dot red, PNc data; dotted brown, PNc mediation model; dashed dark green, PNc self-regulation model).

Figure 6

Block diagram depicting a smoking cessation self-regulation model relating Craving and Cigsmked.

Figure 7

Block diagram of a Hybrid Model Predictive Control approach to design of an optimal, adaptive smoking cessation intervention.