Table 1 Development of the model
From: Health behavior change in advance care planning: an agent-based model
Conceptualization | Logic |
|---|---|
• Based on statistics for a population ages 65+ • Baseline ACP behavior distribution (from literature) o % pre-contemplation o % contemplation o % preparation o % action-maintenance • Cut-points (on 0–100 scale) determine each of TTM stages o Each stage consists of a different (not equally-distributed) point range o Based on different difficulties to move up in TTM stage • Agents move each day | Distributed to fit percentages (0–100) based on TTM Sliders for each of 5 stages to determine starting distribution ACP propensity based on a changing number of points (0–100 scale) per individual; varying cut points to designate Threshold rules for moving up stages Turtle changes color at action stage Each tick equals 1 day Move for at least 5 years |
Dynamic Modeling of Experiences | Logic |
• Personal critical illness o Smaller patch (less likely) o Higher impact factor (one’s own severe illness likely has a greater impact on Death Planning Anxiety) • Loved one’s critical illness/death o Larger patch (more likely to know someone who has had severe illness) o Smaller impact factor (the experiences of others likely have a lesser impact on Death Planning Anxiety) • Advance care planning discussion with primary care provider o Relative small influence, based on non-urgency of the primary care setting | 1 patch for each event (personal illness, loved one’s illness, and primary care interaction) Sliders to indicate degree of impact for each Probability of affecting ACP change when land on patches can vary (sliders 0–100 indicate likelihood) • If gain points, then count points • If count > next TTM threshold, then move to higher stage • If count < next TTM threshold, then stay in current stage If move up stage, then reevaluate current stage • If in Action-Maintenance stage, then turn designated color • If not in Action-Maintenance stage, then retain color |
Dynamic Modeling of Social Interactions | Logic |
• Interactions with other individuals • Recognize level of ACP • Susceptibility (not all agents are impacted by other agents) • At each tick, evaluate any agents on same patch • At each tick, if patch-mate in higher stage, then gain interaction points o If neighbors, then evaluate for higher stage than self o If neighbor at high stage, then probability of assign associated number of points o Susceptibility: slider-based probability at agent level o Each stage associated with a number of points gained by lower stages upon interaction • Local Networks o Observable connections between agents that interact o Agents move at a constant rate, from patch to patch in random directions (in contrast to randomly across entire matrix) • Backsliding (negative social interaction) o Negative social influence can accumulate o With a sufficient accumulation of negative points, agents can cross the threshold back into the previous stage | If interact with neighbor, increase ACP propensity for lesser neighbor Different degrees of disparity will have a different levels of influence If gain points, then count points • If count > next TTM threshold, then move to higher stage • If count < next TTM threshold, then stay in current stage If on same patch, then make connection with agent At each tick, move at random 360° and move forward at designated moving-rate If move up stage, then reevaluate current stage • If in Action-Maintenance stage, then turn designated color • If not in Action-Maintenance stage, then retain color • If interact with neighbor, decrease ACP propensity for higher neighbor |
Susceptibility | Logic |
• Not all agents are impacted by experiences and social interactions | • If land on patch, then probability of gaining points |