PMC

Two Contact Tables, one Impact Table as presented to the experts for elicitation. A graphical interpretation with pie plot was used to visualize the proportions.

Events in the late-life events of the elderly person modify the social support they receive from different actors (A), which impacts their depressive status (B). In turn, this depressive status changes the social support they receive, creating a loop effect (C).

A population is simulated where each individual elderly person is processed. Elderly depression dynamics are influenced by late-life events, social support and personal depressive status, and this influence is quantified by experts' opinions. Each year some elderly people die and new ones are integrated in the population.

The Scenario1 reflects simply the expert's answers. Simulated patterns of elderly depression prevalence by age are generated based on experts' answers, considering several typologies of experts. Empirical Mean and CI 95% of depression prevalence, retrieved from the Belgian Health Interview Survey, are plotted for comparison.

The Scenario 4 introduces a distinction regarding the socio-economic status. Simulated patterns of elderly depression prevalence by age are generated based on experts' answers, considering several typologies of experts. Empirical Mean and CI 95% of depression prevalence were retrieved from the Belgian Health Interview Survey. They are plotted for comparison together with the regression line on the empirical prevalence.

The Scenario 3 introduces the effect of a treatment by psychotherapy. Simulated patterns of elderly depression prevalence by age are generated based on experts' answers, considering several typologies of experts. Empirical Mean and CI 95% of depression prevalence were retrieved from the Belgian Health Interview Survey. They are plotted for comparison together with the regression line on the empirical prevalence.

1) An artificial population of 1000 individuals is initialized. 2) The Individual process is run for each individual. 3) New individuals are added. 4) Individuals are married. 5) The Age path of singles is updated. 6) Individuals die and their corresponding new widows are created. For one simulation, steps 2–6 are performed 35 times. At the end of the simulation, the prevalences of depressed individuals are computed by age.

1) The Historic Variable is checked to see if the individual is widow for only one year; if this is the case, the new widow processes directly step 5. 2) Decision tree leading to the random generation of a Contact with one of the actors (grayed columns of the tables are not used. 3) An Impact is generated at random. 4) This impact is added to the current Depressive State 5) The Age is incremented.

The Scenario 2 introduces a wedding between husband and wives and defines 3 marital statuses: single, married and widow. Simulated patterns of elderly depression prevalence by age are generated based on experts' answers, considering several typologies of experts. Empirical Mean and CI 95% of depression prevalence were retrieved from the Belgian Health Interview Survey. They are plotted for comparison together with the regression line on the empirical prevalence.