Characteristics of the study variables according to gender are presented in Table . Of the 3471 participants, 14.5% reported living alone, with an equal proportion of men and women living alone. The prevalence of antidepressant use in all participants was 6.2% at baseline, and altogether 17.2% used antidepressants at some point during the 7-year follow-up period. Each year, 1% to 2% of participants started to use antidepressants, and the prevalence of incident use during the entire follow-up period was 12.9% (year-by-year figures are provided in Additional file 1
: Annex Table 1). Women had a greater prevalence and incidence of antidepressant use than men. Compared to women, men had less social support at work and a lower occupational grade than women, more often lived in rural areas, smoked more often, and used alcohol more heavily than women.
Characteristics of the study variables
Table presents psychosocial, demographic, and behavioral factors according to living arrangements. Participants living alone experienced a worse job climate, less support in private life, had a higher level of cynical hostility, were more often non-employed, lived in urban areas, were more likely to live in rental accommodation, had worse housing conditions, were more often smokers and used alcohol more heavily than those who did not live alone.
Comparison of psychosocial, socioedmographic and behavioral factors in participants not living alone versus those living alone
When comparing purchases of antidepressants, those living alone had purchased antidepressants more often than those who did not live alone (9.1% versus 5.7%) at the baseline examination. Likewise, during the follow-up period, 25.4% of persons living alone versus 15.8% of others had purchased antidepressants in any year between 2002 and 2008. As shown in Table living alone was associated with a 1.61-fold higher (CI = 1.12-2.25) purchase rate of antidepressants at baseline and 1.81-fold higher (CI = 1.46-2.23) rate during the 7 following years compared with people who did not live alone. When the analyses were restricted to participants who started antidepressant use after the baseline examination (incident use), the association remained similar (OR = 1.71, CI = 1.32-2.21).
Odds ratios (OR) for use of antidepressants in participants living alone compared to participants not living alone
Sociodemographic factors attenuated these associations by 18%, 21%, and 20%. As shown in Table the ORs adjusted for sociodemographic factors were 1.48 (CI = 1.04-2.09) for baseline antidepressant use, 1.64 (CI = 1.32-2.05) for 7-year use, and 1.57 (CI = 1.21-2.05) for incident antidepressant use. Psychosocial factors attenuated the associations by 30%, 12%, and 6%, with the adjusted ORs being 1.43 (CI = 1.03-1.99) for baseline use, 1.71 (CI = 1.38-2.11) for 7-year use, and 1.67 (CI = 1.29-2.16) for incident antidepressant use. Finally, health behaviors had only a marginal effect on the associations between living arrangements and antidepressant use, with adjusted ORs of 1.58 (CI = 1.25-2.21), 1.74 (CI = 1.41-2.14), and 1.64 (CI = 1.26-2.12), corresponding to a contribution of 5% to 9% owing mainly to heavy alcohol use.
There was no statistical evidence to suggest that gender modifies the association between living arrangements and antidepressant use (P-values for gender × living arrangement interactions were 0.618, 0.673, and 0.984). However, the underlying mechanisms could be gender-specific. Thus, we ran the analyses shown for the total cohort in Table stratified by gender to examine whether the contributing factors differed between men and women. Table shows that in men, psychosocial factors had the greatest contributing effect on the association between living arrangements and antidepressant use (attenuations were 31%, 21%, and 15%). In women, sociodemographic factors stood out as the major contributors, showing attenuations of 24%, 27%, and 31%.
Odds ratios (OR) for use of antidepressants in participants living alone compared to participants not living alone, separately for men and women
Finally, we conducted a sensitivity analysis to examine whether the associations were dependent of the length of the follow-up period. We repeated the analysis of living arrangements and antidepressant use for a 2-year instead of a 7-year follow-up period. The association was directionally similar but slightly weaker than in the main analysis (Additional file 2
: Annex Table 2). The contributing factors were similar in magnitude to those in the main analysis, with sociodemographic factors having the greatest contribution and health behaviors the lowest contribution to the greater antidepressant use of participants living alone.