We investigated the mental health of 47 subjects (30 men, mean age 63.8+/-7.7; 17 women, mean age 68.9+/-8.7) with the 60-item General Health Questionnaire (GHQ). All the subjects lived at home in a Japanese rural community and were examined from 2 to 3 years after suffering a stroke. Among the subjects, 18 (38.3%) had GHQ scores of 17 or more, which indicated a mental health problem (MHP). The following variables were included in multiple logistic regression analysis: age, sex (men/women), grade of motor paralysis (no/slight/moderate/severe), side of motor paralysis (no/left side/right side/both sides: in analysis, we used dummy variables), paresthesia (no/yes), rehabilitation (need no rehabilitation or participate in rehabilitation/fail to participate in rehabilitation), social support (not needed or sufficient/insufficient) and overall physical recovery (1/2/3: 1 = 67-100, 2 = 34-66, and 3 = 0-33 on a visual analog scale 100 mm long, 100 meaning full recovery). In univariate analysis all variables except age and sex showed statistically significant associations with MHP. In multivariate analysis, only one variable, overall physical recovery', had a statistically independent association with the status of MHP (Odds ratio 4.39, 95% confidence interval 1.46-13.19). The results of logistic regression analysis indicate that the presence of an MHP is more strongly dependent upon subjective assessment about overall physical recovery after stroke than upon physical impairments and the other psychosocial variables. Therefore, in the community setting, the visual analog scale of overall physical recovery is considered to be a simple, valid method for assessing MHP following stroke.