Background: Non-participants to psychosocial studies have been shown to have higher mortality, and mortality differs between partial and complete responders to psychosocial questionnaires. Yet, there is very little information available directly linking survey response status with completing suicide. Methods: The study population consisted of the participants of the Japanese Public Health Center-based prospective study. Ninety-nine thousand four hundred thirty-nine subjects who returned the 10-year follow-up questionnaire and 31 754 individuals who did not return the questionnaire were included in our analyses. The risk of dying by suicide according to response status was estimated by Cox regression models. Results: There were 358 suicides during 1 128 831 person-years of follow-up (mean follow-up time: 8.6 years). Of those who returned the questionnaire, 53.9% were full responders, 42.8% were partial non-responders, and 3.3% were complete non-responders. The risk of suicide was increased for both complete non-responders [hazard ratio (HR) = 1.84, 95% confidence interval (CI), 0.51, 6.64] and partial non-responders (HR = 1.36, 95% CI, 0.999, 1.84) to the questionnaire as a whole. The adjusting variables explained around 40% of the risk for complete non-responders whereas they did not explain the increased risk of suicide for partial non-responders. The risk of dying by suicide was significantly increased for partial non-responders to the subscale on coping (HR = 1.36, 95% CI, 1.01, 1.85) and for complete non-responders to questions on sleep (HR = 2.07, 95% CI, 1.03, 4.16). Conclusions: Partial and complete non-responders have increased suicide risk compared with full responders. More than one non-responder category should therefore be considered when interpreting data pertaining to psychosocial questionnaires in longitudinal studies.