Objective To investigate factors associated with subsyndromal depression (SSD) in very old persons, and to develop a model for prediction of SSD among very old persons. Methods A cross‐sectional, population‐based study was undertaken on 85‐year‐old persons in Sweden. Data were collected from a postal questionnaire, assessments in the participants' homes and at reception visits. Depressiveness was screened with GDS‐15 (Geriatric Depression Scale), and the results were classified into three outcome categories: non‐depression (ND), SSD and syndromal depression. Data were analysed with binary logistic, ordinal logistic and linear regression. Results With univariate logistic regression 20 factors associated with SSD were identified in very old persons, and the four hypothesized domains—sociodemographic factors, declining physical functioning, neuropsychiatric factors and existential factors—significantly related to SSD. The multivariate logistic model included seven independent factors that increase the likelihood of SSD instead of ND (lower self‐perceived health, life not meaningful, problems with self‐care, use of tranquilizing medication, no contact with neighbours, history of affective disorder and history of stroke). The ordinal logistic and the linear regression models resulted in seven partly different factors for predicting SSD and depressiveness, in the very old. Conclusions The identified markers may help clinicians with the detection, prevention and treatment of SSD in very old persons. The findings indicate the importance of a comprehensive functional approach to diagnosing and treating depressiveness in this population, and the findings might be interpreted as offering support for the coexistence of a dimensional and a categorical view on depressive disorders. Copyright © 2015 John Wiley & Sons, Ltd.