Background: Selective participation in retrospective studies of families recruited after the patient's death may threaten generalizability of reports on end-of-life experiences. Objectives: To assess possible selection bias in retrospective study of dementia at the end of life using family reports. Methods: Two physician teams covering six nursing home facilities in the Netherlands reported on 117 of 119 consecutive decedents within two weeks after death unaware of after-death family participation in the study. They reported on characteristics; treatment and care; overall patient outcomes such as comfort, nursing care, and outcomes; and their own perspectives on the experience. We compared results between decedents with and without family participation. Results: The family response rate was 55%. There were no significant differences based on participation versus nonparticipation in demographics and other nursing home resident characteristics, treatment and care, or overall resident outcome. However, among participating families, physicians perceived higher-quality aspects of nursing care and outcome, better consensus between staff and family on treatment, and a more peaceful death. Participation was less likely with involvement of a new family member in the last month. Conclusions: Families may be more likely to participate in research with more harmonious teamwork in end-of-life caregiving. Where family participation is an enrollment criterion, comparing demographics alone may not capture possible selection bias, especially in more subjective measures. Selection bias toward more positive experiences, which may include the physician's and probably also the family's experiences, should be considered if representativeness is aimed for. Future work should address selection bias in other palliative settings and countries, and with prospective recruitment.