Background Many patients are reluctant to speak up about breakdowns in care, resulting in missed opportunities to respond to individual patients and improve the system. Effective approaches to encouraging patients to speak up and responding when they do are needed.

Objective To identify factors which influence speaking up, and to examine the impact of apology when problems occur.

Design Randomised experiment using a vignette-based questionnaire describing 3 care breakdowns (slow response to call bell, rude aide, unanswered questions). The role of the person inquiring about concerns (doctor, nurse, patient care specialist), extent of the prompt (invitation to patient to share concerns) and level of apology were varied.

Setting National online survey.

Participants 1188 adults aged ≥35 years were sampled from an online panel representative of the entire US population, created and maintained by GfK, an international survey research organisation; 65.5% response rate.

Main outcomes and measures Affective responses to care breakdowns, intent to speak up, willingness to recommend the hospital.

Results Twice as many participants receiving an in-depth prompt about care breakdowns would (probably/definitely) recommend the hospital compared with those receiving no prompt (18.4% vs 8.8% respectively (p=0.0067)). Almost three times as many participants receiving a full apology would (probably/definitely) recommend the hospital compared with those receiving no apology (34.1% vs 13.6% respectively ((p<0.0001)). Feeling upset was a strong determinant of greater intent to speak up, but a substantial number of upset participants would not ‘definitely’ speak up. A more extensive prompt did not result in greater likelihood of speaking up. The inquirer’s role influenced speaking up for two of the three breakdowns (rudeness and slow response).

Conclusions Asking about possible care breakdowns in detail, and offering a full apology when breakdowns are reported substantially increases patients’ willingness to recommend the hospital.