To describe the consequences of medical care interruptions (MCIs) we selected patients with at least two medical encounters between January 2006 and June 2013 in the Dat'AIDS cohort. Patients with any time interval >15 months between two visits were defined as having a MCI, as opposed to uninterrupted follow-up (UFU). Patients’ characteristics at the time of HIV diagnosis and at the censoring date were compared between groups. Cox proportional hazards models were built to assess the role of interruptions on survival (total and AIDS-free). Of 11 116 patients, 824 had at least one MCI. These patients were younger at the time of HIV diagnosis (30 vs. 33 years, P < 0·0001). MCI was less frequent in men having sex with men vs. heterosexual patients [odds ratio (OR) 0·81, 95% confidence interval (CI) 0·69–0·96)], and a centre effect was described. MCI was independently associated with AIDS (OR 2·54, 95% CI 2·10–3·09) and death (OR 2·65, 95% CI 1·94–3·61). At the censoring date, 52·2% of patients with at least one MCI had viral load below detection vs. 85·3% of the UFU group (P < 0·0001). In conclusion, MCIs were associated with patients’ survival and with the proportion of viral loads below detection in our cohort, compromising individual and collective treatment benefits.