Study design: Randomised controlled trial. Objectives: To determine the effectiveness of a sustainable community-based intervention designed to prevent serious complications and death 2 years after discharge in people with spinal cord injury in Bangladesh. Setting: Bangladesh. Methods: A pragmatic randomised controlled trial was undertaken. People who had sustained a spinal cord injury in the preceding 2 years, were wheelchair-dependent, and were about to be discharged from hospital in Bangladesh were recruited and randomised to an Intervention or Control group using a concealed allocation procedure stratified by level of lesion (tetraplegia/paraplegia). Participants in the Intervention group received 36 phone calls and three home visits over the first 2 years following discharge. All participants received usual post-discharge care. Survival status and date of death were determined by blinded assessors 2 years after randomisation. Results: Between July 2015 and March 2018, 410 participants were randomised (204 to Intervention, 206 to Control). There was no loss to follow up. At 2 years, 15 (7.4%) participants in the Intervention group and 16 (7.8%) participants in the Control group had died (hazard ratio from unadjusted Cox model = 0.93 [95% CI, 0.46 to 1.89]; p from log rank test 0.85). There were no clinically important or statistically significant average causal effects of intervention on the incidence or severity of complications. Conclusion: A program of community-based care for people with recent spinal cord injury in Bangladesh involving frequent phone contact and occasional in-person contact with a health professional after discharge from hospital is no better at preventing death at 2 years than usual care.