Timely adherence to clinical and pharmacy appointments is well correlated with favourable patient outcomes among HIV-infected individuals on antiretroviral therapy. To date, however, there is little work exploring reasons behind missed visits or evaluating programmatic strategies to recall patients. For this study we implemented community-based follow-up of late patients as part of a large-scale programme for HIV care and treatment in Lusaka, Zambia. Through a network of local home-based care organizations, we attempted home visits to recall patients using locator information provided at time of enrolment. Between May and September 2005, home-based caregivers were dispatched to trace 1,343 patients with missed appointments. Of these, 554 (41%) were untraceable because the provided address was invalid, the patient had moved or no one was at the home. Of the remaining 789, 359 (46%) were reported to have died. Only 430 (54% of those traced, 32% overall) were contacted directly and encouraged to return for care. The likelihood of patient return was higher among traced patients in crude analysis (relative risk [RR] = 2.5; 95%CI = 1.9-3.2) and in multivariable analysis controlling for baseline body mass index, sex and CD4 + count < or = 50/microL (adjusted RR = 2.3; 95%CI = 1.7-3.2). However, the process was inefficient: one late patient returned for every 18 home visits that were made. Reasons for missed visits were provided in 271 of 430 (63%) of the patients who were successfully traced. Common reasons included feeling too sick to come to the clinic, travelling away from home and being too busy. Despite the availability of free ART in Lusaka, patients face significant barriers to attending scheduled clinical visits. Cost-effective and feasible strategies are urgently needed to improve timely patient follow-up.