Dedicated tracing studies of patients LTFU are an important instrument to improve quality of care as well as outcome evaluation of ART programs in sub-Saharan Africa. In the present study in Lilongwe, Malawi, only 40% of patients LTFU who gave consent could be traced since contact information was not available for many and field tracing attempts outside Lilongwe were infeasible. However, outcomes were successfully updated in 74% of those traced. Among those successfully traced, 41% of adults and 33% of children had died, most commonly in the first six months following ART initiation. Among patients found to be alive, 92% of adults and 72% of children were still taking ART. The availability of a phone contact improved the chances of successful tracing and identifying outcomes in patients LTFU. As a result of this study, we introduced regular ascertainment of contact information for each patient at the first clinic visit, recommended patients living outside Lilongwe to register at ART clinics closer to their homes and are now able to successfully trace 85% of patients LTFU .
A recent systematic review and meta analysis of studies that traced patients LTFU to ascertain their vital status in sub-Saharan African ART programs, which included Lighthouse, showed that on average 46% of those traced were found to have died . The review also noted the scarcity of information on risk factors for successful tracing as well as death in patients LTFU as a major limitation of existing surveys. We identified short follow-up time on ART as the strongest predictor for high risk of death in successfully traced patients. Patients with no follow-up visit or a last visit in the first 6 months showed 50% to 65% mortality as compared to less than 20% mortality in those seen last after 6 months. In Uganda, deaths of patients LTFU were examined depending on the period from their last clinic visit and rates were highest early after the last visit (mortality of 115, 42 and 21 per 100 person years after 1, 2 to 6, and after 6 months since the last visit, respectively) . These findings suggest that an earlier start of active tracing of patients that missed a scheduled visit date, especially when loss to follow occurred shortly after ART initiation, may help to reduce early mortality.
Studies examining associations between baseline characteristics and tracing success or death show conflicting results making targeted interventions to prevent LTFU more difficult. While we and others did not find a strong relation between age, sex and tracing success , in Zambia, a higher proportion of men were untraceable . Advanced disease stage at baseline measured by WHO stage in our study or pre-ART CD4 count  were also not associated with tracing success. With respect to identifying mortality through tracing, Zambian patients who died were older, more likely to be TB/HIV co-infected, had lower BMI, advanced WHO stages and surprisingly, had higher CD4 count . In contrast, advanced WHO stages were not associated with death among our patients LTFU and low CD4 count was associated with higher mortality of patients LTFU in Uganda . Additionally, an earlier analysis of our dataset showed a lower median CD4 count in untraceable patients compared to traceable patients who died, and we hypothesized that many untraceable patients might have died in line with the Uganda findings .
Our study demonstrated that outcomes, including deaths, of patients LTFU can be found in records of other health facilities. Where death registries are operational, such as in South Africa, conclusive evidence on vital status of patients not found by active tracing can be retrieved from record linkage using unique patient identifiers [21, 22]. Therefore, improved information transfer between facilities may prevent costly and time-consuming tracing.
In our study, during first tracing attempts, phone tracing was more successful in revealing outcomes than in-field tracing. The availability of a phone contact appears an effective way of identifying true outcomes of patients LTFU and was associated with mortality. However, tracing methods may ascertain outcomes differently. On the phone, guardians may not want to admit that their patients have left the clinic, but will reveal that the patient has died. Likewise, survivors may be more likely to have changed their phone number leading to a failed phone tracing attempt, but not their home address thereby increasing the likelihood of being found alive through field tracing. Both scenarios could increase the proportion of deaths among successfully traced patients via phone compared to in-person field tracing, as observed in our study. A study from South Africa found that 46% of patients LTFU could not be contacted because of incomplete or missing contact details and highlights the difficulties of phone tracing in patients without a permanent residential address, such as sharing of one cell phone by the entire family and changing of phone numbers due to theft or expiration of the SIM card . However, cell phone coverage is increasing in Malawi and positive experience using phones in patient follow-up may result in wider use . Therefore, more information on advantages and limitations of different tracing methods is needed.
Our findings highlight the importance of maximizing the potential for early tracing, including soliciting of patient consent to be traced and maintaining current contact information for patients who consent to be traced. Tracing was possible in only 40% of our patients LTFU, but when attempted, the patient status was verified in 74%. Nearly all patients LTFU and found to be alive in our study were still taking ART; more than half of them in our facility. This may be partly related to the early initiation of tracing, after being 2-weeks late for appointment, in a substantial proportion of patients LTFU. Patients may accumulate leftover tablets from previous supplies without telling us, receive tablets from family members on ART or from other clinics, without our information- this needs to be explored further. As a result of the present study, the period for patients becoming eligible for tracing was extended to 3 weeks and the list of patients LTFU to be traced is now updated weekly within our routine services as to avoid unnecessary contacting of returning patients.
Our study has limitations. We traced only 40% of patients LTFU who consented. This sample may not be representative of all patients LTFU, as it includes only patients who had given consent for tracing, had provided contact information in Lilongwe or had a phone when living outside, and who were in care by May 2004 and beyond. With increasing delay between LTFU and tracing, patients or guardians may have been more difficult to contact, as contact information that can change over time was not regularly updated. Guardians may further forget details of patients' ART or may report outcomes they perceive desirable by the phone or field tracers, which could result in under-reporting of the outcome "stopping treatment". In addition, we did not analyze the period from LTFU and tracing to provide further information about the importance of this window period for tracing success and outcomes. Finally, few data on children LTFU in our sample prevented us from a separate analysis and more information is needed to examine the specific features in this age group.