Of approximately 11 700 patients who had started ART at the study clinic by the end of September 2008 and could thus have been on treatment for at least 3 months at the time of sample selection, 869 patients met all the inclusion criteria for the intervention as of January 8, 2009, as illustrated in . A random sample of 493 (57%) of these patients was selected for follow-up by the tracer. Over 4 months, the tracer was able to reach or learn the status of 260 (53%) patients. The remaining 233 (47%) could not be located using contact information provided to the clinic by the patients. As shown in , patients whose last reported CD4 count exceeded 350 cells/mm3 were more likely to be located by the tracer than those with lower CD4 counts at their last visit.
Status of patients in intervention.
Proportion of sampled patients located by tracer, by duration on ART and last CD4 count
Among those reached, all but one patient were contacted by telephone only; at the patient's request, one contact was made in person, at the patient's home. The tracer made a total of 883 phone calls in the 4-month period, or an average of 1.8 calls per patient in the sample. Each patient was called between 0 and 8 times. One hundred and forty patients did not provide any contact information and could not be called at all (0 calls). Among the 353 who were called at least once, the average number of calls/patient was 2.5. Of these, 93 patients were not found by phone (calls not answered or answered by someone who did not know the patient or did not know the patient's location or status). The status was determined for the remaining 260.
The status of the 260 patients who could be traced is shown in . Approximately one fifth of the sample was reported by household members to have died. Another fifth told the tracer that they were still on treatment at the same facility; of these, roughly half said that they had not missed any appointments, indicating an error in the site's record-keeping. The other half had interrupted their treatment temporarily, justifying the notation of lost to follow-up in their records, but returned to the clinic during the intervention period.
Status of patients traced during intervention
Nearly a third of the sample (30%) reported that they had transferred to another ART facility, either nearby or in a different province or country. Those transferring locally generally cited transport cost as the reason for transferring, though several said that they were now attending a nearby clinic that was open on Saturdays to avoid conflict with work schedules. Most of those who transferred to more distant sites indicated that they had returned to their original homes after living in Johannesburg temporarily.
Finally, the remaining 27% of patients traced had discontinued treatment for a variety of reasons, as shown in . Approximately a quarter of these patients reported that they had left Johannesburg and had not re-started treatment at a new facility, though some indicated that they still intended to do so. The next largest group – 16% of the total – cited traditional medicine or religious beliefs as their reason for discontinuing treatment. Among these, 4 of the 11 patients said that they believed they had been cured of HIV and no longer needed treatment. A number of patients were unwilling to remain in care due to fear of disclosure of their HIV status to their partners or families or for other family-related reasons. Several others expressed fear of losing their jobs if they took time off to attend the clinic. A variety of other reasons were cited by smaller numbers of patients, as shown in .
Patients' stated reasons for discontinuing treatment
Over the course of the 4-month intervention, the tracer was able to persuade 20 of the 97 patients (21%) who had voluntarily discontinued treatment prior to being contacted by the tracer to return to the clinic. Each of these 20 patients completed at least one clinic visit during the intervention period. An additional 27 patients (28%) said that they would like to return to care and allowed a visit to be scheduled for them but did not complete a visit before the end of the intervention. Several others indicated that the tracer's call would motivate them to resume treatment at a new site, but we were not able to confirm that they did so.
Intervention cost and cost per patient returned to care
The main cost of the intervention was the salary of the tracer, who was employed for 4 months. Costs in 2009 USD are shown in , using the average exchange rate over the intervention period (February–May 2009, R9.36/$1.00).
Intervention costs over 4 months
For the 20 patients returned to care by the tracer, the cost per patient for the intervention was $432/patient. This estimate does not take into account other outcomes of the intervention, such as the improvement in the accuracy of the clinic's records through the ascertainment of the true status of patients recorded as lost to follow-up; the possible return to care by some patients as a result of the intervention after the intervention period ended; and the possible return to care at other sites of patients for whom the tracer's contact provided the necessary motivation. The average cost per patient attempted to be traced in the intervention, including those who could not be found through tracing, was $18.
Because the salary of the patient tracer in this evaluation was equivalent to that of a professional (senior level) nurse at the clinic, we also considered the cost of the intervention had the tracer been a junior enrolled nurse instead. In this case, assuming the same results of the intervention, the cost per patient returned to care would have fallen to $262, and the average cost per patient attempted to be traced would have been $11. We cannot know, however, whether a junior nurse, who would likely have been less experienced than our tracer and lacked training in social work, would have achieved the same results.
If patients who had died, transferred to another site, or been mistakenly reported as lost to follow-up due to record-keeping errors could have been removed from the tracer's original sample prior to starting the intervention, the number of patients on the tracer's roster would have fallen by 39%, and the cost per patient returned to care to $265 using the tracer's actual salary or $161 using a junior nurse's salary.