South Africa’s antiretroviral treatment (ART) programme is one of the largest in the world. However, estimates of the number of patients receiving ART need to be improved. Some claim that official figures overestimate the number of patients on ART. But there are other arguments to suggest that the reported totals may underestimate how many are receiving treatment.

The rollout of free ART in South Africa began in 2004 after the government agreed to fund treatment through the country’s public health service. Delivering ART requires organising drug procurement and distribution, formalising treatment guidelines, and deciding on eligibility criteria. Clinicians have been trained and key indicators for monitoring the programme have also been defined. But standardised clinical records and registers have not been issued for recording and aggregating precise figures for enrolment and retention.

Most healthcare facilities use diaries to track patients who start ART. Clinic ‘events’ are recorded in tally sheets or log books. These events include deaths, patient transfers and patients lost to care. The number of patients remaining in care at each facility is calculated for each month. This is done by subtracting the number of monthly events from the total number of patients who began ART at a facility.

If clinics do not determine the precise level of programme attrition, the treatment totals they report may be inaccurate. Migration between clinics is common: patients, for example, may move elsewhere or they may decide to go to another, new clinic closer to where they live. When patients refer themselves to other clinics, such changes often remain unregistered. This means that in some instances patients may appear in the treatment totals of two different clinics and therefore be counted twice.

Patients may decide to stop taking ART. Others may die. If such events are not recorded, patients will remain registered as ‘receiving ART’. Although the Department of Health states that its treatment totals only include the number of patients currently on ART, many clinics still appear to be reporting the total number of patients they have ever enrolled (i.e. their cumulative treatment totals), instead of making adjustments for programme attrition.

To calculate the number of ‘active’ patients accurately, clinics usually deduct the number of patients who have not been seen in the last three months from the cumulative number of patients seen. The length of this cut-off period is arbitrary but may also bias the reported ART figures. If, for example, a substantial proportion of patients seen in the last three months are no longer on ART, the reported totals will be overstated. But if a substantial proportion of the patients who have not been seen in the last three months are actually still on ART, the figures will be understated. Distortions may also be made worse by clerical errors or delays in the updating of patient registers. Choosing a shorter cut-off period would reduce the risk of over-statement but it would increase the risk of understatement.

An underestimation of patient numbers could also occur if clinics do not report their totals to provincial health departments on time. When this happens, previous totals may be carried forward to the next reporting cycle. Estimates of how many people are receiving ART may be similarly distorted at the provincial level: late submissions may mean that new national ART estimates are based on outdated figures.

Treatment totals reported by the Department of Health usually relate only to public sector ART programmes. Patients who receive ART through the private sector (either through medical schemes or workforce treatment programmes) are not included. Although most NGO ART programmes have been incorporated into the public sector, certain NGOs are still excluded. These exclusions from official Department of Health statistics mean that the official number of ART patients in South Africa may be an underestimate.

Hope for improvement

South Africa’s ART statistics have not always been reliable. But there is reason to hope that the quality of the data is improving. The introduction of standardised clinical records, for instance, and a new three-level M&E system (paper-based data gathering, electronic registers and networked software: see page 3) are helping to improve the accuracy of reporting. An intensive focus on monitoring and evaluation training, as well as new support and accountability measures will help us to improve the quality of the data available.

TIER.Net, a new electronic data collection system, is being implemented at over 2,200 facilities nationally. This software uses algorithms to accurately calculate the number of patients lost to care. The TIER.Net system makes it easier to access patients’ demographic profile details and their clinical characteristics (such as baseline CD4 count distributions and viral load suppression). Completion and retention rates are also monitored. Clinics also use reports to identify which patients have missed appointments. Contact by cell phone or home visits may also be helping to improve retention rates.

If reporting and accountability measures improve, the data we have will be more accurate. This will help to improve resource allocation policies and improve the impact of ART interventions.

Leigh Johnson and Meg Osler work at the Centre for Infectious Disease Epidemiology and Research, University of Cape Town.