Syndicate content

Add new comment

Using viral load and CD4 data to track the HIV response in South Africa

Nicole Fraser's picture



Sergio Carmona and Tendesayi Kufa-Chakezha are guest blog contributers from South African National Department of Health: National Health Laboratory Services and South African National Department of Health: National Institute of Communicable Diseases, respectively.

South Africa has the largest HIV treatment program in the world with over 3 million people currently on antiretrovirals. Every year, millions of VL and CD4 count tests are carried out to check treatment eligibility for new HIV cases (CD4 count) and treatment success in those on antiretroviral therapy (ART). A VL test monitors viral suppression, the goal of ART given to a HIV-infected person.  The CD4 count checks whether the patient suffers from immune deficiency due to low CD4 counts and tracks recovery of the immune system during ART. In 2014, close to half of all VL tests carried out in lower-middle income countries were done in South Africa. In addition, large numbers of CD4 cell counts have been done routinely to predict patients’ risks for opportunistic infections and provide preventive therapy where indicated. While VL and CD4 testing are essential to monitor individual ART patients, the data is also useful in tracking the impact and performance of the ART program as a whole.
 

In 2015, the VL coverage data coming from South Africa’s clinic-operated electronic HIV patient monitoring system (Tier.Net) showed that 46% of patients on ART had received a VL test over 12 months. However, the National Health Laboratory Services (NHLS) had done many more VL tests and sent the results to the over 4,000 public sector HIV treatment facilities. There, the VL results should have been entered into patient files and Tier.Net. This was never accomplished for hundreds of thousands of VL test results due to various constraints (the NHLS test results cannot be automatically merged into Tier.Net and the lack of a unique identifier in most patient records makes linkage of patient data across systems difficult). The same data system challenge prevented a comprehensive view on CD4 count data and therefore limited what could be learned from patterns of immune recovery of ART patients across South Africa.
 
What was done?
 
The NHLS and the National Institute of Communicable Diseases (NICD) engaged directly with data scientists to provide a methodology to estimate actual VL coverage, VL suppression and CD4 count levels. Several organizations worked with the NHLS and NICD, including the World Bank, the Health Economics and Epidemiology Research Office, Boston University and the National Department of Health. A new data record linkage procedure and algorithm were developed, tested and applied. The multi-stage procedure started off with 44 million VL test results and ended with 12.7 million estimated unique patients. A time-bound patient dataset) was created through the use of Big Data analytical methods.
 
What was found?

  • VL test coverage was much higher than previously believed. Using the new patient-linked cohort, 75% of ART patients had received a VL test in the 12 months from April 2014 to March 2015 (not 46%, as in Tier.Net).
  • An additional ~800,000 VL tests had been carried out at a cost of ~USD 30 million, demonstrating the urgency to better capture VL test results in the electronic patient system.  
  • However, while 78% of ART clients who had had a VL test were virally suppressed, only 58% of all ART patients were known to be suppressed.
  • Young ART patients below 25 years and men had poorer VL results. One in 5 ART patients aged 15–24 years (a group with high sexual activity) had a VL >10,000 copies/µl (indicating infectiousness despite being enrolled in the ART program).
  • There were stark geographic differences in VL suppression levels across the country.  Three provinces had 25% or more patients not virally suppressed. There were 200 clinics with VL suppression below 50%, but one-in-30 clinics had 90% or more of ART clients virally suppressed (demonstrating that the “third 90” is achievable).
  • VL results were also geographically clustered, pointing to “above-facility” factors determining the effectiveness of the ART program.
 
 
Regarding CD4, 1 070 900 individuals were included in this biggest ever national analysis of immune recovery. Recovery was lower among older individuals, males, and those with low CD4 counts at treatment start – much in agreement with previous studies in similar settings. CD4 count recovery differed across provinces with KwaZulu-Natal showing highest and Eastern and Western Cape lowest levels among individuals with CD4 follow up data. Virological suppression during follow up and the CD4 count at ART initiation were the strongest predictors of the likelihood and extent of CD4 count recovery – indicating the utmost importance of gaining viral suppression of patients on HIV treatment.

So what?
 
The analysis provides an excellent baseline prior to South Africa’s fast-tracking of treatment scale-up through focused District Implementation Plans from April 2016 (the Universal Test and Treat policy was launched in September). Understanding viral suppression at population level over time is important for any countries putting strategies in place to reach ambitious HIV treatment targets. Following VL trends over time enables program managers to detect where ART guidelines are not well implemented, and where the treatment program is provided at quality and scale.
 
The matched data has already been used to characterize the HIV care cascades in different demographics, and to assess system-wide retention of ART clients in public sector care. It provides a lesson for other national HIV programs that data science can strengthen the evidence base and bridge system challenges.  

While routine data systems tend to be imperfect, innovative data solutions may sometimes mitigate shortfalls. As South Africa moves forward on the progressive path of “HIV treatment for all” and rolls out differentiated chronic care models, data science and data systems engineering can play a transformative role in making it happen. Already, the NHLS has launched a dashboard system for district health authorities and health facilities to view their achievements in VL suppression, CD4 counts and HIV care cascades.
 
Related 
Analysis of Big Data for Better Targeting of ART Adherence Strategies : Spatial Clustering Analysis of Viral Load Suppression by South African Province, District, Sub-District and Facility

Better monitoring HIV treatment adherence in South Africa : using existing data in new ways to guide improvements

Determinants of CD4 Immune Recovery Among Individuals on Antiretroviral Therapy in South Africa : A National Analysis