Locating Labs, South Africa

Selective Analytics produced findings that really help us to improve our HIV programme. Their skill at finding a new way to optimise such a complex problem is remarkable, and they had the ability to communicate their findings clearly to our executive management team. I would recommend them.
Dr Sergio Carmona, Haematology specialist heading the HIV priority programme, NHLS, South Africa

Optimising >3,000 demand and >200 service points

South Africa is undertaking a big programme of improvement to their HIV and tuberculosis (TB) testing and treatment regime. This required a country-level analysis of the current testing arrangement to improve coverage and travel time from hospital to a laboratory. We worked with the National Health Laboratories Service (NHLS) who provide disease testing to over 80% of the population through a national network of laboratories.

Summary of the work

Selective Analytics and its academic partners were initially asked to optimise HIV CD4 testing. This was a complicated location and capacity optimisation problem with multiple criteria involving 3,266 hospitals and clinics and 223 laboratories across the whole of South Africa and required the development of a new location optimisation algorithm. The outputs included:

  • Used scenario planning to investigate different ways of providing the service.
  • Analysis to find the the best mix of laboratory and clinic-level test instruments.
  • Optimising the location of all testing devices against travel distance criteria.

Used advanced location optimisation algorithms

Key in this project was using rapid optimisation plus scenario planning to allow different arrangements of testing devices to be compared against criteria such as coverage, travel distance and cost.

Client Benefits

The client had been looking for a method of analysing this problem for some time and had not found one. They were delighted with our solution which could optimise across the whole of the country as well as provide detailed suggestions for each laboratory placement.

At the detail level our results matched the local optimisations the the NHLS team had done in certain priority areas. This gave them high confidence that our model was producing answers that reflected the actual situation on the ground.