Models of Care: Good Grades and Unanswered Questions

photo by Dominic Chavez
photo by Dominic Chavez

It’s a matter of life and death. Say there are two brothers in their early 30s. Both become infected with HIV at around the same time and both have about the same CD4 level, a measure of how well one’s immune system is working. One brother remains in his village near the Kalahari Desert to help with the family farm and cattle post. The other moves to Botswana’s capital, Gaborone, to work in an office. Both men are treated with the same AIDS drugs under the country’s national treatment program. Yet depending on where they get treatment, one brother has a ten times greater chance of dying than the other. Why?

In Botswana, the HIV prevalence among adults aged 15 to 49 is 25%, the second- highest rate in the world, behind only Swaziland. To address this epidemic, Botswana created a national antiretroviral treatment (ART) program in 2001 to provide free, life-saving drugs to all eligible citizens. After initial delays, the number of patients increased quickly and the program expanded to increasingly rural areas. Because of the fast growth, it was important to monitor and evaluate the results. HAI played a significant role in this process.

“As you’re building a large, new program, it’s hard to have everything run well,” said Dr. Richard Marlink, Executive Director of the Harvard AIDS Initiative (HAI). Marlink was the Principal Investigator of the Models of Care project that evaluated the 2002 to 2010 scale-up of Botswana’s national program. His team asked several critical questions: What are the things associated with better or worse outcomes for patients? Are there different outcomes by geographic region? And lastly, what’s it all costing and will it be sustainable?

Researchers looked at a number of variables, from how fast test results were recorded to how patients fared when nurses rather than doctors prescribed drugs. Now, after several clinical trials and an analysis of a vast quantity of data, Botswana has some answers.

Report Card

“What we showed with the Models of Care project is that the national ART program is doing well,” said Marlink. “Consistently, the program is improving year by year. The mortality rates are going down. Overall, I’d give it a grade of an A or A-.”

Specifically, research showed that as the life of the program increased, the odds of patients’ survival improved. When looking at costs, the team projected that the cost of treatment per patient would reach $430 USD in 2014, compared to $357 in 2011. The total cost of the program is projected to reach $99 million in 2014 when patient enrollment is estimated to be just over 200,000. The primary costs for providing ARTs are the drug themselves, laboratory tests, and personnel.

Researchers were not surprised to discover that overall the national program was doing well. What surprised them was finding a huge difference in mortality rates between different locations. “Some districts have a very low mortality rate and some districts have a very high rate,” said Dr. Mansour Farahani, a research scientist at HAI. “Statistically controlling for everything, you have a wide variation.” Or, as Marlink explained, “to put it bluntly, if you’re in the treatment program, you’re much more likely to die in one location versus another.”

Mortality Graph: Results show there is much greater variation in mortality between districts than between individuals.
Results show there is much greater variation in mortality between districts than between individuals.

Even with a thorough analysis of the data, researchers could not explain the difference in mortality. Some urban hospitals did well, while others lagged far behind. The same was true for rural clinics. “All the worst cases from across the country are referred to Princess Marina Hospital in Gaborone, so you’d expect to see the worst scenario there,” said Farahani. “But they have the best scenario.”

Next Steps

The next challenge is to figure out why the large difference in mortality exists between districts. “Why are things going so well in some places and not so well in other places?” asks Marlink. Why would brothers with similar genetics and a similar disease progression have such different outcomes depending on where they live?

“We think there are contextual factors, not just how the clinic is structured,” said Marlink. “What is the community like? What are the socio-economic factors? What are other disease burdens?”

“It is a very complicated picture,” said Farahani, an expert in improving the effectiveness of health care programs. “We can’t say this is because they don’t have a lab close by, or enough supplies, or enough nurses, or enough training. It’s a combination of things.” He is designing a new study to determine which factors are most important for improving health outcomes.

If HAI researchers can figure out how to improve AIDS treatment on a national scale, their work will have far-reaching consequences, both for the two brothers from Botswana, but also for the millions of men and women currently on ART in Africa. In order to sustain large treatment programs, it’s essential to find the most efficient and cost-effective way to do so.

“If we get more information about the districts and what’s associated with better outcomes, then we can say this model for this type of district seems to work best,” said Marlink. “You have to know what’s actually happening before you can bring the worst up to the level of the best.”

Models of Care was a joint project with the Botswana Ministry of Health and the Botswana-Harvard Partnership and was funded by the African Comprehensive HIV/AIDS Partnerships (ACHAP).

Download a copy of Models of Care report