Thursday, November 22, 2007

The Power of Survey Design

Unaids has finally read The Power of Survey Design, I guess.
U.N. Agency Denies Inflating Cases of H.I.V. Deliberately;
The lower figure is based on newer, more accurate surveys in India and several African countries. The costly, time-consuming household surveys made it clear that previous estimates, gleaned mainly from tests on women in urban clinics, were too high.

Dr. De Lay said that Unaids’s job was to give advice and monitor trends, and that its budget did not come from the money that has poured into the field recently to buy drugs and hunt for vaccines.

Also, he argued, “cooking this data would be almost impossible” because it is gathered by health ministries in each country and overseen by several agencies.

Behind the U.N.’s HIV Numbers;
India’s gyrating numbers shine a light on the questionable nature of earlier HIV estimates. The 5.7 million figure was based on tests of high-risk people: Primarily female sex workers, certain drug users and people attending neonatal clinics or clinics for sexually transmitted diseases. Despite the lack of a broader, nationwide sample, figures around five million were the basis for a U.S. government projection, in 2002, of 20 million to 25 million Indians with HIV by 2010. They also helped spur a $2.8 billion plan to fight AIDS in India, set to kick off just as the far lower — though still formidable — infection numbers were announced.

India’s population — 1.13 billion, about 380 million more than in sub-Saharan Africa — means this revision had the biggest effect on the numbers. But the difference between prior estimates, based on samples, and newer national estimates was no less dramatic elsewhere. National surveys in Kenya, Burkina Faso and Ethiopia also resulted in estimates that were half prior estimates, or less. Realizing that antenatal clinics are not representative of the broader population, UNAIDS will now be multiplying clinic HIV rates by 0.8 when extrapolating from them.

The basic lesson we should learn from this is we should be careful about shooting from the hip without good data,” Gerhard Heilig, chief of the U.N.’s estimates and projections section, told me.

UNAIDS responds to a letter
(this was before the revisions)
Editor -- In response to the March 11 Open Forum article by James Chin, "Myths and misconceptions of the AIDS pandemic": UNAIDS disagrees with Chin's assertions on AIDS data. We work with WHO and many of the world's most-respected epidemiologists, as well as individual countries to monitor the status of the AIDS epidemic. Estimates are not produced in isolation, but in close collaboration with national epidemiologists and governments.

We stand by our scientific approach. The AIDS epidemic continues to evolve and the response must evolve accordingly. Counter to Chin's claims, UNAIDS has underlined the need to "know your epidemic" and to focus HIV prevention, treatment and care programs to reach those at highest risk of HIV infection.

Articles such as Chin's minimize the devastation the AIDS epidemic has had, and continues to have across the world. U.N. efforts have been critical in moving many countries out of denial and into action. We will continue to use the best available data to advocate for the most effective response for the millions in need.

PURNIMA MANE, director
Policy, Evidence and Partnerships
UNAIDS, Liechtenstein

Countdown Clocks Offer a Lot of Drama But Little Information
WHO’s counting?
The revision of the figures is mainly a result of better data-collection methods, particularly in India (which accounts for half the downward revision) and five African countries (which account for another fifth). In India many more sampling points have been established, and in all countries better survey methods, relying on surveyors knocking on doors rather than asking questions at clinics, have gathered data from more representative samples of the population.

Sceptics will feel vindicated by the revision. There has been a feeling around for a while that the older survey methods were biased, and that the inflation thus produced was tolerated because it helped twang the heart-strings of potential donors. However, the structures for collecting and distributing money to combat AIDS are now well established, and accurate data are crucial if that money is not to be misdirected.

Do Figures Lie? AIDS Prevalence, Cost-Effectiveness and Donor Priorities

No comments: