I am a "global health" statistician; I love data, appreciate the value of estimates.

I have struggled with the lack of data for key questions in many areas of the world.

This article is an incredibly important read for all. 1/ https://twitter.com/thenation/status/1334543432858079236

In many ways, my focus has been the antithesis of IHME. While we all advocate for data-driven decisions, our team has focused on local data, driven by the priorities of the communities we serve and analyzed by individuals who understand the nuances behind the numbers. 2/

Both our team and IHME can produce estimates on abx resistance in rural Rwanda; one is based on data from rural Rwanda. The study was *incredibly* difficult to do and it will be difficlt to publish as single-site studies from Africa don't have the appeal of "global" estimates. 3/

Don't get me wrong, I use the IHME estimates. Probably every grant application cites them as I give the magnitude of the problem. But the pedestal for IHME estimates, and the glory given to these types of models, undermines the investment and valuing of "local" data. 4/

About a 12 yrs ago, we submitted a proposal focused on training statisticians connected to health programs around the world to produce high quality estimates (both getting better data, and analytic technques for messy data). The response "we fund IHME to do this type of work." 5/

My reaction then - how did IHME get a monopoly on global health data and statistics? (Even in approaches they weren't really doing.)

We have demonstrated models (others do too) and could do a lot with <<<$600 million. @gatesfoundation, happy to have a meeting to discuss. 6/

Punch line - stop fetishizing IHME (and similar types of modeling procedures). Yes, their work is valuable, but it is not perfect.

And all of this work could be greatly improved by investing in local data and the statisticians who are the stewards of that work. 7/7

PS1 - I tweet this acknowledging the vulnerability of doing so. But in the spirit of scientific discourse, hopefully differing views leads to stronger approaches overall.

PS2 - I gave a talk a few years ago at the Joint Statistical Meetings on this topic, if anyone is interested to hear more.

"Making Inference in Global Health When There Is Limited (Or No) Data"

https://ww2.amstat.org/meetings/jsm/2019/onlineprogram/AbstractDetails.cfm?abstractid=300485