search

ONEcampaign / topic_health_financing

A repository for the Health Financing topic page
MIT License
0 stars 0 forks source link

Discussion: IHME data vs GHED data #6

Open jm-rivera opened 1 year ago

jm-rivera commented 1 year ago

IHME has finally updated their health financing data up to 2020 and their health spending projections up to 2050. I'm going to have a read through the methodology tomorrow to make sure GHED is still the best dataset to be using. I know in past iterations of the IHME dataset they used GHED as their primary source, so will double check tomorrow.

Jorge--if you have time, would be really helpful to get your advice/opinion (if you have any) based on your experience with both databases

_Originally posted by @nupur-parikh in https://github.com/ONEcampaign/topic_health_financing/issues/1#issuecomment-1404072632_

jm-rivera commented 1 year ago

A comment I had included in #1 is here: https://github.com/ONEcampaign/topic_health_financing/issues/1#issuecomment-1405382743

nupur-parikh commented 1 year ago

From my understanding, IHME has 5 main indicators: image

IHME downloaded the metadata for each data point for all the indicators from the GHED website to help decide how each given data point should be weighted from 1 to 5 being applied as inverse variance weights. They created guidelines for the metadata on how to weight them with priority given to factors such as complete, documented source information and penalized factors such as having been derived or estimated. They then adjusted from current local currency to 2021 USD. They used ST-GPR to model the full time series for each variable across a total of 204 countries. They also supplemented government health spending data using the IMFs database of country fiscal measures in response to the COVID-19 pandemic, in order to capture additional government spending in response to the pandemic. The full methodology can be found here on page 46.

nupur-parikh commented 1 year ago

two points:

  1. The methodology annex says they've only done their analysis using GHED up till 2019 but they have included 2020 with no clear methods for this. I've emailed IHME to see if they use the same weighted methodology for 2020 as they did with 2019 data.
  2. The biggest difference is that IHME does not use the "external" source of financing from GHED. Instead they use their own DAH estimates, which I think may be more a more accurate assessment of aid than GHED's external indicator as they collect and audit budgets/reports from all the major development agencies including the development banks, bilateral aid agencies, and health institutions like the GF, Gavi. If we don't use it for this exercise, I'd venture a guess that we will probably need this particular data at some point this year to talk about MDB optimization as it could apply to health. The other main difference would be that the other IHME indicators don't cleanly add up to one GHED indicator so it's not a 1:1 comparison between IHME and GHED_Current_Health_Expenditure_By_Source but rather sums of a few indicators from what I think is GHED_Current_Health_Expenditure_by_financing_schemes