[Use Case]: Support the INSPIRE project's effort to link population health data to geospatial data

[kzollove]

Background

Eight countries, 20 HDSS sites across Anglophone and Francophone Africa

• HDSS sites go to household, collect residence data, demographics, death data, GPS coordinates
• In some cases, have access to data collected at health facilities in the HDSS site catchment area

Iganga HDSS site in Uganda

• Climate data from various weather stations across Africa (Hourly rain and temperature from satellite data)
• Related to Malaria data collected from health facilities

Jim Todd's question

• How are incidence of known risk factors for NCDs (diabetes, obesity) affected by drought or temperature in an area over time?
• Does a drought occurring in this year affect the occurrence of NCDs in following years?

TODO

• What does this kind of analysis involve?
• How can we use the OHDSI toolchain to foster the completion of this analysis?

Depends on:

Datasets

• Scope out SDG indicator database ingestion workload gaiaDB#5
• Data Request: DIVA-GIS (country- and global-level geometry data) gaiaDB#31
• Climate/Drought/Temperature

[kzollove]

INSPIRE met with OHDSI GIS earlier this week. INSPIRE shared more about their HDSS.

Talked about demographic specs, "verbal autopsy" spec. A question is how can we tie environmental data to these person-centric specs... particular interest in malaria cases.

• What kind of geocoding can we do with the available location information from this project

Contemplating a larger model that would subsume some of these specs.. is there a single spec that subsumes what is done in longitudinal research? STAR schema similar to I2B2? How would this be mapped into the OMOP model (visits, "waves", instruments) This will be explored in a workshop later this week. Hopefully settled on a schema by end of the week.

There is an existing method from moving I2B2 and moving it to OMOP... this might be leveraged or adapted to do some of that transformation work for this project.

[kzollove]

RE Geocoding/Administrative areas/boundaries in countries in Africa:

From Duncan Penfold Brown (Aquaya)

I'm coming up to speed with this as well given some of my day-to-day. The short answer from me is that I don't have any/much experience with the address systems of most of these countries, except in passing (ie: I've seen them, but don't know quite how the systems work).

The data I'm more familiar with is the administrative area and community designations eg: Region > District or County > Sub-county; City / town / village / hamlet, etc etc. Most of the data I'm using here, because of its authoritative weight, is from the UN. OCHA maintains the common operational datasets (CODs) for population statistics and administrative boundaries that are pretty useful in that right, but sounds like this isn't what you're after.

Miles and our survey-oriented guys might have better ideas! Otherwise, I'd be at the whims of looking for a reverse geo-coding API and seeing how well it covers Africa, which is always a concern.

---

From Duncan Penfold Brown (Aquaya)

Also, still probably a little left-field if talking about geocoding in a more programmatic/accessible sense, Ghana at least has a well-documented national survey of households. You can scope out what the survey uses to ID households, though I'm not certain they record specific household location.

https://microdata.statsghana.gov.gh/index.php/catalog/110/data-dictionary/F9?file_name=defactopopn_10%_20220828d

---

From Miles Schelling (Aquaya)
I can speak to the use of addresses in Ghana specifically. In terms of how we conduct surveys, we do not include household addresses because most people do not know their address. In Ghana everyone has a "GPS digital address" (e.g., my address when I was living in Accra was: MR5M+9QM, Unnamed Road, Accra, Ghana). These theoretically give a unique identifier to each 5m by 5m plot in the country (see this article). This is an interesting opportunity where address and GPS location are tied. However, in practice this address is rarely used and certainly not included in our survey questions/methodology.

[kzollove]

Designed a Snowflake schema for database for longitudinal studies. Finished design last week and started loading a number of studies into the schema,

Will return to Agnes and Maureen to propose taking data through staging database

• 1 dimension for "resident episode" (analogous to location_histor)
• sub dimension for place-based data
• Would be pushing into OHDSI some of the place-based exposures

Resident episode might have better person location information than site-wide shapefiles.

[jaygee-on-github]

Because we are working with sentry surveillance sites in a federated way, in most cases in our location table, we have longitudes and latitudes. However, that is not always the case. So we would like to do some geocoding in our use when they are missing used on the village in rural areas. We can provide the villages we want geocoded across several countries.

@kzollove, after consultation with Agnes, we want to include SDoH in our use case. More specific we want to include some UN SDG indicators that are to be determined. These indicators are availalble at the country and often at the district within country level across the continent. Using a technique called small area estimation which is actually a decision tree that includes many practices, we want to derive an SDG indicators dataset with place-based locations that re smaller than country districts. In effect we will be creating a "distribution" from a source dataset. To execute thus transformation we will want to follow the decision tree.

After talking with with Agnes we will be assigning a mathematician with GIS background to derive the small area estimates.

I have an interest in the metadata we will use to describe a dataset that is derived from another dataset.

[jaygee-on-github]

We have onboarded our statistician / mathematician who will be creating a catalog entry for social determinants of mental health (SDoMH) in as many as three East African countries.

As part of the onboarding process we have created a concept note for her to follow:

SDoMHs in Three East African Countries v3.pptx

[kzollove]

	○ Next steps for incorporating SDG indicators for mental health at country and district level to make this work at community level using small area estimation.
	○ Add questionnaire data (with SDG indicators and other common variables that go back to the district and country SDGs) to protocol
	○ Will create some synthetic questionnaire data for some ongoing projects to test this
	○ SDGs are by UN and WHO
	○ Jay will give a timeline on this dataset in the future… do we want to capture the transformation of the district or country dataset into this community level dataset, in order to be FAIR… how do we do that capture 
	○ Andrew: suggests RO-Crate for capturing transformation metadata