[META]: Eheza Reports

[anvmn]

Requirements from https://docs.google.com/document/d/1-6ToGLuCbmeI4wKH5IwI6KL-tkjRlimEZYk3HUD_A48/edit:

We need a site that is capable of displaying high-level statistics from the e-heza usage that can also be filtered by date range and demographic areas (geofields hierarchy / health center).

The site will be used by administrators and implementers and will not be used by healthcare workers on the ground. As such, the application will not need to work without web connection and can use a normal web interface with a browser.

We envision several groupings of data:

• demographic data (information on the number of registrants and encounters by type - see demographic report
• ANC Data (information on the number of visuals per patient and the types of diagnoses) - see ANC report and pregnancy report
• Acute Illness data - information about AI diagnoses - see AI report
• Nutrition data - see nutrition report.

Requirements:

• Data needs to be on demand and filter able - can wait for scrips to run.
• Data doesn’t need to be “up to the minute” - monthly would be enough.
• Should be password protected.
• No levels of permission - all users can see all data.
  • Ideally the data would be consumable in a standardised format so that we could provide other interfaces to some eta of the data for other purposes.

[anvmn]

@adamhstewart

Analysed the requirements, and as I see this, there's no need to create dedicated portal. We can have this on our back-end.

that can also be filtered by date range and demographic areas (geofields hierarchy / health center).

This rules of the usage of existing scripts, if we want to be able to provide quick replies. There are numerous variations of inputs, and we can't prepare a data set (using scripts) that can support this. Also, Pantheon times out when response to request takes more than 60 seconds to generate.

A solution I suggest is similar to what we have for aggregated NCDA:

• We'll have a field on Person CT that will store reports data for person.This data will be updated when patient content is updated, using AQ.
• We'll create ELM Application, similar to what we have for aggregated NCDA, where we'll allow HC / Geofields hierarchy selection. After selection is made, the request will be sent to backend, which will pull reports data of all patients that match the selection. This (raw) data will be decoded by client and logic will be applied to create requested output.
• Since data sets can be very large (currently there are 180K patients registered), we won't be able to generate raw data on the fly (due to timeout from Pantheon). To overcome this, we'll:
  • Generate following large data sets offline (daily):
  • Global (1)
  • Province (3)
  • Province + it's Districts (5)
  • Health Centers (49)
  • These Data sets will be stored in txt file at dedicated content type. Every day, we will update the file for each of (currently) 58 large data sets.
  • When request for large DS arrives, we'll find matching node and respond with contents of it's txt file.
• For smaller data sets, we can generate aggregated data on fly.

Ideally the data would be consumable in a standardised format so that we could provide other interfaces to some eta of the data for other purposes.

This implies that back-end will need to generate final results, which is not supported by suggested solution. We had same issue with NCDA, and decided that we'll deal with it if actual requirements will come from customer. I suggest we take similar approach here (and if possible, reject requests for API).

[anvmn]

@adamhstewart

POC demo:

[anvmn]

@adamhstewart

Went through Demographics report script, and found that the scope of Registered Patients part is:

• Acute Illness
• Prenatal
• Nutrition - group and individual So, the registered and impacted patients we report on, have at least one of these measurements.

The scope of Encounters part also includes:

• Home Visit
• SPV

Nutrition report scope is:

• Nutrition - group and individual
• SPV

To determine impacted, it uses the table created for Demographics report, so we have a potential for impacted calculation to be slightly off.

Should we adjust the scope of Demographics patients report? Add SPV? Home encounter? We also have additional types of encounter now (Child scorecard, TB, HIV)?

[anvmn]

@adamhstewart

Isn't impacted defined by patient that had 2 or more encounters? Currently, the label says 'more than 2 visits':

[anvmn]

@adamhstewart

Questions for demographics report:

• When counting Acute Illness encounters, should the postpartum encounters be included?
• Which additional encounters types should be added into scope (currently we have)? Also, should it be same scope for registered patients and Registered patients tables, and Encounters table?

Questions for nutrition report:

• Encounters in scope are Nutrition (groups + individual) and SPV, correct?
• Impacted should be determined using nutrition scope and not demographics scope, correct?

[adamhstewart]

1. Postpartum encounters should be included under ANC, not acute illness
2. All encounter should be counted. I pushed a new PR called fix – reports that is more inclusive, but we should add to that also HIV and TB management.
3. Nutrition report should include all encounters in which nutrition measurements are taken, nutrition, and SPV.
4. Impacted scope for nutrition should only include nutrition encounters to determine if the patient is impacted.