ENABLE EMR Export Loader

Antenatal care visits recorded in the EMR need to reach the DHIS2 ANC tracker, where they drive an SMS programme of lifestyle messages and appointment reminders. Until the facilities push their data directly, their exports travel as CSV files. This tool converts each facility's export into the format the ENABLE mediator accepts and sends the records to the mediator over HTTPS. The mediator, which runs alongside the DHIS2 server, validates every record and imports it into the DHIS2 tracker program.

facility export (.xlsx) ──► incoming/ ──loader.py──► mediator ──► DHIS2 ANC tracker

The loader cleans up what can be cleaned mechanically (like normalizing the GA entries to a single format) and rejects what it cannot. Phone numbers are normalised to the international format the SMS programme requires, free-text gestational ages ("22", "22w+4d", "22 wks") become whole weeks, and duplicate visit entries keep their newest version. A record that cannot be imported truthfully, e.g. by not having a usable visit number, no way to date the pregnancy, or no valid phone, is left out and listed in the conversion report, by medical record number and reason, so the facility can fix the source data. Imports are idempotent: re-sending an export updates the records it already created, it never duplicates them, so running the same file twice is always safe.

These files contain patient data. Keep this folder on an encrypted computer, out of shared drives and email, and delete dumps from processed/ once an import is confirmed. The .gitignore is set up so that neither patient data nor the API key can reach git.

Setting up

The tool needs Python 3.9 or newer and keeps its libraries in a private .venv folder inside this directory. Nothing is installed system-wide and no administrator rights are needed.

On a Mac, open Terminal in this folder and check python3 --version. If that fails or shows a version below 3.9, install Python with Homebrew:

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
brew install python

then close and reopen Terminal. Create the library folder and install:

python3 -m venv .venv
.venv/bin/pip install -r requirements.txt

On Windows, install Python from https://www.python.org/downloads/ (tick "Add python.exe to PATH" in the installer), then in PowerShell in this folder:

py -m venv .venv
.venv\Scripts\pip install -r requirements.txt

Finally, on either system, copy config.env.example to config.env and fill in the mediator's URL and the PulseTech API key. Both come from the integration team, and the key should be treated like a password. Check facilities.csv: one line per facility, where match is a word that appears in that facility's file names and facility_code is its DHIS2 code, provided when the facility is onboarded. Adding a facility means adding a line.

match,facility_code,facility_name
akaki,1029107,Akaki Health Center

Converting and sending

Put the facility's .xlsx file(s) in incoming/ and run, from this folder:

.venv/bin/python3 loader.py        # Mac
.venv\Scripts\python loader.py     # Windows

(Calling the .venv Python directly means there is no environment to remember to activate.) For each file the loader prints how many women were converted and pushed and where the reports are. A file that went through completely moves to processed/ with a date stamp; a file with a problem stays in incoming/ so that, once the cause is fixed, the same command picks it up again. Add --dry-run to convert and write reports without sending anything.

The reports

Each run writes a folder out/<facility code>-<timestamp>/:

File	What it holds
...-report.md	The conversion report: what was imported, and every record left out or adjusted, grouped by consequence. Women not imported (with what to fix), visit rows left out, things imported but worth checking. This should be reported back to facilities to help identify possible data issues.
push-report.txt	The mediator's answer per batch. HTTP status, timing, per-record outcome counts, and any rejection quoted with its reason.
response-*.json	The mediator's full answers, kept for reference.

When something goes wrong

What you see	What it means	What to do
SKIPPED: no line in facilities.csv matches	The loader cannot tell which facility the file is from	Add a line to facilities.csv (ask the team for the facility code)
HTTP 403 — the API key was refused	Wrong or missing key	Check MEDIATOR_API_KEY_PULSETECH in config.env with the team
Mediator unreachable	No connection to the server	Check the internet connection and MEDIATOR_URL; if both look right, tell the team
HTTP 502	The mediator is up but DHIS2 is not responding	Run the same command again later. Re-pushing the same data is safe
FAILED to convert	The Excel file does not have the expected sheets	The export is incomplete or from a different system. Go back to the facility
Many records rejected for the same reason	A data-quality problem at the facility	Send the conversion report to the facility; re-send the corrected export later
Certificate error when pushing	The server's certificate is not trusted	Ask the team; only for a test server should MEDIATOR_TLS_VERIFY=false be set

Trying it out safely

test-data/Test_Facility_Data.xlsx is a synthetic export, including eleven dummy patient records (MRNs 900001–900011, fake phone numbers) carrying one of every known data problem. Copy it into incoming/ and run the loader to see the whole flow, rejections included. 8 of the 11 import, and the conversion report shows every failure category with its reason. The file is matched by the test_facility line in facilities.csv; point its code at a test instance's org unit, never at production. The imported test women are easy to find and delete afterwards: their MRNs start with 900.

Support

For any questions or troubleshooting, please contact me at johan@devotta.no