Manufacturer data feeds an infusion provider can check before sending
A multi-site ambulatory infusion provider had agreed to send recurring data feeds to a biopharma manufacturer. Its records live in WeInfuse, whose reporting runs on an embedded Looker instance with predefined models and no direct SQL access. Greyfield built the four contract feeds in that Looker and a QA routine to run before each upload.
contract feeds: status, dispense, inventory, patient master
status by order and dispense by administration, reconciled
in the QA routine before every upload
At a glance
The challenge
The contract set the required fields, column order, and grain for status, dispense, inventory, and patient master feeds. Reporting ran through predefined Looker models with fixed joins, so picking the wrong model could duplicate or drop rows. Order status lives on the order, while dispense in infusion means an administered dose, so the two feeds could never match row for row. Required fields such as diagnosis, prescriber details, and payer were sometimes blank in the source.
What Greyfield did
Fix the grain of each feed first
Set status to one row per order, dispense to one row per completed administration, inventory to one row per location, NDC, and snapshot date, and patient master to one row per patient. Then chose the Looker model whose base table matched each grain, and added only fields that could not multiply rows.
Rebuild inventory balances
Removed lot from the grain because it split one product into many rows, summed same-day received and removed quantities, and calculated beginning and ending balances in a consistent date order. Helper fields that formulas depend on stay hidden in the report and out of the export.
Reconcile status against dispense
Checked that every Active order appears in the dispense feed with a matching latest dispense date, and that pending and denied orders do not appear there and carry blank ship dates. Separated true status mismatches from repeated-order appointment rows.
QA before every upload
Wrote a pre-upload routine covering uniqueness at the business grain, the date window, blanks in each required field, the exact column set, and cross-feed reconciliation. Source blanks were kept and traced upstream instead of filled with guessed values.
What changed for the client
The provider has four contract feeds built inside its own embedded Looker, each with a written grain, and a QA routine that runs before anything is sent. Feed-to-feed reconciliation and contract acceptance are reported separately, so a status and dispense match no longer hides a blank required field. WeInfuse is a trademark of its owner; Greyfield Data is independent.
contract feeds: status, dispense, inventory, patient master
status by order and dispense by administration, reconciled
in the QA routine before every upload
Have a report or data workflow like this?
Discuss a similar project