HEALTHCARE OPERATIONS
Clinical Productivity
Registered dietitians were logging daily work in shared Excel files. I designed the Elevra workflow that replaced them.
See it
Managers needed the missing day, not another percentage.
Recreated from the shipped workflow, not screenshotted. Switch surfaces, then expand either one full screen.
Elevra Clinical Productivity
Elevra
Sacred Heart Medical
Clinical Productivity Team Submissions
14 team members, August 2026
| RD Name | Reminder | ||||
|---|---|---|---|---|---|
| 2+ days missing6 | |||||
| Marisol Vega | Tue, Wed missing | ||||
| Vanessa Rhodes | 3 days missing | ||||
| Blessing Nwosu | Tue, Thu missing | ||||
| Yuki Tanaka | Wed, Thu missing | ||||
| Mateo Flores | 4 days missing | ||||
| Ian McAllister | Mon, Tue missing | ||||
| 1 day missing5 | |||||
| Priya Kapoor | Mon missing | ||||
| Rachel Dempsey | Tue missing | ||||
| Katie Holbrook | Thu missing | ||||
| Thiago Carvalho | Mon missing | ||||
| Nicole Hartsell | Wed missing | ||||
| Complete week3 | |||||
| Travis Reeves | Complete | ||||
| Megan Pruitt | Complete | ||||
| Finn O'Malley | Complete | ||||
The problem
The workbook recorded a value. It could not tell the manager who owned the day.
Daily productivity lived in shared Excel files, one per cost center and month, edited in circulating copies.
Shared workbookTwo writers, one cell, one file per cost center
2:02Y. Tanakaoverwritten4.5 hrs
2:41R. Dempseythe file keeps this one2.5 hrs
Owned recordTwo writers, one owner, every write attributed
Yuki TanakaRD, Inpatient adult
- Day
- Wed, 13 August 2026
- Hours
- 2.5
- Care setting
- Acute medical, 4 East
Revisions
- 2:41 PMSet to 2.5 by R. DempseyFlagged: not this record’s owner.
- 2:02 PMCreated at
4.5by Y. Tanaka, owner
One RD. One day. One owned record.
Evolution
One record supported different responsibilities.
The RD still owned the daily record. Managers could inspect its state and exceptions, but did not author the work on the RD’s behalf.
Two states of the same record, recreated from the shipped product. Every name and number shown is demo data.
One RD, one day, captured inside Elevra's existing Assessments engine, with the AI field sitting above the fields it actually feeds. The AI layer itself was still in early testing.
The evidence layer
Three decisions shaped what shipped.
Use the platform that already owns assessments.
Clinical Productivity became a workflow inside Elevra's existing Assessments infrastructure, not a new platform.
The pain was shared files, fragmented records, and manual aggregation.
Only show the fields this care setting requires.
Selecting a care setting is what creates its activity fields; an unselected setting has none to fill.
Most of the source schema is irrelevant on any single day.
What it cost
A shorter-looking form can still be incomplete: fields were missing from the polished interface.
Care setting
Activity records
Inpatient adultAcute medical, 4 East
Unit rule awaiting confirmation
Inpatient pediatrics
Outpatient adultNutrition clinic, Suite 210
Outpatient pediatrics
Flag only what a rule can explain.
The manager surface uses day-level status, with explicit rules for data-quality exceptions.
Managers were reconciling the monthly result by hand.
What it cost
Day-level detail creates a wider table and needs careful weekend and PTO handling.
Rachel Dempsey, Wed
Mandate
Design owned the workflow, not every production layer.
I designed the module end to end and documented the gaps.
Design owned
What I led
Daily record states, RD and manager role behavior, interaction model, documented gap inventory.
Engineering owned
What they built
Form.io configuration, production data model, RBAC implementation, APIs and reporting datasets.
The evidence layer was the part built for this: day-level status, deterministic exception flags, and a recreated interactive surface. Every figure above is that surface, running.
Outcomes
Production status is real. Product impact remains mostly open.
Shipped
The module reached production, covering daily logging, the manager grid, drill-down, and edit states.
Directional
Centralized records and deterministic exception flags point toward a safer alternative to concurrent workbook edits. The AI layer remained in early testing, separate from the shipped workflow.
Not measured
No outcome number is claimed here. Adoption, reconciliation time, notifications, Power BI, and cross-module behavior were not measured. The first thing I would instrument: the time from a missing submission appearing to the manager resolving it.
Two things I’d change next time.
Field gaps, earlier
I would put the field-gap inventory beside the interface earlier: the surface already looked coherent while required fields were still absent, and that gap is a delivery risk precisely because the surface looks ready.
Policy before polish
I would force the time-unit, edit-window, and role-switching decisions into a policy workshop before visual polish. The UI can display the answer; it cannot responsibly invent one.
More work