Healthcare Clinical SaaS ✦ −42% Time 10 months
(01) CASE STUDY
⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯
BEAM HEALTH · 2023
PRODUCT / UX DESIGN

Designing for the
twelve seconds
between patients

Beam Health makes EHR-adjacent tooling for outpatient clinics. I led the redesign of the charting flow, the screen clinicians live in, and cut 42% off the average chart-completion time while passing every HIPAA and audit review.

−42%
Less clinician time per chart
after the redesign
ROLE Senior product designer
SCOPE End to end
INDUSTRY Healthcare, clinical SaaS
YEAR 2023
TEAM Two designers, 22 engineers
app.beamhealth.io/chart/encounter/e-9043 after
Schedule Charting Orders Search patient or MRN… DR
MR
Maria Reyes
MRN 40‑992154 FVISIT Follow-up · HTNENCOUNTER Today 2:14 PM
Chart · 3 of 4 sections
Encounter · SOAP
Subjectivehistory, symptoms
Objectivevitals, exam
3Assessmentdiagnosis
4Planorders, follow-up
Assessment
In the order you think — not the order the auditor reads
Primary diagnosis *
Essential hypertension  · I10
From last visit & today’s BP: continue lisinopril 10mg, recheck in 2 weeks. Accept Edit
Clinical impression
BP improved from 152/94 to 138/86 on current therapy. Patient adherent, tolerating well. No new symptoms.
Secondary diagnoses
Add a diagnosis…
Optional — you can submit without this. Nothing will block you mid-chart.
Today’s vitals
BP
138/86
HR
72
Temp
98.4°
SpO₂
98%
Context
Last visit6 wks ago
Active medsLisinopril 10mg
AllergiesPenicillin
Care gapsA1c due
● Saved · 2s ago Save & close Continue to Plan →
(02) Context

The screen between
a doctor and their
next patient.

Charting is the most-used screen in any EHR-adjacent product. A clinician sees twenty to thirty patients a day. Each visit ends with charting. Save thirty seconds per chart and you give a clinician back an hour.

Beam's charting flow had grown by accretion. A dropdown here, a modal there. By the time I joined it was a fourteen-field form clinicians had to navigate while talking to a patient, or in the twelve-second handoff between rooms.

Company snapshot
STAGE Series B healthtech
USERS 8,000 clinicians
COMPLIANCE HIPAA, HITRUST, SOC2
INTEGRATIONS 4 major EHRs
CHART VOLUME ~120k per day
(01)
High-stakes UX

A bad chart is not a bad form fill. It is a missing diagnosis or a billing error. Mistakes here carry clinical and financial weight.

(02)
Compliance is the floor

HIPAA, audit logging and clinical-content rules constrain every interaction. Design without legal in the room is design that gets rolled back.

(03)
Twelve-second windows

Clinicians do not have minutes to chart. They have seconds, often standing, often interrupted. The interaction model has to assume hostile conditions.

(03) The problem

A fourteen-field form,
in twelve seconds,
while standing.

Chart-completion time was the company's most-watched non-clinical metric, and it had been creeping up for three releases. Diagnosis was hard. Clinicians completed charts late at night, in batches, off the floor. The screen was failing them, quietly, and asynchronously.

Shadowing fourteen clinicians across four specialties gave a sharper picture. The charting screen was demanding full attention at the moment clinicians had the least to give.

8.2 min
Average chart-completion time.
Category baseline was around 4.5 min
Failure mode 01
Modal interruptions

Required-field errors fired as modals mid-charting and broke train of thought. Clinicians described mentally starting over after each one.

Failure mode 02
Hidden defaults

Smart defaults existed but were buried two levels deep. Most clinicians did not know they existed. They typed everything from scratch.

Failure mode 03
Audit-shaped UI

The form was structured for the auditor, not the clinician. Field order matched the chart's legal structure, not the encounter's natural flow.

"I chart at 9pm at home. The screen is the last thing I see before I close the laptop and the first thing I dread in the morning."

Internist, shadowing study
app.beamhealth.io/chart/legacy/e-9043 before
Encounter chart · Maria Reyes14 required fields · page 1 of 2
Diagnosis code (ICD-10) *
Required — enter code first
Billing modifier *
Select…
Place of service *
11 · Office
Encounter type *
Established patient
CPT level *
Select…
Referring provider
Chief complaint *
HTN follow-up
History of present illness
Scroll down to section 9…
!
Required field missing
You must enter a valid ICD-10 diagnosis code before continuing. Unsaved entries in this section may be lost.
Go back
FIG. 01 — The original chart: audit-ordered (codes first, history last) with modal errors that interrupt mid-thought.
(04) Role

Senior product designer.
Ten months.

I led the charting redesign as the senior designer, paired with one junior designer and a clinical informaticist who was embedded in the team. The compliance officer reviewed every flow.

My remit covered discovery, shadowing and diary studies, interaction design, accessibility, and the validation plan that satisfied both clinical and audit stakeholders.

Responsibilities
DISCOVERY Shadowing, diary, audit log analysis
DESIGN Flows, forms, compliance UX
VALIDATION Usability, A/B, field trials
HANDOFF Specs, audit logs, rollout plan
Q1
Shadowing + diaries
Q1
Flow rearchitecture
Q2
Concept + prototypes
Q2
Compliance review
Q3
Field trial in three clinics
Q4
Rollout + measurement
(05) Initiatives

Three bets.
One direction.

Three structural changes drove the result. Each one took a piece of the form's cognitive load off the clinician.

(01)
Encounter-first form
Reordered fields to match the natural flow of a patient encounter, not the audit document.
↳ −42% chart time
(02)
Inline assistance
Smart defaults surfaced where they were needed, not buried in settings.
↳ 3.2× default usage
(03)
Non-blocking validation
Form errors became gentle inline cues, not modals. No more starting over.
↳ Cognitive load ↓
Deep dive
Encounter-first form

Field order is a
design decision.

The single highest-impact change was reordering the form. The original chart was structured for the auditor reading it three months later. Diagnosis codes first, history last. Clinicians filled it in a different order. History, then observations, then assessment, then plan. The encounter itself.

Every clinician we shadowed scrolled up and down to fill the form, because the form's structure did not match how they thought. Reordering looked trivial in Figma. It took four months of compliance review.

The research
14 SHADOWING SESSIONS, 4 specialties
6 WEEK DIARY STUDY, 22 clinicians
AUDIT LOG ANALYSIS, 8 weeks of data
3 CLINIC FIELD TRIAL, 60 days
01
SOAP-aligned section order

Sections were reordered to mirror SOAP, subjective then objective then assessment then plan. Clinicians stopped scrolling. Time to first field dropped from twenty-two seconds to six.

Trade-off. Required updating every audit template downstream. Four months of compliance work, validated case by case.
02
Inline non-blocking validation

Modal errors were replaced with gentle inline cues. The form never blocks the clinician mid-thought. Errors surface at submit, with a clear path to fix.

Outcome. "Feels less like a form" was the most common verbatim from the field trial.
03
Smart defaults at the field, not in settings

When a field had a likely value, based on patient history, specialty or previous chart, the value surfaced inline as a one-tap accept. No settings menu required.

Trade-off. This needed a per-specialty config, which engineering pushed back on. The field trial numbers won the argument.
app.beamhealth.io/chart/legacy/e-9043 before
Field order matches the legal documentcodes first · history last
Diagnosis code (ICD-10) *
Enter code…
Billing modifier *
Select…
CPT level *
Select…
Place of service *
11 · Office
Chief complaint *
HTN follow-up
History (field 9)
Scroll down…
Before · the clinician scrolls up and down to think8.2 min / chart
app.beamhealth.io/chart/encounter/e-9043 after
Encounter · SOAP
Subjective
Objective
3Assessment
4Plan
Assessment
Subjective → objective → assessment → plan. The encounter’s order.
Primary diagnosis *
Essential hypertension  · I10
Continue lisinopril 10mg, recheck 2 weeks.Accept
Clinical impression
BP improved to 138/86 on current therapy. Adherent, tolerating well.
Codes & billing fill in automatically from your assessment.
After · the form matches how clinicians think4.7 min / chart
(07) Impact

The work,
measured.

−42%
Chart time per encounter
From 8.2 min to 4.7 min
60 DAY FIELD TRIAL
3 CLINICS, 22 CLINICIANS
3.2×
Smart-default usage
INLINE BEAT BURIED
IN SETTINGS
0
Compliance regressions
4 AUDIT REVIEWS
FULL PASS
+18 NPS
Clinician NPS
Post-rollout
60 DAYS POST
FULL ROLLOUT
8,000
Clinicians on the new flow
PHASED ROLLOUT
10 WEEKS
NOTE. Some metrics are directional, based on data available at the time. Where possible, figures reflect controlled comparisons. Otherwise they are stakeholder-validated estimates.
(08) Outcomes

What changed.
What it earned.

BeforeAfter
Audit-ordered form, codes firstEncounter-ordered, SOAP
Modal validation interrupts thoughtNon-blocking inline cues
Smart defaults buried in settingsInline, one-tap accept
8.2 min average chart time4.7 min average chart time
No clinical-flow researchShadowing, diary, field trial
Single rollout, fingers crossedPhased rollout with rollback gates
−42%
Less clinician time per chart, after the redesign
(09) Reflection

What I would
do differently.

What worked
Shadowing beats surveys

The audit-versus-encounter ordering insight was visible in five minutes of shadowing and invisible in six weeks of survey data. Be in the room.

Compliance as a co-author

Bringing the compliance officer into design reviews, not just sign-off, surfaced constraints early and produced better designs faster.

Field trial before rollout

The three-clinic trial caught two regressions and validated the metric movement. Worth every week.

What I would do differently
Earlier engagement with EHR vendors

Some integration constraints surfaced late. Earlier vendor conversations could have widened the design space.

Better baseline instrumentation

The 8.2 minute baseline came from log analysis, not a clean experiment. Cleaner instrumentation would have made the result harder to dispute.

Plan for off-floor charting

A non-trivial fraction of charting happens at home, late at night. I designed for the floor and treated home charting as out of scope. It should not have been.

The clinician finishes their day at six, not nine. That is the metric. Everything else is downstream.

Artur Lopez Zarytskyi · Beam Health · 2023
← Back to portfolio Get in touch