A GLP-1 chart that actually works: titration ladders, plateau alerts, side-effect tracking, adherence signals

The case was the first post in this series. The readiness checklist was the second. This is the chart.

What follows is a walk-through — six panels, in the order a clinician opens them on a Wednesday afternoon, with the AI patient overview sitting on top of all of them. The point is concrete, not aspirational. GLP-1 patient tracking is no longer a free-text SOAP exercise. It is six structured views that have to live next to each other, because a metabolic intervention does not fit in one note.

Most generalist EHRs miss the same four things in this workflow: titration ladders, plateau alerts, side-effect scoring, and injection-site rotation. The chart below treats those as first-class data, not workarounds.

The gap between a generic chart and a metabolic chart

If you have run a weight-management practice on a generalist EHR, you already know what is missing. Dose changes live in free-text. Weight goes into a vitals field that does not graph. Nausea ends up in a "patient reports" sentence three notes back. Injection sites are a rotation diagram tucked into a paper folder.

Every one of those is a structured-data problem disguised as a charting problem.

This isn't science fiction anymore. The pieces — discrete fields, longitudinal graphs, rule-based flags, document OCR, retrieval over chart data — exist today. The work is putting them in one chart and letting them talk to each other.

1. Titration ladder panel

The first panel sits at the top of the metabolic chart. It is a vertical ladder of the patient's drug, current rung in solid, future rungs in outline, last-rung-date stamp underneath. The ladder is drug-specific and label-linked. For injectable semaglutide, the default adult weight-reduction path starts 0.25 → 0.5 → 1.0 → 1.7 → 2.4 mg weekly; current Wegovy prescribing information also allows 1.7 mg maintenance and, for adults who tolerate 2.4 mg and need additional weight reduction, escalation to 7.2 mg weekly. For tirzepatide, current Zepbound prescribing information starts at 2.5 mg weekly for 4 weeks, then increases in 2.5 mg increments after at least 4 weeks toward 5, 10, or 15 mg maintenance, so the visible rungs are 2.5 → 5 → 7.5 → 10 → 12.5 → 15 mg weekly.

[screenshot: titration ladder panel]

Underneath the visual: days since last rung, expected next-rung date, and a single-line reason field if the clinician held escalation. Every rung change is a discrete event with a date — not a sentence in a note that an audit reviewer would have to hunt for.

The AI does two small things here, and only two. It flags the chart when a label-based 4-week step or clinic-defined review interval has passed. And it raises a soft hold-suggestion when the side-effect panel below shows a worsening trend with each rung. It does not escalate dose. It does not auto-prescribe. The clinician sees the flag, makes the call, and the chart records the decision and the reason. (This is the same pattern the BALANCE readiness checklist walks through under item 6.)

2. Weight trajectory chart

The second panel is a single line graph: weight over time, with an expected-trajectory band overlaid in a softer color. The expected band is not a universal promise of 0.5–1.5% body weight lost per week. It is a configurable comparison aid anchored to medication, dose, baseline status, and clinic protocol. In trial-level terms, STEP 1 reported mean body-weight change of -14.9% at 68 weeks with semaglutide 2.4 mg, and SURMOUNT-1 reported -15.0%, -19.5%, and -20.9% at 72 weeks with tirzepatide 5 mg, 10 mg, and 15 mg. Each visit adds a point. Each lab and home-scale upload adds a point too, if the patient has the portal turned on.

The interesting line is the band, not the data points.

When weight is flat for four consecutive weeks, within a clinic-defined tolerance band, the chart auto-tags as "review weight trend" and surfaces the tag on the patient overview. The AI does not diagnose plateau. It flags four weeks of flat weight for clinician review, which is a different and much smaller claim; clinical trial analyses define plateau over longer intervals, including a 12-week definition in a SURMOUNT post-hoc analysis. The clinician decides whether to hold dose, escalate dose, switch agent, intensify lifestyle support, or refer.

At week 12, or the clinic's chosen review interval, the chart computes responder status. A ≥5% body-weight loss after three months is a common anti-obesity medication response benchmark, and the Endocrine Society guideline uses that threshold when discussing whether to continue or discontinue medication. That is not the same thing as a BALANCE continuation rule. Current CMS BALANCE materials define model eligibility, coverage criteria, and monitoring expectations; the State Medicaid RFA does not set a universal 5% patient responder cutoff. In Oli, responder status is a configurable clinic or payer field: responder, not-yet-responder, or continue-with-rationale.

3. GI side-effect scoring

The third panel is the one most generalist charts simply do not have. Every visit captures a structured 0–3 score for nausea, a count for vomiting frequency, a Bristol scale entry for stool form, and a 0–3 score for abdominal pain. Injection-site reactions get a yes/no plus a free-text qualifier. Those fields map to the adverse-event categories clinicians already watch in Wegovy and Zepbound labeling: nausea, vomiting, diarrhea or constipation, abdominal pain, and injection-site reactions. None of these are paragraphs. They are discrete fields.

[screenshot: GI side-effect scoring grid trended over time]

The trend graph runs underneath the panel — six rows, one per metric, each a sparkline against the titration ladder above it. When nausea climbs from 1 → 2 → 3 across the last three rungs, the visual is unmissable. When constipation spikes only after a specific dose change, the dose change is the column that lights up.

Patients can fill the same fields between visits from the portal. Anything they log shows up in the next-visit pre-read with the date stamp. We all know how rarely a patient remembers to mention week-three nausea by week six. The capture has to be where the side effect happens, not where the appointment happens.

4. Injection-site rotation

The fourth panel is a body map — abdomen, thighs, upper arms — with site markers stamped by date. Patients with the portal can self-log the site immediately after the injection; clinicians can edit it at the next visit. The map renders the last eight weeks by default, with a toggle for the full history.

The AI flag here is narrow: any site reused within seven days lights up in amber. That is the entire rule. Rotation is a label-aligned administration habit, not a diagnosis. Wegovy instructions tell patients not to use the same spot each time, and Zepbound labeling says to rotate injection sites with each dose. The reminder exists to prevent crowding and make local reactions easier to interpret; it does not claim that a reused site caused an absorption problem.

5. Labs cadence

The fifth panel is schedule-driven, but the schedule is a clinic protocol rather than a universal GLP-1 label rule. A common default is baseline at therapy initiation, follow-up around three months, then every six months, with tighter monitoring when diabetes, kidney disease, liver disease, severe GI symptoms, or payer criteria require it. The cadence is owned by the chart, not the clinician's memory. When a patient is due, the chart drafts the order set, surfaces the to-be-ordered prompt at check-in, and watches for the result coming back from the lab.

Trend views graph HbA1c, lipid panel, ALT/AST, and creatinine longitudinally. Those fields line up with the comorbidity and eligibility categories clinics already have to document for metabolic care and BALANCE planning, including type 2 diabetes, kidney disease, cardiovascular disease, and MASH. Each trend has a clinically reasonable band overlay — not a diagnostic call, just a visual that makes a drift look like a drift instead of a number-of-the-day. Out-of-range results route to the inbox the same day.

The AI's role here is small and useful. New lab results get a one-paragraph synthesis when they arrive — what changed, what stayed steady, what to flag in the next visit. It feeds into the patient overview below. (This pattern shows up in the bariatric metabolic workflow walkthrough too.)

6. AI patient overview

The sixth panel is the one a clinician actually opens first on visit day. It is a one-page synthesis generated from the chart plus any OCR'd documents the patient uploaded — prior bloodwork, an outside cardiology note, a sleep-study summary. Six lines, in this order:

1. Where the patient is on titration, days since last rung, next expected step.
2. Weight trajectory state — on-band, review weight trend, responder/non-responder.
3. Side-effect trend — what is rising, what is steady, what is improving.
4. Most recent labs and the AI synthesis paragraph.
5. Adherence signals from the patient portal — injection logs, side-effect entries, scale uploads, missed-message flags.
6. Recommended talking points for today's visit. Not a prescription. A check-list of what is unresolved.

The pre-visit read

Six lines, in this order, generated automatically before the clinician opens the chart on visit day. Status pills are gentle — most are "ok", one or two read "review" when something asks for a second look. Nothing is red unless something genuinely needs to be.

The clinician keeps total control of the chart. The overview is a reading view, not a writing view — every edit to the chart still goes through the clinician, gets logged, and the prior version stays accessible. (Audit trails on every AI-generated note are item 11 of the BALANCE checklist for a reason.)

What the overview gives back is the evening. A clinician walks in already knowing the state of the patient, instead of clicking through five tabs in front of them. Pajama-time charting drops because the structured capture during the visit already populated five of the six panels above. The note writes itself out of the visit, not on a laptop at 8:00 PM.

What this looks like in your own clinic

Day 6 in this series argued the case. Day 7 walked the readiness checklist. Today's piece is the chart those two posts were pointing at.

The honest invitation is the simplest one. Oli is $19.95/month flat per practitioner, all-in. The 30-day free trial runs on real data, not a sandbox demo. A weight-management clinic can spin up an account on a Wednesday afternoon, load three composite patients, and model a full GLP-1 visit — titration ladder, weight curve, side-effect scoring, injection-site map, lab cadence, AI patient overview — in one session. If the chart holds up under your actual workflow, you have the answer. If it doesn't, you have lost an afternoon.

That is the close. Not a demo booking. An afternoon with the actual chart.