· 2026
Emergency department clinical decision support tool
A decision support tool that gathers the fifteen most-used clinical scores from daily emergency-department practice into a single file, with interpretations calibrated to local hospital context.
The problem
In the emergency department, clinical scores are tools used in the middle of a case — while labs are pending, before deciding on imaging. HEART for chest pain, Wells for suspected PE, CURB-65 for pneumonia, Alvarado for suspected appendicitis — all of them need to be within reach of the deciding physician. In practice it doesn’t work that way: every time, you open another browser tab to another calculator site, fill in the fields, read the result, navigate back. Tapping out a score on your phone with one finger, keeping one eye on the patient and one on the screen — the practical reality is itself a source of error.
On top of that, most online score calculators give a single generic answer: “low risk,” “high risk.” But the disposition for a low-risk HEART patient depends on what the hospital actually has. In a hospital with a cardiology service, “outpatient cardiology follow-up” is a meaningful recommendation; in one without — like ours — the same recommendation hangs in the air. Same score, same result, different action.
What I built
A single-file HTML toolkit. Opens in a browser, no install, no server, no internet connection required — it runs on hospital computers, restricted networks, locked-down user accounts.
It contains the fifteen scores most often used in the day-to-day practice of an emergency department: HEART, CHA₂DS₂-VASc, HAS-BLED, ABCD² (TIA), Wells (PE), PERC, Wells (DVT), CURB-65, SIRS/sepsis, Glasgow-Blatchford (upper GI bleed), Alvarado, the Canadian CT Head Rule, PECARN (pediatric head injury), NEXUS (cervical spine), and the Ottawa Ankle Rules. Categories run down the left sidebar: cardiovascular, cerebrovascular, pulmonary/VTE, sepsis, GI, trauma/neuro, orthopedic.
For each score, not just a number but the recommendations that follow from it — and the recommendations are calibrated to local context. A high-risk HEART result lists the standard recommendations (“urgent cardiology consultation, early invasive strategy, coronary ICU admission”), then adds: “At our hospital, all of these recommendations mean transferring the patient.” The HEART score’s troponin field shows the hospital’s own reference value (upper limit 100) inline.
What was technically interesting
Technically, the tool is simple. The real work is in the accuracy of the content and the calibration to local context — and that’s not a software problem, it’s a clinical responsibility.
Two technical decisions are worth naming, though:
The first is keeping score definitions as data. Each score’s fields, options, points, and interpretation logic live in a single JavaScript object. Adding a new score, or updating an existing one’s interpretation when a new guideline comes out or the hospital’s resources change, means touching this data object rather than the HTML template. The presentation code doesn’t know about the scores; it just renders whatever definition it’s handed.
The second is being fully offline-capable. Hospital computers are restricted: some are offline, some block specific sites, some don’t permit installing a new version every day. A single-file HTML tool bypasses all of that — a user copies the file to the machine once, and from then on needs neither the network nor an update. The same file will run five years from now.
This toolkit also became the building block for the ED consultation and documentation generator I later built — its scoring drawer inherited these fifteen score functions wholesale.
Outcome
It sits on hospital-computer desktops, on personal phones, in browser tabs opened at the start of a shift. I don’t track which scores get used how often (an analysis I haven’t done in its own right), but the ones most commonly opened in daily use — HEART, Wells, Alvarado, CURB-65 — match the ones I most often need in my own practice.
The version hosted on this site is anonymous — no personal data is stored, no server is involved, no usage records are kept. It’s downloadable; it’ll run in other departments too, but the recommendations are calibrated to our hospital. Anyone wanting to use it in their own department should adjust the recommendations and threshold values that begin with “At our hospital…” to fit their own institution’s resources.