A simple 5-question checklist that predicts which patients are most likely to return to the ER within a week. Toggle the factors below to calculate.
intime → outtime)Background
The Discharge Severity Index is a quick bedside checklist for emergency rooms. Right now, when patients leave the ER, everyone gets roughly the same instructions. The DSI helps doctors and nurses figure out who needs extra follow-up and who is safe with standard care.
About 1 in 7 patients who leave the ER end up coming back within a month. Many of those return visits could be prevented with a follow-up call or check-in—but until now, there was no simple way to decide who needs that extra attention.
At discharge, you check 5 simple things: the patient's age, heart rate, oxygen level, how long they were in the ER, and how many medications they take. Add up the points (0–7) and the score tells you their risk level (1–5).
Based on the score, each patient gets a follow-up plan that matches their risk: a phone call, a video visit, a home nurse visit, or simply a reminder to see their regular doctor. Higher risk = more follow-up.
Variable Definitions
The DSI was derived from MIMIC-IV (Massachusetts General Hospital EHR data). These operational definitions reflect how each variable was extracted.
Definition: Time from ED registration (intime) to physical departure from the ED (outtime).
Includes: Waiting room time, active treatment, and boarding delays (time spent in the ED after a disposition decision while awaiting a bed or transport).
Does not start at: Nurse triage, first provider contact, or attending assignment.
Does not end at: Disposition decision (discharge/admission order). It ends when the patient physically leaves the department.
The authors acknowledge LOS is “an imperfect measure of medical complexity” because it includes systems-level delays.
Definition: The count of home medications the patient reports taking, documented during medicine reconciliation at ED admission.
Source: MIMIC-IV medrecon table—staff ask the patient what medications they currently take on arrival.
Does not include: Medications administered during the ED visit, or prescriptions written at discharge.
“Active” means: Medications the patient reports currently taking at the time of the ED encounter. No specific lookback window is applied—it reflects the patient’s self-report.
Scoring
These 5 factors were identified by analyzing nearly 230,000 ER visits. Each factor is weighted by how strongly it predicts a return visit—the stronger the link, the more points it gets. Maximum score: 7.
1.3× more likely to return (OR 1.29)
1.7× more likely to return (OR 1.67)
1.4× more likely to return (OR 1.36)
2.9× more likely to return (OR 2.87). Measured from ED registration to physical departure.
2.7× more likely to return (OR 2.65). Counted from medicine reconciliation at ED admission—medications the patient reports taking at home.
Risk Stratification
The score maps to 5 levels, from highest risk (Level 1) to lowest (Level 5). Each level comes with a suggested follow-up plan. Validated on a separate group of ~57,000 patients.
| DSI Level | Score Range | Risk Multiplier | 7-day Return Rate | Prevalence | Suggested Follow-up |
|---|---|---|---|---|---|
| 1 Highest risk | 6–7 | 14.6× | ~1 in 20 | ~1.4% | Video/phone check-in within 24h, home nurse visit, remote monitoring |
| 2 High risk | 5 | 11.7× | ~1 in 25 | ~10.4% | Nurse calls within 48h, medication review, video visit |
| 3 Moderate risk | 3–4 | 8.4× | ~1 in 33 | ~25.3% | Follow-up plan before leaving, see your doctor within 1 week |
| 4 Low risk | 1–2 | 3.5× | ~1 in 100 | ~46.5% | Standard go-home instructions, schedule a doctor visit if needed |
| 5 Lowest risk | 0 | 1.0× (baseline) | <1 in 200 | ~16.3% | Routine care, no special follow-up needed |
Context
Several scoring tools exist in medicine, but the DSI is the first designed specifically for the moment a patient leaves the ER. Here's how it compares.
Evidence
The DSI was published in a peer-reviewed emergency medicine journal and featured on ALiEM, a leading medical education site.