LACE Score vs HOSPITAL Score: Which Readmission Risk Model Actually Works in Your EHR

Two readmission risk models dominate hospital workflows: LACE and HOSPITAL. Both predict 30-day readmission. Both are well-validated. Choosing between them is a question of data availability and operational fit, not which is “better.”

LACE: the operational classic

LACE (Length of stay, Acuity of admission, Comorbidity, ED visits) was developed by van Walraven et al. in 2010. The score:

• L — Length of stay (1-7+ points)
• A — Acute / non-elective admission (3 points)
• C — Comorbidity (Charlson index, 0-5+ points)
• E — ED visits in the prior 6 months (0-4+ points)

Total ranges 0-19. Scores ≥10 are high-risk. The advantage: every component is available in any EHR with discharge data and a problem list. LACE works with the data you already have.

HOSPITAL: the lab-driven alternative

HOSPITAL (Donzé et al., 2013) replaces some LACE inputs with lab and admission detail. Components:

• H — Hemoglobin at discharge
• O — Oncology service discharge
• S — Sodium level at discharge
• P — Procedure during admission
• I — Index admission type
• T — Number of admissions in past year
• AL — Length of stay

HOSPITAL outperforms LACE in some validation studies, particularly for medical (vs surgical) populations. The cost: HOSPITAL requires lab values at discharge, which may not be standardized in your EHR's discharge dataset.

How to choose

Three practical considerations beat statistical c-statistic comparisons:

1. Data availability. If your EHR doesn't consistently capture sodium and hemoglobin at discharge in a structured field, HOSPITAL falls back to imputation, and imputed scores degrade rapidly. LACE's inputs are universal.
2. Population mix. HOSPITAL was validated primarily on internal-medicine populations. If your readmission penalty exposure is heavy on AMI, CABG, and THA/TKA (HRRP's surgical conditions), LACE generalizes better.
3. Workflow integration. Whichever score gets shown to the discharging physician at the right moment is the score that drives behavior. A C-statistic 0.02 better doesn't matter if the dashboard shows up after discharge.

What we recommend

Default to LACE unless you have specific evidence HOSPITAL fits your population. The marginal predictive gain rarely exceeds the operational cost of maintaining the more data-intensive model. Either way, the score is only useful if it flows into the discharge workflow with enough lead time to change the discharge plan.

Vizier's readmission module computes both LACE and HOSPITAL automatically from EHR data and surfaces the high-risk cohort to the care management workflow. Connect via your EHR connector.

The honest scoring caveat

Both models miss social determinants — housing instability, transportation, social support, food insecurity. A patient with LACE 8 and significant social risk often readmits more than a patient with LACE 12 in a stable home. Layer SDOH screening data on top of either score and the predictive performance improves materially. That layer is becoming standard in 2026 analytics platforms; it should be part of any vendor evaluation.