oasis-score

name	oasis-score
description	Calculate OASIS (Oxford Acute Severity of Illness Score) for ICU patients. Use for mortality prediction with fewer variables than APACHE/SAPS, or when lab data is limited.
tier	validated
category	clinical

OASIS Score Calculation

The Oxford Acute Severity of Illness Score (OASIS) is a parsimonious severity score that achieves comparable predictive accuracy to APACHE using fewer variables. It does not require laboratory values, making it useful when lab data is missing.

M4Bench Use

In M4Bench, target concept tables listed in the task configuration are removed or unavailable in the agent database. Use this skill as procedural guidance and derive the requested output from available source or intermediate tables; do not rely on a precomputed target table or bundled SQL script.

When to Use This Skill

• Mortality prediction when lab data is incomplete
• Quick severity assessment with minimal variables
• Real-time severity scoring (no lab turnaround time)
• Research requiring a validated, simple severity metric
• Comparison with APACHE/SAPS scores

Score Components (First 24 Hours)

OASIS sums 10 component scores. The cutoffs below are aligned with the M4Bench/eICU labels, which follow the eICU implementation adapted from Johnson AEW et al. (Crit Care Med 2013, Table 2). For min/max-bound continuous variables, use the eICU evaluation order shown below because first-match ordering affects component labels when both low and high extremes occur in the first 24 h.

Pre-ICU Length of Stay

Range (hours)	In minutes	Score
< 0.17	< 10.2	5
0.17 – 4.94	10.2 – 296.4	3
4.95 – 24.00	297.0 – 1440.0	0
24.01 – 311.80	1440.6 – 18708.0	2
> 311.80	> 18708	1

Age (years)

Range	Score
< 24	0
24 – 53	3
54 – 77	6
78 – 89	9
≥ 90	7

Glasgow Coma Score (worst observation)

Range	Score
3 – 7	10
8 – 13	4
14	3
15	0

Heart Rate (min^-1)

Evaluation order	Bin	Score
1	min < 33	4
2	max 33-88	0
3	max 89-106	1
4	max 107-125	3
5	max > 125	6

Mean Arterial Pressure (mmHg)

Evaluation order	Bin	Score
1	invasive min < 20.65	4
2	invasive min 20.65-50.99	3
3	invasive min 51-61.32	2
4	invasive min 61.33-143.44	0
5	invasive max > 143.44	3
6	noninvasive min < 20.65	4
7	noninvasive min 20.65-50.99	3
8	noninvasive min 51-61.32	2
9	noninvasive min 61.33-143.44	0
10	noninvasive max > 143.44	3

Respiratory Rate (min^-1)

Evaluation order	Bin	Score
1	min < 6	10
2	min 6-12	1
3	min 13-22	0
4	max 23-30	1
5	max 31-44	6
6	max > 44	9

Temperature (deg C)

Evaluation order	Bin	Score
1	min < 33.22	3
2	min 33.22-35.93	4
3	max 33.22-35.93	4
4	min 35.94-36.39	2
5	max 36.40-36.88	0
6	max 36.89-39.88	2
7	max > 39.88	6

Urine Output (mL/day)

Evaluation order	Range	Score
1	< 671	10
2	671-1426.99	5
3	1427-2543.99	1
4	2544-6896	0
5	> 6896	8

Mechanical Ventilation

Status	Score
No	0
Yes (any invasive ventilation in 24 h)	9

Elective Surgery

Status	Score
eICU/APACHE elective-surgery signal present and not admitted from ED	0
Non-elective, ED-source, missing elective-surgery signal, or otherwise non-elective	6

The OASIS total is the sum of the 10 component scores. Report the component scores alongside the total so implementation differences are auditable.

Direction-aware Scoring (eICU labels)

For the eICU OASIS task, follow the eICU-code evaluation order in the tables above. This ordering is intentionally dataset-specific and differs from the MIMIC OASIS task labels: if both a low and high extreme occur in the same first-24h window, the first matching row in the component table wins.

Key conflict rules for eICU labels:

• Heart rate checks the low extreme first, then high-heart-rate bins.
• MAP checks invasive BP mean first; noninvasive BP mean is used only if no invasive BP bin matches.
• Respiratory rate checks low/normal bins before high respiratory-rate bins.
• Temperature checks low-temperature bins before fever.
• Urine output uses the eICU boundaries shown above.

GCS uses the worst (minimum) value only. Pre-ICU LOS, Age, Urine Output, Mechanical Ventilation, and Elective Surgery are scalar — no min/max distinction.

Critical Implementation Notes

1. No Laboratory Values Required: OASIS uses only vital signs, urine output, and administrative data — no labs needed.

2. Mechanical Ventilation: Any invasive ventilation during the first 24 hours scores 9 points; otherwise 0.

3. Elective Surgery: Follows the eICU/APACHE fields used by the M4Bench/eICU labels. Stays with an elective-surgery signal and no ED-source admission score 0 points; ED-source, non-elective, missing, or otherwise non-elective stays score 6 points.

4. Missing Data Handling: Per task instruction, treat missing data as normal (score 0). When a component cannot be computed because the underlying observation is absent, score that component as 0 before summing. Output ALL ICU stays in the result, even those missing some or all source measurements.

   • Special case: the ventilation indicator is binary (yes/no) — absence of any ventilation record is interpreted as "no ventilation", contributing 0 points (not missing).

Mortality Probability

OASIS predicts in-hospital mortality through a logistic model (Johnson et al. 2013):

oasis_prob = 1 / (1 + exp(-(-6.1746 + 0.1275 * oasis)))

This probability is informational only and is not required by the M4Bench output schema for this task.

Source Data Required

OASIS requires the following clinical/administrative data per ICU stay:

• ICU admission and discharge times; hospital admission time (for pre-ICU LOS)
• Admission type (elective vs non-elective)
• Service or specialty assignment at the time of ICU admission (to identify surgical admissions)
• Patient age at admission
• First 24-hour vital signs: heart rate (min, max), mean arterial pressure (min, max), respiratory rate (min, max), temperature (min, max)
• First 24-hour Glasgow Coma Score (worst observation)
• First 24-hour urine output total
• Invasive mechanical ventilation status during the first 24 hours
• Note: when the database represents pre-ICU LOS as a signed offset from ICU admission (e.g., negative minutes back to hospital admission), take its absolute value before applying the scoring bins above.

Advantages Over APACHE/SAPS

• Simpler to calculate (10 variables vs 15-17)
• No laboratory data required
• Can be calculated earlier in admission
• Similar predictive accuracy

References

• Johnson AEW, Kramer AA, Clifford GD. "A new severity of illness scale using a subset of Acute Physiology And Chronic Health Evaluation data elements shows comparable predictive accuracy." Critical Care Medicine. 2013;41(7):1711-1718.