Clinical background
In a mass casualty incident, pandemic or natural disaster, the number of critically ill patients can quickly exceed the available intensive care resources. Prioritising patients for the allocation of intensive care beds and mechanical ventilation requires an assessment instrument that predicts mortality and the need for ventilatory support while remaining feasible when laboratory capacity is limited. The full SOFA score requires four laboratory values (platelets, bilirubin, arterial oxygen tension and creatinine) and is therefore difficult to use when laboratory logistics are poor or patient flow is high.
The modified SOFA (mSOFA) score was developed to address this. By omitting the platelet count, replacing the arterial PaO₂ with the pulse oximetry-based SpO₂/FiO₂ ratio and replacing serum bilirubin with clinical observation of scleral icterus, the laboratory requirement is reduced to a single value: creatinine, which can be measured with a point-of-care device at the bedside [1]. The score is therefore not a diagnostic instrument but a triage tool for allocating resources under extreme load.
Calculating the modified SOFA score
mSOFA scores five organ systems, which are then summed:
where = respiration (SpO₂/FiO₂ ratio), = liver (scleral icterus), = circulation (MAP/vasopressors), = CNS (Glasgow Coma Scale), = kidney (creatinine in mg/dL).
Respiration (SpO₂/FiO₂ ratio): >400 = 0; 316–400 = 1; 236–315 = 2; 151–235 (with respiratory support) = 3; ≤150 (with respiratory support) = 4. With nasal cannula oxygen, the FiO₂ is estimated by multiplying the flow in litres by 0.03 and adding 0.21 [1].
Liver (scleral icterus/jaundice): Absent = 0; Present = 3. The component is binary and has no level for mildly or moderately raised bilirubin.
Circulation (MAP/vasopressors, µg/kg/min): MAP ≥70 mmHg = 0; MAP <70 mmHg = 1; dopamine ≤5 or dobutamine at any dose = 2; dopamine >5, or adrenaline/noradrenaline ≤0.1 = 3; dopamine >15, or adrenaline/noradrenaline >0.1 = 4.
CNS (Glasgow Coma Scale): 15 = 0; 13–14 = 1; 10–12 = 2; 6–9 = 3; <6 = 4. In the derivation study the highest GCS during the 24-hour period was used, not the lowest, and this performed better than the lowest GCS [1].
Kidney (creatinine, mg/dL): <1.2 = 0; 1.2–1.9 = 1; 2.0–3.4 = 2; 3.5–4.9 = 3; ≥5.0 = 4. The thresholds are given in mg/dL; in SI units, 1.2 mg/dL corresponds to about 106 µmol/L and 5.0 mg/dL to about 442 µmol/L.
The derivation cohort consisted of patients in a medical, surgical and trauma intensive care unit at LDS Hospital, Salt Lake City, during 2006: 718 patients (58% men), 30-day mortality 17.3% [1]. A prospective observational study was then carried out in a corresponding 24-bed intensive care unit at Intermountain Medical Center, Murray, Utah, during 2008: 1,770 patients (56% men), 30-day mortality 10.5% [1]. The modelled outcomes were all-cause 30-day mortality and the need for mechanical ventilation on days 3 and 5.
Interpretation in practice
The triage protocols on which mSOFA rests use three priority bands, originally developed for pandemic influenza and then applied to mSOFA [1]:
| mSOFA on day 1 | Triage priority | Mortality in the derivation cohort |
|---|---|---|
| 0–7 | High priority for intensive care | 4% |
| 8–11 | Intermediate priority | 31% |
| >11 | Low priority; consider not starting or not continuing intensive care | 58% |
A score of 0–7 means the patient will very probably survive and should be given high priority for resources. At 8–11 the mortality is substantial and the patient probably needs individual assessment; the score alone should not determine allocation. At >11 the mortality approaches 60% and the protocols aim to reserve intensive care resources for patients with a better prognosis. This is, however, an ethically problematic position, and a mortality of 58% means that 42% survive, so the score alone should not be the basis for withholding intensive care.
The score performs best on day 1. On repeated measurement on days 3 and 5 its discrimination falls, which means that the trend over time is more informative than a single score.
Validation and performance
In the prospective cohort (n=1,770), the day 1 mSOFA performed comparably to the full SOFA in predicting 30-day mortality: AUC 0.84 (95% CI 0.82–0.85) versus 0.83 (95% CI 0.81–0.85), p=0.33 [1]. Performance was likewise equivalent for predicting the need for mechanical ventilation on day 3: mSOFA AUC 0.82 (95% CI 0.80–0.84) versus SOFA AUC 0.83, p=0.43 [1]. On day 3 the AUC fell to 0.79 for mSOFA and 0.78 for SOFA, and by day 5 to 0.74 and 0.72 respectively, with no statistically significant difference [1].
Subgroup analysis showed that mSOFA performed best in trauma patients (AUC 0.84) and medical patients (AUC 0.82), while surgical postoperative patients showed greater variability [1].
External validation in other populations or resource-limited settings is lacking. The only published study outside the derivation group is a paediatric triage study in which mSOFA was compared with clinicians' judgement; there, clinical judgement outperformed all the scoring systems, and no instrument showed acceptable discrimination for separating patients who needed intensive care interventions from those who did not [3].
A separate modelling study of SOFA-based ventilator triage during the COVID-19 pandemic, in 205 intubated patients at three hospitals in New York, showed that 57% of patients would have been identified for withdrawal or withholding of ventilation under the protocol [2]. Of these, 24% survived to discharge. The study therefore questions whether SOFA-based triage protocols (of which mSOFA is a simplified variant) have sufficient predictive precision to justify life-and-death resource decisions.
Limitations
mSOFA has been validated only at a single centre within Intermountain Healthcare in Utah, in a mixed medical, surgical and trauma intensive care unit [1]. Its generalisability to other health systems, other patient populations and above all to the resource-limited settings in which the instrument is intended to be used has not been studied.
The liver component is binary (0 or 3) and captures only clinically visible jaundice. A patient with moderately raised bilirubin but without scleral icterus scores zero, which may underestimate liver involvement. Documenting scleral icterus also requires a systematic routine and training of staff, which in the study required specific instruction [1].
The SpO₂/FiO₂ ratio is a surrogate marker for PaO₂/FiO₂ and may underestimate hypoxaemia at high FiO₂ values because of the sigmoid oxyhaemoglobin dissociation curve. At low saturations (<70%) the pulse oximeter is also less reliable, which may affect scoring in precisely the most severe cases.
For sedated or paralysed patients, the study recorded the actual GCS irrespective of sedation, and the highest GCS over 24 hours was used [1]. This differs from conventional SOFA practice, where the lowest GCS is traditionally used, and makes the CNS component difficult to interpret in patients sedated for ventilation.
The paediatric triage study found that clinicians' judgement outperformed mSOFA and other scoring systems [3]. This underlines that mSOFA is not a substitute for clinical judgement, and that its role in triage should be supportive rather than decisive on its own.
References
- Grissom CK, Brown SM, Kuttler KG et al. A modified sequential organ failure assessment score for critical care triage. Disaster Med Public Health Prep 2010;4(4):277–84. PMID: 21149228
- Cuartas PA, Tavares Santos H, Levy BM et al. Modeling outcomes using sequential organ failure assessment (SOFA) score-based ventilator triage guidelines during the COVID-19 pandemic. Disaster Med Public Health Prep 2022;16(4):1–4. PMID: 35152936
- Sweney JS, Poss WB, Grissom CK et al. Comparison of severity of illness scores to physician clinical judgment for potential use in pediatric critical care triage. Disaster Med Public Health Prep 2012;6(2):126–30. PMID: 22700020
