A single smartphone facial photograph contains measurable prognostic information regarding a patient’s risk of mortality upon arriving at an emergency department, according to a Nature report.
Rapid risk stratification remains essential in clinical settings. However, standard vital signs, laboratory tests, and triage scores frequently fail to capture complete risk at presentation. Researchers examined 27,660 smartphone facial photographs gathered from patients admitted to Danish emergency departments, where the median patient age was 73 years and 52 percent of the cohort were male.
Using this visual dataset, investigators trained deep neural networks to identify mortality risk directly from facial features. The resulting face-derived scores demonstrated strong discrimination capabilities for mortality outcomes.
The algorithmic facial scores achieved area under the curve values of 0.80 for three-day mortality and 0.77 for 30-day mortality. These predictions outperformed traditional clinical indicators, including standard triage levels, early warning scores, vital signs, routine blood chemistry, and estimated facial age.
| Mortality Window | Area Under Curve (AUC) | Primary Comparator |
|---|---|---|
| 3-Day Mortality | 0.80 | Outperformed vital signs and early warning scores |
| 30-Day Mortality | 0.77 | Outperformed blood chemistry and triage level |
Patients who presented with high early warning scores alongside facial scores in the highest quartile experienced a thirty-fold increase in 30-day mortality compared to low-risk patients. Among patients with low early warning scores, the highest facial score quartile showed a substantially elevated risk, with an odds ratio of 13.39.
Nature, founded in the autumn of 1869 and first circulated by Norman Lockyer and Alexander MacMillan, has published peer-reviewed research across the natural sciences as a weekly international journal. In 2007, it received the Prince of Asturias Award for Communications and Humanity alongside Science.
These findings indicate that unstructured facial photographs encode clinically relevant prognostic details. Such automated visual assessments may eventually complement existing triage and patient monitoring protocols in acute care environments.
Worth a look
- Occupational Solvent Exposure Linked to Increased Parkinson’s Disease Risk
- Athens Hospitals Remove Heart-Extending Tumor from Girl in 15-Hour Surgery
- Robert Pattinson Trolled by UK Government After James Bond Reveal (newsylist.com)
- Sofascore Adds Auto-Updating Live Handball Scores for Top European Leagues (archyde.com)