Five variables separated older emergency department patients by delirium risk
Delirium remains one of the easiest geriatric syndromes to miss in the emergency department, yet the downstream consequences are hard to ignore: longer stays, more complications, and higher mortality. In a prospective cohort from a tertiary hospital in Bangkok, investigators built and internally validated a simple clinical risk score designed to flag older adults who were most likely to develop delirium during an emergency department encounter.
The cohort was deliberately practical. Patients aged 65 years and older who remained in the emergency department for more than 4 hours were screened within 12 hours of arrival using the Thai version of the Confusion Assessment Method for the ICU. That timing matters because delirium in the emergency setting often evolves quickly, and many patients arrive with pain, infection, metabolic derangement, dehydration, polypharmacy, or occult organ dysfunction already in motion. A bedside tool that depends only on readily available information has obvious appeal when staff are stretched and cognitive fluctuation can be mistaken for fatigue, dementia, or general frailty.
Among 380 enrolled patients, 102 had delirium, a prevalence of 26.8%. The final model retained five independent variables and translated them into a point-based score ranging from 0 to 118. Age 70 to 79 years carried 1 point. Age 80 to 89 years, chronic kidney disease stage III to V, or prior cerebrovascular accident each carried 12 points. Age 90 years or older carried 29 points, cirrhosis carried 24 points, and dementia carried 41 points.
Dementia dominated the score, and for good reason. Baseline neurocognitive vulnerability leaves little reserve when acute illness, unfamiliar surroundings, sleep disruption, and medication exposure collide. Cirrhosis likely marks another biologically coherent pathway: hepatic dysfunction can amplify neuroinflammation, alter ammonia handling, and worsen cerebral vulnerability to even modest physiologic stress. Kidney disease and prior stroke sit in a similar high-risk zone, reflecting reduced physiologic reserve, vascular brain injury, and a greater chance that acute metabolic changes will tip the brain into inattention and disorganized thinking.
Discrimination was strongest at the bedside extremes
The score’s overall performance was solid, with an area under the receiver operating characteristic curve of 0.79 and a 95% confidence interval of 0.74 to 0.84. In practical terms, the model separated higher-risk from lower-risk patients well enough to support triage decisions, close observation, and earlier delirium-prevention measures. Delirium occurred in 17.97% of patients assigned to the low-risk category, defined as 0 to 24 points, compared with 57.65% of those in the high-risk category, defined as 25 points or more.
The spread between those groups is clinically meaningful. A nurse or physician who knows a patient sits in the high-risk band can respond differently within minutes: reassess cognition more often, review medications for deliriogenic agents, look harder for hypoxia or infection, normalize hearing and vision aids, and reduce avoidable environmental stress. In the emergency department, where throughput pressures can crowd out cognitive surveillance, risk stratification creates a way to assign attention before overt delirium becomes obvious.
A brief score is only useful if it can be remembered and applied without chart fishing. This model meets that standard better than many prediction tools because every included variable is already available early in care or can be elicited rapidly from history and routine assessment. Age, dementia, cirrhosis, kidney disease, and prior stroke are not exotic markers. They are the exact kinds of features emergency clinicians already weigh when deciding who needs closer monitoring, but the score gives those impressions a clearer structure.
Why the model fits the biology of delirium
Delirium is not a single disease. It is a final common pathway for acute brain dysfunction, often triggered when baseline vulnerability meets an acute precipitant. Age increases vulnerability through reduced cognitive reserve and higher multimorbidity burden. Dementia lowers the threshold for inattention and disorganized thinking. Kidney disease and cirrhosis can intensify metabolic instability, toxin accumulation, and inflammatory signaling. Prior stroke leaves structural brain injury that can make even mild systemic stress produce disproportionate neuropsychiatric effects.
Seen through that lens, the score is less a statistical exercise than a map of cerebral fragility. It does not require laboratory turnaround, imaging, or specialized neuropsychological testing. That simplicity is a strength in emergency care, where the first useful signal is often the one that arrives early enough to change behavior.
Still, the model should be read as an internal validation study from a single tertiary center rather than a finished universal tool. A score that works well in one hospital can lose calibration elsewhere if case mix, staffing, language, screening timing, or baseline frailty differ. The retrospective registration date also means external users will want to see independent replication before adopting the score as a formal triage standard.
What the score can change in practice
The strongest near-term use is prevention. Older adults classified as high risk could be prioritized for delirium precautions: orientation support, sleep protection, hydration review, pain control, avoidance of unnecessary restraints, and careful medication reconciliation. Even without a formal delirium diagnosis at presentation, the score can prompt more disciplined observation during the hours when symptoms may still be subtle.
The score may also help clinicians distinguish benign confusion from early delirium in crowded emergency settings. A patient with dementia is not automatically delirious, but a patient with dementia who also has kidney disease, cirrhosis, or prior stroke should trigger a lower threshold for repeated cognitive assessment. In that sense, the tool functions as a reminder that delirium risk is cumulative, not binary.
The article leaves one practical question unresolved: whether using the score prospectively will actually reduce delirium-related harm, shorten stays, or improve detection rates compared with usual care. The present data show that the score separates risk groups well. They do not yet show whether acting on that separation changes outcomes. External validation in other emergency departments, ideally with impact analysis, will determine how far this bedside tool can travel beyond the derivation cohort.