Exceptional mortality prediction by risk scores from common laboratory tests

Area of Science:

• Clinical Chemistry
• Biomarkers
• Predictive Analytics

Background:

• Complete blood count (CBC) and basic metabolic profile (BMP) components are established risk predictors.
• Underutilized CBC and BMP components may hold independent prognostic value.
• Current risk prediction models often overlook many available laboratory parameters.

Purpose of the Study:

• To develop and validate a comprehensive risk score for all-cause mortality prediction.
• To assess the predictive ability of a score incorporating all CBC and BMP components.
• To determine if underused laboratory parameters contribute independent risk information.

Main Methods:

• A large cohort (N=71,921) was used for training, with independent validation cohorts (N=47,458; N=16,372; N=2558).
• Risk scores were calculated using multivariable modeling of all CBC and BMP components.
• Age- and sex-adjusted weights were applied to predict 30-day, 1-year, and 5-year mortality.

Main Results:

• Exceptional predictive performance was observed across all time points in training and test populations (AUCs ranging from 0.82 to 0.90).
• Risk scores demonstrated strong associations with mortality (P <.001) in validation cohorts.
• Risk score thresholds effectively stratified mortality risk in diverse patient and general populations.

Conclusions:

• A risk score integrating all CBC and BMP components is highly predictive of all-cause mortality.
• This score offers a cost-effective method for comprehensive risk assessment in clinical settings.
• The findings highlight the value of both commonly and infrequently used laboratory test components.