Risk adjustment for injured patients using administrative data

Area of Science:

• Trauma research
• Health services research
• Predictive modeling in healthcare

Background:

• Effective risk adjustment is crucial for population-based studies involving injured patients.
• Existing methods require refinement to accurately assess outcomes in diverse patient populations.

Purpose of the Study:

• To evaluate the predictability of hospital outcomes for injured patients.
• To compare the utility of different injury classification systems and demographic factors in risk adjustment.

Main Methods:

• Utilized National Hospital Discharge Surveys data from 1996-2000.
• Employed International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnoses to categorize injury severity, mechanisms, and comorbidities.
• Constructed regression models using weighted survey data, incorporating injury classifications, age, and sex to predict mortality, length of stay (LOS), and discharge to long-term care (LTC).

Main Results:

• Increased Abbreviated Injury Scale (AIS) score and Injury Severity Score (ISS), or decreased ICD-9-CM Injury Severity Score, were associated with higher mortality, prolonged LOS, and more frequent LTC.
• Injury mechanisms (penetrating, burn, vehicle) and comorbidities influenced LOS and LTC.
• Demographic factors like age and sex demonstrated significant associations with outcomes; males had shorter LOS and less frequent LTC than females.

Conclusions:

• Hospital outcomes following injury are predictable using readily available data including age, sex, and standard diagnosis groupings.
• Anatomical injury scales provide comparable predictive performance to other methods when adjusted for key patient and injury characteristics.