High resolution data modifies intensive care unit dialysis outcome predictions as compared with low resolution...

Area of Science:

• Health Data Science
• Clinical Informatics
• Biostatistics

Background:

• Electronic health records (EHRs) offer high-resolution clinical data, surpassing traditional administrative databases.
• Detailed clinical information in EHRs enables advanced analytics and better control for confounding variables.

Purpose of the Study:

• To compare clinical research analysis using low-resolution administrative data versus high-resolution EHR data.
• To evaluate the impact of detailed clinical variables on sepsis research outcomes.

Main Methods:

• A parallel cohort of intensive care unit (ICU) sepsis patients requiring mechanical ventilation was analyzed.
• Two databases were used: Nationwide Inpatient Sample (NIS) for low-resolution and eICU Collaborative Research Database (eICU) for high-resolution data.
• Mortality was the primary outcome, with dialysis use as the exposure of interest.

Main Results:

• Low-resolution analysis showed dialysis use increased mortality (NIS: OR 1.40; eICU: OR 2.07).
• High-resolution analysis, after adding clinical covariates, found no significant association between dialysis and mortality (OR 1.04).
• The addition of detailed clinical variables significantly improved confounder control.

Conclusions:

• High-resolution clinical data from EHRs are crucial for accurate statistical modeling in health research.
• Findings from studies using administrative data may require re-evaluation with detailed clinical datasets.
• EHR data enhances the ability to control for confounders, leading to more reliable research findings.