Why Recommended Visit Intervals Should Be Extracted When Conducting Longitudinal Analyses Using Electronic Health...

Area of Science:

• Biostatistics
• Epidemiology
• Health Informatics

Background:

• Electronic health records (EHRs) generate rich longitudinal patient data.
• Irregular assessment times in EHRs, often linked to patient health, can bias disease course estimation.
• Ignoring the informative nature of assessment timing can lead to inaccurate conclusions.

Purpose of the Study:

• To enhance disease trajectory estimation using physician-recommended visit intervals from EHRs.
• To characterize the patient assessment process and assess sensitivity to non-random assessment.
• To improve the rigor of analyses involving irregular longitudinal data.

Main Methods:

• Leveraging physician-recommended intervals as a key piece of information within EHRs.
• Characterizing the assessment process and investigating assessment not at random (ANAR).
• Applying the proposed approach to a juvenile dermatomyositis (JDM) cohort study.

Main Results:

• Recommended intervals explained 78% of the variability in patient assessment times.
• Assuming earlier visits due to disease worsening shifted the estimated disease trajectory downward.
• Demonstrated sensitivity of results to departures from the assessment at random (AAR) assumption.

Conclusions:

• Physician-recommended intervals are crucial for improving the rigor of longitudinal data analysis.
• Recommended intervals allow assessment of the AAR assumption's plausibility and result sensitivity.
• Studies with irregular longitudinal data should incorporate recommended visit intervals into their analyses.