biostudies-literatureCui HThe Mark and Debra Leslie fund for AI in HealthcareNational Science Foundation Graduate Research Fellowship Program577https://www.ebi.ac.uk/biostudies/studies/S-EPMC12475073biostudies-literatureUnknown8(1)Electronic health records (EHRs) contain rich longitudinal information for clinical decision-making, yet LLMs struggle to reason across patient timelines. We introduce TIMER (Temporal Instruction Modeling and Evaluation for Longitudinal Clinical Records), a method to improve LLMs' temporal reasoning over multi-visit EHRs through time-aware instruction tuning. TIMER grounds LLMs in patient-specific temporal contexts by linking each instruction-response pair to specific timestamps, ensuring temporal fidelity throughout the training process. Evaluations show that TIMER-tuned models outperform conventional medical instruction-tuned approaches by 6.6% in completeness on clinician-curated benchmarks, with distribution-matched training demonstrating advantages up to 6.5% in temporal reasoning. QuNPJ digital medicineTIMER: temporal instruction modeling and evaluation for longitudinal clinical records.PMC12475073NSF 23-605Alsentzer EUnell AFries JACui HChen BShah NHKoyejo SfalseTIMER: temporal instruction modeling and evaluation for longitudinal clinical records.Electronic health records (EHRs) contain rich longitudinal information for clinical decision-making, yet LLMs struggle to reason across patient timelines. We introduce TIMER (Temporal Instruction Modeling and Evaluation for Longitudinal Clinical Records), a method to improve LLMs' temporal reasoning over multi-visit EHRs through time-aware instruction tuning. TIMER grounds LLMs in patient-specific temporal contexts by linking each instruction-response pair to specific timestamps, ensuring temporal fidelity throughout the training process. Evaluations show that TIMER-tuned models outperform conventional medical instruction-tuned approaches by 6.6% in completeness on clinician-curated benchmarks, with distribution-matched training demonstrating advantages up to 6.5% in temporal reasoning. Qu2025-01-01T00:00:00Z2025 Sep2026-06-03T22:46:40.211Z2026-05-02T03:11:39.814ZS-EPMC124750734100689810.1038/s41746-025-01965-9