Supplementary Data 2: Anonymized sample of an automated MedFactEval evaluation report (HTML format).

Factual evaluation of the single-prompt-gpt-4o summarizer by a LLM jury

Jury members: claude-3.5-sonnet-v2, llama4-scout, deepseek-r1

Main results

A. Presence of important facts

Score: 48/100 (95% CI: 37-58) (higher values better)

Feedback regarding the important facts missing:

Hospital course summaries frequently omit critical clinical facts due to recurring patterns: (1) Incomplete documentation of ongoing interventions, where removal of devices is noted but retention is excluded (Patient 0: JP drain left at discharge), or partial treatment details are provided without key parameters (Patient 0: DVT prophylaxis mentioned, but anemia/Hb 8.8 omitted). (2) Failure to explicitly link related conditions, such as hypoxia attributed to pleural effusion without causal phrasing (Patient 1) or AKI in CKD not documented despite urinary retention management (Patient 4). (3) Procedural/timeline vagueness, omitting surgical steps (Patient 2: adhesion takedown), medication specifics (Patient 6: insulin type/dose discrepancies), or temporal patterns (Patient 5: POD1 drain output trends). (4) Diagnostic ambiguity, where findings are implied but not stated (Patient 8: MRSA nares vs. sputum source) or critical terminology is avoided (Patient 17: “delirium tremens” vs. “alcohol withdrawal”). (5) Discharge plan omissions, including follow-up frequency (Patient 12: TPN without 4-week timeline) or diet/medication specifics (Patient 24: full liquid diet assumed but not explicitly stated). These patterns highlight systemic gaps in specificity, causal attribution, and comprehensive documentation across diverse clinical scenarios.

B. Hospital course summaries contradicting important facts

Score: 12/100 (95% CI: 5-18) (lower values better)

Feedback regarding the inconsistencies:

Hospital course inconsistencies commonly arise from direct therapeutic or diagnostic contradictions, omission of critical details, and temporal or interpretive mismatches. For example, Patient 0’surgical drain status conflicts between “removed” (summary) and “left at discharge” (fact), while Patient 6’s insulin regimen discrepancies (Levemir 16 BID vs. Lantus 15 BID) reflect medication/dose errors. Diagnostic contradictions include Patient 12’s gallbladder fistula (summary: “no definitive fistula” vs. fact: “found”), Patient 15’s principal diagnosis (summary omits viral gastroenteritis), and Patient 21’s fracture timing (summary: “new” vs. fact: “pre-existing”). Omissions are evident in Patient 9’s missing cipro/finasteride documentation and Patient 29’s unmentioned deferral of cancer management. Interpretive mismatches occur in Patient 26 (SMA stenosis attributed to gastroparesis vs. mesenteric ischemia) and Patient 18’s hypokalemia (fact: “resolved” vs. summary: ongoing supplementation). Temporal inconsistencies also appear in Patient 18’s “persistent fever” (fact) versus “afebrile at discharge” (summary). These patterns highlight errors in documentation specificity, reconciliation of imaging/clinical findings, and medication/diagnosis attribution.

Detailed results

A. Presence of important facts

Patient Fact 0 Fact 1 Fact 2 Majority Vote
claude-3.5-sonnet-v2 llama4-scout deepseek-r1 claude-3.5-sonnet-v2 llama4-scout deepseek-r1 claude-3.5-sonnet-v2 llama4-scout deepseek-r1 F0 F1 F2
Patient 0 Yes Yes Yes No No No No No No Yes No No
Patient 1 Yes No No Yes Yes Yes Yes Yes Yes No Yes Yes
Patient 2 Yes Yes Yes No Yes No No No No Yes No No
Patient 3 Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes Yes
Patient 4 Yes Yes Yes Yes Yes Yes No No No Yes Yes No
Patient 5 Yes Yes Yes Yes Yes Yes No No No Yes Yes No
Patient 6 No No No No No No Yes Yes Yes No No Yes
Patient 7 No No No No No No Yes Yes Yes No No Yes
Patient 8 No Yes No No No No Yes Yes Yes No No Yes
Patient 9 No No No Yes Yes Yes No No No No Yes No
Patient 10 Yes No No Yes Yes Yes No No No No Yes No
Patient 11 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Patient 12 Yes Yes Yes No No No No No No Yes No No
Patient 13 Yes Yes No Yes Yes Yes No Yes No Yes Yes No
Patient 14 Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes
Patient 15 No No No Yes Yes No No Yes No No Yes No
Patient 16 Yes Yes No No No No No No No Yes No No
Patient 17 No Yes No Yes Yes Yes Yes Yes Yes No Yes Yes
Patient 18 No Yes No Yes Yes No Yes Yes No No Yes Yes
Patient 19 No No No Yes Yes Yes No Yes No No Yes No
Patient 20 Yes Yes Yes Yes Yes No No No No Yes Yes No
Patient 21 No Yes No No No No No No No No No No
Patient 22 Yes Yes Yes No No No No No No Yes No No
Patient 23 Yes Yes Yes Yes No No No No No Yes No No
Patient 24 Yes Yes Yes Yes Yes No No Yes No Yes Yes No
Patient 25 Yes Yes Yes No No No No No No Yes No No
Patient 26 Yes Yes Yes No No No No Yes No Yes No No
Patient 27 Yes No No Yes No Yes No Yes No No Yes No
Patient 28 Yes Yes Yes No Yes Yes No No No Yes Yes No
Patient 29 Yes Yes Yes No Yes No No No No Yes No No

All Missing Facts and Explanations

1. Patient 0, Fact 1:

2. Patient 0, Fact 2:

3. Patient 1, Fact 0:

4. Patient 2, Fact 1:

5. Patient 2, Fact 2:

6. Patient 3, Fact 1:

7. Patient 4, Fact 2:

8. Patient 5, Fact 2:

9. Patient 6, Fact 0:

10. Patient 6, Fact 1:

11. Patient 7, Fact 0:

12. Patient 7, Fact 1:

13. Patient 8, Fact 0:

14. Patient 8, Fact 1:

15. Patient 9, Fact 0:

16. Patient 9, Fact 2:

17. Patient 10, Fact 0:

18. Patient 10, Fact 2:

19. Patient 12, Fact 1:

20. Patient 12, Fact 2:

21. Patient 13, Fact 0:

22. Patient 13, Fact 2:

23. Patient 14, Fact 0:

24. Patient 15, Fact 0:

25. Patient 15, Fact 1:

26. Patient 15, Fact 2:

27. Patient 16, Fact 0:

28. Patient 16, Fact 1:

29. Patient 16, Fact 2:

30. Patient 17, Fact 0:

31. Patient 18, Fact 0:

32. Patient 18, Fact 1:

33. Patient 18, Fact 2:

34. Patient 19, Fact 0:

35. Patient 19, Fact 2:

36. Patient 20, Fact 1:

37. Patient 20, Fact 2:

38. Patient 21, Fact 0:

39. Patient 21, Fact 1:

40. Patient 21, Fact 2:

41. Patient 22, Fact 1:

42. Patient 22, Fact 2:

43. Patient 23, Fact 1:

44. Patient 23, Fact 2:

45. Patient 24, Fact 1:

46. Patient 24, Fact 2:

47. Patient 25, Fact 1:

48. Patient 25, Fact 2:

49. Patient 26, Fact 1:

50. Patient 26, Fact 2:

51. Patient 27, Fact 0:

52. Patient 27, Fact 1:

53. Patient 27, Fact 2:

54. Patient 28, Fact 1:

55. Patient 28, Fact 2:

56. Patient 29, Fact 1:

57. Patient 29, Fact 2:

B. Hospital course summaries contradicting important facts

Patient Fact 0 Fact 1 Fact 2 Majority Vote
claude-3.5-sonnet-v2 llama4-scout deepseek-r1 claude-3.5-sonnet-v2 llama4-scout deepseek-r1 claude-3.5-sonnet-v2 llama4-scout deepseek-r1 F0 F1 F2
Patient 0 No No No Yes Yes Yes No No No No Yes No
Patient 1 No No Yes No No No No No No No No No
Patient 2 No No No No No Yes No No No No No No
Patient 3 No No No No No Yes No No No No No No
Patient 4 No No No No No No No No Yes No No No
Patient 5 No No No No No No No No No No No No
Patient 6 No Yes No Yes Yes Yes No No No No Yes No
Patient 7 No No Yes No No No No No Yes No No No
Patient 8 No No Yes No No Yes No No No No No No
Patient 9 Yes Yes Yes No No No No No No Yes No No
Patient 10 No No Yes No No No No No No No No No
Patient 11 No No No No No No No No No No No No
Patient 12 No Yes Yes No No No No No No Yes No No
Patient 13 No No No No No No Yes Yes Yes No No Yes
Patient 14 No No No No No No No No No No No No
Patient 15 Yes No Yes No No No No No Yes Yes No No
Patient 16 No No No No No No No No No No No No
Patient 17 No No Yes No No No No No No No No No
Patient 18 Yes Yes Yes No No Yes Yes Yes Yes Yes No Yes
Patient 19 Yes No No No No No No No No No No No
Patient 20 No No Yes No No Yes No No No No No No
Patient 21 No No Yes Yes Yes Yes No No No No Yes No
Patient 22 No No Yes No No Yes No No Yes No No No
Patient 23 No No Yes No No Yes No No Yes No No No
Patient 24 No No No No No No No No No No No No
Patient 25 No No No No No No No No No No No No
Patient 26 No No No No No Yes Yes No Yes No No Yes
Patient 27 No No Yes No No No No No No No No No
Patient 28 No No No No No No No No Yes No No No
Patient 29 No No No Yes Yes No No No Yes No Yes No

All Inconsistent Facts and Explanations

1. Patient 0, Fact 1:

2. Patient 1, Fact 0:

3. Patient 2, Fact 1:

4. Patient 3, Fact 1:

5. Patient 4, Fact 2:

6. Patient 6, Fact 0:

7. Patient 6, Fact 1:

8. Patient 7, Fact 0:

9. Patient 7, Fact 2:

10. Patient 8, Fact 0:

11. Patient 8, Fact 1:

12. Patient 9, Fact 0:

13. Patient 10, Fact 0:

14. Patient 12, Fact 0:

15. Patient 13, Fact 2:

16. Patient 15, Fact 0:

17. Patient 15, Fact 2:

18. Patient 17, Fact 0:

19. Patient 18, Fact 0:

20. Patient 18, Fact 1:

21. Patient 18, Fact 2:

22. Patient 19, Fact 0:

23. Patient 20, Fact 0:

24. Patient 20, Fact 1:

25. Patient 21, Fact 0:

26. Patient 21, Fact 1:

27. Patient 22, Fact 0:

28. Patient 22, Fact 1:

29. Patient 22, Fact 2:

30. Patient 23, Fact 0:

31. Patient 23, Fact 1:

32. Patient 23, Fact 2:

33. Patient 26, Fact 1:

34. Patient 26, Fact 2:

35. Patient 27, Fact 0:

36. Patient 28, Fact 2:

37. Patient 29, Fact 1:

38. Patient 29, Fact 2: