# Medical Record Review Prompts for Expert Witnesses > Seven prompts for medical expert witnesses: opinion basis tables, contrary evidence, methodology support, opposing expert analysis and cross exposure. Canonical page: https://medrecords.ai/content-hub/prompt-library/medical-record-review-prompts-for-expert-witnesses/ --- [Content Hub](https://medrecords.ai/content-hub/) › [Prompt library](https://medrecords.ai/content-hub/prompt-library/) › Medical record review prompts for expert witnesses Prompt library ## Medical record review prompts for expert witnesses Medical record review prompts for expert witnesses build the opinion basis table and, in the same pass, the contrary evidence inventory. Run them as a pair and compare the lengths. A short contrary list is usually a prompt problem rather than a record that happens to agree with you. By [Ahmed Jemaa](https://medrecords.ai/authors/ahmed-jemaa/) , Co-Founder & CEO of Medrecords AI · Published 8 Sep 2026 The job Form an opinion the record actually supports, know every record that cuts against it, and be ready to defend both under cross-examination. Produces Expert report, deposition preparation, response to an opposing expert. Spine priority S5, S7, S9, S10, S12, S16, S18, S19, S21. Run order Spine, then E1 and E2 as a pair, then E6, then E3 and E7 for preparation. E4 and E5 when the opposing report arrives. ### The prompts Run these after the spine prompts marked priority above. Placeholders in angle brackets need filling in. E1 #### Opinion basis table For each opinion to be offered, list every record supporting it with verbatim quotation and citation. One table per opinion. E2 #### Contrary evidence inventory For each opinion, identify every record that contradicts or complicates it. Quote and cite each. Do not omit or soften. Run E1 and E2 together, always, and compare their lengths. E1 and E2 are a matched pair, and an E2 that is much shorter than E1 is more likely a prompt problem than a record that happens to agree with you. E3 #### Strongest challenge Identify the strongest record-based challenge to each opinion, and the record evidence available to answer it. E4 #### Opposing expert basis Extract the opposing expert's stated opinions and the factual basis given for each. Locate the underlying records for each stated basis. Identify any stated basis with no supporting record. The third sentence is where this earns its place. An asserted basis with no record behind it is the most useful thing you can find in an opposing report. E5 #### Unaddressed record check Identify records in this set that the report does not address, and assess whether any of them are material to the opinions offered. Run against the opposing report. Then run it against your own, before it is final. E6 #### Methodology support For each opinion, identify the documented findings the methodology relies on and confirm each is present in the record with citation. Flag any inferential step not supported by a document. This is the admissibility prompt. Every flagged inferential step is a question you will be asked. E7 #### Deposition exposure Identify the record facts most likely to be used on cross: inconsistencies, gaps, contrary findings, and timeline problems. For each, give the record citation and the available response. **Lane failure mode: the fabricated citation, at maximum cost.** An expert report is the worst possible place for failure mode 5.1. Every citation in a report gets Tier 1 verification, opened and read at the page, with no sampling. A quotation you have not personally confirmed is a representation you cannot make. **Professional responsibility.** Two federal provisions bear on this directly. Rule 26(a)(2)(B) requires a written report containing, among other things, "a complete statement of all opinions the witness will express and the basis and reasons for them" and "the facts or data considered by the witness in forming them." Whether an AI-assisted review process falls inside "facts or data considered" is not settled, and it is a question for your retaining counsel rather than for you. What is settled is that the underlying records are facts or data considered, and they are disclosable whether a machine or a clerk pulled them. Rule 702, as amended in 2023, is the sharper instrument. The proponent must now demonstrate to the court that it is more likely than not that each requirement is met, which the amendment makes clear is a Rule 104(a) admissibility question rather than a matter of weight for the jury. Two of the four requirements land on record review: subsection (b), that the testimony is based on sufficient facts or data, and subsection (d), that the opinion reflects a reliable application of the principles and methods to the facts of the case. The practical consequence is that an unverified citation is not a tidiness problem. It is a defect in the "sufficient facts or data" showing, raised at a stage where the court decides rather than the jury. This is why E6 exists and why this lane carries a Tier 1, every-citation, no-sampling rule. The defensible posture is unchanged and simple to state: the opinion is yours, every record supporting it was verified at the source by you, and you can describe how you reviewed the file. State practice varies, and not every state follows the amended federal standard. ### Lane failure mode: the fabricated citation, at maximum cost An expert report is the worst possible place for failure mode 5.1. Every citation in a report gets Tier 1 verification, opened and read at the page, with no sampling. A quotation you have not personally confirmed is a representation you cannot make. ### Professional responsibility Two federal provisions bear on this directly. Rule 26(a)(2)(B) requires a written report containing, among other things, "a complete statement of all opinions the witness will express and the basis and reasons for them" and "the facts or data considered by the witness in forming them." Whether an AI-assisted review process falls inside "facts or data considered" is not settled, and it is a question for your retaining counsel rather than for you. What is settled is that the underlying records are facts or data considered, and they are disclosable whether a machine or a clerk pulled them. Rule 702, as amended in 2023, is the sharper instrument. The proponent must now demonstrate to the court that it is more likely than not that each requirement is met, which the amendment makes clear is a Rule 104(a) admissibility question rather than a matter of weight for the jury. Two of the four requirements land on record review: subsection (b), that the testimony is based on sufficient facts or data, and subsection (d), that the opinion reflects a reliable application of the principles and methods to the facts of the case. The practical consequence is that an unverified citation is not a tidiness problem. It is a defect in the "sufficient facts or data" showing, raised at a stage where the court decides rather than the jury. This is why E6 exists and why this lane carries a Tier 1, every-citation, no-sampling rule. The defensible posture is unchanged and simple to state: the opinion is yours, every record supporting it was verified at the source by you, and you can describe how you reviewed the file. State practice varies, and not every state follows the amended federal standard. *Issues to raise, not answers. Requirements vary by jurisdiction, and none of this is legal advice.* #### Run these on a file you already have Send one real record set. You get back a cited chronology, and you review it, you revise it, you sign it. [Test a file](https://medrecords.ai/test-a-file/?src=content-hub-prompt-library-medical-record-review-prompts-for-expert-witnesses) · [Book a demo](https://medrecords.ai/demo/?src=content-hub-prompt-library-medical-record-review-prompts-for-expert-witnesses) Last verified: 2026-09-08 · [← All 71 prompts](https://medrecords.ai/content-hub/prompt-library/)