BREAKING THE BOTTLENECK: A NOVEL HUMAN+AI PROCESS FOR GENERATING FULLY-ANNOTATED, FOUNDATIONAL-DRAFT PRO MANUSCRIPTS FROM SOURCE MATERIALS IN UNDER 48 HOURS
Author(s)
Walter Bender, ., Alex Sterling, MD, PhD, Meghan A. Berryman, PhD.
Sorcero, Inc, Washington, DC, USA.
Sorcero, Inc, Washington, DC, USA.
OBJECTIVES: The timely dissemination of patient-reported outcomes (PRO) is both an economic and ethical imperative, yet PRO manuscripts are a critical bottleneck in the publication pipeline. Reliance on third-party vendors often results in lengthy turn-around times; generic AI tools have problems with trust, because of risks of bias, factual inaccuracy, inconsistent quality, and lack of regulatory compliance. To address this, we developed a bespoke Human+AI collaborative process engineered for the rapid generation of scientifically robust, secondary PRO manuscripts.
METHODS: The input to the process are unstructured source materials, including CSP, SAP, and TFL. The process combines a hybrid retrieval-augmented generation (RAG) framework, which grounds all AI-generated output in verified source documents, with a proprietary definitive AI scoring rubric (DAISTM) to ensure data fidelity, elimination of hallucinations, and adherence to industry and regulatory guidelines, including CONSORT-2025, CONSORT-PRO, GPP22, and data privacy laws. A core component of the process is a mandatory review by a subject-matter expert (SME) to ensure scientific integrity before the draft is delivered.
RESULTS: The process has been applied repeatedly to deliver a fully-formed foundational draft, complete with abstract, plain-language summary, source-traced data points, full tables, design explanations for figures, and a transparently constructed and auditable reference list (including an associated RIS file). These high-quality, foundational-draft secondary PRO manuscripts are generated in under 48 hours. Trust is enhanced because the process reproduces results consistently and includes source traceability, immutable audit trails, transparency and explainability, and benchmarking and metrics.
CONCLUSIONS: This capability directly addresses the challenge of timely dissemination of PROs to decision-making bodies by accelerating the high-quality generation of these critical, value-evidence publications. To formally quantify its impact on time-to-publish and quality, a rigorous comparative validation against traditional human-only and generic LLM methods is currently underway, with initial results expected in early 2026.
METHODS: The input to the process are unstructured source materials, including CSP, SAP, and TFL. The process combines a hybrid retrieval-augmented generation (RAG) framework, which grounds all AI-generated output in verified source documents, with a proprietary definitive AI scoring rubric (DAISTM) to ensure data fidelity, elimination of hallucinations, and adherence to industry and regulatory guidelines, including CONSORT-2025, CONSORT-PRO, GPP22, and data privacy laws. A core component of the process is a mandatory review by a subject-matter expert (SME) to ensure scientific integrity before the draft is delivered.
RESULTS: The process has been applied repeatedly to deliver a fully-formed foundational draft, complete with abstract, plain-language summary, source-traced data points, full tables, design explanations for figures, and a transparently constructed and auditable reference list (including an associated RIS file). These high-quality, foundational-draft secondary PRO manuscripts are generated in under 48 hours. Trust is enhanced because the process reproduces results consistently and includes source traceability, immutable audit trails, transparency and explainability, and benchmarking and metrics.
CONCLUSIONS: This capability directly addresses the challenge of timely dissemination of PROs to decision-making bodies by accelerating the high-quality generation of these critical, value-evidence publications. To formally quantify its impact on time-to-publish and quality, a rigorous comparative validation against traditional human-only and generic LLM methods is currently underway, with initial results expected in early 2026.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
MSR198
Topic
Methodological & Statistical Research
Topic Subcategory
PRO & Related Methods
Disease
No Additional Disease & Conditions/Specialized Treatment Areas