Integrating stakeholder perspectives in modeling routine data for therapeutic decision-making

This paper proposes a stakeholder-inclusive conceptual framework that aligns the diverse priorities of clinicians, industry, patients, and statisticians with methodological requirements for modeling routine health data, utilizing multistate models to generate robust, transparent, and clinically meaningful real-world evidence for therapeutic decision-making.

The Gist

Imagine you are trying to build a giant, high-tech map of a city to help people decide the best route to take. This map isn't drawn from scratch; it's built using millions of tiny, everyday notes left behind by people walking, driving, and living their lives (this is what the paper calls "routine health data").

The problem is that everyone who needs this map wants to see something different, and if you don't listen to them, the map becomes useless.

The Cast of Characters (The Stakeholders)

Think of the people involved in making this map as four different groups standing around a table, each holding a different pair of glasses:

1. The Doctors (Clinicians): They wear glasses that zoom in on clarity. They need the map to be easy to read so they can instantly tell a patient, "If you take this path, here is exactly what happens next." They don't care about complex math; they care about whether the route makes sense for a human body.
2. The Pharma Companies: They wear glasses focused on rules and proof. They need the map to be legally perfect so they can show regulators, "See? This route is safe and effective." They are worried about following the rulebook to the letter.
3. The Patient Groups: They wear glasses focused on trust and privacy. They want to make sure the map doesn't reveal who they are (privacy) and that it includes their personal feelings and stories (like "I felt tired," not just "my heart rate dropped"). They want to know their voice is heard.
4. The Statisticians (The Map Makers): They wear glasses focused on accuracy and logic. They are the engineers ensuring the map doesn't have holes, that the roads connect correctly, and that no one is tricked by a trick of the light (bias). They worry about the math being perfect.

The Problem

In the past, these groups often worked in silos. The engineers built a mathematically perfect map that was too confusing for doctors. The doctors drew a simple map that didn't satisfy the rulebook. The patient's story was left out entirely. The result? A map that no one could fully trust or use effectively.

The Solution: A "Group Huddle" Framework

The authors of this paper suggest a new way to build the map: Start with a group huddle.

Before drawing a single line, they propose a framework where everyone sits down and says, "Here is what I need to see."

• The doctors say, "Show me the journey, not just the destination."
• The patients say, "Include how I feel along the way."
• The companies say, "Make sure it follows the law."
• The statisticians say, "I'll make sure the math holds it all together."

The Magic Tool: The "Multistate Model"

To make this work, the paper suggests using a specific tool called a Multistate Model.

Think of this not as a straight line from Point A to Point B, but as a dynamic video game.

• In a boring old map, you just go from "Sick" to "Cured."
• In this new "video game" map, you can see every level you pass through. You can see the player getting stuck in a "Side Quest" (a temporary symptom), moving to a "Bonus Level" (feeling better), or taking a "Shortcut" (a new treatment).

This tool is special because:

• It satisfies the Doctors because it shows the whole story, not just the final score.
• It satisfies the Statisticians because it handles complex movements without breaking the math.
• It satisfies the Patients because it captures the ups and downs of their real life.
• It satisfies the Companies because it provides a rigorous, rule-abiding record of the journey.

The Bottom Line

The paper concludes that if you want to use real-life data to make life-or-death decisions about medicine, you can't just let the mathematicians build the map in a basement. You have to invite the doctors, the patients, and the regulators to the drawing board from day one.

By using this "group huddle" approach and tools like the "video game map," we can create evidence that is scientifically strong (the math works), clinically useful (doctors can use it), and socially fair (patients feel safe and heard). It turns a confusing pile of data into a clear, trustworthy guide for everyone.

Technical Summary

Technical Summary: Integrating Stakeholder Perspectives in Modeling Routine Data for Therapeutic Decision-Making

1. Problem Statement

The increasing reliance on routinely collected health data to generate Real-World Evidence (RWE) for therapeutic decision-making faces a critical challenge: the misalignment of priorities among key stakeholders. Currently, analyses often suffer from fragmentation because they fail to explicitly account for the divergent needs of:

• Clinicians: Who require clinical relevance and interpretability.
• Pharmaceutical Industry: Who prioritize regulatory compliance and robust RWE generation.
• Patient Advocacy Groups: Who demand transparency, the inclusion of Patient-Reported Outcomes (PROs), and strict privacy protection.
• Statisticians: Who focus on bias control and methodological rigor.

Without a structured approach to integrate these perspectives, generated evidence risks being scientifically sound but clinically irrelevant, or transparent but methodologically flawed, ultimately undermining the utility of routine data in decision-making.

2. Methodology

The authors developed a stakeholder-inclusive conceptual framework through an interdisciplinary approach:

• Data Collection: Insights were gathered via an interdisciplinary workshop involving representatives from the aforementioned stakeholder groups.
• Literature Synthesis: The framework was further supported and validated by targeted literature examples.
• Framework Construction: The core methodology involves mapping specific stakeholder priorities to corresponding methodological requirements and identifying analytical strategies capable of integrating these diverse needs into a unified modeling approach.

3. Key Contributions

The paper's primary contribution is the formalization of a framework that bridges the gap between social/clinical expectations and statistical methodology. Key contributions include:

• Priority Mapping: A clear delineation of how distinct stakeholder groups prioritize different aspects of data quality and analysis.
• Operationalization of Perspectives: A guide on how to translate abstract stakeholder needs (e.g., "transparency" or "clinical relevance") into concrete statistical modeling choices.
• Role Definition for Statisticians: A nuanced definition of the statistician's role, ranging from providing consultation for standard analyses to adapting advanced methods or developing novel methodologies for complex, specific research questions.

4. Results

The study identified specific priority clusters and demonstrated how a Multistate Model (MSM) serves as a versatile methodological tool to satisfy these competing demands:

• Stakeholder Priorities Identified:

  • Clinicians: Interpretability and clinical relevance.
  • Industry: Regulatory compliance and RWE generation.
  • Patient Groups: Transparency, PRO inclusion, and privacy.
  • Statisticians: Bias control and rigor.

• Methodological Solution (Multistate Models):
  The framework illustrates that Multistate Models effectively operationalize these requirements by:

  1. Capturing Dynamic Trajectories: Modeling disease progression as a series of transitions between states, which aligns with clinical reality.
  2. Integrating Intermediate Outcomes: Allowing for the inclusion of PROs and other intermediate endpoints valued by patients and clinicians.
  3. Graphical Interpretability: Providing visual representations of disease pathways that enhance transparency and understanding for non-statisticians.
  4. Rigorous Bias Control: Offering a structured statistical environment to control for confounding and selection bias, satisfying the need for methodological rigor.

5. Significance

This work provides a foundational methodological guidance for the design of routine health data analyses. Its significance lies in:

• Holistic Evidence Generation: It shifts the paradigm from siloed analysis to a collaborative approach where research questions are aligned with intended perspectives from the outset.
• Robustness and Transparency: By integrating diverse needs, the resulting evidence is more likely to be scientifically rigorous, clinically meaningful, and socially acceptable.
• Scalability: The framework positions Multistate Models not just as a specific statistical technique, but as a flexible tool to operationalize stakeholder integration, potentially applicable across various therapeutic areas and regulatory contexts.

In conclusion, the paper argues that successful therapeutic decision-making using routine data requires a deliberate, framework-driven integration of stakeholder perspectives, with Multistate Models serving as a key technical vehicle to achieve this synthesis.