National Insurance Institute of Israel
Fair Research Organization
n clinical trials, outcomes are often clear. In real-world practice, they rarely are.
- Why did this particular patient respond so well, and what does that imply for expected benefit in routine care?
- Why did another patient with apparently similar characteristics not respond at all, and is this a signal of underperformance or of modifiable implementation gaps in routine care?
- Was this adverse event truly drug-related, and to what extent should it affect risk communication and payer confidence?
- Or was it predominantly driven by underlying disease risk and system-level factors beyond the product itself?
Real-world evidence (RWE) has expanded our visibility into treatment performance and is now central to payer and health-system decision-making. Yet it has also exposed a gap: we can describe outcomes, but we often struggle to explain them proportionally and coherently at the individual level, in a way that is both scientifically credible and operationally useful for access discussions.
Across therapeutic areas, identical outcomes can reflect very different causal realities, and therefore very different implications for value, implementation, and accountability between the product and the health system.
When the Same Outcome Means Different Things
Consider obesity management with a GLP-1 drug. In real-world cohorts, weight loss commonly ranges between 4 % and 12 % at six to 12 months. Suppose two patients both achieve 12 % weight loss at nine months. On paper, the outcome is identical. In practice, the underlying drivers may differ substantially.
- One patient maintained high adherence, engaged in structured follow-up, and received lifestyle counseling embedded in a multidisciplinary care pathway.
- The other had irregular follow-up, inconsistent adherence, and minimal behavioral support.
The pharmacology is the same. The context is not. For Medical Affairs and Value & Access, this distinction matters: the same “headline” outcome may represent a predominantly pharmacological effect in one case and a fragile, context-dependent result in another.
In practical terms, identical headline outcomes can correspond to very different underlying value stories. In a highly structured program, payers may reasonably expect this level of benefit to be reproducible at scale; in a fragmented setting, the same outcome may be difficult to generalize without investing in adherence support and care coordination.
Now consider biologic therapy in psoriasis. Patients with moderate-to-severe psoriasis carry elevated baseline risks of serious infections and cardiovascular events, even in the absence of biologic exposure. When a patient develops a serious infection during biologic treatment, the immediate reaction may be to attribute the event to the drug.
But is the event predominantly drug-driven? Disease-driven? Or the result of multiple interacting factors, including comorbidities, concomitant treatments, and background systemic inflammation?
Traditional pharmacovigilance tools often rely heavily on temporal association. They are less equipped to account for baseline disease risk or multifactorial causation. As a result, safety events may be overattributed to the product, insufficiently contextualized, and communicated in a way that can unintentionally undermine payer and prescriber confidence.
In both effectiveness and safety, the challenge is the same: We observe an outcome, but we lack a shared, proportionate language to distinguish what is reasonably attributable to the medicine from what reflects patient behavior, care pathways, or underlying disease risk. This is precisely where an attribution lens can support more balanced communication, more nuanced value discussions, and more defensible benefit-risk narratives.
From Description to Attribution
Medical Affairs operates at the intersection of science, communication, and access. The goal is not only to report outcomes but to interpret them responsibly in ways that inform prescribing, contracting, and safety decision-making. An attribution-oriented approach reframes the question. Instead of asking only “What happened?,” it asks, “Why did this happen in this patient, under these conditions, and what does that imply for future decisions?”
Concretely, this means structuring interpretation around a small set of recurring questions:
- What were the relevant contributing factors behind this outcome (for example, pharmacology, patient behavior, care pathway, background disease risk)?
- How strong is the expected association of each factor with the outcome, based on available evidence and clinical judgment?
- How might the observed result reflect the interaction between the medicine and its real-world context, rather than either one in isolation?
In obesity treatment, this means recognizing that real-world weight loss reflects more than molecular effect alone. Persistence, early response, behavioral support, and structured care pathways often amplify or attenuate the realized benefit. This does not dilute the drug’s value; on the contrary, it clarifies how pharmacological efficacy is translated into real-world effectiveness and which parts of that effectiveness are realistically modifiable through system-level interventions.
A simple, nontechnical way to make this visible is to speak in proportional terms. For a patient achieving 12 % weight loss, Medical Affairs might communicate that the outcome plausibly reflects a major pharmacological contribution (for example, “around two-thirds”) and a meaningful contextual share (for example, “around one-third” linked to adherence and structured care), instead of attributing the entire result to one single driver. This kind of language preserves scientific caution while giving payers and clinicians a more intuitive sense of how much of the observed value is dependent on implementation.
In psoriasis pharmacovigilance, attribution-based reasoning highlights that adverse events occur within a background of elevated systemic inflammation and comorbidity burden. Incorporating baseline risk into interpretation reduces the risk of oversimplified “drug-or-not” conclusions and supports more balanced safety communications with regulators, payers, and clinicians. Rather than presenting events as binary (drug-related or not), an attribution lens encourages gradations of contribution, which can be discussed in terms of relative likelihood and context.
This article refers to this as a “practical attribution lens”: a pragmatic way to make visible the relative contributions of pharmacology and context behind individual real-world outcomes, without requiring complex mathematics. For Medical Affairs and Value & Access teams, adopting such a lens turns real-world data from a static description into a more actionable explanation that can guide communication strategies, access negotiations, and post-approval evidence generation plans.
Importantly, this attribution lens does not require complex modeling to be operationally useful. It begins as a structured way of thinking and communicating about outcomes in proportional terms. In settings where more precision is needed, quantitative methods can support this reasoning—but the first step is conceptual clarity rather than mathematical complexity. The underlying logic is consistent with multifactor attribution principles used in other high-accountability domains, such as medical-legal evaluation and risk analysis, where outcomes must be transparently partitioned across interacting drivers.
Why This Matters for Medical Affairs
Medical Affairs teams are increasingly asked to engage in evidence interpretation, not merely evidence transmission. They sit at the interface between clinical data, payer expectations, and real-world practice, and are often the first to be challenged on “why” a medicine appears to perform differently across patients, sites, or health systems. A structured attribution lens offers practical advantages in this role.
Clearer external communication: When discussing heterogeneous outcomes with healthcare professionals or payers, proportionate explanations build credibility. They acknowledge complexity without undermining confidence in the therapy by showing how pharmacology, behavior, and care organization jointly shape results rather than competing for the entire “share” of the outcome.
Better internal alignment: Cross-functional discussions between Medical, Market Access, Health Economics, and Pharmacovigilance teams often stall when narratives diverge: one group emphasizing product performance, another highlighting system gaps, another focusing on safety signals. A shared causal framework, using the same attribution language, enables more coherent dialogue and more consistent messaging across functions and geographies.
Identification of actionable levers: If persistence and structured follow-up meaningfully influence outcomes, then patient education programs, adherence support, and care pathway optimization are not ancillary activities—they are extensions of therapeutic impact. An attribution lens helps Medical Affairs prioritize which levers are likely to yield the greatest incremental benefit, and to explain this to both internal decision-makers and external partners.
Stronger medical-legal and regulatory defensibility: By documenting that interpretation of outcomes explicitly considers baseline risk, contextual factors, and plausible contribution shares, Medical Affairs can support more robust benefit-risk narratives. This is particularly relevant in complex safety cases, where oversimplified attribution can either unfairly penalize a product or under-recognize meaningful risk.
Implications for Value & Access
Payers increasingly evaluate therapies based on real-world performance and long-term system impact, rather than trial results alone. If effectiveness varies depending on adherence, care structure, and patient support, then value discussions must distinguish between intrinsic pharmacological performance and implementation-dependent amplification. This distinction is critical for fair and sustainable access decisions.
In practical terms, this helps distinguish intrinsic product performance from implementation-dependent performance—an increasingly relevant distinction as payers scrutinize real-world evidence more closely.
For outcome-based or risk-sharing agreements, an attribution lens can help clarify which portion of an observed real-world outcome is plausibly related to the pharmacological effect of the medicine and which portion reflects implementation factors such as adherence, care pathways, or behavioral support. Although payer contracts typically evaluate outcomes at the population level, explicitly recognizing these components can help set more realistic expectations regarding what a medicine can deliver under minimal versus optimized implementation conditions.
In practice, observed results reflect the joint effect of the medicine and the care context in which it is used. Framing outcomes in this way may support more nuanced discussions of “success” in value-based agreements, particularly when interpreting heterogeneous real-world outcomes.
At the health-system level, attribution-based interpretation reframes underperformance. Instead of asking “Does the drug underperform in the real world?”, the more productive question becomes “Under which conditions does the drug realize its full potential, and what would it take to create those conditions at scale?” This shift enables more constructive engagement with health systems, aligning discussions about access with conversations about care pathways, adherence infrastructure, and data quality.
Finally, integrating attribution language into payer and HTA interactions can help align expectations over time. As new real-world evidence accumulates, Medical Affairs and Value & Access teams can revisit initial assumptions about contribution shares, update them transparently, and co-create adjustments to coverage or implementation strategies, rather than reacting to raw outcome metrics in isolation.
Rethinking Safety Communication
In chronic inflammatory diseases, baseline risk complicates safety interpretation. For psoriasis, systemic inflammation contributes independently to cardiovascular and infection risk, meaning that serious events can occur even in the absence of biologic exposure. Without accounting for this baseline, adverse events may appear disproportionately linked to treatment, especially when they occur soon after initiation.
A more structured attribution approach explicitly asks how much of the observed event is plausibly explained by underlying disease, comorbidities, and concomitant therapies, and how much incremental risk is reasonably attributable to the product. Rather than relying primarily on temporal association, Medical Affairs and Pharmacovigilance can frame safety narratives in terms of relative contribution and context, making room for gradations instead of binary labels.
This has several practical consequences. It encourages more proportionate risk communication, avoiding both under-recognition of genuine signals and overattribution that may unnecessarily erode prescriber and payer confidence. It strengthens medical-legal defensibility by documenting the reasoning behind attribution, including explicit consideration of baseline risk. It also enhances regulatory dialogue by offering a transparent, repeatable way to discuss complex cases where multiple factors are at play.
Ultimately, an attribution lens allows safety teams to communicate not only whether an event could be related to treatment, but also how strongly and under which conditions, and with what implications for monitoring, risk minimization, and access.
Toward a More Mature Real-World Evidence Culture
Real-world evidence has matured significantly over the past decade, but much of its interpretation remains average-based and siloed by function. Medical Affairs is uniquely positioned to help the field move from description—“what we observe on average”—toward explanation: “why different patients and systems see different outcomes, and what we can do about it.”
By adopting a more explicit attribution mindset, teams can explain heterogeneity more coherently, connect clinical outcomes to system-level determinants, and integrate effectiveness and safety narratives within a single causal framework. This fosters a culture in which variation is not treated as noise to be averaged away, but as a source of insight about how medicines interact with real-world practice.
In day-to-day work, this can translate into small but meaningful changes: framing slide decks around “shares” of contribution instead of single-point estimates, structuring cross-functional reviews around a common set of attribution questions, and embedding attribution thinking into payer discussions, pharmacovigilance assessments, and post-approval evidence plans. Over time, such practices can elevate the quality of scientific exchange and support more resilient, trust-based relationships between industry, payers, regulators, and clinicians.
Three Practical First Steps for Medical Affairs Teams
- Reframe outcome discussions in proportional language, distinguishing pharmacological contribution from contextual amplification.
- Explicitly integrate baseline risk into safety narratives, especially in chronic diseases with elevated background event rates.
- Align cross-functional reviews (Medical, Market Access, Pharmacovigilance) around shared attribution questions rather than isolated performance metrics.
These shifts do not necessarily require new analytical infrastructure. In most organizations, the necessary elements—baseline risk estimates, treatment effect data, and contextual variables such as adherence or care pathway characteristics—are already generated through routine real-world evidence activities. The main change lies in how these elements are interpreted and communicated.
As Medical Affairs continues to evolve from data presenter to evidence interpreter, the ability to explain why outcomes occur—not merely what occurred—may become one of its most strategic capabilities. An attribution lens offers a practical way to build that capability step by step within existing RWE and Medical Affairs processes.
Although a fully integrated framework for individual-level attribution in real-world medicine is still emerging, several elements of this approach are already being explored in pharmaceutical research. For example, analyses of real-world outcomes with GLP-1 receptor agonists such as semaglutide have illustrated how identical observed results—such as substantial weight loss—can arise from different combinations of pharmacological effects and contextual amplifiers including persistence, care intensity, and behavioral support.
Similarly, ongoing collaborative work with clinical teams in Israel is exploring attribution-oriented approaches in real-world analyses of biologic therapies for psoriasis, where treatment response reflects the interaction between pharmacology, baseline disease risk, and care context.
More broadly, related reasoning already underlies several established practices in pharmaceutical evaluation. Pharmacovigilance assessments routinely interpret observed safety events relative to expected background incidence, while health technology assessment increasingly integrates treatment effects with baseline risk and contextual modifiers when evaluating net treatment benefit.
The framework proposed here builds on these analytical traditions. Rather than introducing a new regulatory methodology, it translates these attribution principles into an operational interpretive language that Medical Affairs teams can use when communicating heterogeneous real-world outcomes across pharmacovigilance, regulatory, payer, and scientific dialogue contexts.
These exploratory applications suggest that attribution-based interpretation may represent a natural next step in the evolution of real-world evidence analysis.