SPEAKER:
John Flemming, VP Commercial Insights & Analytics, OptimizeRx
KEY TAKEAWAYS:
- Predictive modeling demonstrated 82% accuracy forecasting when GLP-1 patients reach second-line eligibility windows
- Coordinated DTC and HCP engagement yields a 17 – 45 increase in conversion rates compared to single-channel activation strategies
- More precisely targeted campaign may exhibit lower performance on conventional reach and frequency metrics despite improved clinical and commercial effectiveness
- Claims-based triggers typically occur after the optimal clinical intervention window has elapsed
- Confirmed patient-physician occur after should be prioritized over impression volume as the primary leading indicator of engagement effectiveness
The GLP-1 therapeutic category has driven a measurement-focused competitive dynamic. The presence of multiple brands, overlapping patient populations, and historically high levels of DTC investment has led many commercial teams to adopt similar strategies-expanding reach, broadening audience definitions, and increasing frequency thresholds. However, the primary metrics being optimized in these approaches demonstrate limited alignment with the clinical decision points that ultimately determine whether a patient initiates or converts to a given therapy.
John Flemming, VP Commercial Insights & Analytics at OptimizeRx, made that case at Reuters Events: Pharma USA. "We're so focused a lot of times on frequency and reach, that we're missing the opportunity of timing," he told the Philadelphia audience. "And that's really what we need to start looking at from a KPI perspective: timing, not so much reach and frequency."
That isn't a tactical refinement. It's a measurement philosophy change with structural consequences for how campaigns are built, evaluated, and defended internally, consequences most organizations haven't yet reckoned with.
Claims Data Confirms What Already Happened
Standard trigger-based approaches are inherently delayed, as they rely on retrospective claims data to confirm that a patient has already reached a clinical milestone. In contrast, a predictive modeling framework aims to identify patients prior to milestone attainment, enabling earlier intervention at a more clinically actionable point in the care pathway.
Claims-based activation is inherently retrospective. A submitted claim confirms that a clinical event occurred days or weeks earlier, anchoring any dependent targeting model to a past patient states rather than current clinical need. In context of chronic disease management, where the interval between worsening biomarkers and a prescriber's decision to escalate therapy is both limited and clinically significant, that latency represents not a marginal inefficiency, but a fundamental structural disadvantage.
In GLP-1s specifically, this dynamic has implications that are not yet reflected in current marketing investment strategies. Novo Nordisk, Lilly, and a growing field of expanding set of competitors are targeting the same population of patients with Type 2 diabetes patients who are approaching second-line therapy eligibility. These patients can often be identified in advance through clinical indicators such as rising A1C levels and BMI trajectories, which precede prescribing decisions by several weeks. The organization that identifies this eligibility window earliest does not merely achieve incremental efficiency; it establishes presence within the prescriber’s consideration set before competing brands have initiated activation.
The alternative framework described by Flemming positions current-state targeting as "fundamentally binary, reliant on ICD-10 codes inclusion or exclusion to construct audiences, with limited temporal precision in reaching patients at clinically relevant moments. In contrast, predictive modeling functions as a form of "decision intelligence", enabling the identification of when a patient is approaching therapy eligibility and when a specific healthcare provider is most likely to initiate prescribing.
OptimizeRx validated the approach using ROC-AUC analysis across patients already receiving GLP-1 therapy, retrospectively mapping the clinical milestones that preceded their treatment escalation and training models on those feature sets. The result models demonstrated "nearly 82% accuracy" in forecasting eligibility windows in advance of their occurrence, supporting more proactive and temporally aligned intervention strategies.
To illustrate the operational implications of this precision, Flemming described a representative patient persona; Julie, Type 2 diabetes, diagnosed eighteen months prior, BMI and A1C levels trending out of target range. The critical question for the model is not whether Julie will eventually require a second-line therapy, but when that need will become clinically actionable.
Rather than delivering repeated exposures over an extended period, the predictive framework identifies a define intervention window – for example, signaling that the patient is likely 30-days from a therapy escalation or readmission risk. Activation can then be times to that interval, aligning outreach with the point of highest clinical relevance.
The distinction between these approaches is substantive: one relies on sustained, undifferentiated exposure, while the other enables temporally precise engagement aligned to patient-specific disease progression.
Better Campaigns Will Look Like Failure
The primary barrier to predictive targeting is not model performance, data infrastructure, or organizational readiness for change. Rather, it lies in the structure of existing marketing measurement frameworks, which are often misaligned with clinically precise strategies and may systemically penalize campaigns that are otherwise effective.
The underlying cause is historical rather than analytical. Core pharmaceutical marketing KPIs, particularly reach and frequency, originate from broadcast media paradigm in which agencies were compensated on volume, inventory was scarce, and advertisers relied on scalable proxies for effectiveness in the absence of individual-level conversion data. Although digital channels have since enabled deterministic targeting, this measurement framework has persisted. It has become embedded within procurement structures, agency incentive models, and internal performance dashboards that continue to treat impression volume as a surrogate for impact, despite its limited relevance to clinically meaningful outcomes.
The challenge extends beyond institutional inertia to structural fragmentation. As Flemming noted, DTC and HCP brand teams “typically operate in organizational silos”, limiting the ability to coordinate engagement across the patient care continuum. And what we need to think about is how we break those silos down so that we engage the patients. Effective activation requires aligning outreach to the point of care journey when a patient becomes clinically eligible for a given therapy, an objective that is difficult to achieve within disconnected operating models.
These structural separations propagate downstream: siloed teams drive siloed budgets, which in turn reinforce siloed measurement frameworks. Under these conditions, even highly performant predictive models are constrained in their ability to deliver coordinated impact, as they cannot reconcile divisions that are embedded at the organizational level.
The adoption paradox is direct: predictive targeting inherently reduces scale in favor of precision. As Flemming noted, more precise identification of patients within specific points in their care journey necessarily results in "When you're doing a predictive model, your reach is going to be smaller reachable audiences, prompting a need to redefine performance metrics beyond traditional downstream conversion measures”. In practice, this creates a structural misalignment in evaluation. A campaign reaching 40,000 patients at a clinical relevant moments will underperform on conventional media metrics, such as reach, impressions, and delivery volume – relative to a campaign reaching 400,000 patients indiscriminately. Different stakeholders interpret the results through legacy frameworks; marketing leadership observes reduced scale, procurement registers lower delivery volumes, and agency models reflect decreased billable activity. Despite superior alignment with clinical decision-making and likely higher conversion efficiency, the predictive campaign is systematically characterized as underperformance within existing measurement paradigms.
This is where most precision-targeting initiatives stall, not in the model development, but in the first performance review. The proposed alternative metric is conceptually straightforward: "success is defined by whether predictive activation contributes to a verified patient–physician interaction. In this framework, confirmation that a patient and HCP have engaged in a clinical conversation represents a meaningful early indicator of impact”.
A verified patient-physician encounter is a leading indicator of conversion in a way that impression volume does not. However, operationalizing this metric requires a fundamental redefinition of success across all stakeholders involved in campaign evaluation. Aligning on such a shift entails changes to measurement frameworks, incentive structures, and reporting standards—an organizational transformation that is often more complex than deploying the predictive model itself.
Discover more on this topic at Pharma Customer Engagement USA 2026 (October 27-28, Philadelphia) - where commercial, marketing, medical, data and AI pioneers converge. Explore the agenda here.
Synchronization Produces Non-Linear Returns
"We ran five overlap analyses with activation. And what we typically see is when both patients and their HCP are exposed to media using a predictive, synchronized model, they are 17 to 45 times more likely to convert to a brand of interest compared to non-synchronized or single-channel approaches."
The measurement methodology is designed to ensure rigorous attribution. OptimizeRx runs three parallel analyses: isolated DTC performance, isolated HCP performance, and the synchronized overlap cohorts. Patients exposed to both channels are excluded from the single-channel analyses to prevent double-counting and conversion rates are calculated at the patient level across all cohorts to enable direct comparability. The reported ~82% model accuracy in predicting eligibility windows is what makes the synchronized activation strategy. By identifying the clinically relevant moment in advance, the model enables both DTC and HCP channels to activate against the same patient population at the same point in the care journey, ensuring temporal alignment of engagement.
On channel selection, EHR activation emerged as the primary vehicle for HCP engagement. "For HCP, what we find is really EHR is the most effective because it's within the workflow and it's within that patient journey." Point-of-care EHR activation ensures the message reaches the prescriber while the patient is present. On the DTC side, connected TV (CTV) emerged as the most effective channel in OptimizeRx's analyses despite being among the most expensive, while programmatic delivered broader reach with less temporal and clinical precision. The distinction reinforces a broader theme; channel effectiveness is less a function of scale than of alignment with the timing and contest of decision-making.
The budget implication follows a clear trajectory. If synchronized activation at clinically relevant moment delivers conversion multiples that isolated channel investment cannot replicate, the ROI associated with an integrated approach becomes increasingly difficult to ignore. Under these conditions, the case for unifying DTC and HCP budgets within a single predictive targeting framework begins to outweigh the organizational inertia that sustains team-level silos.
The Infrastructure Hiding a Larger Strategic Opportunity
During Q&A, Flemming briefly acknowledged that the same predictive models developed for commercial GLP-1 activation have a broader applicability beyond brand marketing. "Taking those diagnosed patients and looking at their prior patient journey up to the point of diagnosis can really help health systems understand those milestones and react to them prior to those patients coming into the hospital." He moved past it quickly. The implications deserve more attention.
The predictive infrastructure described patient journey modeling, clinical milestone identification, population-level eligibility forecasting, is functionally indistinguishable from population health management tooling. The same model that tells a brand when to activate a DTC campaign also tells a health system which patients are thirty days from a preventable hospitalization. That dual-use capability matters in a market where most IDNs have closed their doors to pharmaceutical sales representatives and where formulary access increasingly depends on demonstrating value beyond the molecule.
A pharma company that offers predictive patient journey modeling as a clinical intelligence partnership isn't presenting as a promotional entity seeking formulary access. It's presenting as a population health collaborator reducing system costs. That repositioning may prove more durable than any single marketing campaign, because it addresses the structural reason health systems have closed their doors: the perception that pharma's interests and clinical interests are misaligned. The methodology is still maturing, as Flemming acknowledged, "as we continue to create these models, not only are we learning, but the models are learning as well", which means current performance is a floor, not a ceiling.
Organizations that act early to reform measurement frameworks will establish timing advantages that cannot be replicated through incremental increases in spending. While reach and frequency will likely remain visible within reporting structures, they will no longer function as the primary indicators of performance or decision-making.
To get you highlights of Pharma USA 2026 faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.
Discover more on this topic at Pharma Customer Engagement USA 2026 (October 27-28, Philadelphia) - where commercial, marketing, medical, data and AI pioneers converge. Explore the agenda here.
