Precision recruitment: Fixing the future of clinical trials

AI models deployed within the organization made it easy for the team to identify trial-naïve patients across eight countries. The inputs were diverse and imaginative. Self-submitted calorie tracking and exercise data suggested late-day hunger spikes and declining activity levels, location data inferred adult status based on weekday movement patterns, credit card trends and youtube.tv viewing data hinted at a shift toward more sedentary hobbies, and cell phone location data confirmed commuting patterns and proximity to site, while EMR data ensured no red flags against the protocol’s strict inclusion/exclusion criteria.

The result? A list of patients perfectly matched to the study requirements, with a high probability of successfully adhering to the complex protocol.

Automated outreach was triggered – emails, app notifications, even voice calls. The team waited. And then… nothing.

Spam filters blocked half the messages. Potential participants who had opted out were protected from receiving the messaging under CAN-SPAM and GDPR. Many of the remaining contacts were outdated, scraped from long-abandoned social media accounts. Those who did receive the outreach were already digitally exhausted, bombarded by a previous barrage of irrelevant trial invitations that drove recipients to opt out of future notifications. The AI had found patterns rather than people, and the signals that the team thought that it had identified were statistical artifacts instead of clinical hallmarks.

The above hasn’t happened yet, but we think it is the inevitable and slightly bleak reality of a world where data-driven recruitment is deployed without discipline. Saturation, irrelevance, and erosion of trust threaten to undermine the very promise of digital transformation in clinical trials. In this article, we explore how to avoid that future and how to properly build a digitally enhanced recruitment pipeline.

Across the industry, organizations are realizing the potential that AI has to improve patient recruitment. These initiatives are structured around a long-standing industry challenge: how do you more quickly identify the right patients, in the right place, at the right time? With the public more digitally connected than ever before, we now have the tools to properly make sense of the increasingly diverse data available.

Putting people first, not number-chasing

Finding the “hidden patient” has been the holy grail for many clinical trial organizations. We know that there are orders of magnitudes more eligible patients within a trial site than a study team will ever be able to identify, engage, and enroll, with the notable exception of certain rare disease trials. Consequently, solutions to help uncover potential participants are highly sought after.

AI is proving to be highly skilled at identifying patterns that have previously eluded us (for instance, one application of Natural Language Processing increased the eligible population by 3.5x what was believed to exist), so it is only natural that organizations are looking to quickly apply AI tools within their most challenging and pivotal trials. In the rush to develop and deploy new capabilities that allow for never-before-seen levels of mass patient identification and outreach, teams may inappropriately prioritize scale and automation over having relevant engagement through an experience-backed campaign design. We risk overwhelming patient populations with inappropriate and poorly targeted outreach, triggering data opt-out requests and disengagement, which permanently closes off communication channels. This is unacceptable for an industry that has consistently struggled with trust and public buy-in, as these individuals are unlikely to re-engage, leaving behind a ‘grey zone’ – a future cohort which is technically visible in the data but unreachable in practice.

For clinical research to fully realize the potential of AI-enabled recruitment, we must focus on the human experience at the center of the study and not be tempted to stray into poor practices associated with an aggressive digital lead-generation funnel. Copying approaches from other industries’ wholesale may already be adversely impacting clinical research. One study experienced a 97.8% drop off between an individual consenting to share their contact information and ultimately enrolling in the study. We must apply lessons from consumer user experience research, planning and deploying communications with appropriate timing and tone, and via the correct channel. This will make any outreach and subsequent follow-up more credible and respectful, even when participant identification is supported by AI tooling.

Patients trust people, not platforms

Research is clear that trusted channels matter. Provider apps like Epic’s MyChart, the NHS app, and pharmacy networks offer familiar, secure channels for outreach outside of the site environment. These channels can also be supplemented by credible voices, which we’ve seen be successful in our experience helping a large pharma company incorporate Digital Opinion Leaders (healthcare stakeholders with a strong social media presence) into their patient engagement strategy.

The use of Digital Opinion Leaders (DOLs) is still narrow in its application. We believe there is also untapped opportunity for companies to engage DOLs who can speak authentically within disease communities to help bridge the trust gap and connect their followers to clinical research opportunities. These techniques shouldn’t simply be mechanisms to attract interest but should be applied throughout the recruitment and enrollment journey to provide assurance and credibility to patients.

Personal care starts with personal connection

AI can provide additional recruitment benefits outside of patient identification through the personalization of outreach. We have observed the use dynamic content creation and Natural Language Processing help direct-to-consumer brands communicate more relevant and targeted messages to their customers. For example, Sephora, the global beauty retailer, implemented an AI-powered feature called “Virtual Artist” in its mobile app and website. Powered by computer vision and machine learning, the tool lets users upload selfies and digitally try on thousands of makeup products in real time. The tool led to millions of virtual product try-ons and increased conversion rates and average order values for users who engaged with the “Virtual Artist”.

Within a patient recruitment context, insights from the source datasets must be appropriately enhanced so that patients understand why they’ve been contacted, how they were identified, and how they can provide feedback if the outreach is inappropriate, ultimately creating a learning loop for the digital tooling that informs future contact efforts in real time. Organizations can leverage their historical experience and insight within a therapeutic area to ground communications in expert messaging, building confidence and clarity.

Finally, the clinical trial industry needs to be thinking now about how to approach “re-opt-in” processes if potential patients end up disengaging over time. Under existing regulations in the US and EU, re-consenting to marketing communications must be led by the participant. One way to navigate this may be through the separation of the clinical trials brand from that of the pharmaceutical organization – as we’ve started to see with ThermoFisher’s Trialmed and ICON’s Accellacare. This would allow broad-reaching awareness and reengagement approaches to be developed, providing consistent, positive touchpoints with patient populations over an extended period – allowing trust to gradually be re-established.

Building a study that people want to join

Transforming the way that clinical research organizations approach patient recruitment isn’t just a technical challenge – it requires strategic and ethical use of AI, clear governance, and a recognition that speed without trust is self-defeating. The tools now available to clinical development teams are powerful, but they must be used wisely. Years of experience have shown us that patient recruitment is more than just a numbers game. As we harness the ability to derive novel insights from disconnected data sets, the organizations that succeed will be those that scale without alienating.

The opportunity is clear: to build a recruitment ecosystem that patients want to be part of, one that respects their time, their data, and their experience. Now is the moment to get it right.

This article was first published in Clinical Leader.