It is estimated that approximately 30% of the world’s stored data is generated in the healthcare industry—and that’s according to a 2012 Ponemon Institute survey, so it may be even more than that at this point. Unfortunately, much of that data is unstructured, which can make it a bit difficult to understand and analyze—even for computers (though improvements made in artificial intelligence, machine learning, and natural language processing are helping). And beyond healthcare data is the ever-growing amount of information being collected on consumers and their habits, preferences, location, and more—all of which can be helpful to marketers. But the question is how can marketers take all of this information that is available and make it useful. To turn it into actionable insights that will allow them to better serve their customers. To find out, PM360 asked 10 experts in data and analytics:
- How can pharma marketers effectively mine through the collection of Big Data available to them in order to get the insights they need to more effectively reach their customers?
- What sources should pharma look to in order to gain the most meaningful data about their consumers? What data can best inform them about how to deliver more personalized, data-driven campaigns in terms of channel preference, digital habits, content, etc.?
- How can pharma implement predictive models and analytics to make smarter decisions about how it approaches its customers? What are the best ways that pharma marketers can use predictive analytics right now?
- How can pharma collect, analyze, and interpret data fast enough to make real-time decisions and adjustments in its marketing or sales approach? What tools, services, or strategies can be used to make this happen?
- How can pharma marketers use data and analytics to ensure they are delivering a campaign that will resonate with their target audience?
Newton’s third law states “For every action, there is an equal and opposite reaction.” In many ways, this applies to 21st century marketing and our ability to process and manage the overwhelming amount of data available. To help ensure data capture can be leveraged during the entire lifecycle and that marketers can adjust their multichannel efforts in real time, we suggest the following tips:
1. Baseline Profile: Consolidate all the various sources of patient information into a universal sign-up form. This allows a user to enter first name, last name, and email address (the minimum information needed to instantiate a profile) and allows a real-time lookup to see if the information is a duplicate. This simple change prevents the flood of duplicate profiles and allows the organization to track exactly where the source came from.
2. Progressive Profiling: Equally critical is the need to continuously add information to the baseline profile. Known as progressive profiling, it means that every interaction with a user is recorded and stored in a customer relationship management system.
3. Big Data/Artificial Intelligence Engines: These monitor all interactions and develop behavior triggers, for example, 55% of the visitors to a website click on a call-to-action (CTA) button and proceed to conversion. For the 45% who ignore the CTA, the trigger could be as simple as having a pop-up or chatbot to prompt for improved conversion.
4. Predictive Analytics: As more behavior triggers are defined, the AI tools will also provide recommendations and threshold alerts to the pharma marketing team, which is looking at geographic and demographic sales as well as other influences. And, by using robust content management and social tools, pharma marketers can be predictive and use Newton’s third law as a guide: Try an action, look for the equal and opposite reaction.
Predictive analytics has been one of the most buzzed about concepts across industries and it has applications in the healthcare arena. A fast follower to that concept is real-time analytics or “next best action.” That said, there are challenges with data acquisition, harmonization, and validation to enable these types of techniques. Healthcare data scenarios are often fraught with technical issues, privacy considerations, and institutional barriers. One of the biggest hurdles: The strong desire to leverage these techniques without having a problem in mind.
When to Use Predicative Analytics
Where there is a need to examine drivers of behaviors that ultimately can be measured, there’s a case for predictive analytics (and potentially real-time reporting as a follow-on). For example, which patient types are more likely to use a certain protocol of treatment? The next issue: There are so many potential sources of information that could proxy for drivers of behavior. The best models consider things such as formulary and coverage status, reimbursement hub engagement, prescribing behaviors, patient co-morbidities and history, consumer behaviors, practice settings, engagement with manufacturers, clinical research, publications—the list can be long. Successful projects need all this information to be tagged and loaded into an environment where they are accessible and linked together where possible.
Once a successful model has been built, the real work begins—to socialize, drive understanding and adoption, explore all potential updates or improvements, and then iterate on the model. In some cases, there is the interest and appetite to continuously update the model, and that is when “next best action” scenarios take flight. Adjustments in infrastructure are needed to make the model updating on a more real-time cadence. Then the challenge becomes finding the right way to care and feed a robust complex model that’s updating every few hours, and partnership in the early stages is crucial to develop and maintain a complex evolving model.
When healthcare stakeholders talk about “unstructured data,” they usually refer to data that is not natively standardized against accepted medical vocabularies. EHR data and raw digital voice-of-the-consumer (VoC) data are two examples. The value of these data sources is immense—while insurance claims data gives us a baseline view of the treatment pathway, EHR data reveals clinical reasons for switching, and VoC data illuminates patients’ decision-making processes, allowing researchers to dig into the “why” behind the “what.”
Non-standardization presents a challenge when working with these data sources. EHR data is used by HCPs to inform treatment or maintain history—as opposed to, say, informing automated payment decisions. A patient may suffer from “Diabetes” or “Type 2 Diabetes” or “T2D,” with “A1C” or “HbA1c” levels of “5.2%” or “5.2.” To a physician, this may suffice to manage treatment; to the life sciences analyst, it’s a major headache.
Dealing with Unstructured Data
Open source text-mining techniques to address issues like this have long been available, but the challenge is not so much about sheer mechanics as it is context and process. A STEM graduate can run an algorithm. Extracting meaningful insights from unstructured data requires having the personnel to know which healthcare-tuned taxonomies and ontologies are appropriate, which text analytics approaches makes sense, whether outputs are correct, that “5.2” probably means “5.2%,” and finally to make sense of the data and draw valuable and actionable conclusions.
This points to a fundamental challenge in life sciences analytics—creating and cultivating a team of both therapy area/domain talent and tech-leaning analytics talent, and ensuring they effectively work together to extract value from any data source, structured or unstructured. Unstructured data isn’t a challenge simply because it’s messier, but because interpretation requires a human touch and human expertise—unstructured data demands plenty of both.
Pharma marketing can drive business outcomes by taking an integrated, data-driven, and data-centric approach. The challenge is to find relevant data that is timely and usable. Data is only as good as the question or challenge the brand is trying to solve (e.g., digital marketing ROI).
Repositioning a Mature Brand
One example: How a mature Rx allergy eye drop invested in market research, physician behavioral segmentation, and tracking and tagging digital assets to help reposition the brand and increase awareness. The brand began by analyzing the data and speaking to the customer. Using physician behavioral segmentation and advisory boards, the brand identified 50,000 physician targets out of a universe of 120,000. This data allowed marketing to focus on only those clinicians with the highest opportunity to increase brand awareness and ultimately script lift. The “Ah-Ha” moment was when the market research data uncovered a key insight: High-potential prescribers didn’t recognize the benefits of the brand over generics.
Adjusting in Real Time
Using these insights allowed the brand to develop and serve highly targeted reminder messages through digital channels about the product benefits. The brand purchased the email addresses of those segmented 50,000 physicians and tagged and tracked every digital tactic (e.g., keywords on the website, eBlast, and banner ads). By monitoring physician responses through tagging and tracking, the brand was able to adjust in real time by further tailoring the messaging to keep the physician engaged.
This action increased website traffic and call to action activities such as registering for CRM, downloading a coupon, or requesting a sample. Furthermore, the allergy eBlast had a CTR average of 16%, and programmatic delivered over 262% more impressions to the targeted audience while dramatically decreasing cost per click by 80%.
1. Define your hypothesis: Before jumping headfirst into analytics, a company must determine which problem it’s trying to solve. It may be tempting to throw large quantities of data into a model and see what happens, but doing so runs the risk of finding correlations that are not relevant or useful. Forming a hypothesis first can produce more meaningful and actionable results.
2. Gather data: Pharma companies must gather appropriate data to plug into predictive models. For example, when working with external vendors on marketing campaigns, it’s important to ask for data that is as detailed as possible. If you don’t have that data, then you can’t measure the campaign and could possibly skew the impact of other activities within your model. Fully forming the problem you are trying to solve in step one will assist in acquiring the right data.
3. Identify the best model: Analysts can choose from many analytical techniques, and each has its own strengths and weaknesses. Once a company determines the best data to use, it must identify the correct analytical technique to deliver the best results. For example, selecting the proper clustering technique in a physician segmentation is important to determine which to target.
4. Take action: Commercial teams should use the results of their predictive models to arm their sales forces with actionable insights. For example, if a certain segment of physicians is likely to be receptive to a particular message, the sales force needs to know how to easily identify this group. While the predictive models themselves may be intricate, their outputs should be intuitive and easy to explain. The value of predictive analytics models is lost if the insights cannot be implemented.
While 30% of the world’s stored data may be generated in healthcare, 50% of that data is rendered unusable—a result of a confluence of common factors. Data is often stored in locations across the enterprise, it arrives in multiple formats and models, and its complexity often varies by the source.
Broadly, commercial pharma executives are wrestling with these challenges caused in large part by investments made in historically manual approaches to data management and analytics. Because healthcare data is constantly changing and increasing in volume and complexity, organizations must invest in real-time data analytics to ensure that when an insight is needed, it’s being driven by the most up-to-date and accurate data available.
Single Source of Truth
With the overload of data, analytics platforms can help pharma marketers by aggregating this data into a single source of truth, ensuring that all departments have a unified view that allows teams to work toward the same goal and plan accurately. Data is a key component to reaching customers, but transforming it into real-time, actionable insights is the differentiator for effectiveness and efficiency.
Ensuring that clinical and commercial data is fed to marketers allows them to more quickly adjust strategies to reach the right healthcare providers and patients for successful communication and commercialization—and ultimately accelerate the access to therapy to patients who need it the most.
In today’s life sciences industry, pharma and biotech companies need solutions that will help them streamline and optimize their business processes to make better real-time decisions and adjustments in marketing or sales approaches. The key is cross-departmental collaboration, coupled with a platform approach to leverage the tremendous amounts of data that a typical sponsor generates while conducting clinical trials and during the commercialization phase after approval.
Technology advances, such as state-of-the-art data analytics platforms that incorporate the latest techniques of deep learning and artificial intelligence (AI)—coupled with advanced data ingestion and harmonization engines—can deliver game-changing results for life science companies. Add to that the ever-increasing trove of real-world data (RWD) and advances in machine learning platforms, and sponsors have a chance to move closer to the streamlined knowledge-based enterprises they need to become—the ones that are poised for future success.
Tracking the Patient Journey
AI-enabled patient journey tracking solutions monitor the pathways patients travel, enabling the evaluation of treatments, disease progression, outcomes, and total economic costs. Such state-of-the-art solutions leverage RWD to provide decision-making insight into how a given disease is treated and/or managed, lines of therapy, treatment options and outcomes, cost of therapy, treating physicians, disease progression, and complications. De-identified medical transcription data is a great RWD source for real-time decisions because is it available as a daily feed instead of quarterly like most claims data sets.
Pharma and biotech companies can understand the clinical and commercial journey of patients, so they are better equipped to assess/compare which pathways and treatments tend to lead to better outcomes. By efficiently and effectively collecting, analyzing, and interpreting this critical data, they are well-positioned to evaluate the impact of their marketing and sales approaches and make real-time decisions to enhance their success.
You can use artificial intelligence and machine learning to predict any action or behavior as long as you have what is called a “training set” of data that includes related behavior that precedes it. With the appropriate training set, you can determine which segments of patients are most likely (1) to be diagnosed, (2) to get treatment, (3) to stop treatment, or (4) do well on treatment, just to name a few. The key is in determining what behavior or action you want to predict. And, pharma marketers should choose events they know they can positively impact.
New Model for Sales Forces
The most direct and proven point of intervention is at the time of treatment decision. Predictive analytics is being used today to foresee when a patient is likely to start or change therapy. This intelligence allows brand teams to move away from reach and frequency models for their sales forces. So instead of prioritizing HCPs based on their past prescription writing, reps can call on HCPs right before a patient presents who could benefit from their brand. Thus, allowing their reps to deliver a relevant and timely message—one that an HCP is likely to appreciate.
Customer access and compliant engagement has been an ongoing challenge for pharma. Customer data and related compliance information is often scattered in various systems. To organize incoming and existing data so that it can be efficiently used in data-driven decision making and to more effectively reach customers, it’s crucial that pharma marketers first start with a reliable data foundation, or “single source of truth.” That means bringing data from all internal, external, and third-party sources together and breaking down silos among various applications and teams.
Once a reliable data foundation is created, pharma must make sure there are processes and solutions in place for continuous data organization and collaborative data curation. Additionally, it’s important to uncover relationships and affiliations across data entities, such as HCPs, HCOs, patients, payers, and health plans. Graph technology can help with that.
Turning Data into Insights
Building on the reliable data foundation, pharma should invest in systems that help visualize this information for account, field, and medical teams using business user-grade, data-driven applications. Technologies such as advanced analytics and machine learning can then be used to provide relevant insights such as influence scores of a KOL or recommendations about content and timing for the next detailing—and ultimately measure and correlate the impact of actions on plans.
When data from all sources and formats is brought together into a single customer profile view, with relevant insights presented in the context user’s objective, pharma teams can let information drive better decisions—helping them understand markets, drive revenue opportunities, and reduce compliance risk. Today, modern data management technologies are offering such capabilities, and forward-looking pharma companies are adopting this approach, recognizing the value the technology provides in enabling them to understand customers and drive intelligent and compliant customer engagement.
Identify the right patients for a specific therapy: Patient finding is often a rate-limiting step to patient care. Time constraints and limited resource availability may compromise clinicians’ ability to identify patients with a difficult-to-diagnose disease. Even when patients are accurately diagnosed, selecting the most appropriate therapy for an individual patient can be challenging. By using advanced, AI-driven clinician segmentation, pharma marketers can develop hypertargeted messages to drive changes that optimize patient care.
Optimize disease management by leveraging local clinical leaders to drive the diffusion of innovation: While targeting the right clinicians with the right messages is essential, behavior change is unlikely to occur unless the message is delivered by the right messenger. Objectively identifying local clinical influencers, who are responsible for driving clinical innovation at the local level, is a difficult task to achieve at scale without using predictive analytics. In spite of their ability to impact the care of many patients, these clinicians are routinely overlooked by pharma marketers because they tend not to be high-volume prescribers.
Maximize patient adherence: Even when patients are placed on an optimal therapeutic regimen, poor adherence and abandonment remain a perennial challenge. Given the volume of clinical, demographic, and behavioral information currently available about each patient, predicting patient nonadherence is no longer an unattainable ideal, but rather an obtainable goal. Predictive analytics can be leveraged by marketers to serve as a catalyst to partner with clinicians to prevent issues before they arise, and take steps to maintain patients on essential therapies.