Leveraging Individual Patient Data

In building simulation models of pharmaceutical markets over the last fourteen years, one of the biggest changes has involved the availability of data for very large blinded samples of individual patients. In Europe this trend moved somewhat ahead of the United States. But now, the availability of data from managed care organizations, hospitals and large physician practices has made use of these data common in the United States as well. Moreover, Ventana has noticed that the analysis of these data yields important insights that are independent of the results of aggregate market simulation models that the data also support.

Simple persistence (or adherence) to therapy

One of the more innovative parts of the 1997-98 model described in the Marketing New Pharmaceutical Products example was its method of tracking patient flow. At the time, pharmaceutical marketing was heavily oriented toward traditional primary research focused on small physician/patient panels and drug utilization. The Ventana model tracked these things, but it also tracked flows of patients as they moved onto and off of therapy. To do this, it was necessary to infer the nature of these patient dynamics. Ventana was able to indirectly measure them from the types of prescription data widely available at the time – so-called NRx and TRx data sold by industry data vendors.

Some managers were uncomfortable with indirect measures of patient persistence on therapy. They were used to direct measurement of important market parameters. Consequently, they made a special purchase of blinded data on individual patients. The data confirmed Ventana’s inferences. This gave credibility to Ventana methods, but more important, it suggested the use of a new source of information about patients. Indirect measurement can only work in certain situations, and more details of treatment are knowable with the patient-level perspective.

Two years after Ventana showed the marketing value of the patient perspective, data vendors were providing a new syndicated data product that replicated Ventana’s initial approach to patient-level data. This was probably an independent development motivated by different factors. The important fact is that blinded patient-level data is now in common use by pharmaceutical companies, and the current challenge is interpreting what it means correctly. The data afford important advantages, but it can be misinterpreted if used without reference to other types of data, and Ventana has seen significant evidence of this. (The interpretation issues are similar to those discussed in Improving Survey Data. While the vagaries of human motivation and memory are not relevant to this type of data, issues of patient sampling as well as the data rules used to aggregate individual patient histories introduce significant interpretation problems.)

Defining persistence, adherence and compliance for episodic or intermittent conditions

Over the last decade, Ventana has continued to develop its patient-level data approach, treating increasing more complex patient flows. The pharmaceutical information services industry has moved faster to profile even more complex treatment scenarios.

The key difference between what the industry does and what Ventana does is that data vendors largely restrict their offering to one type of data, and they sell it by therapy area. Data vendors allow their customers to define the data rules, but they typically set bounds on what is possible and they presume expertise about the implications of data rules. Often, this expertise is not in evidence. Because Ventana builds systemic models of markets, we cannot isolate our attention to one kind of data. Our focus is on the consistency of data across many data types ranging from patient and physician survey/opinion data through aggregate data on markets (e.g., incidence/prevalence of disease and national script data) to patient-level detail. For us, understanding inconsistencies between data sources is critical, and insight emerges from identifying and explaining inconsistencies.

Ventana analysis has surprised, but has ultimately won over thought leading physicians. In particular, explicit recognition of periods when patients were not on therapy has proven important in understanding the dynamics of drug treatment and the magnitude of unmet medical need. Extant patient-level data tends to focus on the dynamics of treatment, largely ignoring the characteristics of periods when patients are not using prescribed medicine. Inattention to this blind spot can (and has) caused medical experts to both overestimate and underestimate the size of unmet medical need depending on the dynamics of a particular disease and therapy. For example, the nature of the medical condition makes the requirements for—and interpretation of—patient-level data on schizophrenia and migraine patients fundamentally different.

Ventana continues to work with industry data vendors to improve their offerings relating to patient flow. We compete and cooperate in the growing information services industry.

Synthesis of disease progression over lifetimes

Ventana’s most ambitious projects involving patient-level data require the synthesis of virtual life histories from the fragments of data on real individuals that are available. Typically, in the United States it is difficult to get large samples of patient histories (say hundreds of thousands or millions of patients) for more than 5-6 years. People change health plans and physicians too often to make long-term data linkages reliable. In Europe, much longer histories are available, but they are not yet close to providing birth-to-grave coverage. For conditions like diabetes or the umbrella condition “metabolic syndrome” (diabetes, hypertension, dyslipidemia and obesity), the evolution of conditions and treatment over decades is necessary to understand it properly. Using methods similar in philosophy to indirect measurement, Ventana has used models to transform patient-level information for the limited windows of observability on real patients into longer histories on virtual patients. Profiling and trying to explain these histories using simulation models has changed one pharmaceutical company’s views on metabolic syndrome. Where the data can support it, Ventana recognizes this as an important research area. Currently, the state and pricing of most vendor data in the United States makes such research prohibitively expensive.