91黑料

  1. Home
  2. Insights
  3. Blog
  4. Validated tokens: The key to following the patient journey

Validated tokens: The key to following the patient journey

Increasingly, sponsors are finding value in integrating clinical trial data with real-world data and precision medicine data to follow patients through the continuum of care, beyond their trial participation. Fortunately, there are dozens of primary and secondary data sources that can be brought to bear on obtaining a complete and longitudinal view of the patient. However, since no one supplier has all the data, data from disparate sources has to be linked at the patient level in a way that is consistent, trustworthy, and protects patient privacy. 

Being able to do so in a blinded and encrypted way makes it possible to study healthcare outcomes across the spectrum of clinical development, from early phase to commercialisation. Ultimately, this supports the development of more effective, safer drugs.

Tokenisation as the solution

Tokenisation – the process of replacing Personally Identifiable Information (PII) with a unique encrypted de-identifier – is straightforward enough when applied to clinical trial data in a single dataset. However, it is more complex when drawing on data from multiple sources.

Essentially, tokenisation employs a de-identification engine that assigns a unique and random string of characters, or tokens, to a single patient identity. With the application of such a consistent identifier, one can confidently link data sets together where the same patient exists – all in a completely HIPAA-compliant environment – for use in analytics. And patient data can be tracked over time, for example as a patient moves from trial participation into real-world usage, post launch. 

Unlike with traditional encryption techniques, tokenised data can be made indecipherable and irreversible. Through tokenisation, it is impossible to associate sensitive health data with an identifiable individual because the key to the linkage between them is stored in a secure database or “token vault.” Thus, tokenised data on its own holds no intrinsic value for unauthorised users who might gain access to the data. 

The need for validation

Conventional wisdom assumes that because tokens are generated using actual patient data that they are automatically valid and trustworthy. However, this assumes that the PII used is of high quality and that the tokenisation engine is run correctly. These are faulty assumptions. In reality, to be trusted, tokens must be validated as being correctly linked to all the relevant patient data. 

The process of validating tokens involves matching them against a curated patient master – in other words, against patient data that has been selected, classified, cleansed, and formatted to reduce incorrect or unusable values. A patient match rate between the token and the patient master of 95% or better within an individual data source is the desired standard; anything less suggests an issue with the quality of the token. (Note: patient match rates should not be confused with patient overlap rates which refer to the number of common patient identities in two or more datasets after business rules have been applied.) 

Thus, since the quality of the token is directly related to the accuracy and completeness of the PII, maintaining a comprehensive patient master requires access to a vast array of trusted data sources.  By investing in multiple rich data sources and an integration scheme using validated tokens, sponsors can unlock the full potential of healthcare data to drive innovation and improve patient outcomes. 

Linkage challenges: A real-life example

A clinical trial site that applied tokenisation was achieving a patient match rate of 0%. In other words, there were literally no matches between the patients in the trial and those in the sponsor’s patient master. When the site was asked to confirm what was being entered for the key PII fields, it was found that they were not entering complete details for the patient’s first name. After the site corrected the first name entries, the patient match rate between the trial database and the sponsor’s patient master increased to the expected match rate of 95% or better.

In this section

Connect with us