The internet has long been recognized by the pharmaceutical industry as a source of patient reported outcomes (PRO) data, however, methods of harnessing its information have been flawed due to the complexity of conversations unfolding online. Internet forums and blogs allow patients to interact, asking questions and providing each other with advice. As such, patients are reporting and deciding upon millions of treatment choices online every day.
With its widespread adoption, social media can serve as a largely untapped source of information for healthcare organizations. By understanding and managing the nature of internet discussions, pharmaceutical companies benefit from access to real-world, unbiased PRO data. That being said, identifying appropriate ways of collecting, managing and analyzing these data can be difficult.
Contrary to traditional data sources, the majority of patient-reported information within internet discussions is unstructured, making it particularly difficult to analyze using conventional techniques. More specifically, the complex nature of human communication means that methods such as keyword identification cannot link individual pieces of information to understand the motivation and sentiments behind the patient’s actions.
Advances in health-focused natural-language processing technology have enabled the evaluation of patient healthcare outcomes reported within social media. The technology facilitates the analysis of PRO data with the aim of identifying the decisions patients make, and why they make them, by combining different pieces of information and putting them into context. This allows patients’ actions, such as changing from one therapy to another, to be linked to underlying reasons.
Given that patients only discuss what matters to them, the insights gained have the potential to influence drug development decisions relating to new medicines that could significantly impact patients’ lives. Additionally, as regulators increase pressure to demonstrate true patient benefit, the analysis of internet discussions with advanced language processing can provide accurate and quantifiable real-world PRO data.
The accompanying article that goes more in depth can be found
