Intern - Patient Centered Outcomes

Intern - Natural Language Processing (NLP) for the evaluation of saturation of text interviews

Office based, Lyon - 6 month duration

Objectives of the internship

Context: In the medical and pharmaceutical drug development field assessing the quality of a treatment closely related to assessing its efficacy and its tolerability. To do so, a physician could report the treatment outcomes based on a medical visit. To complement the picture, patient experience (perceived efficacy, tolerability, but also well-being, satisfaction)  can be captured by asking directly the patients about the outcomes they face (patient-reported outcomes – PRO). PRO has become a standard approach when it comes to work on rare diseases where the treatment impact on patients’ lives is as much important as treatment effectiveness. As the treatment outcomes could be totally different from one disease to another, disease specific PRO questionnaires are usually developed and used. As a first step in PRO instrument development, exploratory (also called concept elicitation) interviews are organized with a representative sample of patients affected by the disease of interest. Then, those interviews are analyzed to create a conceptual framework/model to display the concepts of importance and how they are linked to each other’s. Based on this framework, questionnaires (items, response options) are developed and then undergo statistical tests to prove their validity and reliability and sensitivity to change.

Approach: When the scientists analyze text interviews, they have two main objectives. The first one is to create a conceptual model with every themes/categories patients spoke about. The second objective is to know when they reach the point at which analyzing more interviews will not bring new information to the model, and thus could  confirm no more interviews are needed (i.e. enough information collected from text interviews) enabling to develop a robust and consistent conceptual model. The point at which adding another text interview to the analysis do not give more information than the one were collected is named saturation.

Objectives:

• To conduct a literature review on Natural Language Processing (NLP) to understand how NLP is functioning and what are the methods that could be adapted to analyze text interviews;
• To prepare the text data in order to be able to apply NLP methods;
• To find the most appropriate methods to analyze conceptual model saturation
• To prepare a quotation tool which could retrieve from text interviews some quotations illustrating given concepts

Additional information:

Study design

We propose to apply NLP to transcripts of patient interviews conducted as part of qualitative research to identify text characteristics indicative of concepts relating to the value of treatment and outcomes, specifically constructions that indicate value/lack of value to the patient. Since these transcripts have previously been analyzed using standard qualitative methodology both the interview texts and the thematic analyses are available. 

Study population and sample size

The precise study populations and sample sizes will vary according to studies chosen for inclusion in the analysis. Currently the ICON team has access to over 30 sets of transcripts relating to projects in the following areas: Rare diseases, Allergies, Cardio-vascular diseases, Dermatology, Endocrinology, Gastroenterology, Hematology, Infectious Diseases, Musculoskeletal disorders, Nephrology, Neurology, Metabolic disorders, Obstetrics, Oncology, Ophthalmology, Otorhinolaryngology, Pain, Respiratory, Urology.

Connaissances et aptitudes recherchées chez le stagiaire / Knowledge and abilities required for the intern

Knowledge : programming skills in Python or willingness to learn

Abilities : creative approach to problem solving, flexibility, team work