What can social networks teach us about drugs ?

Social networks have become an important source of information about the medicines we consume. Our patients share their experiences and talk with each other about their illnesses, treatments, and course of care on discussion forums, Facebook, and Twitter. In doing so, they create a vast source of information that can no longer be ignored.

Why use social networks to study medicines ?

To study a pathology or therapeutic class from the patient’s point of view

Patients’ experiences in dealing with their pathologies and treatments are crucial for understanding disease management.

Analysing data from social networks provides qualitative information untainted by the measuring biases typically seen in studies and surveys.

1.

Permanent observatories and cross-disciplinary studies.

2.

Extraction and analysis of pertinent information.

3.

Snapshot of disease management from the patient’s perspective.

4.

Identification of unmet medical needs and the challenges patients face.

5.

Organisation of information regarding a medicine or therapeutic class.

To analyse adherence in a new way

In the case of both primary and secondary adherence, social networks provide qualitative information on the reasons why patients stop their treatments or never start it.

Poor adherence has become a public health issue. Our analyses make it possible to identify where the greatest gains can be achieved and then assess the impact of adherence improvement strategies.

1.

Extraction of messages relating to the cessation or modification of a given treatment.

2.

Identification of the reasons given by Internet users for stopping the treatment and data coding.

3.

Detection algorithm for factors that predict the cessation or modification of treatment.

To analyse the use of medicines and health products

There are many recommendations regarding the proper use of medicines. Analysing use via social media makes it possible to quickly spot misuse and off-label use, and monitor these cases over time.

1.

Extraction of messages relating to the cessation or modification of a given treatment.

2.

Identification of the reasons given by Internet users for stopping the treatment and data coding.

3.

Detection algorithm for factors that predict the cessation or modification of treatment.

4.

Identification of medical concepts expressed by Internet users and coding using MedDRA terminology.

5.

Detection algorithm for medical concepts compatible with the MA.

Voir le graphe

Example of a drug dependence signal discovered during the surveillance of a central nervous system medicine.

Longitudinal monitoring of CNS medicine use.

To anticipate and analyse pharmacovigilance signals

There is growing interest among medicine sector stakeholders in using social networks to conduct pharmacovigilance. It is now possible to detect early safety signals via social networking sites by using confirmed mathematical, semantic, and numerical algorithms.

Our algorithms also make it possible to code patient language according to the MedDRA dictionary.

1.

Extraction of messages containing the name of the medicine and information about taking it (noise filtering).

2.

Identification of medical concepts expressed by Internet users and coding using MedDRA terminology.

3.

Adverse event detection algorithm.

4.

Use of standard signal detection methods (Bayesian and frequentist).

Over 30% of signals in our database are identified an average of 75 days before their inclusion in international institutional pharmacovigilance databases (FAERS and VigiBase).

PPV 95 %

SE 81 %

Specificity 56 %

AUC 0,84

Voir le graphe

Chi-squared and PRR distribution for medicine X.