Finding KOLs – Identifying the Right Expert in a Sea of Possibilities
KOLs are important – everybody agrees on that. A well-known speaker at a conference, the right clinicians involved in a trial, or a recommendation by a highly regarded physician, can make a world of difference whether one develops a drug, medical device, or diagnostic.
The challenge is to identify the person(s) with the necessary expertise and experience, in the right therapeutic area, geography, and institution in a sea of possibilities. The basis for this process is a comprehensive and up-to-date database containing not only the basic information about all possible experts but also information about their relevant activities: from clinical trials they are involved in to patents they have filed, talks they have given to competitors they work with. H1insights’ databases, Ada and Curie, contain more than one billion data points.
Big Data in KOL Identification
Big data is the basis and at the same time a challenge in KOL identification: How does one go about narrowing the amount of possibilities to just a handful of candidates given all that data? The process H1insights works through with their clients is outlined here:
· Define the objective: the process starts with the simple but critical question “What is the objective?” This question helps to clarify why a client is looking for a KOL and answers questions such as: Do they need a “generic” cardiologist or somebody who has studied a certain pathway? Are they looking for a KOL with experience in one field who also needs to be an expert on another field? Based on this starting point the next step is to…
· Develop a keyword list for the database search: keywords cover information such as drugs, disease states, and indications that are relevant to the search. Experience shows that seven to eight keywords yield the best results. Proprietary algorithms break those keywords down and return relevant data. What sounds straightforward is actually an involved process that takes into consideration things like differences in publishing frequency between disciplines. Cardiologists, for example, generally publish more than epidemiologists. When searching for a KOL in epidemiology, metrics relevant in their field of expertise need to be applied, as opposed to broad averages. The outcome of this step is a long list of pre-classified leads that need to be scored in the next step.
· In KOL scoring, the pre-selected KOLs are matrixed against a set of typically about 50 criteria. Criteria often include: How often does the KOL publish? Where does the KOL publish? How often is the paper cited? Is this KOL the main, or co-author? Which clinical trials are they involved with? Do they have speaking engagements? Where are their speaking engagements and how often do they take place? Which grants have they received? In discussion with the client a weight gets assigned to each criterion, recency of data is considered and that information is fed into the algorithm, which returns a score and a trend for each individual on the list.
· Making sense of the score: scores can be contentious when created by a black box. To allow the clients to look into how and why KOLs received the score they have, H1 provides clients with the underlying data in a searchable format allowing the clients to work with the data and do their own mapping, e.g. by specifically looking for experts in certain geographies, specific institutions, or searching for those who have worked with specific companies before. The data created during the scoring and mapping steps can also be loaded into the client’s CRM system.
· Understanding trends: another piece of useful information is the trend analysis that H1 provides for each expert. The trend captures, for instance, the rate of change in publishing in scientific journals, relative to the length of one’s career. This trend helps identify rising stars – the up and coming KOLs in their area of expertise - a very sought-after group.
Big data is the basis for KOL identification, but more than one billion data points without the tools to make sense of all these pieces of information are useless. Using proprietary machine learning algorithms and the client’s customized requirements, H1 is able to boil down all that data into a scored list that is easily searchable, shows trends, and allows for identifying the most relevant KOL every time.