What does Neoastis do?
Neoastis provides topic modeling and natural language processing services for sales and marketing.
For the pharmaceutical industry Neoastis provides medical insights, automated research and opportunity detection.
What happens if the jointly defined search terms do not provide useful key topics? Can be readjusted? How and how often?
The offer contains 5 search terms. PAI first works with 3 keywords. Then there are one or two iterations with 1 or 2 varying search terms (after elaboration of the topics and feedback from the customer on usability). We will contact you to discuss the intermediate result. Only then do we start the next iteration. Ideally, we will receive a briefing in advance on where and in what sense to pay attention to certain topics or the label.
As with manual research, there is still a residual risk. As with the search for gold, we cannot guarantee that interesting topics will be found.
I have a fixed theme for my product (e.g. duration of therapy) from the company and don't know how to stage it, because everything has already been said about it. Can PAI help to transport an old topic in a new way?
Yes, that can work. For example, we would search for therapy duration, know from the briefing what has been done so far and then specifically search for terms that are new in this context. Experience has shown that this approach works better than a "blind" search.
Can PAI also be applied to full studies to find interesting topics that are not covered in the abstract?
Theoretically, PAI could also be applied to full studies if the corresponding text material is available. The problem is copyright. Abstracts are public, studies generally not.
How is relevance defined and weighted?
Is a single case report equally relevant compared to a study of 20,000 patients? Since it is not common to specify the study scope in abstracts, this has to be done manually afterwards. I.e., if you read the abstract, you can exclude this thread. There is a "relevance score" for relevance, which is defined together with you (e.g. study scope, citation, journal ranking, publication year, etc.). We adapt the relevance score individually for each project (e.g. certain authors give a higher score), e.g. in order to weight the results of Austrian authors higher.
Can PAI ensure that an issue is relevant in my geographic market?
Yes, this is done manually. We have the option of restricting by author (e.g. only a list of Austrian authors). Whether a topic is relevant in the geographic market, however, is not represented in data, but rather depends on the prevailing opinion. However, we can map it indirectly:
For this PAI searches for an opinion and a contrary opinion. See also "Relevance score" in question 4.
Does PAI find important topics in abstracts that are actually about something else?
We control this by selecting the topics. The algorithm finds related studies that have a certain conceptual proximity. The initial search is an AND link: i.e. PAI searches for abstracts that treat e.g. heart failure AND atrial fibrillation. Hidden topics are, by the way, the algorithm's speciality: if certain terms are used in only a few studies, this topic gets a higher score and would appear as an identified topic.
(more technical: Such an exceptional study, which uncovers a connection that no other study shows, has low entropy and would therefore be shortlisted.)
For my product, Medical and the sales force are in exchange with Key Opinion Leaders. Can PAI really find new topics here?
If we look for exactly one product, we probably won't find much that's new. Because the results are often product studies, most of which are well known. Therefore, a search in the indication area is often more promising. Although we have been able to identify topics that were previously unknown for individual products, the risk of finding nothing here is higher
How do I exclude the possibility that the found topics do not move in the offlabel area (i.e. not FKI-compliant)?
You can control the topics using search terms and an exclusion list. The latter is done manually. Here the results are compared with the label. The strength of PAI lies above all in the indication, only in the use of the results would be offlabel to treat. Of course, a purely product-related search delivers many offlabel results.
Are there certain advantages or disadvantages if I call in a key opinion leader (or other specialists for the respective indication) for a PAI project?
There are different options here: cooperation with Medical, Commercial or with a KOL. It is also possible to work directly with the KOL on site in a centre and to integrate the KOL into the treatment.