Bridging the gap between patients and scientists to extract (lactose) intolerance correlations

Online health support groups allow for communities of people with a similar diagnosis to communicate. For example, dedicated Facebook groups for lactose intolerant people allow patients to ask questions to one another and to share personal experiences. However, to our knowledge, the information in such fora is not yet utilized to align research topics with what the patients themselves are struggling with. We argue that doing so is important to bridge the gap between scientific research and patient’s personal experiences, as well as to identify valuable research gaps.

In the context of lactose intolerance health communities, a notable matter in such groups is that some patients still, after eliminating lactose out of their diets, have bothersome gastrointestinal symptoms. There are suggestions that a potential cause of persisting gut symptoms might be another intolerance co-occurring alongside someone’s lactose intolerance. We aim to search for and possibly identify research gaps between what patients struggle with on this matter, compared to what has been researched.

To identify intolerance correlations in the largest public Facebook group for the Dutch lactose intolerant community, we performed relation extraction with regexes. The regexes were created after visual inspection of the test data and extended based on human intuition regarding sentence formulation. Multiple types of relations between lactose intolerance and other possible intolerances were extracted from free text, ranging from mentions such as "lactose and ... intolerance" to given advices regarding persisting symptoms to look into “reactions to …”. The engine searches for words on the “…” spots that match with a list of the potential intolerances a person can have. The relation extraction engine leads to suggestions of intolerances co-occurring with lactose intolerance, ordered by prevalence.

To find out whether or not the co-occurrences have been investigated in medical studies, we mined scientific PubMed publications. We examined how often lactose intolerance is referred to in combination with another intolerance. A PubMed article fit into our profile if it mentioned each of the following three words at least once throughout the entire text: the word ‘lactose’, the word ‘intolerance’ and lastly the word of a (by the fora) suggested intolerance.

Results showed that there are multiple knowledge gaps regarding the potential relation between lactose intolerance and soy, gluten, sugar, nut or food additive intolerances. These intolerances were often discussed within the investigated online health community, whilst their association with lactose intolerance was rarely addressed in PubMed publications. In particular, a noteworthy trend was spotted between soy and lactose intolerance.

There is a patient need for future work to investigate intolerance correlations. It is especially meaningful to investigate the correlation between soy and lactose intolerance. Method-wise, this study opens the door for machine learning approaches to address the matter of co-occurring intolerances through online support groups. Lastly, we hope to inspire future work to combine patient fora with previous scientific work. We consider it important to bridge the gap between patients and scientists because ultimately, the medical scientific research field tries to better patients’ lives.