Show HN: Medical Research Summarizer – Complex research as simple summaries | Hacker News Reader

Show HN: Medical Research Summarizer – Complex research as simple summaries

summarizer.hdphealth.com

We've built a system that allows us to translate difficult to read research abstracts and convert them into 4 to 8 bullet point descriptions that are readable by the common person.

It's like Blinkist, but for complex research.

Our backend consists of a ton of med school students, PhD's, and researchers.

I myself had an autoimmune disorder for 10 years, well I still have it, but I've been in remission for the last 2 years. Making sense of the research out there is a tall order without an organic chemistry background. When I was first diagnosed, I ended up spending hours trying to read research I barely understood, but that was relative child's play for a PhD to clarify.

It's called Ulcerative colitis and has no known cure. Some of the most intriguing research that I'd wish I had been cognizant of and had a way to read is this article(Http://bit.ly/fodmapfree) on FODMAPS (Fermentable Oligosaccharides, Disaccharides, Monosaccharides, and Polyols Saccharrides) and their impact on those with Irritable Bowel Syndrome.

Had I known, I would have experimented with something like it to see outcomes. Keep in mind, there's no such thing as an FDA approved diet.

Do you not worry that people who are not researchers, who are providing essentially the "TL;DR" of a disease, will provide incomplete or potentially misleading advice?

In addition to what rf1331 said, to what extent is hdphealth liable for a case where e.g. someone reads that treatment X is beneficial for their case, but in fact someone got the facts slightly wrong and it made their case much worse?

https://www.hdphealth.com/2016/02/16/the-research-on-biopsie...

Notice the disclaimer at the bottom.

As well, we link to the research article itself in every scenario.

At some point we should try to add a "Share with your doctor" button.

Thanks for the feedback.

I think a niche certainly exists for your product...

If you could manage to land an ad in AARP-The Magazine for instance, I think you'd be pleasantly surprised at the response...that demographic is of an age where many common ailments begin to show up...

I'd caution you to give the idea of adding a "Share with your doctor" option very careful thought before implementing it...

Some physicians don't mind patients that attempt to research their ailments online prior to an appointment...others dread seeing patients arrive with printouts from the internet...

This is anecdotal from my wife's side of the family, which includes 2 physicians, a dentist, and a physical therapist...

Having said that, times are changing...

Good luck!

Thanks. We've had an overwhelming response even from just the tech community submitting articles on all sorts of things ranging from hypopharangeal cancer and beyond.

Of course we do.

This is we try to adhere to common translational research guidelines. The focus is not on opinions but understanding and characterization of the research out there.

Over time, we'll publish these guidelines. As well, in each summary there's a disclaimer that states this is not medical advice. Any action should be discussed with an experienced medical licensed professional.

I would rather have incomplete knowledge of a topic than none. The toughest part of learning something, like so many other cases, is in getting started.

Also, I can't be sure pending research into this, but I would be surprised if these reports ended up worse than the typical news story attempting to cover such things.

I think the more important part of that sentence is "or misleading".

"We don't know what happens if you drink cyanide." is different than "You should drink cyanide."

I think I've addressed your concern below.

I think it's amusing that your backend is still human, because really you're just shifting the task onto others, not necessarily solving it.

We've "solved it" per se on Experiment by getting good at training other scientists to get good at this. Unfortunately, that's also still pretty technically hard, but we've developed some editorial heuristics.

Like a real NLP + UpGoerFive approach would be awesome and a real value to the science community, but probably very difficult. :)

   "but probably very difficult."

Unless I've really missed something, "probably" here means with near certainty enough beyond our current technical capabilities that predicting a timeline is folly.

It's getting closer

Data61 (ex-NICTA) in Australia have a mostly-working thing that parses legislation, compiles it[1] and then checks privacy policies and shows violations[2].

[1] Yes, based on conversations I've had I believe they actually turn it into a byte-code like representation.

[2] http://www.csiro.au/en/Research/D61/Areas/Data-for-decisions...

It is not worth the engineering effort at the current moment in time.

I really think your on to something here. People must be advocates for themselves and keep up to date in the areas of research where their medical problems exist if they want be sure they are getting the best care and treatments.

Absolutely! Patient education is one of the biggest issues around treatment in my opinion

This is generally an awesome idea. The value add is really a jumping off point that can equip you with enough understanding to begin to read about the issue overall.

An additional glossary for terms most important to that research area would also be helpful.

Adding that to our to do list. good idea.

Interesting!

Random thought: If this catches on in a big way, have you thought of how the system will deal with the fundamentally political nature of some areas of medical research? Real effort is going into trying to subvert this process at the source, so why not the summary?

We'll deal with that if it really catches on, but if that happens I'd say it's a better problem to have than no one wanting it at all :)

If you have any ideas we'd love to hear them though!

Citation graph analysis might be useful, but I'd have to think about it a bit.

I suspect you can make a nearly risk free bet that if there is a significant amount of money being made that depends on the results of a contentious issue in the existing science, there are deep problems in the related literature.

How about CRISPR? There is a tremendous controversy right now over who should get recognized (Nobel? Patents?) for CRISPR's invention

Looks interesting. I'm curious about the ways a user can visualize the summarized information.

Is it just via paragraphs (like in the one example on your site), bullet points or some other manner? Does each paragraph equal one sentence or do you look for hard-to-understand terms and only summarize those?

On a side note, if any readers know of all the best/varied ways to summarize information ('manually', not via algorithms), I'd be very interested in seeing examples, particularly of news articles and research.

Essentially we get individuals with medical expertise to make an abstract that is readable by every day people. Some of these you'll read "the common carotid artery" but in laymen's terms is "blood vessel in the neck", we make this new research more accessible to the every day person

I'll be interested to see how this works - I submitted something, and I'll post my thoughts when I get your summary.

One question: How do you deal with paper links that go to non-open access publications? Have you paid for the usual journal subscription bundles? Relying on crowdsourcing?

We'll either purchase it or look for someone with access :) The crowdsourcing certainly helps!

Where is the privacy policy?

Information about diseases people search for is valuable to a lot of third parties.

Definitely need to add that.

At the moment, we do not make that information public externally.

Really interesting idea!

Are the summaries entirely free? If so, what's in it for the summarizers?

We pay them.

We get great SEO value out of publishing the long tail of summarized research.

What is the pay?

$5/summary.

It ends up being about $15-$25/hr.

Interesting. Did you guys pivot from helping people find clinical trials?

The clinical trial recruitment is actually the business model behind this. We're using this as a way to attract patients to our brand.

Reply on news.ycombinator.com