lundi 15 janvier 2018

Mnemosyne status update, 5 years of use



Mnemosyne project status update

14 January 2018

I think it is time I summarize my experience with Mnemosyne again. It has now been approximately a year since I last blogged about this spaced-retrieval program.

I now have 5950 cards with 1-20 questions each (new cards usually contain 1-4 questions). They are spread out in the following main categories:

·         Anatomy and physiology (408 cards)
o   Contains pictures of diseases (Lyme, Mumps, Rubella), questions about hormones (serotonin, aldosterone, etc.), body parts and organs (bones, arteries, heart, muscles, etc.), systems (infectious, inflammatory, etc.), normal biological values (K, Na, Hb, TP-INR, heart rate, etc.).
·         Health (1438 cards)
o   Contains questions broadly related to health (e.g. definitions, public-health, motivational interviews, cyber health, Swiss health system, health promotion, history of medicine, perspectives, laws, etc.). Also contains cards about techniques, physiology, research and diseases which I didn’t bother to classify correctly. Admittedly all decisions are arbitrary and many cards deal with multiple domains.
·         Social work (262 cards)
o   Contains questions related to social work (e.g. sexuality and intellectual handicap, challenging behaviors, behavior management, conduct disorder, how to plan activities, health of migrants, how to behave with deaf-blind people, social work associations, etc.) and education (dyslexia, dyspraxia, ADHD, learning difficulties).
·         General knowledge (139 cards)
o   Contains questions that I couldn’t fit in other categories (e.g. what is copyright, gamification of activities, Chinese words to distinguish, English words to pronounce, DNA and crimes, etc.).
·         Education/teaching (165 cards)
o   Contains questions related to education (e.g. effective ways to use homework, inclusion, evidence-based learning strategies, educational institutions, spaced learning, dyslexia, thinking skills, autism, giftedness, growth mindset, Montessori, etc.).
·         Pathologies, symptoms and diagnosis (809 cards)
o   Contains questions related to diseases or diagnosis (e.g. diabetes, obstructive sleep apnea, schizophrenia, pneumothorax, addiction, heart failure, etc.). Most of them contain at the very least some key items such as: definition, prevalence/incidence (in general population), etiology, signs and symptoms, diagnosis (criteria, thresholds, diagnostic options, LR+/LR-), differential diagnosis, treatment (pharmacological, non-pharmacological, surgical, therapy, 1st/2nd line), prevention, prognosis and complications.
·         Psychology, sociology and culture (105 cards)
o   Contains questions related to psychology (e.g. emotional intelligence, coping strategies, cognitive development, adolescence, culture, validation, labelling, couple theories, neuromyths, Kahnehman’s ideas, etc.)
·         Scientific research (2006 cards)
o   Contains questions related to research and evidence-based medicine/practice (e.g. study designs, how to flaw/bias results, biases, how to read a meta-analysis, the GRADE approach, how WHO guidelines are made, steps in doing a systematic review, peer-review, reporting guidelines, key flaws/errors in the literature, blinding, randomisation, why pre-register studies, designing randomized trials, how to teach EBM, definitions, future of health research, criticisms, myths, managing trials, story of health research, etc.).
·         Drugs and substances (309 cards)
o   Contains cards on drugs and substances (e.g. paracetamol, lidocaine, warfarin, iron, melatonin, epinephrine, adenosine, pantoprazole, fennel, etc.). Some cards are meant to link names to substances (e.g. Sintrom -> warfarin) and others focus on contra-indications, side-effects, details on how to use, dosage or antidote mainly.
·         Techniques and protocols (204 cards)
o   Contains cards related to techniques (e.g. clinical examination of ear/nose/eye/etc., dressing a wound, wound care, transfusion, drawing blood, administering oxygen, catheter care, EKG, CPR, suicidal risk, etc.). Those may ask about the steps/how to, the equipment required, the indications/contra-indications or risks.
·         Untagged and retracted cards which are no longer used (238 cards)
o   Contains cards I no longer deem useful (e.g. psychanalysis, nursing theories, techniques I never use in practice and don’t feel will be needed, etc.), cards which were too complicated (20+ questions in a single card or so) and, I notice, cards which shouldn’t have ended up there (wrong decision in retrospective? Error?).
·         Other

It looks like I have roughly doubled the amount of active cards in a year, with about 1800 new cards on research, 500 new cards on diseases and 500 new cards about health.

I usually answer 60-80 cards per day (often before going to bed, as part of my routine), which takes approximately 1 hour. This is twice as many cards as I used to do a year ago in the same amount of time (but cards are now much shorter with less questions). I will get 50-80 cards per day in the following month (it seems). I do not use cramming functions.


I have added what looks like 30 cards on average (1-70) every 7-10 days (I often add cards during weekends, which explains the 7 days intervals) with a spike 6 months ago (a transition period where I had more free time). I usually write the contents of cards during the week on a Word document I keep open and add them to Mnemosyne in bulk afterwards. While Mnemosyne keeps telling me I probably shouldn’t add 15 cards at once, I never care about the warnings and haven’t had problems so far. It is sometimes quite funny to see how by the time I add the card in Mnemosyne I have entirely forgotten the answer.


My statistics tell me I have answered cards almost every single day (350/365 days?) in the past year with, on average, 90% retention/correct answers. It’s actually not entirely correct since there are often many questions on a single card and I may grade it as “correct” if I miss a question I don’t think is important or if I read the question differently than expected (at which point I also sometimes restart Mnemosyne since “skip” isn’t an option). I also sometimes offer different answers than those expected and may decide that both options are correct. I wish Mnemosyne displayed statistics over many years, which could perhaps allow me to see long-term patterns (decline in memory with age?).


When looking at overall grades, almost all (4622) cards are grade 3 “correct but with much effort”, a decent amount (1498) are grade 4 “correct with some effort”, a few (58) are grade 2 “barely correct answer” and almost none (7 and 1) are grade 1 and 5. The numbers don’t add up to 5950 as untagged and retracted cards are also included in the analysis (there is no option to keep more than one card category). One hypothesis may be that I do not make very difficult or very easy cards. Perhaps I only try to learn things that aren’t too difficult. My personal opinion is that I simply almost never consider that I know something really well (grade 5, “correct without any difficulties”) and have a hard time considering I know things well (grade 4). Also, I may be afraid to grade cards incorrectly and not see them again and grade 3 seems like a safer option than grade 4 and 5.

Most of the cards feel very easy to answer with a couple rare exceptions (e.g. I keep confusing some drugs with very similar names such as alosetron/ondansetron/olansetron, Chinese words are still confusing and questions with many answers such as telling all the 10+ complications of a disease can be a pain). 

I am now trying to learn and remember the certainty associated with results of studies (such as “low certainty drug X reduces mortality with disease Y”) but haven’t yet found it satisfying as doubts still linger and I repeatedly make errors. Having read the full study doesn’t seem to make a difference.
Learning likelihood ratios associated with diagnostic tests is also quite challenging at times (was it LR+3, LR+3.4, LR+10?) and it is interesting to see how I now tend to assess my answers with margins of errors (e.g. the answer is LR+ 5, if I thought it was LR+3, LR+4, LR+7 I would still judge my answer reasonably close and correct as it wouldn’t make much of a difference in practice).

I wonder if re-reading research books would lead to more new cards (new insights, new/better/different understanding) or if it would be a loss of my time. I may have to try one day.

I still wonder how to best keep the cards up to date (e.g. new treatments for disease X, treatment Y now considered useless and risky) without spending too much time doing that. How often should I check if things have changed? I will likely run into trouble if one day I learn that something I have learned is wrong (e.g. pharyngitis is actually a cancer) since I may have used various wordings to talk about the same thing and related questions may be spread out over many different cards. Advanced searching options may then perhaps be of help, especially with 5000+ cards.

One may then think about creating a centralized system which automatically updates cards with new current knowledge (depository of knowledge?). Kind of like Wikipedia automatic updates when new Cochrane systematic reviews are published. Could be neat, my cards would be up to date all the time. Or super-dangerous if those managing the database are conflicted, biased or make errors. Also can be difficult with interpretations of questions/answers, we may then need basic templates which get re-used or training to all understand them the same way.

I remain a bit concerned about the future of the Mnemosyne project considering how few people seem to be working on it. If I plan to keep using such programs for at least the next decade I may have to make plans in case development is halted. 

An option to "divide" a card in two unique cards could be useful, particularly when there are, say, 5 questions in a card and I know 4 very well and 1 not at all. Making a fresh new card makes it look as if I had never seen the question before.

I do keep regular backups of my cards in case I lose access to my laptop (broken, stolen, lost, not working, etc.). The thought of losing years of effort isn’t pleasing but I’m not really afraid of that, it’s a possibility I acknowledge.

Some last thoughts... Am I using Mnemosyne in an efficient/effective way? I don’t know. Is it really helping me remember things? I don’t know. Is it harming me? I don’t know. Maybe one day I may have answers. In the meantime every day is an experiment.