One medical student's simple approach to becoming a doctor
Evidence-based Dating: the Johnson-Purdy Nomogram

Evidence-based Dating: the Johnson-Purdy Nomogram


So, I’ve never really understood dating. I have a complete inability to figure out what is going on. Anybody who knows me well, knows this to be the case. I do, however, like to think that I understand the basics of evidence-based medicine. I can often figure out what is going on and I often get a bit too excited about the topic.

As such, it recently fell to my good friend Heather Johnson to explain the world of dating to me in terms that I might understand. Luckily, she often knows just how to do that. Enter evidence-based dating (EBD) and the Johnson-Purdy nomogram:

Screen Shot 2014-05-02 at 1.49.24 PM

For those who are unfamiliar with the Fagan nomogram, it is a tool used in clinical medicine to facilitate understanding of the result of a diagnostic test in the context of a single patient. Clinicians create an initial impression based on clinical gestalt or risk stratification tools, then they apply a diagnostic test. The result of that test either increases or decreases the likelihood of a patient having a disease. The degree to which the test will change your opinion depends on the operating characteristics of the test at hand.


How does this tool apply to dating?

Let’s say you meet someone. Maybe you find him/her attractive. Maybe you have a witty or meaningful conversation. Based on this interaction, you feel that this individual is a potential romantic partner. We can call the initial impression the “pre-test dateability.”

But how are you to know whether or not you are likely to have a successful romantic relationship with this person? Not only are there many, many variables involved, but first impressions can be misleading. You need more data.

Though Facebook stalking and subtly questioning mutual friends (i.e. gathering collateral history) can be useful, this approach will only get you so far. In order to truly determine whether the two of you have a shot of making this whole relationship business work, you actually need to spend time with them. This means actually going on a date.


The High Sensitivity Date

All dates are not created equal. Different activities have their own properties – sensitivities and specificities if you will. Take for example going for coffee, a standard (or so I’m told) first date activity. This encounter has a high sensitivity. It is the dating world equivalent of the CRP in the presentation of joint pain. You are unlikely to miss any potential suitors with this initial screening date but if you can’t stand each other for the short time it takes to finish a cup of joe then there probably isn’t much of a future. If it was tedious as hell then the possibility of a relationship has been effectively ruled out. SN(out) for the win!

Unfortunately, the first date coffee has a high false positive rate due to its poor specificity. You might enjoy an afternoon coffee with many people even if you don’t want to date them in the long run. Let’s take a minute to review the formula for a positive likelihood ratio:

Likelihood Ratio

A high sensitivity combined with a low specificity does not make for a strong positive likelihood ratio, and as such the post-test dateability will not change much from the pre-test dateability. Luckily, this is an iterative process and we can apply further testing.


The High Specificity Date

On date number two you may opt to go to dinner. The stakes are somewhat higher as you are committed to the activity for at least the length of the meal; the more time you spend together, the more likely it is that the conversation will turn to discriminating topics. Having the dinner at someone’s house as opposed to a restaurant will further increase the specificity of the encounter.

Screen Shot 2014-05-02 at 3.08.10 PM

A word to the wise – I have been told it is possible to be too specific at this early stage in the dating game. The couple who travels together can do almost anything together, so if the two of you survive a day-trip then things bode very well indeed. That being said, many individuals in successful relationships will be less-than-enamored with their partner after several hours in a car. I have been instructed to beware of the early false negatives that come along with high specificity dates!

As we know sensitivity is often compromised for increased specificity. Highly sensitive tests are used for initial screening then more specific tests can be applied to rule in conditions- SP(in). The same basic principles of EBM may be applied to dating. The more dates you have with a person the less sensitive and more specific the encounters need be.


The High Sensitivity and High Specificity Date

Based on information from my available resources, this date does not exist. If it does, please share your wisdom.

The ideal date would be highly sensitive and highly specific putting it in the top left corner.

The ideal date would be highly sensitive and highly specific putting it in the top left corner.










The Bottom Line

In EBM, the right evidence needs to be applied to the right patient in the right way. In the case of likelihood ratios a test’s result must always be interpreted in the context of the pre-test probability. The same would seem to hold true for EBD. At the end of the day though practice never really makes perfect. You can theoretically understand EBM and EBD until you are blue in the face, but each new patient and each new prospective partner will bring something unexpected to the table that does not fit into one’s current understanding of medicine, the evidence or life as you know it. Like most, uncertainty makes me a bit uncomfortable but I have been told to “just go with it.” I’ve also been told not to “over think it” but I think that ship has sailed.

So, I guess epidemiology and dating aren’t so different after all, except that medical students spend a lot more time and energy on one. I’ll just let you guess which.


Stay tuned for some more evidence-based approaches to life.


Thanks to Heather Johnson (@hnj26) for co-authoring, and to Heather Murray (@HeatherM211) for reviewing this post.

The following two tabs change content below.
A fourth year medical student with interest in all things medicine but especially emergency medicine, rural medicine, medical education and social media in health care. While not clerking it out, she is usually running, playing guitar or planning an outdoor adventure.

Latest posts by Eve Purdy (see all)

15 Responses to Evidence-based Dating: the Johnson-Purdy Nomogram

  1. This piece reminds me of this TED talk…

    Once you know the game, you can play it waaaaaaaay better. :D

    This applies to medicine, exams, life in general.

    Thanks for making me laugh (quite literally) out loud.

  2. OK- stretching my EBM brain (since dating brain barely existed and is now defunct) – but doesn’t prevalence play a role here somewhere? Low prevalence of date-ability affects sensitivity and specificity? In theory (cough, cough, cough) that was possibly what happened during my engineering degree… or so, I read, in a book, somewhere, maybe. :)

    • Eve Purdy says:

      That’s a great thought, though as it turns out though sensitivity and specificity do not change with prevalence :( – they are characteristics of the test no matter what the number eligible prospects is in a given population.

      What does change with prevalence however is the positive predictive value. That’s why we are more a fan of the likelihood ratio which is not affected by the prevalence of date-ability.

      The only way to change the sensitivity and specificity of the test would be to change one’s standard. To increase the sensitivity, lower them and to increase specificity, raise the bar.

  3. Right- I knew that and should have not rushed my answer. Was busy having traumatic flashbacks. :)

  4. OK- brain engaged now. I remember this confusing me before, and it’s not totally straightforward so I looked it up like I should have done before. In some cases – sensitivity and specificity CAN be affected by prevalence. Here: or . As I recall, (perhaps wrongly?) this may vary depending on the test- some might be straight “yes/no” tests that are less likely to be affected by prevalence, but for other tests, some judgment is required as to where to set the cut off, which is what then gives the sensitivity and specificity. In my world, I remember this when we look at scores on cognitive tests – someone has to decide what the cut off is for positive and negative screens, and that decision may well be affected by prevalence. Now, before I embarrass myself further, I’ll sign off and get back to other matters … and I am no where near clever enough to figure out how to apply this to dating.

    • Eve Purdy says:

      You are very correct! At some point, someone smart has to choose cutoffs to make it a binary test. This is where the receiver operating curve comes in .

      I think it is very fair to say that depending on the population the threshold of cutoff for a positive vs. negative test could change. For example if you are around many eligible prospects (let’s say… at the med-law-mba-nursing mixer, or at the gym, or while engaging in activity that you enjoy) then you would be willing to settle for a lower sensitivity and increase specificity. But if you happen to be in a desert in the middle of nowhere then you would arguably need sensitivity. But shifting along the ROC curve requires that one shifts his/her cutoff standards- a very personal and serious choice.

  5. Sorry I didn’t comment earlier, got a bit busy! Here’s just a few quick thoughts.

    As a physical scientist, I tend to think of romantic relationships as chemical bonds. (I was just working at a hospital, where I was for the 1st time introduced to ROCs!) The analogy holds fairly well.

    Attraction decreases exponentially with distance- both for electrons and humans (ie long distance relationships rarely work out). Touch is very important in romantic relationships- it helps the pair bond solidify, and hormones released upon touch modify neural chemistry. (I’m pretty sure there’s a few papers on this, but I am too lazy to look them up.)

    There must be significant orbital overlap- symmetry requirements- for the bond to form between atoms. That is, I like to have a lot in common with a guy to date him. Now that I’m a bit older and out of grad school, I tend to only date guys that know more quantum than I do. This is useful for fending off engineers/lawyers/MDs in bars, and selects for physicists; who generally find to be extremely amusing. (My past BFs have mostly been physics PhDs. As a chemist, I find it handy to have people who know more linear algebra than I do around.)

    Bonds can be ionic, with one highly electronegative atom hording all the electrons; or covalent, with the electrons evenly shared and delocalized over both atoms. I think everyone’s seen relationships where one partner clearly has all the power- an ionic situation- and ones where each partner contributes equally in a more covalent situation.

  6. Gina E says:

    hilarious read, and to add some anecdotal commentary :)

Leave a Reply