These Two Tests Can Immunize You From The Social Media Freakshow
With the offshoot that you will understand your medical tests better
It’s all mathematics
— Yasiin Bey (formerly Mos Def)
My friends often send me lab tests they received from various health facilities for me to interpret.
It hardly helps to tell them that the information they share is like giving me three pixels and asking me to paint the picture for them. The easier option is to reassure them that they are okay, because most of the time, they are.
Tests have features about them that even medical doctors find difficult to interpret. This realization was made clear after reading Gerd Gigerenzer’s Risk Savvy.
I was in my final year in medical school, reading books I shouldn’t. While my classmates carried clinical texts, I pursued my variegated intellectual appetites. Gigerenzer’s book, however, introduced me to a feature that is hardly stressed enough in medical schools — tests.
Patients and clinicians will insist on the relevance of tests, but not ask further questions about them. Those who do are a rare breed. In my internship, I encountered one such professor. To give him his roses, Prof. Chakaya takes keen note of the sensitivity and specificity of tests before interpreting them. Most of the other clinicians quickly jump to a diagnosis; they try to prove they are right rather than question whether the test is reliable.
A test can be as simple as attaching leads to a chest and checking the heart rhythm. Countless times, I have been called by a worried nurse, only to find that they place more emphasis on what the monitor relayed than on the patient’s clinical state. That reaction they portray is no different from how individuals react to social media posts.
Kristy Scott recently filed for divorce. Frankly, I didn’t know who she was. It became more than a coincidence when I saw two people post it. Then three. So I went online. A single search was enough. Social media personality. Digital Creator. Filmmaker. Forbes Top 30 under 30.
Tells you a lot about me — hardly ever on social media, with good reason. It’s the same reason I don’t usually watch the news. My razor: If the news is important, it will get to me; if not, I don’t bother seeking it. It has never failed me. Now, back to Kristy.
The reactions from other social media influencers were somewhat expected. To keep your audience entertained, your reaction needs to be extreme. A video, preferably, interspersed with reactions that you would never do in real life. Then ending with a conclusion that should seem obvious. After the Kristy-Desmond announcement, the general conclusion was: I am never getting into a relationship ever, or, I am keeping my relationship private.
These conclusions, taken from one-off samples, are often validated in the comments section.
It’s hardly different from other crazy incidents. I recall the craze of Sanpaku eyes. It was so big that the cast of Too Early For Birds included it in their script for effect.
It’s hardly ever enough to convince anyone that most of what is shared on social media is an extreme version of reality. You will not find the happily married constantly sharing their life online. And the reality is, marriage will not always be rainbows and daisies. Content consumers will more likely agree with what influencers say and spread the gospel to their lives and those of their friends.
It is not surprising that social media influencers are the most vocal ones regarding mental health, with only a small fraction giving helpful advice. The concept creep of words such as depression and toxic applies easily to many situations, so much so that arguing against it is one step away from being labeled an enemy.
I, therefore, want to try something new. A more practical approach. It’s easy. And if you have read this far, you should understand that I have baptized you with insight and tolerance, enough to understand the simple math that follows. Even kids understand it. I don’t think you’re a kid. Additionally, you will get an invaluable lens into medical tests. Practically, we’ll be hitting two birds with a single stone. As Yasiin Bey raps, it’s all mathematics.
Natural frequencies
Now, I want us to think about a test and what a positive one means. Consider the COVID-19 test. Say someone actually has the disease, but they don’t know if they have it at all. That is one half of the story. A test is needed to confirm these suspicions that they indeed have the disease. That’s the other half. The test results, therefore, need to be interpreted in context.
Maybe not the best analogy, but these two parts are similar to flipping a coin. You, the coin-flipper, know that heads is up. You then ask someone to guess if the coin is indeed heads. That is what medical tests do.
Let’s assume someone has an early-onset breast cancer. They then visit a clinic because they noticed a lump in their right breast. Since they are below 40 years, they go for an ultrasound. Let’s break down the numbers in this fictional world we have created.
1. The prevalence rate of breast cancer is 10%. It means in a village of 100 people, 10 of them have the disease.
2. The sensitivity of the test is 80%. Sensitivity is a test of true positive — that is, if you have the disease, there’s an 80% chance that the test will pick it. It is therefore called a true positive test.
3. The chance that the test will come back positive, even though someone does not have the disease, is 10%. These special individuals are known statistically as false positives.
When clinicians are presented with these figures and asked the probability with which they think the patient has the disease if the test comes out positive, they will often give a figure that is wildly out of tune with the truth. Most will say 80% because of the sensitivity. This group will not stop to wonder why the prevalence rate was given. A mistake. Others will say 8%, by making a product of the sensitivity and the prevalence. This is the cue for you to pause, calculate, and find if you’re any different from the clinicians.
The only other clinician I have ever known to ask for something else is now dead. He has been for years now. It was Professor Hassan Saidi. He was my mentor when I was breaking into science and research. He is the only person, to date, who has ever asked me what the difference is between sensitivity and positive predictive value. As I continue to live on this blue planet, no other person has paused to ask this about a patient whose diagnosis, of breast cancer, in this case, will shape their lives from that point onwards.
Some labs will be so keen as to mention the sensitivity and specificity at the bottom. But these hardly help if we don’t know the prevalence rates. Enter natural frequencies.
A natural frequency ties two events into a single number. The coin toss and the question to confirm if indeed the coin shows heads or tails. These are two distinct events. In our case, the two events are:
We have a positive test, and
The chances that the patient has the disease.
If we’re to use the percentages, it becomes difficult. Especially since we’re dealing with a single individual. Statistics needs numbers. Applying them to an individual is difficult. And that’s the key. Rather than use a single person, let us imagine a bigger number: 1000 is a good starting point.
The prevalence is 10%. It means that out of the 1000, 100 have the disease. The remaining 900 don’t.
The sensitivity of the test is 80%. Among the 100 who get tested, 80 will return positive. These are the true positives. The remaining 20 will come out as negative (false negatives).
Of the remaining 900 who also get tested and who will turn out positive, even though they don’t have the disease, is 10%. That is, 90 individuals. The remaining 810 will be negative. That’s the true negative. The test of true negatives is called specificity. However, the 90 individuals who came out positive will have the false diagnosis — false positives.
Take a moment to consider the two samples. False positives (90) are more than true positives (80). The question we ask clinicians is similar to the question we ask the coin-tosser. If the coin turns out to be heads and is indeed heads, then what we have is the positive predictive value. In the patient’s case, it answers the question of whether the patient has breast cancer if the test comes back positive.
To get the positive predictive value, we take the true positives (80) divide it by the sum of all the positives (false and true positives — 90+80 = 170), and then get the percentage by multiplying it by 100.
This result is very different from the early responses of 80% and 8%. A 47% predictive value means out of 100 people who test positive, 47 have cancer. The remaining 53 don’t. Imagine that, 53! That’s a big number. That’s 53 people who believe they have the disease but don’t have it. More than 50% of all who test positive will have their lives changed. Seated behind the desk is a clinician who will interpret it as it comes — that the patient has the disease.
Fortunately, clinicians have several redundancy systems in place to be cock-sure about certain test results. An ultrasound is not the definitive test for cancer. A tissue sample of the growth must be viewed under a microscope by a specialist pathologist to confirm that it is cancer. That is the basis for several tests — redundancy. A single test is not usually enough. Something as simple as your blood pressure falls under the same category. The cuff acts as the test, which is to be done at least three times, when one is calm, to confirm a diagnosis of hypertension.
At this point, you may ask why I included the prevalence rate. Let’s use another example to relay the relevance. All the other figures can remain the same except for the prevalence. This time around, let the breast cancer have a prevalence not of 10%, but 1%.
In the same village of 1000, only 10 will have the disease. The remaining 990 don’t.
A sensitivity of 80% means 8 of the ten will test positive — the true positives. The remaining 2 will not.
Of the 990 remaining, 10% will still test positive for the disease. That’s 99.
Notice the difference. From a single test, with unchanged sensitivity and specificity, when the prevalence rate goes down, the ones who suffer the sad news from the clinicians are those who don’t have the disease. True positives are 8. False positives are 99.
The positive predictive value now changes:
Of 100 people who test positive, those who likely have the disease despite the positive test are only 7.
The conclusion is very clear: the sensitivity and specificity of a test need to factor in the prevalence. In its absence, the interpretation will be off the mark.
Natural frequencies are an easy way to understand the interpretation of a result. If the calculation is too difficult, the simple rule of thumb is to ask: What is the prevalence rate? This can be easily gotten from a quick online search. The more geographically specific the prevalence rate, the more accurate the interpretation.
If the prevalence is low, a positive test needs to be substantiated with other tests, as it has low positive predictive value. If the prevalence is high, the positive predictive value starts to gain more weight in interpreting the result. Also, having other tests to substantiate a positive one is helpful. For instance, an HIV diagnosis. Several tests are needed before you can label anyone positive.
This was one of the biggest issues I had with the COVID-19 testing. Those who would go to get tested are unlikely to have the disease. By extension, it means the prevalence rate, based on the foot traffic to a health facility, is likely to be lower than documented. It would also mean that the people without the disease walking into health centres or purchasing home-based tests are likely to experience false positives. However, the upside was that, unlike cancer, one can take the medication and quarantine for less than a month and get a retest done. Other diseases, such as cancer, cannot be very forgiving.
Now that you understand how tests can be biased in their interpretation based on prevalence and positive prediction values, we can step into social media.
Social media encourages more false positive diagnoses
It’s 1 universal law, but 2 sides to every story
— Yasiin Bey
As a rule of thumb, 99% of the content shared on social media is contributed by 1%. In essence, it’s the freaks who post. I’m one of the freaks who writes daily.
Taking this into perspective, it means the prevalence of all the shared content is not likely to be true by a wide margin.
Important disclaimers — journalists and investigators might also report their findings regularly. When the Nobel committee announces the year’s winners, it cannot be refuted. Context is therefore necessary.
Nevertheless, not every day do we have the Nobel laureates announced. For the remaining days, we have others taking over. In particular, influencers. In reality, the 1% of active creators or posters on social media is a generous number. It could be smaller. But let’s work with 1%.
Recall the marked difference between the prevalence rates of 1% and 10%. In the former, the false positives were significantly more than in the latter. If social media has only 1% posting, then the numbers are likely to be skewed in this manner.
A reaction by an influencer will mostly give the wrong impression or conclusion that is out of touch with reality. Several of them made their reactions about the Kristy and Desmond split, and concluded that long-standing relationships don’t exist anymore. But with what we have just discussed, we can see how deeply flawed these conclusions are.
But let us run it even further. Celebrities have levels. Among comedians, for instance, there are Dave Chappelles and Kevin Harts. These are mega stars compared to someone like Taccara Williams (and don’t get it twisted, I love Tacarra. If she were to come to Kenya, I’d get her tickets without batting an eye). Thus, among influencers, the 1% who talk about the 1% of celebrities takes our prevalence rate even lower. It becomes 0.01%. It means that the positive predictive values are extremely dismal.
Regardless, these figures will not show up on the screen when consuming their highly sensationalized videos. I saw one of the Kristy-Desmond reactions and laughed. Influencers can be hilarious. It’s in their name and job description. That is the point. The message this particular influencer finished off with was that relationships are a farce. Loyalty only resides in the past. We now know that, from a numerical point of view, it is poorly interpreted.
Freya India passionately described the fear Gen Zs have of getting into relationships. While she expresses worry, she tries to encourage her readers to take the risk. Relationships are, after all, bets. You make the wager that someone will try to show up regularly at home and offer support for as long as you live. These stories are not nearly as engaging as the extreme versions of gossip, rage, and humour we see on social media.
A decision as important as that of picking your life partner should not be dampened by a single social media post. I have shared the numbers to help you think it over. It does not, nevertheless, mean that you should stick around in an abusive relationship. The tests of prevalence rates and positive predictive values are a dose you need to constantly remind yourself about to become immune to the misinterpretations that social media can pretend to share.
Like vaccine shots, it always helps to think of the small percentage of creators who contribute to our online experience, especially on something as sensitive as a relationship. Once a post goes up about a breakup, one may begin to wonder if indeed their relationship is any different. If celebrities are doing it, why shouldn’t I? That, right there, is self-diagnosis in the absence of a verifiable and reliable test.
Tests such as medical tests are internationally recognized and approved because they impact individual lives. They have to meet the standards of reproducibility for them to acquire widespread use and reliability. Diagnosing your relationships doesn’t have that advantage. It means that one becomes a victim of our brain’s evolutionary wiring — confirmation bias.
We tend to confirm rather than disconfirm. It is easier. Disconfirming by definition leaves one with the burden of proof. Our minds were not fashioned to find the hard truth. It was fashioned for easy heuristics, for ease over drudgery, for the short ways rather than the rigorous and smart option. No reliable test.
Why that is the case, I argue, is that organisms seek to avoid annihilation and not just to reproduce. Better to avoid getting into a relationship and getting your heart broken. Better lock yourself in a room and enjoy porn than step out on a Friday evening and interact with women who would potentially become your partner. Better to fight online, distanced by technology, than confront individuals face to face, where you risk a punch in the face.
Social media posts may augment this tendency. We are, furthermore, induced by stories more than numbers. As Roger Penrose once confessed about numeracy, this piece may lose readability just because I included numbers, and by virtue of it being a long-ish essay. In contrast, a TikTok viral video showing the reaction of a power celebrity couple will gain more views. Numbers are a recent innovation. Stories have and will always be there. I cannot deny reality.
It’s why I compare these numbers to a vaccine shot. They keep you immune for as long as there are no posts that you encounter hyping a gossip, or a wave of influencers celebrating or mocking the split of a couple. The situation may be worse now that social media is awash with bots who know just how to hijack attention.
And this last one takes me to mental health.
Mental Health
I don’t like how easy it is for anyone to raise the topic of mental health and have immediate backing or “justifiable” reasons for acting the way they did.
You can be sad without being depressed. You can be loved without being love-bombed. You can be disappointed without feeling like the world is against you. You don’t always have to believe that you are a victim.
In a space of low prevalence, false positives are more than true ones. Mental health is rather new. We don’t have any tests for it. Psychiatrists don’t even know much of what they try to figure out, as much as they wish for their patients to improve. Therapists are taking the role of parents. AI is becoming an easy go-to solution when you have a bad day. These are all recipes for self-diagnosing, hijacking our confirmation bias.
Loneliness is lucrative. From an economist’s point of view, anyone who makes a killing from this crisis would not want it to end. The pathologization pandemic then continues, making it difficult to confidently say that the mental health status of individuals is rising, as opposed to the ease of self-diagnosis, layering up the already worse state of overdiagnosis.
Presently, because we do not have such reliable tests, we cannot know the acumen of the one sitting opposite you, or even whether the AI that you consult has a high sensitivity or specificity. And even if they did, our prevalence rates may be difficult to substantiate and defend. We may have high prevalence rates of loneliness and poor mental health, but we lack the tests to back this up with high sensitivities and specificities to diagnose individuals. As a result, we have poor positive prediction values.
This is a recipe for making a tonne of money for anyone who wishes to make the most of our current situation. My country was recently listed as the one with the most viewership on TikTok. I wonder what impact that may have on the young generation. The older ones might already know how the real world operates, but those who have been brought up by rectangular screens might not.
A clear heuristic would be to remember the rates I had earlier mentioned — the 1% of creators who post and the 99% who consume. It is easy for an influencer to do just that, to influence you, but in the wrong way, intentionally or otherwise. In the absence of these validating tests, it may be useful to use the rule of thumb.
Since most of us self-diagnose, we can just as well self-vaccinate. Using natural frequencies is an easy way. For those more interested in understanding the numbers game, the book by Gerd Gigerenzer may be helpful. For those who can’t flip pages, a reminder may be:
1. If the prevalence is a single digit, then take the conclusions with a pinch of salt. Often, the freakshow on social media lies here.
2. If the prevalence is double digits, several other independent tests might clarify the initial diagnosis. They must be reliable tests. Medicine lies in this province.
3. If there is a tonne of self-diagnosis, then you are better sticking with rule number 1 than 2.
We need to intentionally build resilient systems. Resilience is a stoic practice. Gurwinder’s blog is a fantastic example that can occasionally guide your online experience and create mental tools to screen ourselves from the manipulative forces that lurk behind most online platforms.
We are all vulnerable, myself included. So we need to remind ourselves often, as stoics do. I would recommend works by Stoic Philosophy and Ryan Holiday’s Daily Stoic.
A low prevalence of behaviour such as cigarette smoking is good. However, when it comes to matters as serious as your diagnosis, the low prevalence of the same disease should be taken with little emphasis, unless it is supported by other tests. It’s all mathematics.
What I’m trying to say is…
We may have a way to shield ourselves from social media with the double benefit of protecting our mental stability and reducing visits to medical facilities.
Additionally, we may have a way to save our cognitive well-being by getting a cleaner picture of reality. Natural frequencies are one such option. There may yet be others. What I love about natural frequencies is how a single fact can get a completely different look based on a different lens.
Dana Meadows listed 12 leverage points. Numbers, she would say, are the least useful areas to improve a system. The most important is transcending paradigms, and that is what natural frequencies do so beautifully — taking numbers and viewing them in a new light.
The power does indeed lie within us.
This song inspired some of the lines used in this article. Source — YouTube