Predicting Correctly Seems More Often Luck Than Skill. I’m Making Some Healthcare Predictions Anyway.

by Jonathan A. Handler, MD, FACEP, FAMIA

As of this writing (2025), there are about 8 billion people in the world. Let’s say that every person guesses the likelihood that “the stock market” (let’s say the S&P 500) will be up or down from one quarter to the next. And let’s say that this is done purely by luck — a coin toss performed each quarter by each person in the world. What are the odds of getting all but two of the quarterly results correctly guessed (up vs. down) for eight straight years (correct for 30/32 quarterly results)? Answer: about 1 in 8 million. So, it might not be unreasonable to expect about 1000 people in the world to look like unbelievable financial geniuses of the highest order, simply through luck.

So, when I hear that someone has made some correct predictions, were they lucky or good? Baseball Hall-of-Famer Lefty Gomez reportedly said he’d rather be lucky than good. However, if you are putting your trust in someone’s predictions, you might prefer knowing if they are actually good at predicting or just waiting for their luck to run out. In my case, I’d like to think I’m good, but:

1. I’ve made wrong predictions.
2. For my correct predictions, how would I know if they were lucky or good?

So, my recommendation is to assume that any of my past successes at predicting were all luck and read no further. If you do read further:

1. I declare all these predictions are for entertainment purposes only (at least my own) and you should not trust them.
2. Also, don’t shoot the messenger! Please don’t mistake a prediction of what I think might happen for what I want to happen. These two may or may not be the same.
3. If there are apparently definitive statements below (e.g., “this will happen”), they are really just guesses and may not actually happen.
4. Some will read this and say, “There’s nothing new here!” I’m not claiming there is. Almost everything that could ever happen has already been predicted by someone, but which of those predictions are correct? The fun is in making predictions and then seeing what happens.
5. Absolutely nothing here constitutes medical or health guidance or advice. For example, much of the research cited here is preliminary (or very preliminary), and associations often do not mean causations. References and links in this post may have counter-examples that were not included in the post. This is not an objective and comprehensive review of the literature. There are many additional reasons that nothing here should be in any way construed as medical or health advice or guidance. This is just me making guesses about the future. TRUST NOTHING WRITTEN HERE!

So, now for some fun.

Prediction 1: GLP-1 agonists (and/or other drugs with similar impacts) will radically reduce healthcare costs and improve clinical outcomes

GLP-1 agonists (“GLP-1s”) have been reported to have possible efficacy in some patients for treating obesity, diabetes, hypertension, and even smoking, drinking, and other addictions. These conditions sit at the foundation of primary care, and they contribute to many other health issues that GLP-1s also may treat or prevent in some patients. Those include: cardiovascular disease, some cancers, COPD, sleep apnea, Alzheimer’s dementia, opiate overdose, fatty liver disease, and kidney disease. These latter conditions comprise a substantial portion of specialty care.

It has been claimed that as many as 80% of strokes and heart attacks may be associated with therapeutic inertia — failing to start or change treatment when needed. This comes as no surprise to me. How often are patients told repeatedly by their primary care providers to diet, exercise, and/or quit smoking as a key part of their treatment? And how often do patients struggle to be successful with those efforts? Reportedly, between 50% and 90% of people gain back all their weight in the ensuing years after initiating a diet. Smoking cessation programs have reportedly high failure rates, aerobic exercise programs have high dropout rates within six months, and an estimated 80% of US adults do not get recommended levels of exercise. For many reasons, people often struggle to successfully follow this advice. Some are successful, but many are not. Now we have medicines that may help.

As with virtually all medications, GLP-1s may have potentially serious side effects. For example, some may have major gastrointestinal complications from the GLP-1s (and there may be other serious side effects).

Based on this, if clinicians, patients, and regulators believe that benefits substantially outweigh risks for a substantial number of patients, I predict that primary and specialty care will be radically different in a GLP-1-influenced future, once prices for the drugs come down. Price drops may happen soon, as generic versions have already been approved. The need for a shot may limit their use today, but pill versions have been in the news recently, with at least one currently available and others in development. Note that this is not a recommendation for or against these drugs, just a prediction about how they might affect our healthcare future.

If the upsides prove not to be outweighed by the downsides of the GLP-1s, and as prices come down, there may be tremendous economic pressures leading payers (including the government) to encourage their use. If reductions in nationwide disease burden and associated healthcare costs can result from these medications, that may prove very compelling. This may lead to over-the-counter availability and/or semi-automated prescribing (e.g., via AI chatbots). If all the chronic conditions described above become significantly less common due to GLP-1 drugs (and similar medications), then many clinicians (primary care, cardiology, oncology, etc.) may see their practices change dramatically.

Will this reduce the demand for doctors and other providers? Or will our healthcare providers in the future be just as busy, but serving patients with other types of problems, perhaps even in a more timely fashion? That probably depends on how some of my other predictions below turn out. 🙂 But in any case, unless new data changes what seems to be the perceived risk/benefit tradeoff for these drugs, I predict the GLP-1s will dramatically transform primary care, population health, much of specialty care, overall healthcare outcomes (improving them), and overall healthcare costs (lowering them).

Prediction 2: The Germ Theory of Disease “Plus” will revolutionize treatment for most “not-germ” diseases

My friend and mentor told me long ago that he suspected many, maybe most, diseases felt to be non-germ-related will be found to be caused by germs (e.g., virus, bacteria, fungus, parasite), at least in part. I investigated and agreed. Over and over, for decades, one disease after another has been shown to be caused by, or at least strongly associated with, one or more germs. Based on this trajectory, I predict that most of the diseases we now consider “primary” or “idiopathic” (fancy terms for “we don’t know the cause”), or autoimmune, will be found to have a germ-based component.

Some may find this hard to believe. If so, consider the following…

In the 19th century, Ignaz Semmelweis showed that, when students washed “death particles” (i.e., germs) off their hands in-between working on cadavers and delivering babies (all without gloves), maternal mortality after delivery dropped dramatically. Those death rates dropped from 18.4% in April 1847 down to 2.2% in June and then to 1.2% in July, and maintained similarly low rates thereafter. He published this work and was widely ridiculed by disbelievers. Ultimately, Semmelweis allegedly had a nervous breakdown. He was placed in a sanatorium where he was beaten by the guards, dying 14 days later. Ironically and tragically, this champion of antisepsis died of sepsis from a gangrenous hand wound (potentially from his beating by the guards).

We used to think that stomach ulcers were most commonly caused by stress and spicy foods. Drs. Marshall and Warren performed a series of studies demonstrating that ulcers unrelated to NSAID (anti-inflammatory) medication usage were very often caused by the germ Helicobacter pylori. Their work was labeled “preposterous,” but now we recognize that this germ is probably the most common cause of these ulcers, and antibiotics have become a mainstay of therapy in many of these cases.

Cervical cancer is now known to be commonly caused by the human papilloma virus, and we have a vaccine to help prevent it. Some oral, esophageal, penile, anal, and lung cancers are also caused by, or strongly associated with this virus. Bell’s Palsy (a type of facial paralysis) was once felt to be “idiopathic” (unknown cause), but now we know a very common cause is viral infection. Both thrombotic thrombocytopenia purpura and hemolytic uremic syndrome were considered idiopathic, but infection with a variant of E. coli has been strongly implicated as a cause. Certain infections are now suspected as potential factors in the development of autoimmune diseases such as lupus, rheumatoid arthritis, and type 1 diabetes.

The microbiome — the set of germs living in and on us — is now recognized as contributing substantially to our health. Associations have been found between certain germs and heart disease, Parkinson’s disease, ALS (“Lou Gehrig’s Disease”), obesity, Crohn’s disease, Alzheimer’s dementia, schizophrenia, and kidney stones (to name a few). Many of those germs are either considered part of the microbiome or appear to have effects on the microbiome. Wide ranging therapies for treating various health conditions by affecting the microbiome have been proposed, including antibiotics, probiotics, and even fecal transplants, all commonly aiming to replace or counterbalance “bad” germs with better ones.

Despite well over one hundred years of research finding that germs cause health disorders previously thought unrelated to germs, resistance to this concept remains. What happens when I note to colleagues that many “non-germ” diseases actually have demonstrated associations with specific germs, and then propose that most such diseases will likely be found to have some germ associated with their cause? Most can barely contain their disbelief, often manifested in three stages. First, they express unawareness of the existing data. Then they doubt the validity of the studies. Finally, with a chuckle, they suggest that these examples are outliers that do not represent a larger trend. Well, as Han Solo once said, “Laugh it up, fuzzball!” I’ll try to hold back from saying “I told you so” if I’m proven right. 🙂

Obesity is at the heart of many health issues prevalent in the US and other parts of the world. For too long, too many have blamed the patients themselves for “lacking the willpower” to adhere to a more healthy lifestyle. Now we recognize that infections and the microbiome may have an important role in obesity, playing a large part in explaining why one person seems to ravenously but gain no weight while another has to watch every morsel. Microbiome modification may change all that. If my prediction about GLP-1s fails to come true, it may be because microbiome-based therapy has made GLP-1s unnecessary.

Note that I wrote “Germ Theory (Plus)” as the heading of this section. That’s because it seems likely to me (based on research I have seen) that germs alone do not fully explain the development and clinical course of germ-related diseases. Rather, it seems research suggests that genetic factors, other biological factors (many of which are related to gene expression), and/or other environmental exposures (e.g., toxins) also may play a role in determining whether a germ causes disease, as well as the clinical course of that disease for someone who has it. Why do two people get infected with the same germ, and one develops cancer and the other doesn’t? Evidence I have seen suggests that genetic/biological factors and non-germ exposures (e.g., smoking, pollution, etc.) may affect the likelihood that a germ will cause certain diseases, and vice versa (germs may affect the likelihood that other exposures, like smoking, and/or genetic factors lead to some diseases). For example, in lung cancer, smoking is associated with changes in the lung microbiome (germs in the lung), the microbiome germs can release toxins that affect genes, and the germs may affect the immune system’s ability to prevent or fight cancer. The effects of genes and germs may be two way: genetic factors may affect the content of the microbiome, and in turn, the microbiome may affect genes and their function. In other words, germs, other environmental factors, and genetic/biological factors all appear to affect one another. This complex interaction may determine the likelihood of developing a disease. With that said, I predict that germ-related treatment (whether antibiotics, probiotics, antivirals, vaccines, germ transplants, or other) will become a critical aspect of care in our near-future.

I predict that, in the not-so-distant future, most diseases that we now consider “idiopathic,” autoimmune, or “lifestyle-based” will be recognized as having a germ-related cause and will be prevented or treated with germ-related therapy. Achieving that will require serious research to identify germ-disease relationships and effective therapies. That work could take decades unless we find an accelerant. And I predict one of those accelerants will be molecular medicine. So the correctness of my next prediction is key to the correctness of this one.

Prediction 3: “Molecular medicine” will dramatically enhance logistical access to care

It took me a while to identify an umbrella term for biologic therapies (i.e., those involving DNA, RNA, and proteins) and “-omics” (i.e., analysis of DNA, RNA, and proteins, perhaps other things). Thank you, ChatGPT, for suggesting “molecular medicine.” (I hope it’s right!) We’ve been hearing about this for years, and it has never seemed to arrive. But, the future has arrived! I’ll provide some examples.

A mainstay of preventive medicine is colon cancer screening. For most people I’ve known over the years, this has meant spending many hours doing “colon prep” (often drinking as much as a gallon of a laxative until what comes out the bottom is clear liquid), followed by an invasive procedure in the hospital or surgicenter requiring anesthesia and recovery time. If the prep is performed improperly, the procedure may need to be rescheduled. Plus, it is expensive. Many people, perhaps especially many in rural areas, do not live physically close to a gastroenterologist who performs this procedure, and not every gastroenterologist is equally skilled at identifying precancerous polyps. Therefore, even though many (probably most) consider colonoscopy the gold standard for colon cancer screening, many patients choose not to get it.

However, the good news is that colon cancer screening is increasing. Although colonoscopy screening growth has significantly diminished, home-based cancer screening using stool DNA analysis (the stool is collected at home and sent to the lab for analysis) has been rising rapidly (these trends per NIH). In a different study of over 1.4 million first-time colon cancer screenings that involved either colonoscopy or DNA-stool screening, the trend was far more dramatic. In 2018, it was 97% colonoscopy, 3% DNA stool testing. By the end of 2023, it was under 70% colonoscopy and over 30% DNA stool testing. My noting of these trends is not a comment on the appropriateness of one test over another, rather it is a recognition of the changing patterns of care. DNA testing is already here, and it’s changing healthcare, at times replacing an invasive test done in the hospital with a non-invasive one done at home.

Many have wondered whether AI will replace doctors (I wrote a blog post about this), but many have assumed that AI won’t soon replace clinicians that perform invasive procedures. Recent developments suggest this assumption might not hold, because robots can reportedly automatically draw blood and autonomously do at least parts of some surgical procedures. However, I suspect the bigger and faster impact on invasive procedures will come from: 1) GLP-1s and/or microbiome modification reducing the need for many procedures, and 2) molecular medicine replacing many procedures. We’ve seen #2 start to happen with colonoscopy, and I predict it will happen with many more procedures going forward.

I predict the rise of AI use in molecular medicine (known as “bioinformatics”) will lead to even more dramatic enhancements in care. Access to care will improve as AI-driven, home-based molecular medicine diagnostics and therapies replace those done in hospitals and by medical specialists. Bioinformatics is used today in some “GWAS” and “PheWAS” studies, which are studies comparing clinical data to genetic data, to identify (for example) associations between genetic factors and disease. Bioinformatics is also being used to design or discover new medications and predict their effects.

In 1973, Schwaber and Cohen described using “hybridomas” to make antibodies. In 1975, Jerne, Köhler, and Milstein extended this work, winning the Nobel Prize in 1984 for “discovery of the principle for production of monoclonal antibodies.” Monoclonal antibody therapy (a form of molecular medicine) has been accelerating ever since, and is now “having a moment.” Have you seen generic drug names that end in “mab”? That stands for “monoclonal antibody.” It seems I can’t turn on the TV these days without seeing an ad for a mab like Keytruda, Dupixent, Skyrizi, or Humira (those are brand names for various mabs).

Interesting side note: Jerrold Schwaber, of “hybridoma” fame above, is the brother of Ken Schwaber, one of the founders of Scrum agile software development.

During the COVID-19 epidemic, in the span of about a year, starting from scratch, companies developed and released home-based antigen (molecular medicine) tests to diagnose COVID-19, mRNA-based (molecular medicine) vaccines to prevent it, and monoclonal antibody therapies (molecular medicine) to treat it. The era of molecular medicine is not some distant future — it is here today! And, it’s been proven that it can be safely developed and released with lightning speed and offer incredible effectiveness.

Now, molecular medicine seems poised to be turbocharged by AI (bioinformatics). Imagine that AI can auto-analyze your “-omics” (DNA, RNA, and/or proteins) to make a precise diagnosis and to determine the specific therapies most likely to be both safe and effective. We may soon arrive at a future where patients less often suffer the dual pain of 1) months-long waits for specialist appointments, and then 2) the trial and error approaches common in medicine (i.e., “this works for most people, let’s try it and see if it works for you”). Piling on, the potential for CRISPR + AI seems even more disruptive, since CRISPR enables precise gene-editing that may treat or create treatments for medical conditions.

Putting it all together, we might imagine this:

1. A patient at home will talk to an AI bot about their concerns.
2. The AI bot will send an AI-controlled drone to deliver a testing kit to the home.
3. The patient will put some body fluid (like saliva or blood) into the testing kit.
4. An AI-controlled drone will pick up the testing kit and deliver it to a lab.
5. An automated set of “-omics” machines will process the fluid’s DNA, RNA, and/or proteins.
6. AI will analyze the data to generate a “diagnosis” (which may be provided as a list of one or more abnormal -omics causing the problem instead of being provided as a medical term like “lupus”).
7. AI will design a personalized safe and effective CRISPR-driven treatment.
8. An AI-powered gene-editing robot will use CRISPR to create the treatment.
9. An AI-powered delivery drone will deliver the treatment to the patient’s home.
10. An AI bot will empathically and sensitively discuss and explain the diagnosis and treatment to the patient.
11. The patient will administer the treatment, which may be as simple as a nasal spray in some cases.
12. The patient’s issue (e.g., cancer, autoimmune disease, etc.) will be both correctly diagnosed and completely cured in a matter of days.

Everyone is talking about “agentic AI” (AI-based process automation) in medicine, and this might be an ultimate example. As unrealistic as this imagined future might sound, we may already be much closer than you might think:

• Patient History-Taking AI Bot: Studies of AI chatbots have already shown “great potential” for history taking.
• AI-Automated Drones: Amazon now says it can deliver 60,000 different items by drone, often within 60 minutes of ordering; and drones are already used in some parts of the world for delivering medicines.
• DNA Analysis Informing Care: DNA analysis is already used in cancer care to help make precise diagnoses and target therapy.
• CRISPR-based therapy: In 2023, for the first time ever, the FDA approved a CRISPR-based gene-editing therapy, in this case to treat patients with severe sickle cell disease. Just this year (2025), it was reported that CRISPR was successfully used to create a DNA-based cure for an infant with an incurable disease. Also this summer, it was reported that a man with long-standing type 1 diabetes received CRISPR-modified islet cells and the cells were making insulin for months afterward.
• AI-Designed Therapy: Just a few years ago, DeepMind’s AlphaFold AI shocked the world with reportedly near-perfect prediction of protein folding. This and other AI may have the potential to predict and create new drugs with specifically desired effects. AI has now designed proteins, built using CRISPR, that precisely control genetic machinery.
• AI-Powered Gene Editing Robot: AI-automated CRISPR robots have already been developed or at least are in development.

Note that I specified “logistical access to care” in my prediction because I do not know how quickly prices for molecular medicine will fall, or how quickly it will be covered by payers. If molecular medicine costs to patients are high, then even if logistical access to care improves, financial access to care may remain low. With that said, there are indications that much of this may accelerate with generative AI, and that could affect the speed at which prices fall.

Conclusion

The biggest unknown in all this, as I see it, may be the impact of current and future federal funding and regulation on these predictions. However, as Bill Gates once reportedly noted, we tend to overestimate the change that will happen in two years and underestimate the change we will see in ten. With that in mind, in two years, some will likely turn around and say “See! The world isn’t transformed as you claimed.” In ten years, some may say, “See! The world was transformed, but not exactly in the way you described it!” Yes, a lot of what I wrote, maybe all of it, may be proven wrong. And if any or all of it is proven right, can we really know if that was due to luck or skill? And for every correct prediction, many will say it was obvious anyway. Prediction seems like a no-win endeavor, but that didn’t make it a less interesting exercise for me.

And… if I am proven right, then I can make one more prediction: I will probably claim it was all due to my knowledge and prediction skill, and not at all from luck. 🙂

All opinions expressed here are entirely those of the author(s) and do not necessarily represent the opinions or positions of their employers (if any), affiliates (if any), or anyone else. The author(s) reserve the right to change his/her/their minds at any time. Nothing written here or anywhere else on this site should be considered as medical guidance or advice.