Ben Locwin, Healthcare Science Advisors07.14.15
Appraising trends is always an interesting endeavor. Most of the time, the really interesting part isn’t even the trend data, or lack thereof, but the meta-appraisal itself—that is, what is going wrong within the review of the data. People always tend to see patterns within the data that aren’t really there, and much of this is hardwired into our brains because we’ve evolved to be pattern-seeking and correlation-drawing machines. Those brain architectures, which could recognize patterns in the wild and draw correlations—even if often plagued with false positives—were the ones, which became the most successful. Unfortunately, cognitive biases and wishes-to-be-true also short-circuit people’s objective and accurate appraisal of trend data.
When I’ve advised venture capitalists and Wall Street investors, I have always used great care to explain many of the logical fallacies that people fall into when looking at corporate performance and market variability. For example, recency bias has people tending to over-weigh things that have happened very recently, at the expense of taking full stock of a larger picture in context.
Two Points Do Not Make a Trend
While line plots are often used to demonstrate performance of some sort, called in these cases ‘run charts’, two points don’t make a trend. Nor perhaps do three points make a trend—after all, two points can be connected by a line (linear fit), but what about the third point? Does it indicate error in the model or a next point in time? What’s the confidence interval around each of these points? A recent run of performance is not indicative of future possibility, and the reason Walter Shewhart developed the control chart ninety years ago was to put statistical context on the accumulation of data. For example, a couple of recent impressive performances does not forecast a ‘rising performer,’ nor the ‘next big thing.’ In just the same way that there really is no ‘hot hand’ in sports, and a couple payouts on a slot machine doesn’t mean that it’s ‘hot’ or that one should keep playing on it.
In the world of Chaos theory, it’s easy to see that systems are enormously sensitive to changes in initial conditions. The end result of miniscule changes to factors in a process can lead to wildly different outcomes. This makes reliably hedging on trends a difficult proposition, as it’s not just the new and hot trends having value that determines their long-term success. There’s a huge impact from different, often unmeasured, factors, such as the adoption rate of particular trends, marketing differentiation, word-of-mouth impact across the industry, and the effect of primacy within a niche—being first to market, and also the effect of being on the ‘leading edge’ rather than the ‘bleeding edge’.
Chaotician DJ Patil observes the difficulty in predicting and objective-setting in our contemporary market by calling it the ‘next two-hours era,’ meaning that effects are so variable that pinning anything down as a prediction for any length of time is virtually impossible. Physicist Niels Bohr commented that, “Prediction is very difficult, especially about the future,” to capture just how unknown, and unknowable, upcoming events and in our case, trends, really are.
Social media has rapidly accelerated the agility with which the public can change allegiance and direction. It used to be that, when information dissemination was slower and much more compartmentalized within regions and market segments, that market resistance to fluctuations was more robust. But now well-placed advertising, social commentary, or public response to corporate missteps can swirl into a maelstrom of market change within hours that is agnostic to region or market segment. But it should also be duly noted that simply being present on social media such as Twitter and Facebook does not make it a corporate strategy. Seeing which trends are surviving in the wild, and how they’re arranging in terms of prioritization in the industry’s consciousness can help with allocating an appropriately-scaled response if it is a market-changing trend or plan for adoption if it is an uptake trend.
Some Hot and Upcoming Trends
Clearly, if we look at ad campaigns, mobile devices, and social usage, these have dramatically changed the landscape of direct-to-consumer (DTC) marketing of drugs, which in turn has created a bullwhip effect in reverse where the tail, being the consumer, wags the dog of pharma and healthcare providers (HCPs).
It used to be entirely that physician visits would lead to a differential diagnosis and perhaps a treatment, in a very logical cause-and-effect fashion. Now this has been a relationship that has become inverted, where the consumers hear about therapies that they want to try, and they approach their HCPs to get that treatment. This has subtly, and not-so-subtly, changed prescribing practices and the outcomes of office visits. In a recent Pricewaterhouse Coopers survey, about half (52%) of clinicians are comfortable using mobile devices for healthcare to monitor vital signs. Within that group, there isn’t equal comfort with this technology when looking across the various clinicians’ age ranges. This will lead to some differential variability.
Another trend to note that is occurring between the patients and the researchers is that of genomics and their place in pharma and healthcare. We know that genetics and associated demographics have a vast impact on the existence of diseases and progression of diseases. But many in the public—estimates range from about half to well over 90%—don’t want their genetic profiles to be part of their healthcare. This will seriously alter the landscape of diagnostics in the next couple of decades, and allow disease progress to occur that otherwise wouldn’t need to all because of public misperception of the power of the information. However this could also be due to information security concerns. This other market trend shows that an overwhelming proportion of patients and healthcare consumers are more concerned about their data not being secure than they are about receiving appropriate care. This continues to be a challenge with electronic health records (EHR), where they can be broadly accessed in some cases by providers other than a patient’s own selected clinician.
And the more that big data and population-level healthcare statistics and treatment statistics are in the clinical and public consciousness, the more of a demand for scientific evidence of treatment there has been and will continue to be. The 3 Ps of the pharma equation: the patient, the physician, and the payer (insurance) have been asking more-and-more for clinical trial data and efficacy data, and I don’t see this trend diminishing. In fact, all evidence points to it continuing to grow into more of an issue.
Which brings up a philosophical point within big data: A scientific analysis using statistical data requires pre-specifying an a priori hypothesis before conducting the experiment and analyzing the data. Big data is essentially looking at volumes of post-hoc data to look for correlations. This goes against the very nature of how and why the null hypothesis significance testing (NHST) method of double-blinded randomized controlled trials (RCTs) was developed in the first place. Looking at data after the fact without an initial hypothesis to test isn’t really scientific.
Health and wellness and preventive care will continue to grow, and interestingly, we should see changes to the distribution of diseases and disorders as large-scale public adoption of dietary and health trends along with better clinical care continues. For example, as more people live longer and avoid certain diseases and disorders that had high mortality or morbidity in the past, certain other conditions, which were only of academic importance or were rarely seen in the oldest members of the population in the past now become practical realities for many families and society at large. These geriatric syndromes will continue to grow in importance, impact, and treatment need in our society.
As the pharma industry adapts, much of it will be indirect adjustment to market signals caused by these trends, but understanding where those signals arose from will help businesses to respond more effectively.
References
Lorentz, E. N. (1993). The essence of chaos. Seattle, WA: The University of Washington Press.
Ben Locwin
Healthcare Science Advisors
Ben Locwin, PhD, MBA, MS writes the Clinically Speaking column for Contract Pharma and is an author of a wide variety of scientific articles for books and magazines, as well as an acclaimed speaker. He also provides advisement to many organizations and boards for a range of healthcare, clinical, and patient concerns.
When I’ve advised venture capitalists and Wall Street investors, I have always used great care to explain many of the logical fallacies that people fall into when looking at corporate performance and market variability. For example, recency bias has people tending to over-weigh things that have happened very recently, at the expense of taking full stock of a larger picture in context.
Two Points Do Not Make a Trend
While line plots are often used to demonstrate performance of some sort, called in these cases ‘run charts’, two points don’t make a trend. Nor perhaps do three points make a trend—after all, two points can be connected by a line (linear fit), but what about the third point? Does it indicate error in the model or a next point in time? What’s the confidence interval around each of these points? A recent run of performance is not indicative of future possibility, and the reason Walter Shewhart developed the control chart ninety years ago was to put statistical context on the accumulation of data. For example, a couple of recent impressive performances does not forecast a ‘rising performer,’ nor the ‘next big thing.’ In just the same way that there really is no ‘hot hand’ in sports, and a couple payouts on a slot machine doesn’t mean that it’s ‘hot’ or that one should keep playing on it.
In the world of Chaos theory, it’s easy to see that systems are enormously sensitive to changes in initial conditions. The end result of miniscule changes to factors in a process can lead to wildly different outcomes. This makes reliably hedging on trends a difficult proposition, as it’s not just the new and hot trends having value that determines their long-term success. There’s a huge impact from different, often unmeasured, factors, such as the adoption rate of particular trends, marketing differentiation, word-of-mouth impact across the industry, and the effect of primacy within a niche—being first to market, and also the effect of being on the ‘leading edge’ rather than the ‘bleeding edge’.
Chaotician DJ Patil observes the difficulty in predicting and objective-setting in our contemporary market by calling it the ‘next two-hours era,’ meaning that effects are so variable that pinning anything down as a prediction for any length of time is virtually impossible. Physicist Niels Bohr commented that, “Prediction is very difficult, especially about the future,” to capture just how unknown, and unknowable, upcoming events and in our case, trends, really are.
Social media has rapidly accelerated the agility with which the public can change allegiance and direction. It used to be that, when information dissemination was slower and much more compartmentalized within regions and market segments, that market resistance to fluctuations was more robust. But now well-placed advertising, social commentary, or public response to corporate missteps can swirl into a maelstrom of market change within hours that is agnostic to region or market segment. But it should also be duly noted that simply being present on social media such as Twitter and Facebook does not make it a corporate strategy. Seeing which trends are surviving in the wild, and how they’re arranging in terms of prioritization in the industry’s consciousness can help with allocating an appropriately-scaled response if it is a market-changing trend or plan for adoption if it is an uptake trend.
Some Hot and Upcoming Trends
Clearly, if we look at ad campaigns, mobile devices, and social usage, these have dramatically changed the landscape of direct-to-consumer (DTC) marketing of drugs, which in turn has created a bullwhip effect in reverse where the tail, being the consumer, wags the dog of pharma and healthcare providers (HCPs).
It used to be entirely that physician visits would lead to a differential diagnosis and perhaps a treatment, in a very logical cause-and-effect fashion. Now this has been a relationship that has become inverted, where the consumers hear about therapies that they want to try, and they approach their HCPs to get that treatment. This has subtly, and not-so-subtly, changed prescribing practices and the outcomes of office visits. In a recent Pricewaterhouse Coopers survey, about half (52%) of clinicians are comfortable using mobile devices for healthcare to monitor vital signs. Within that group, there isn’t equal comfort with this technology when looking across the various clinicians’ age ranges. This will lead to some differential variability.
Another trend to note that is occurring between the patients and the researchers is that of genomics and their place in pharma and healthcare. We know that genetics and associated demographics have a vast impact on the existence of diseases and progression of diseases. But many in the public—estimates range from about half to well over 90%—don’t want their genetic profiles to be part of their healthcare. This will seriously alter the landscape of diagnostics in the next couple of decades, and allow disease progress to occur that otherwise wouldn’t need to all because of public misperception of the power of the information. However this could also be due to information security concerns. This other market trend shows that an overwhelming proportion of patients and healthcare consumers are more concerned about their data not being secure than they are about receiving appropriate care. This continues to be a challenge with electronic health records (EHR), where they can be broadly accessed in some cases by providers other than a patient’s own selected clinician.
And the more that big data and population-level healthcare statistics and treatment statistics are in the clinical and public consciousness, the more of a demand for scientific evidence of treatment there has been and will continue to be. The 3 Ps of the pharma equation: the patient, the physician, and the payer (insurance) have been asking more-and-more for clinical trial data and efficacy data, and I don’t see this trend diminishing. In fact, all evidence points to it continuing to grow into more of an issue.
Which brings up a philosophical point within big data: A scientific analysis using statistical data requires pre-specifying an a priori hypothesis before conducting the experiment and analyzing the data. Big data is essentially looking at volumes of post-hoc data to look for correlations. This goes against the very nature of how and why the null hypothesis significance testing (NHST) method of double-blinded randomized controlled trials (RCTs) was developed in the first place. Looking at data after the fact without an initial hypothesis to test isn’t really scientific.
Health and wellness and preventive care will continue to grow, and interestingly, we should see changes to the distribution of diseases and disorders as large-scale public adoption of dietary and health trends along with better clinical care continues. For example, as more people live longer and avoid certain diseases and disorders that had high mortality or morbidity in the past, certain other conditions, which were only of academic importance or were rarely seen in the oldest members of the population in the past now become practical realities for many families and society at large. These geriatric syndromes will continue to grow in importance, impact, and treatment need in our society.
As the pharma industry adapts, much of it will be indirect adjustment to market signals caused by these trends, but understanding where those signals arose from will help businesses to respond more effectively.
References
Lorentz, E. N. (1993). The essence of chaos. Seattle, WA: The University of Washington Press.
Ben Locwin
Healthcare Science Advisors
Ben Locwin, PhD, MBA, MS writes the Clinically Speaking column for Contract Pharma and is an author of a wide variety of scientific articles for books and magazines, as well as an acclaimed speaker. He also provides advisement to many organizations and boards for a range of healthcare, clinical, and patient concerns.