Epidemiologists put all their ducks in a row all the time. Ducks have been programmed by their DNA to do this; it is only since 1854 that epidemiologists have done the same thing – and it is quite an achievement.
It was a little over a year ago (3/11/2020) that the World Health Organization declared COVID-19 a pandemic. Since then, epidemiologists have been lining up people by the calendar day (or week) each got sick with COVID-19 and they have been counting all these new cases over time ever since.
This exercise produces an iconic epidemic curve (showing US COVID-19 cases) like the one below.
The figure shows new cases rising, falling; plateauing, then ascending, then descending etc. similar curves have been produced for nearly 170 years. 
The COVID-19 epidemic curve may become one of the most recognizable images in future museum displays chronicling the year 2020.
Epidemiological scientists start with these curves to which sophisticated models are applied to answer basic “Why” questions: Why did the curve go up and why did it go down, and up again and down yet again? After careful analysis they recommend to public health officials what the evidence shows that may slow or halt the epidemic. It is not a perfect science, but it is a useful and practical one.
This tried-and-true method has great power in our current crises, and in past centuries perhaps even greater power. Indeed, in the mid-19th century, epidemiology was the foundational science behind the development of sanitary sewers. A seemingly simple solution, sanitation became one of the most effective health interventions of all time—dramatically decreasing epidemics of water-borne diseases such as cholera. 
Trajectory takes this powerful scientific method and expands the epidemiologic universe, virtually
In Trajectory’s universe “virtual epidemics” are created — a plethora of them can be created quickly where each is based on unique defined population. The number of these produced is dependent the number of co-factors (also called co-variates in statistical lingo) in the data set, the size of the population, and the availability of computer processing power.
Based on two US patents and engineered with smart computer programming. We start, for example, with women diagnosis with pregnancy.
Trajectory forces every case in a defined population to occur on the same date (called “time zero”). If a classic epidemic curve were produced it would look like an inverted T: 100% of the cases would occur at time zero and no cases before or after.
Trajectory goes beyond this to create a “virtual epidemic” – The Trajectory Curve ™ – of the defined population and one or more co-factors.
Below is depicted the virtual epidemic of the cost of care of women diagnosed as pregnant (in the center of the chart). The chart tracks the total cost of care (measured as percent above a threshold value, termed “Percentage Above Tipping Point” in the graph) for each of these 819 women and shown in trajectory months (duration: 28 days) before and after the first diagnosis of pregnancy (generically called the “event date” in the Trajectory universe and situated at Time Segment #1 on the horizontal axis)
Looks a lot like a classic epidemic curve doesn’t it? But this is a display depicting a virtual epidemic of the cost of care of pregnant women over trajectory time.
Trajectory also allows the software user to adjust this curve by other co-factors. The chart below shows the average resource use of pregnant women over trajectory time, stratified by those with and without “complications of pregnancy” (as defined by AHRQ’s CCS software.). The top/red line refers to pregnant women with complications; the middle/blue line to all pregnant women; and the bottom/green line to pregnant women without complications.
Classic epidemiological methods – and other sophisticated analytic tools — are then employed to understand the reasons for the rise and fall of the co-factor over trajectory time.
After aligning the data by Trajectory time, Trajectory’s artificial intelligence automatically selects co-factors that predict the event (pregnancy in this case) or some other event along the trajectory (like early hospitalization). The prediction is then validated “in real-time” to assess its accuracy.
Additionally, Trajectory’s unique ability to put “all the ducks in row” is ready-made to evaluate the impact of an intervention. Its AI automatically selects an equivalent reference group and evaluates the difference between the intervention and the reference group to assess impact.
In chart above, the event date was the date each person was identified as being eligible for case management. The virtual epidemic shows both the average cost of care in the intervention group and the reference group (i.e. those without case management) before and after the event date.
Trajectory evaluates the equivalence between the two groups using the left side of the figure; when deemed to be similar, Trajectory then calculates the difference in the average cost of care between the two populations on the right side of the figure. This is the impact, or the “return” in an ROI model. When coupled with an estimated per-person cost per month of the intervention (the “investment”) an ROI is easily calculated.
The Advantage of Having All Your Ducks in a Row
“Survival analysis,” often seen in oncology research, employs a similar philosophy, but standard survival analyses does not track negative time, moreover, variables to stratify are manually selected, and standard survival analyses does not track thousands of co-factors in negative or positive time to enable a smarter choice for stratification.
Trajectory’s AI can act as a discovery engine to learn about promising findings and then use that knowledge to test their impact on people.
Here are cases of Trajectory’s benefit in transforming historical data (claims, EMR, registries, etc.) into “virtual epidemics” – all illustrations are from the real world.
- The trajectory of a case management population in a Medicare Advantage plan utilizing insurance claims data. 
- The paper shows the often spurious practice of using a simple pre-intervention post-intervention design to evaluate program impact. The pre-post method often ignores regression-to-the mean, a phenomena that is so apparent in many defined populations after data is transformed into a “virtual epidemic.”
- The role of telephonic case management centered on the date each person started case management compared to an equivalent reference group without case management.
- The impact of a remote alert device on inpatient stays, based on the day of the install, compared to a reference without the device.
- Rapid comparison of the trajectories of all major diagnostic groups to each other to identify opportunities for special population health management interventions. Done in minutes, not hours or days.
- Using diagnosis of COVID-19 as the event date, examining the role of mental health as a factor in predicting COVID-19 – early in the epidemic–while adjusting for baseline physical health condition (follow-up study in progress on post-COVID-19 diagnosis).
- The role of quality improvement program on radiology where the date of certification of provider organizations was the sentinel event.
- Predicting the individuals likely to be super-utilizers and enrolling them in a case management program
- Displaying a classic epidemic in trajectory time: By defining an “event date” as identical for person, e.g., 2/29/2020 as in the first figure above, the software can do “classic epidemiology” and track cases over time too.
- The biggest advantage of all is by transforming data in this way a report/study can be done in hours rather than weeks or months. The efficiency of Trajectory of mind-blowing: Evidence show that 5 or more studies can be done in the time it typically takes to do one study using standard statistical software. Trajectory improves productivity.
Trajectory takes the concept of an epidemic: counting new cases over time of a specific condition to track an epidemic and applies it to situation when an epidemic does not exist. The results are high quality reports and studies that can be accomplished rapidly. Where one study/report was done with current analytics studies, one can now do five or more.
Here is a thought experiment: Imagine all the heart attacks occurred on exactly the same day? The available data would be sliced and diced and one could assess possible causes and solutions to help prevent a catastrophic event like this from ever occurring again. This is what we imagined at Trajectory (and filed and received patents for the idea!)
By translating this experiment into a software program, the epidemiological paradigm is vastly expanded to all diseases – in epidemic status or not.
If fact, one may be able to imagine this for a multitude of events that come over the people (the underlying etymology of the words epi and demo), health related or not (e.g. warranty repair on an automobile).
Virtual epidemics can be accomplished when historical data, powerful computing is available, and creative/expansive thinking is available.
The possibilities of new discoveries are virtually endless. The efficiency of the Trajectory system — and its ROI — is off the charts.
© 2021 Thomas Wilson. All Rights Reserved
Here is the transformation by patented Trajectory© algorithms of traditional data into a virtual epidemic (Sample of Cancer Patients)
 Courtesy of Michael Shaw at https://toons.to/go/michaelshaw. Original cartoon appeared in The New Yorker, February 10, 2014: 74.
 From Washington Post, March 14, 2021.
 Wilson T. The Dr. John Snow Solution circa 1854: The Value of Rapidly Obtained Real-World Evidence Then and Now. July 16, 2019 https://schoonerstrategies.com/the-dr-john-snow-solution-circa-1854-the-value-of-rapidly-obtained-real-world-evidence-then-and-now/
 Adler RL. Medical Firsts: From Hippocrates to the Human Genome 1st Edition 2004. Hoboken, NJ. Wiley & Sons, Chapter 15: 101-8.
 Trajectory algorithms are protected by US patents 7,685,011 & 7,685,012; other patents pending. Trajectory(r) is a registered service mark with the US Patent and Trademark office (serial number: 76399931).
 https://www.hcup-us.ahrq.gov/toolssoftware/ccs10/CCSCategoryNames(FullLabels).pdf. See CCS single level 181.
 New Research Confirms RadSite’s Robust ROI. March 27, 2013. https://radsitequality.com/radsite-news/new-research-confirms-radsites-robust-roi/
 Wilson TW, Sullivan J. Mental/Behavioral Health as a Predictor of Initial COVID-19 Diagnosis: Results from the Colorado All Payer Claims Data Set to June 30, 2020. http://dx.doi.org/10.2139/ssrn.3807198