Updated: Why Your Next Doctor May Be an Algorithm

Updated: Why Your Next Doctor May Be an Algorithm

August 13, 2018

Virginia had been diagnosed with kidney failure 18 months earlier. Due to arteriosclerosis, life-long high blood pressure and many other factors, kidney transplant was not an option. On this day, the doctor that recommended dialysis for Virginia completed his last exam and stated tersely, "I guess you are going to die," as his last goodbye, and stormed out of the hospital room. Virginia's husband had just gathered his composure as she commented in understated fashion that the doctor was not happy with her decision. After months of research and discussion with dialysis providers, Virginia had decided over a year earlier that life on dialysis was not for her even though it was the doctor's recommendation. On that day, about 48 hours before her death, Virginia was actually much more at peace with the decision she had made than the doctor, whose advice she declined.

Since Virginia's death, many medical advancements have evolved. The growth of medical algorithms and machine learning for patient care is one major category of medical advancement. In general, machine learning is the process by which computers refine an analytical model from new data, applying trends to continually improve predicted results. The core power used by predictive computer software is mathematical algorithms. Over the past decade, hundreds of medical-specific algorithms have been created. Machine learning for medical treatment takes patient data over a a wide range of medical big data and applies it to the current patient's diagnosis.

The idea of your doctor being an algorithm is not the fodder of dystopian-future movies. It's already here. The real question is in what form will it evolve. Here are just two examples:

  • Vivify Health’s algorithms drive patient-facing applications that improve care management after a patient leaves the hospital. Read more here.
  • Algorithm's are outperforming dermatologists in recognizing skin cancers in blemish photos. Read more here.

Patient-facing applications using medical algorithms are already on many medical advisory websites. As discussed in The Atlantic article above, getting FDA approval and AMA backing can be a long process. However, these techniques are science backed and doctor vetted. The general patient population already searches every ache and pain through Internet engines. These ad hoc inquiries by the public are infinitely less powerful than medical predictive models because of, among other good reasons, the tiny sampling of data that can be accessed by someone in this manner. Here are three major reasons an algorithm may eventually become your doctor of first diagnosis.


Healthcare costs have become 17% of US GDP, roughly double that of the average western industrialized nation. As our appetite for medical input has risen, so has expensive medical solutions. The cost of these solutions have priced many out of the market and driven patient premium and out-of-pocket payments to increase. One way to begin to reduce these costs is by using medical algorithm-based applications as the first line of patient input and diagnosis.

Data processing power

There are two big knowledge types being applied during patient diagnosis. One is the application of data, research-based knowledge doctors receive in school and with which they must stay up to date. The other is judgment applied based upon experience in the application of that knowledge. Machine learning is eminently more powerful in gathering and applying the former. Now let's look at the latter.

Emotions and agendas

As Virginia learned, doctors are humans with their own emotions and agendas. Whether the doctor's judgment has financial, religious, personal moral and ethical influences and pressures, some medical professionals will always have difficulty parsing their own leanings with the decisions of the patient. Quality of life factors can only be decided by the patient's own conscience and situation. The medical algorithm is a data and diagnosis reporter, as well as outcome predictor. Virginia may have benefited from an impartial analysis when making her decision as well as during her final days, rather than an infuriated medical professional with his own agenda and brusque bedside manner.

For an interesting article about emotional agendas in healthcare, read the following:


For further information about the power of predictive analytics in business and medicine, see the Industry Applications course.