How I am becoming a clinical algorithm
My work as a clinician centered on laboratories growing human embryos that would eventually make its way into the woman’s uterus. This technology that is routinely used to achieve pregnancies in infertile couples has a downside. They often resulted in superfluous embryos because only a limited number – which varied by national regulations- could be placed into the woman’s uterus. These rules served to avoid multiple pregnancies that affect the health of the mother and babies. When faced with excess embryos – which happens most of the time- we have to make choices and select out the embryos that we believe will give the couple a happy childbirth. This selection process, first practiced in the late 70s, has continued to evolve ever since and researchers continue publishing profusely on the subject. As a professional focused on continuously improving outcomes, I would constantly scan the literature, attend workshops and be in correspondence with experts around the world seeking better ways to improve pregnancy rates. We would then implement in our work what we judged from published research to be useful iterations of the selection protocols. The embryo selection process typically incorporates microscopic observations of the embryo made at various times during the 3 or 5 days of embryo culture in the laboratory, and includes parameters such as size and shape of cells, extracellular debris and speed of cell division. These criteria have continued to deepen and expand over the years, but researchers have never reached consensus on the one best method to do it. However, during the last few years, vendors developed time-lapse video technologies to record the growth of embryos and automatically capture the changes. These vendors and their supporting scientists claimed that the volume of data collected in the videos would provide far more insight than is possible with manual observations. Explained simply, video would take away the need to make multiple single point observations during the 3 to 5 days of in-vitro embryo growth and we would no longer be limited to data that is visible only during our manual observations. The evidence for the benefits of this technology was limited and I was one of the majority that did not believe that it was a useful addition to the armamentarium of fertility care, especially when it came with a hefty price tag. While the debate continued I since moved on into the world of digital transformation. Recently, to my delight, I learnt that the new sources of embryo growth data especially the data from videos had made its way into a commercially available machine algorithm. Debate about its usefulness is ongoing as is so common with innovation, but it seems only a matter of time, as we see with progress in driver-less cars and other fascinating fields of technology endeavors that it will become standard clinical practice. The big picture I am witnessing is that a part of my former profession is slowly but precisely being substituted by algorithms. In fact, no professional’s work is ‘safe’ from this or other kind of automation. The last great frontier, which surgeons or even my former self might find comforting, are tasks requiring high level of manual dexterity, but as we all know robotic medical procedures are not unheard of even today. I do not think I would have seen the pace of change from within my clinical world. My change in profession helped me get this outside-in perspective and I am glad I can now play a role to facilitate this change for reasons I will describe below.
Let me now point out the positive fallout of this digital transformation in professional work. An inherent aspect of knowledge work, as you have read in my own experience above, is the continuous exploration and research geared toward improving outcomes. In addition, part of what professionals do is to devise effective ways to adapt new research into routine practice. Therefore, it is likely that the work of clinicians may become more of managing and improvising the algorithms that would take over their work. They could re-allocate their intellectual capabilities to refining algorithms for their unique contexts and devote their bedside time to the most complex patients; while less trained workers –like non-clinical patient navigators – powered by these algorithms, take up the routine work involving what I will cautiously label as the simpler patients. This re-allocation will help scale the impact of what every clinician does. It will reduce cost of delivery and it fill some of the emerging skill shortages. Healthcare transformation will thus be at two levels – one, in the standardized execution of clinical knowledge through algorithms and second, in the way algorithms become the tool for efficient re-allocation of work across the system. Digital platforms today allow you to manage algorithms as colleagues, partners or products depending on what perspective you take. From a technology standpoint, you can now create and manage algorithm deployment separately from the context in which it is used. You can share them and you can monitor their performance. You can continue to develop them as you learn from using it in the field. Progress in such technology is timely and well aligned with the economic pressures that are forcing healthcare to re-think how they operate.