Becker's ASC Review

19 Executive Briefing: Sponsored by: A fter decades of painstaking effort to digitize massive amounts of information into Electronic Health Record databases, it has finally become possible to build targeted applications of artificial intelligence and machine learning. AI and ML can enable health systems to catch up with other industries that harness data at scale to make millions of decisions each day that are smarter, faster and maintain the integrity of the core operational processes. Every patient has a unique "flight path" through the health system on any given day in terms of the sequence of encounters that they must navigate including labs, clinic appointments, procedures and treatments. At every step, patients encounter delays caused by staff shortages, add-ons, cancelations and an insufficient set of resources (machines, rooms, chairs, etc.). The current approach of scheduling patient appointments as though they were as simple as booking a conference room or a tennis court has completely fallen apart. Let us take a simple example to illustrate the possibilities: A visit to a specialist in an ambulatory clinic. The patient calls the specialist's office to make an appointment. She is given a slot, which could be for the same day or the next day but more likely is 3-6 weeks out. The patient makes plans to get the lab work completed — either in advance of the appointment at a neighborhood lab facility or by arriving an hour or two before the appointment with the specialist. Having arrived at the ambulatory clinic where the specialist practices, the patient fills out some forms, and perhaps sits in the waiting room for a short period of time. The average wait time currently stands at 18 minutes and 13 seconds, defined as the time interval between checking in at the front desk to being placed in an exam room. The patient is taken back to an exam room where she waits again for a nurse and then again for the specialist who is typically "running behind" that day. The specialist finally enters the exam room and speaks with the patient for a few minutes before heading out to see another patient in the adjoining exam room. After another agonizing wait in the exam room, a nurse or an assistant enters with the final instructions for wrapping up the appointment and sending the patient onto the next step of their journey — perhaps to an infusion appointment or a procedure. Although excellence in clinical care is the biggest driver of patient satisfaction, excessive waiting time is one of the biggest drivers of patient dissatisfaction. At a cancer institution that we work with, an elderly gentleman opted out of his scheduled infusion treatment after waiting an hour past his scheduled appointment time — he told the front desk staff that since he only had a few months left to live, he would rather spend time with his grandchildren than continue to wait in the waiting room. This example is a powerful reminder of the importance of using the best analytic methods that we can harness to minimize unnecessary wait times across the entire health system. What if by using AI-based software in that same clinic, resources were allocated just a bit differently? What if we could predict key operational parameters such as the volume of patient arrivals within each 15-minute window of each day, the mix of the various appointment types that would be represented in the incoming arrivals, the time spent by a specific provider with a patient for each type of appointment that they encounter, the probability of an add-on, cancelation, late arrival or no-show for every minute of every day. The patient arrives at the lab 30-45 minutes before her scheduled appointment with the specialist and is immediately placed in a chair — blood is drawn and sent off for analysis while the patient walks over to the ambulatory clinic to see their specialist. Upon arrival, at or shortly after their scheduled appointment time, the patient is placed in an exam room, the nurse takes their vitals and gathers all of the relevant information including results from the bloodwork, which are now available in the system. A few minutes later, the specialist comes in after reviewing the updated information and engages the patient in a discussion regarding their treatment plan. After the specialist leaves the room, the patient promptly checks out. At no point during the entire visit, did the patient wait more than a few minutes between steps and the entire visit took 30 minutes instead of 90-120 minutes. Every specialist practices medicine in a unique way based on skills and habits that they have acquired over decades. Some talk a lot to their patients, others don't; some write notes as they go while others keep it in their head and write notes at the end of their day; some practice without a support team while others have a team of nurses, medical assistants and residents helping them throughout the day. Every ambulatory clinic is unique in terms of the availability of resources (exam rooms, support staff, etc.) and the way that those resources are allocated (e.g., pooled to be shared by all specialists who are in clinic that day or dedicated to a specific specialist for the day). Current systems and scheduling guidelines attempt to gloss over this complexity by establishing "standards" e.g., new patients will get a 60-minute slot while return patients will get a 30-minute slot. Harnessing Artificial Intelligence for Ambulatory Clinics By Mohan Giridharadas, Founder and CEO, LeanTaaS

• Cover