On March 12, 2025, we published our initial comments on the problems of access and timely appointments. We discussed the depth of this problem and its impact on both quality and hospital revenue. And we discussed the concept of a “loyal patient” and the importance of a world-class appointment system to assure that patients remain within your care ecosystem. All of this was a good start, but we believe the appointment and access issue requires deeper analysis.
To accomplish this higher-end analysis, we have been collaborating with Professor Dan Adelman. Professor Adelman is the Charles I. Clough, Jr. Professor of Operations Management at the University of Chicago Booth School of Business. He is a leading expert on business analytics, and he leads the Healthcare Analytics Laboratory at Chicago Booth. Professor Adelman and his team have been especially focused on the appointment problem, asking the essential question: Why are hospitals unable to efficiently fill their appointment slots, even in the face of very significant demand?
In the Healthcare Analytics Laboratory at Chicago Booth, Professor Adelman and his team have been working on solutions. The analysis has been inspired by how airlines fill seats on airplanes, and how world-class manufacturing plants efficiently produce automobiles. The goal is to develop a new generation of analytical tools to help hospitals and healthcare systems improve operational performance.
The Healthcare Analytics Laboratory has identified six operational drivers of appointment access problems. But at their core is a critical concept that is missing in the management practices of hospitals. In fact, its absence is one reason why solving appointment access problems for most hospitals is a game of “whack-a-mole,” rather than an improvement effort organized around fundamental operating principles.
Recent research published in JAMA Internal Medicine found that 19% of deceased patients were marked alive in the EHR of a major academic health system in California an average of 1.5 years after death, with 80% of those having outstanding appointments or encounters after death.[1] Hospitals generally do not have mechanisms to easily and automatically determine when patients stop receiving care, either due to death, moving away or even simply because treatment is completed. Counter-intuitively, the absence of mechanisms to effectively track and manage patients whose care has ended may be the underlying reason why you feel ineffective in solving your appointment problems.
We call the sequence of appointment visits for a given condition, in a given specialty, a provider episode. It begins with a first visit for a condition, over time with potentially multiple visits, and then ends with a final visit. A patient may be in multiple provider episodes simultaneously, in different specialties—for example, a cardiology patient may also be under the care of a pulmonologist. There are challenges in calculating provider episodes, but let’s set those aside and see how provider episodes are the missing key to solving access issues.
Operationally, a new patient visit is not just hospital and clinic volume, but a commitment to the patient that you will make available clinician capacity in the future, in the form of return visits. Not an infinite number of return visits, however: Eventually, a patient’s care journey is finished, and that should free up capacity for new patients to be seen.
Let’s now look at the drivers of the appointment problem through the lens of provider episodes:
- Cancellations and no shows. We have found that patients’ tendencies to cancel or no-show can depend on where the patient resides in the provider episode and the patterns they exhibit within it. For example, patients may be more likely to cancel visits at the end of the provider episode. This may be because care at that point may be “if needed” depending on the patients’ health status. Or consider this: Patients who cancel or no-show once may be more likely to cancel or no-show again in the same provider episode. Understanding patient behavior within provider episodes, you can design targeted interventions to pre-emptively avoid losing patient encounters and clinically important follow-ups.
- Appointment schedule templates. Hospitals routinely guess, for each provider, how many new versus return patient slots they should have in their schedule template. This results in an imbalance, where the supply of new versus return slots does not match the actual demand for new versus return slots. This causes delays for patients trying to obtain appointments. With analytics around provider episodes, the target balance is easy to compute by considering the number of return slots each new patient needs over time, on average. Furthermore, this balance can be adjusted dynamically. Extending the concept of a “patient panel” in primary care, as patients leave the care of a provider episode, this should immediately trigger the opening of new patient slots. Such a “smart” mechanism would be impossible to deploy without tracking provider episodes.
- Visit patterns. By comparing provider episodes for the same condition across multiple providers, you will likely discover variation in practice patterns. Some clinicians may have significantly more encounters than other clinicians for the same condition or have longer patient visits. Such variation, in and of itself, may not be a problem, as providers naturally have different practice patterns with the goal of best serving their patients. On the other hand, variation may point to opportunities to free up appointment slots for new patients through the sharing of best practices. Such variation could not be detected without tracking provider episodes.
- Capacity imbalances. With an understanding of provider episodes, together with analytics around referral patterns, you can consider the impact of increasing (or decreasing) capacity within one specialty on the throughput of other specialties. For example, you can forecast that hiring another cardiologist will result in new pulmonology appointment demand, which you cannot meet with your existing staffing. Or you can determine areas where increasing capacity would have the greatest impact on increasing new patient volume. You can assess the impact of deploying advanced practice providers. With such planning tools, you can anticipate and alleviate the critical bottlenecks that inhibit patient flow and timely access, especially for new patients. Such tools have existed in airlines and manufacturing for many years, but they have not been brought to bear in hospitals.
There are other operational drivers of appointment access problems, less tied to provider episodes, but also amenable to analytics.
- Dynamic release and overbooking practices. Airlines fill seats using a practice known as “yield management,” whereby seats on a given flight are protected for last-minute customers with greater urgency to fly (and hence greater willingness to pay). As the departure date approaches, if demand doesn’t materialize, these seats are released for other customers to purchase. This practice is like the concept of “dynamic release” in hospitals where appointment slots are reserved for certain types of patients, but as the appointment date approaches, they are released so a broader set of patients may book them. However, unlike airlines, which have developed sophisticated analytical tools for yield management, these tools do not generally exist in hospitals so release times are guessed at, rather than optimized. In many cases, appointment slots should not be restricted at all.
Likewise, to fill seats, airlines sometimes overbook flights, meaning they sell more seats than are available on the plane. While this practice may seem random or haphazard, it is carefully optimized against data on future cancellations and no-shows. Another idea, borrowed from the airlines, is that of “flying stand-by.” Within limits, hospitals can allow patients who don’t have same-day appointments to nonetheless arrive at the doctor or clinic and wait to be seen within a block of time. With good analytics to anticipate no-show and late cancellations, you can ensure that a patient will have a high statistical likelihood of being seen without overwhelming clinical staff and be able to instantaneously fill a slot that would otherwise have gone unused. - Waitlist management. There are currently no good methods for optimizing the use of appointment waitlists. If you know the probability that an individual patient will accept a waitlist offer for an earlier appointment slot, then it is easy to calculate the number of offers needed to ensure the earlier slots fill with high probability. The trouble is, there may not be enough patients on the waitlist. Or perhaps the wrong patients are being given waitlist offers. Here is an opportunity for AI and machine learning to help by predicting which patients are most likely to accept offers.
Dynamic release of appointment slots and waitlist offers are not typically integrated. Thus, when appointment slots are released to accept different patient types, this may not trigger waitlist offers, but it ought to. When a patient accepts a waitlist offer, their current appointment slot becomes available. It should be immediately offered to other patients, but there currently does not exist process automation that would deploy and optimize such a sequence of appointment slot offers to move all patients to earlier times, like dominoes falling. Such a practice would multiply the impact of moving one patient to an earlier slot, to moving many patients to earlier slots all at once, significantly reducing time-to-appointment across the board.
Based on the above-described research, below are a few observations to improve your appointment process:
- Start a task force to improve appointment access. The drivers we’ve discussed cut across many areas of the hospital organization, such that one person does not have visibility or control over all of them. By establishing an Appointment Task Force, you would raise urgency and empower leaders to work collaboratively toward solutions.
- Learn best practices from the airline industry. To solve appointment access problems, you need strong operations research solutions with great technology and the best software. You need to start filling appointments as efficiently as airlines fill airplanes.
- Deploy AI/predictive analytics. It is going to be critical to predict a patient’s likelihood to cancel or no-show, or to accept a new appointment from the waitlist. You need to understand the behavior of your patients toward appointments to know how to intervene to improve access and to build business logic. Solutions here are complex and data heavy. The inevitable solution is likely AI-driven and dependent.
- Appoint an “end of care” czar. To fix your appointment access problems, you need to know who is being cared for and who is not. Otherwise, you cannot manage around provider episodes. To achieve this, you will need new business processes that integrate novel data sources, adjust clinician prompts in the EHR and even do patient outreach.
- Sanitize, standardize and synthesize appointment data. You will need to apply the same level of discipline that you now apply to medical data except now to appointment data. To optimize access using any of the analytic approaches we’ve discussed, you will need clean and complete data on referrals, new versus return visits, patient behavior patterns and preferences, end of care points, etc.
As we have said before and will say again, the quality of your appointment process is going to be directly tied to the quality of your clinical results and your ability to improve and obtain positive financial results. When the appointment process does not work efficiently and appropriately, both your patients and your hospital organization suffer. The appointment problem currently seems intractable, but by applying smart technology, high-end analytics and contemporary business processes, your hospital will be able to fill appointments as efficiently as the airlines have mastered the art and science of filling passenger seats.
[1] Neil S. Wenger et al., “Consequences of a Health System Not Knowing Which Patients Are Deceased,” JAMA Internal Medicine, vol. 184, no. 2 (Feb. 2024): 213 – 214.
