Mirror, mirror on the wall, will the patient show at all?

In the aftermath of the COVID-19 pandemic, Ardent Health — which operates 30 hospitals and more than 200 care sites in six states — faced a confluence of challenges, including staffing shortages, surging patient demand and an escalation in patient no-show and late cancellation rates. These no-show trends align with findings from 2022 MGMA Stat polling, which revealed that almost half (49%) of medical groups reported increased patient no-show rates since 2021.¹

Patient no-shows pose a significant challenge to healthcare providers. Missed appointments limit access to timely medical attention for all patients, potentially leading to patient dissatisfaction and suboptimal healthcare outcomes. Additionally, no-shows negatively impact provider revenue, further straining healthcare resources.

To mitigate these negative consequences, Ardent adopted a multifaceted approach, employing a combination of technological and communication-based strategies. These measures encompass:

1. Implementing automated patient appointment reminders, facilitating easy cancellation and rescheduling, and enhancing patient control over their healthcare schedules.
2. Introducing an online patient portal, enabling seamless appointment scheduling and cancellation, further strengthening patient autonomy and flexibility.
3. Integrating digital check-in functionality through the patient portal, streamlining the patient arrival process and reducing in-person wait times.
4. Carrying out proactive patient outreach via phone calls the day before scheduled appointments, serving as a final reminder and fostering patient engagement.

Despite their best efforts, markets were still plagued by no-show rates ranging from 7% to as high as 18% in some specialties in July 2022. Ardent needed an innovative solution promising improved outcomes and chose predictive analytics to mitigate the loss of access and revenue due to patient no-shows.

In 2017, Ardent adopted Epic as its sole EHR platform, unifying its use across all markets. This decision enabled the use of Epic’s continuously evolving capabilities, including its advanced predictive analytics features. Epic’s no-show predictive model is a black-box algorithm that uses appointment history and characteristics to calculate the no-show probability of an appointment. This no-show probability is available next to each appointment, prompting the front desk staff to take appropriate action. Health systems use these predictive models to decrease the no-show rate by calling patients to confirm their appointments, a strategy Ardent was implementing. Therefore, the organization began focusing on the outcome rather than relying on patient behavior modification. By leveraging predictive analytics, Ardent could identify patients with a high likelihood of no-shows and proactively schedule additional patients in those slots, optimizing appointment utilization and maximizing patient access to care.

Ardent’s providers had previously employed overbooking strategies to decrease the negative impact of patient no-shows. However, the predictive model emerged as a replacement for random overbooking with a superior, data-driven statistical model that predicts no-show events.

Ardent developed a Power BI report highlighting the significant advantage of using the predictive model to target appointments for overbooking. It summarizes the opportunity to overbook predicted no-show slots and the total volume of all overbooking, including random overbooking. Ardent shifted focus from measuring the model’s no-show prediction success to quantifying its accuracy in identifying overbooking opportunities compared to random overbooking.

In August 2023, Ardent embraced a no-show predictive model to optimize access. Empowered by this advanced tool, Ardent has strategically realigned its approach, now overbooking slots with a higher no-show probability.

Key considerations for successfully implementing a no-show predictive model to optimize overbooking strategies

1. Identify provider champions to engage their peers in discussions about the model and emphasize the need for long-term data evaluation. AI and machine learning algorithms are trained on substantial data volumes and refine their performance over time. Evaluating the model’s accuracy based on short-term data, such as a single day, week or month is insufficient.
2. Secure provider consent for the high-probability no-shows that may be overbooked daily, along with a pledge to accommodate both patients if they arrive.
3. Proactive scheduling based on predictive insights optimizes appointment utilization. Decreasing the no-show rate can remain a secondary goal.
4. Data hygiene and process standardization are crucial for accurate predictive modeling. For example, if one clinic cancels an appointment if a patient calls within 24 hours, another clinic treats it as a no-show, and a third clinic reschedules the appointment, these differing practices can affect the model’s predictions of no-show probabilities.
5. Create a no-show policy across clinics to ensure a standardized process.

Editor’s note: Michael Williams, Manager of Operational Analytics and Power BI Developer, Ardent Health Services, contributed to this article.

Note:

1. Harrop C. “Patient no-shows pose concern amid medical practice staffing challenges, consumer price hikes.” MGMA. Aug. 3, 2022. Available from: https://www.mgma.com/stat-080222.