Prehabilitation—interventions developed to increase a patient’s preoperative physiological reserve to help them cope with the stresses of surgery and recovery, improve postoperative outcomes, and facilitate return to function—is a relatively new field of study and a rapidly growing research area.^1-5 As prehabilitation interventions evolve, exciting new developments are being driven by the integration of artificial intelligence (AI)–powered technology.⁶ With the advent of cutting-edge AI solutions such as wearable trackers, chatbots, and predictive modeling, patients and health care providers are witnessing a paradigm shift in the delivery of personalized health care (Figure). In this article, we will review how these technological advances can be leveraged for the delivery of prehabilitation-focused interventions and highlight anticipated challenges of AI integration, with particular attention to the field of urology.

Perhaps one of the most pressing challenges in prehabilitation is to ensure the acceptability and feasibility of exercise intervention completion, even in the most vulnerable patients. The lack of standardized prehabilitation protocols and the heterogeneity of patients and their disease course further complicate this issue,⁷ requiring a robust solution based on large, diverse datasets. Modern AI technology utilizes large-language models and natural language processing to analyze vast amounts of disparate patient data—including but not limited to radiographic images, medical history, and surgical history—and capture complex, nonlinear relationships that may otherwise have been missed in such highly dimensional data. Using patterns obtained from this analysis, AI can personalize interventions according to patient characteristics and preferences⁸ and characterize personalized risk profiles to anticipate patient outcomes.⁹ For example, researchers utilized recurrent neural networks to predict eligibility for deep inspiration breath-hold radiotherapy treatment of patients with left breast cancer—these candidates were selected based on their functional capacity for breath holding, as well as anatomic and other clinical features.¹⁰ The model produced a binary result, easily allowing providers to discern which patients would likely be able to successfully receive the treatment. Similarly, when it comes to selecting patients for prehabilitation, diverse data incorporating functional capacity and clinicopathologic features can be combined to tailor the most suitable adjuvant intervention for each patient’s unique needs, to maximize successful treatment.

While candidate selection is essential, appropriate selection of the prehabilitation intervention is equally important. Traditional models of prehabilitation focus on multiple supervised exercise sessions, which demand resources (eg, physical facilities, equipment, and supervising trainers), are geographically limiting, and offer little flexibility.¹¹ Unsupervised sessions conducted at home, popularized during the COVID-19 pandemic, avert many of these issues but cannot ensure adequate adherence or intensity. Poor patient adherence, ineffective resource allocation and disparities in equity can all impede a successful prehabilitation intervention.⁷ There is clearly a need for technology that monitors, reports, and responds to patients’ physical activity. Some examples are already ubiquitous—wearable fitness trackers and smartphone applications aid the general public in tracking health metrics and achieving fitness goals. This technology has the power to respond dynamically to human behavior—for example, prompting one to stand up and walk around after a period of inactivity. Waller et al found that a prehabilitation program using a smartwatch and smartphone application for exercise and nutrition counseling resulted in increased exercise adherence and improved functional outcomes.¹² Wearables can go beyond simple tracking—some, such as a smart brace for joint issues, can provide dynamic feedback and appropriately adjust the type of activity being performed.¹³ Others can automatically send tracker-obtained data to a health care team, allowing for remote patient monitoring without the need for in-person visits.^14,15 Tracker data such as patterns of active vs sedentary time or duration of tolerable exercise can be used to train algorithms on patterns of patient behavior to optimize adherence or develop interventions that are more likely to be sustainably adopted.

Following prehabilitation and surgery, AI-powered technology can continue to aid patients and physicians. The aforementioned preoperative predictive analytics can be applied postoperatively as well—AI-based remote monitoring can assist in the early detection of cancer recurrence or complications following treatment.¹⁶ When patients undergoing radical cystectomy were given wearable activity monitors, lower postoperative step counts were associated with longer length of stay and higher rate of postdischarge readmissions.¹⁷ This information can help medical teams prepare for potential challenges and help surgeons convey this personalized set of possible complications when obtaining informed consent. In urology, AI-powered technology has demonstrated benefit in patient education. Cakir et al found that ChatGPT, a natural language-processing platform that generates responses to prompts, was able to accurately and sufficiently answer more than 95% of potential questions about urolithiasis.¹⁸ In this same way, one might apply generative AI under the supervision of urologists to help patients understand more about their medical condition, goals of prehabilitation, and care pathway.

AI also presents inherent challenges. Research studies often do not disclose their code, rendering algorithm assessment difficult. The opaqueness of neural networks, commonly referred to as the “black box” effect, hinders comprehension of their decision-making processes and raises concerns about potential biases. Additionally, questions arise concerning patient data ownership and confidentiality. Effective oversight is essential to ensure the use of high-quality data and equitable representation of diverse patient groups in algorithmic inputs. When integrating AI into health care, rigorous validation is imperative to mitigate the risks of harm and biased outcomes stemming from inadequately trained data. Upholding patient safety and data privacy is of paramount importance. Furthermore, it is vital to be mindful of the potential financial implications of AI technology—a comprehensive cost-benefit analysis is necessary for prudent resource allocation.¹⁹

AI-powered technology has the power to transform the field of prehabilitation for those undergoing major surgical or oncological interventions. It can help identify appropriate patients for prehabilitation, improve adherence via prompting, and serve as an informative tool for patients and physicians via digital health applications and chatbots. While the possibilities are exciting, there is much work to be done, from ascertaining validity and accuracy to ensuring lack of bias, before such technology can be widely utilized. For now, the push for AI technology in prehabilitation is a promising new frontier that is ready for exploration.