Abstract

Hospitals have widely adopted health information technologies (HIT), such as clinical decision support systems and computerized provider order entry, to improve care quality, streamline clinical workflows, and enhance patient experience. Advances in real-time data and analytics have extended this adoption to operational artificial intelligence (AI) for predictive planning, scheduling, and workflow management. Yet, unlike AI embedded in patient portals, telehealth platforms, or point-of-care clinical decision tools, operational AI often works “behind the scenes,” raising an intriguing question: Can AI that patients do not directly encounter still shape their care experience? This question matters because patient experience depends not only on clinical accuracy but also on direct interactions with care providers. By improving staffing and demand prediction, scheduling, routine-task automation, and workflow optimization, operational AI may reduce care providers’ computer-mediated workload and redirect their attention to listening, explaining, and responding empathetically to patients. Although recent studies have documented positive effects of specific AI tools on patient experience, large-sample, hospital-level evidence remains limited. Given the resource intensity of hospital planning, scheduling, and routine workflows, and the role of patient experience as a core care-quality indicator, the effects of operational AI warrant systematic investigation. We examine the association between five AI-enabled operational functionalities — staffing-need prediction, patient-demand prediction, staff scheduling, routine-task automation, and administrative/clinical workflow optimization — and patient experience. We construct a two-year hospital panel linking lagged 2022–2023 AI and telehealth measures from the American Hospital Association to 2023–2024 patient-experience outcomes from the Centers for Medicare & Medicaid Services, yielding 2,873 hospital-year observations from 1,783 hospitals. We focus on two outcomes: a communication score averaging nurse communication, doctor communication, staff responsiveness, and medication communication, and an overall hospital rating score based on the percentage of patients who rated the hospital 7 to 10. We logit-transform both percentage-based outcomes and estimate hospital fixed-effects models, controlling for hospital characteristics, care volume, staffing intensity, telehealth functionality count, and year effects. We find that the total number of AI functionalities is positively associated with the overall rating score but not significantly associated with the communication score. Functionality-level analyses further show that the patient-facing value of operational AI varies by use case. Staffing-need prediction and staff scheduling are positively associated with communication but not rating scores, whereas patient-demand prediction, routine-task automation, and workflow optimization show no significant associations with either outcome. These null results may reflect uneven implementation or insufficient integration, rather than limited value, motivating further analysis of whether operational AI creates greater patient-experience benefits when paired with patient-facing HIT, such as office-visit telehealth. Overall, this study contributes to emerging IS research on AI in healthcare by shifting attention from visible, patient-facing AI tools to less visible operational AI systems that may indirectly shape patients’ care experience.

Download

Share

COinS