Artificial intelligence (AI), now an integral part of healthcare services and presents numerous opportunities. Customized treatment plans, clinical decision support systems, predictive analysis for disease prevention, patient engagement and education, quality improvement, and error reduction are some of these opportunities. In the context of delirium prevention, risk assessment, and treatment planning, the AI-supported system AI-AntiDelirium is designed to standardize the approach to delirium management in alignment with the PADIS guidelines. A randomized controlled trial evaluating the effectiveness of this system found that the workload of nurses decreased, facilitated early diagnosis and prevention of delirium, and recommended evidence-based and individualized delirium interventions. A systematic review concluded that AI applications did not significantly impact the length of hospital stay and emphasized the need for further research. Also, AI platforms contributed to positive results in reducing anxiety and depression in patients. Furthermore, systematic reviews have demonstrated that AI-based chatbots are effective in alleviating symptoms of depression and anxiety. However, the literature includes a limited number of patient education programs specifically designed to prevent or manage delirium through AI-based approaches. Notably, there is a lack of studies comparing the effectiveness of AI-supported educational interventions with those delivered directly by nurses.

The goal of this clinical trial is to develop a structured AINurse and Human Nurse orientation training program for intensive care unit (ICU) patients and compare the effects of these training programs on ICU patients' delirium-free days, level of anxiety and depression, and length of stay in the ICU. Hypotheses of the study:

H1: Patients who receive the structured AINurse patient orientation training program will have longer delirium-free days than patients who receive Human Nurse orientation training.

H2: Patients who receive the structured AINurse patient orientation training program will have lower levels of anxiety and depression than patients who receive the Human Nurse orientation training.

H3: Patients who receive the structured AINurse patient orientation training program will have shorter lengths of stay in the intensive care unit than patients who receive the Human Nurse orientation training.

Researchers will compare the AINurse patient orientation training program and the orientation training program provided by human nurses in terms of patients' delirium-free days, level of anxiety and depression, and length of stay in the ICU.

• Those in the intervention group will receive the AINurse orientation training program twice daily for 3 days.
• Participants in the control group will receive face-to-face structured orientation training from researchers twice daily for 3 days.
• Delirium-free day assessment, anxiety and depression will be evaluated for patients in both groups over 3 days.
• The length of stay in the intensive care unit will be monitored for patients in both groups.

The study is planned to be conducted in the adult intensive care unit of University Hospital, a tertiary care hospital in Türkiye. The general intensive care unit (GICU) of University Hospital has 16 beds and the cardiovascular surgery intensive care unit (CVSICU) has eight beds. The hospital where the study will be conducted has a total of 1950 patient admissions annually, 1500 in the GICU, and 450 adult patients in the CVSICU.

This study is designed as a single-center, randomized controlled experimental study. Although there are no similar studies available for calculating the sample size for this study, a pilot study with 10 patients will be conducted. Following the pilot study, the sample size will be calculated using G-Power analysis.

Recruitment, randomisation and blinding:

A computer-assisted simple randomization method will be used to assign eligible participants to study groups. The randomization process will be performed by a statistician independent of the researcher and communicated to the researcher. The website "https://www.random.org/integer-sets" will be used for this purpose.

Study Intervention:

The study consists of four stages;

Phase 1: Development of the Structured Orientation Program:

Structured orientation training programs will be developed by the researchers for both AI Nurse and Human Nurse interventions based on the literature. While the AINurse training program will be implemented in the experimental group, the Human Nurse training program will be implemented in the control group. The content of the structured training program for both groups will be the same. The training content will be sent to the opinion of 10 experts who are experts in intensive care and AI, and changes will be made in the training program.

In the AINurse training program, the training text will be written in a patient-friendly language to enhance comprehensibility. The Narrative Orientation Experience model will be used not only to deliver information but also to support the patient's psychosocial adaptation to the intensive care environment. Within this method, key information such as time, place, care environment, relatives, and daily routine will be presented in the form of a scenario format in a "starting the day story" in which the patient is addressed by name to foster a sense of personal connection and orientation.

Phase 2: Development of AI-Nurse orientation program/software:

The AI-Nurse orientation program to be implemented in the intervention group will be developed by a researcher with expertise in AI and engineering, who is also part of the study team. During the development phase, the content of the program will be structured by the researchers to be compatible with individual patient information. Voice-based orientation training will be delivered using the Google Cloud Text-to-Speech (TTS) API, which enables the conversion of written text into natural-sounding speech. The TTS engine allows for customization of voice parameters, including gender (female and male), speech tempo (slowed), and tone (calm). For Turkish language output, the use of 'tr-TR-Wavenet-A' or 'tr-TR-Standard-B' voice models is recommended. The training content can be personalized, with relevant information about the patient, weather, and family-related data will be automatically retrieved from the database. The AI Nurse system will be developed to work via mobile devices (e.g., an Android tablet). The patient will listen to the audio training through headphones and, if necessary, will be able to read the text on the screen.

The training program will include approximately 10 minutes of audio narration and will be administered twice daily over a period of three days. Existing orientation practices in the hospital will be observed and documented. During this observation, all elements of nurse-patient interaction, such as whether the nurse introduces herself/himself, whether she/he gives day information, and whether she/he provides information about the environment, will be detailed. The control group will receive a researcher-designed orientation training program of similar duration and frequency, comprising 10-minute sessions delivered over three consecutive days.

Phase 3: Piloting the AI-Nurse orientation training program: The developed AI-Nurse training program will be conducted as a pilot study with 10 patients hospitalized in the ICU. A sample calculation will be made as a result of the pilot study. Patients included in the pilot study will not be included in the sample.

Phase 4: Implementation of structured orientation training to intervention and experimental group:

In the experimental group, the structured AI-Nurse orientation program will be listened to twice daily (10:00 and 14:00) through a headset connected to a structured audio recording application. The content will include orientation to time and environment, but will exclude any personally identifiable information such as the patient's name. Following each orientation session, the daily news will be played with the help of the television in the patient's room (at 10:30-14:30), aiming to enhance environmental awareness and perception of real-time audio streaming. The TTS engines to be used will be able to provide real-time audio streaming. This intervention will be implemented by the same researcher over a three-day period. Anxiety and depression levels, as well as the CAM-ICU score, will be assessed both before and after the intervention. After three days of structured AI-Nurse training, the patient's discharge time from the ICU will be followed and recorded. In the control group, the structured orientation training will be given face-to-face by the Human-Nurse researcher. The patient in this group will also be monitored for signs and symptoms of delirium for three days. Anxiety and depression levels will be measured prior to the intervention, while post-intervention evaluations will include the number of delirium-free days, anxiety and depression scores, and length of stay in the ICU.

Data Collection Instruments: Data will be collected using the sociodemographic characteristics form, a clinical patient information form, the Confusion Assessment Method-Intensive Care Unit (CAM-ICU), and the Hospital Anxiety and Depression Scale (HADS). In the initial stage, patients will be screened for the presence of delirium using the CAM-ICU.

Inclusion Criteria:

• Intensive care unit confusion assessment scale (CAM-ICU scale) score of 0-2
• Age 18 years or older
• Hospitalized in the intensive care unit for at least 24 hours
• Glasgow coma scale score of 13-14-15 points
• Richmond Agitation Sedation Scale (RASS) score between -1 and +1
• No hearing problems

Exclusion Criteria:

• Intensive care unit confusion assessment scale (CAM-ICU scale) score of 6-7
• Any psychiatric illness or impaired brain function
• Defined hearing loss
• Advanced dementia
• Younger than 18 years of age
• Richmond Agitation Sedation Scale (RASS) score outside the range of -1 to +1
• Who use sedative medication
• History of surgery or disease around the ear