Current and Potential Applications of Ambient Artificial Intelligence

The fourth industrial revolution has transformed our daily life, with the introduction of new digital technologies.¹ This revolution has led to the integration of digital and physical worlds, created value, and impacted every sector of the economy. Ambient artificial intelligence (AI) is one such technology that is enabled by the fourth industrial revolution and has a great future potential for augmenting health care delivery.²

The need for digital interventions and augmentation in health care delivery and clinical medicine has never been greater. The postpandemic, current endemic society has crippled health care institutions around the world. Physician burnout and resignations are at an all-time high.³ Physician shortages in the United States and around the world have also peaked. Ancillary clinical staff are suffering from the same conditions. Digital technologies in clinical medicine and in health care delivery are needed now more than ever.

Cook et al,⁴ described ambient intelligence as a presence of digital environment that is sensitive, adaptive, and responsive to the presence of people. Ambient intelligence integrates human-centric computer interfaces, secure systems and devices, and technologies that assist with sensing, reasoning, and acting.⁴

As illustrated in Figure 1, ambient intelligence works at the intersection of Internet of Things devices and sensors placed in the user’s surrounding environment, pervasive computing, AI, machine learning, and human-computer interaction. In addition to machine learning, ambient intelligence also uses knowledge graph–based technologies.⁵

Figure 1.

Ambient AI: technology integration. AI, artificial intelligence; HCI, human-computer interaction; IoT, Internet of Things; ML, machine learning.

Ambient sensing exists as the result of the use of various ambient sensors, such as video cameras, depth, thermal, radio, acoustic, and wearable sensors (eg, in smart watch). The perception data from the ambient sensors, when integrated with various AI solutions, can help to design early warning systems to prevent adverse safety events, improve and support decision making, provide clinical workflow and operational efficiencies, and relieve administrative burden for physicians.

Current and Potential Applications

Ambient AI applications in health care delivery, as shown in Figure 2, can be leveraged at various touch points, such as outpatient clinics, hospital, and patient’s home/daily living space to augment health care delivery, by benefiting patients and their direct care providers/family members, clinical care providers, and enterprises.

Figure 2.

Ambient AI applications in health care delivery. AI, artificial intelligence; EHR, electronic health record; ICU, intensive care unit.

Use of Ambient AI in Outpatient Clinics (Point of Care)

Ambient clinical intelligence: This is a conversational AI application that integrates ambient voice sensing technology and virtual assistant function, to automate and streamline the visit documentation into the electronic health record (EHR), and data retrieval from the EHR by a physician, during a patient’s clinic encounter.⁶ This application can be used at various point-of-care settings, including outpatient offices, and can also be integrated with virtual visit (Telemedicine) platforms. The touchless voice activated virtual assistant helps to decrease the administrative burden of physicians, enables better physician-patient interaction, and improves patient satisfaction and experience.

Ambient AI technology is being used in health care to create virtual nurse assistants that can provide personalized care to patients. These virtual assistants can be integrated into patients’ homes, hospitals, and clinics, providing real-time assistance to patients, monitoring their health status, and communicating with health care providers.

For example, in a hospital setting, virtual nurse assistants can monitor patients’ vital signs and alert the health care providers if there are any changes in their condition. They can also provide patients with information about their medications, assist with scheduling appointments, and even remind patients to take their medication.

In a home setting, virtual nurse assistants can provide patients with reminders to take their medication and follow their treatment plans. They can also monitor patients’ health status and alert the health care providers if there are any concerning changes.

One example of a virtual nurse assistant is Sensely’s “Molly,” a conversational AI platform that provides patients with personalized health coaching and care management. Patients can interact with Molly through a smartphone application or a smart speaker, and the platform uses natural language processing to understand patients’ needs and provide personalized support.

There are various benefits of ambient AI–enabled virtual assistants. The virtual nursing assistants can lower health care costs by reducing the need of in-person nursing staff visits and preventing hospital readmissions by timely triage and enabling care coordination with the physicians.⁷ Physician facing virtual assistants in the form of digital scribe can help in decreasing physician burnout.⁸ Overall, ambient AI is transforming the health care sector by providing new ways to deliver personalized care to patients, and administrative assistance to physicians.

Use of Ambient AI for Inpatient Hospital Care

There are two important hospital spaces where ambient AI has great utility. These spaces are intensive care or critical care unit (ICU) and operating rooms.

Use of Ambient AI in ICU: In ICU, ambient AI has 3 main applications as mentioned further

• i.First is to prevent data fatigue and augment the workflow of clinicians. Mayo clinic built an EHR interface for clinicians in the ICU called Ambient Warning and Response Evaluation. This ambient intelligence–based application enabled filtering of meaningful data from the vast volume of data entering into the EHR, delivering real-time context specific, high value clinical information to the physicians, augmenting timely clinical decision support, and preventing data overload.⁹
• ii.Second is monitoring of patient mobilization. Ambient sensors installed in ICU rooms can help in evaluating patient movements, detect use of external assistance, and interactions with physical space such as sitting on a bedside chair. Ambient intelligence can be used in future to study the relationship between patient mobilization, length of stay, and patient recovery.¹⁰
• iii.Third is infection control. To prevent hospital acquired infections, ambient sensors can help to monitor hand washing activities.¹⁰ Chen et al,¹¹ have studied the use of computer vision and depth sensing sensors to measure health care worker-patient contacts and personal protective equipment adherence within hospital rooms. The authors concluded that using the computer vision and depth sensing, we can estimate potential hand hygiene opportunities at the bedside and estimate adherence to personal protective equipment.

Insights gained from research studies using Ambient intelligence can guide changes in behavior of hospital staff, lead to better infection control strategies in hospital units, lower morbidity rates, and reduce length of stay, resulting in better patient outcomes.

Use of Ambient AI in Operating rooms

Use of ambient cameras acquired video during surgery, and computer vision has potential in evaluation of surgical skills of the surgeons, facilitate timely feedback for refining surgical skills, which can improve technical efficiency, and decrease the complication rate.¹² Another potential use of ambient AI is the use of ambient cameras to automate surgical tool count, for preventing accidental retention of surgical instruments inside the patient and its associated complications.

Use of Ambient AI at Home (Daily Living spaces)

In daily living spaces, ambient AI has 4 potential uses as mentioned further.

• i.Ambient assisted monitoring and fall prevention: ambient sensors in living spaces for the older population can help to monitor daily living activities, impairment of which can be associated with increase in risk of fall, with its adverse health consequences.¹³ Ambient intelligence systems with integrated contactless sensors in daily living spaces of older populations can help in prompt detection of fall, provide alerts to the caregivers, facilitate emergency response, and provide timely intervention to prevent morbidity and mortality.
• ii.Ambient assisted rehabilitation and disease surveillance: ambient sensors can be used for gait analysis, which can help in designing optimum home rehabilitation programs for patients recovering from stroke. Ambient sensors can also be used at home, for patients with Parkinson disease. Yang et al,¹⁴ studied the use of AI-based system that integrates contactless ambient radio sensor for detecting Parkinson's disease, predicting disease severity, and tracking disease progression over time using nocturnal breathing. Authors in this study found that AI system using ambient sensor can identify people who have Parkinson's disease from their nocturnal breathing and can accurately assess their disease severity and progression.
• iii.Ambient assisted living: Assisted living technologies based on ambient intelligence are called ambient assisted living tools, which can be used for improving quality of life for elderly and people with functional diversity.⁵ These may include voice assistants and smart home solutions that can help with automated task performance for people with motor disability, timely medication reminders, and provide adaptations for patients with cognitive impairment. Social robots can help to address loneliness among the older population by providing social company and promoting social engagement with their relatives.
• iv.Ambient assisted working: with an increase in aging workforce that is challenged by physical and cognitive limitations and chronic health disorders, there is an emerging role of Ambient Intelligence–Ambient Assisted Working. These ambient assisted working tools and solutions can use wearable and environmental sensors, cyber-physical systems, and AI to provide flexible workplace adaptations and offer support and adaptation to older working population in a variety of workplace scenarios, helping them to fulfill different work-related activities.¹⁵

Ambient AI applications in health care delivery can create value by promoting quadruple aim, which includes improved clinician experience, improved patient experience, lower cost, and better outcome, as shown in Figure 3.

Figure 3.

Value creation with ambient AI. AI, artificial intelligence.

Challenges and Barriers in Adoption

As we look forward to designing, developing, and scaling ambient AI applications in health care delivery, the following challenges need to be considered and optimally addressed as applicable to the use case, keeping human-centered AI, and multistakeholder perspectives in mind.

Data Privacy, and Security

Various ambient sensors may capture a large volume and variety of personal/private data, which the patient may not be comfortable in sharing. There is risk of inappropriate sharing of data without the patient’s knowledge and use of data beyond its original intended use or by third parties without informed consent.

Fairness and Bias

Ambient AI applications used in clinical practice should be fair and bias free. The potential for bias is a recognized challenge for the implementation of AI applications in health care.16, 17, 18 Ambient AI applications may also be subjected to these biases, including bias in data sets (input data) and algorithmic bias. Beyond the computational and statistical biases, 2 important categories of bias that are overlooked are human bias and systemic bias.

Human bias includes cognitive bias that affects decision making of the individuals involved in designing and developing the ambient AI solutions. The systemic bias is operational at the level of the entire health care institution/enterprise, which has practices or norms that result in the favoring or disadvantaging of certain social groups.¹⁹ Attention should be given to addressing bias and ensuring fairness, while designing, developing, and deploying the ambient AI applications.

Transparency and Explainability

For patients and physicians to trust ambient AI applications, it is important to ensure data and algorithmic transparency. There should be clarity regarding data set composition, mode of data collection, and process of annotation and how the data will be used and shared. Model outputs should be interpretable and clinically relevant. An important challenge is ensuring explainability, which implies the ability of the ambient AI system to clearly explain to the end users its prediction and decision-making process.²⁰ In context of explainability it is important to mention about Explainable AI (XAI), which is a research field that aims to make AI systems results more understandable to human.²¹ Explainability and transparency, help in developing trustworthy ambient AI applications for physicians and patients.

Ethical Use

Ambient AI applications should not be unethically used causing harm or adversity to the patients. Important stakeholders should be included in the design and development phase to ensure ethical implementation. Ambient AI governance framework should include ethical use as an essential component.²²

Legal Liability

Increase in number of ambient sensors and the volume of data associated with continuous ambient sensing introduces legal challenges, such as who is responsible when there is delay in response or when there is lack of timely response to the Ambient AI system?²³ Additionally, who bears the liability if there is an adverse clinical outcome due to an error in the output of the ambient AI system?

End User Resistance

Patients may not be comfortable with use of ambient AI applications due to concern about loss of privacy given the thought of being continuously monitored.¹⁶ There may be additional concerns about security and unauthorized sharing of personal and sensitive data, which may be used to their disadvantage. Physicians may not be comfortable using these systems, due to lack of familiarity, technical knowledge, and trust.

Human-Centered Approach

During design and development of the ambient AI applications, lack of human-centered approach, especially in older population who may have physical, mental limitations, and lack of comfort with technology use, can be a challenge for successful adoption.²⁴

Future Directions

Looking ahead in the future, consideration should be given to the following:

• 1.Use of human-centered design, customizing the ambient AI application to the specific needs of the end user, and the context of use, will help in providing good user experience and promote adoption success.²⁴ This will help to address the challenges of the human-centered approach in adoption.
• 2.Design of smart hospitals with innovative use of ambient AI applications to improve patient throughput in outpatient clinics, increase productivity of physicians and nurses by decreasing administrative workload, data fatigue, and prevent adverse patient outcomes by using early warning systems.²⁵
• 3.Development of collaborative communities involving physicians and important stakeholders across the health care ecosystem to promote clinical validation of ambient AI applications and sharing of best practices. This may help to address the end user resistance among physicians in adopting these systems.
• 4.Promotion of proper data stewardship, development of optimum governance framework, integrating privacy maintaining strategies, and informed consent with transparency regarding data use and sharing. This will help address the challenges of fairness, bias, transparency, trustworthiness, and ethical use of ambient AI applications.¹⁶

Conclusion

Ambient AI is a promising emerging technology, which is in the very early phase of adoption but has great potential to become mainstream over the next 10-15 years. The use of ambient sensing technology and connected intelligence networks in the era of 5G will provide another dimension to a patient’s health care journey, transform the health care delivery of tomorrow and enable us to practice intelligence-based medicine.

To gain optimal value from this technology, attention is needed to ensure human-centered approach, robust data privacy and governance strategies, multistakeholder collaboration, and shared decision making by physicians and patients.

Potential Competing Interests

The authors report no competing interests.