Article Type : Research Article
Authors : Uday Jain
Keywords : Digital Health; Anesthesiologists
Personal Digital
Health Assistants (PDHA) are applications that can run on virtually any
computer or mobile device including a smart phone/ personal digital assistant
(PDA). A PDHA acquires, stores, and analyzes health related information of an
individual. It usually communicates with
remote servers of a large organization which can connect it to various
resources. PDHAs are an integral part of telehealth. Their utility has increased
many fold since the start of the pandemic. The systems are increasingly more
complex and are involved in all aspects of care. The PDHA are usually utilized
by patients or their caregivers with assistance from professionals. Many
automated systems can be utilized free of charge. Some are available as part of
a prepaid health plan. Increasing number of patients that anesthesiologists
care for, utilize a PDHA. Additionally, anesthesiologists may utilize PDHAs for
diagnosing and managing patients as discussed below.
Artificial intelligence
(AI) has substantially increased the power and utility of PDHAs and telehealth.
Large medical databases have been accumulated by some medical insurers,
national health services, and medical institutions. These have been used to train
AI systems to mimic diagnoses and management decisions of physicians. These
systems utilize neural networks and machine learning. Some utilize branching
logic or statistical inference. AI based systems are superior to older keyword
or rules-based systems in which each symptom is associated with a particular
disease with an assigned probability. Training of these systems is usually
ongoing and requires input from medical and information technology
personnel.
The patient can consult
with the clinician via text, voice, or video on a HIPAA compliant connection.
Widely utilized tele-conferencing services like Zoom have HIPAA compliant
versions. Symptoms and other information may be entered by selecting from a
list. Many systems also communicate by natural language text messages. Some
allow natural language voice input and integration with digital voice
assistants such as Amazon Alexa, Microsoft Cortana, and Google Assistant.
Natural language input may be parsed into a list of statements the system
understands. The user can choose the desired statements. Integration with
services such as Google Cloud allows for real-time captioning and language
translation. Some systems are available in multiple languages. Voice and video
consults can be live, i.e., real time, or asynchronous, i.e., recorded. Instead
of projecting the physician’s image on the patient’s screen, some systems have
the option of projecting a file photograph, an avatar, or another image. Some
patients are more willing to reveal their mental and other health information
in that setting. Voice recognition may allow these systems to listen in on the
physician patient consult and suggest differential diagnosis and management.
This can be done in real time or off-line. Some systems allow digital input of
data acquired by various physiological sensors and monitoring systems.
One of PDHA’s tasks is
to develop a differential diagnosis based on symptoms and other information. In
this role it has been referred to as a symptom checker. In the perioperative
period, anaesthesiologists usually care for patients with established
diagnoses. Many patients are familiar with their chronic conditions such as
hypertension, diabetes, and cardiovascular disease and are receiving care.
However, in the settings of preoperative clinic, perioperative surgical home,
pain clinic, and intensive care unit, anaesthesiologists may be called upon to
diagnose and manage their patients. Like the public, anaesthesiologists can perform
Google and other web search of the relevant information. Every day, over one
billion of Google searches are health-related [1]. However, such a search may
not integrate all the information that is known or can be elicited, and is
usually not adequate for clinical management. PDHAs provide diagnosis decision
support that is superior to internet searches. After the symptoms and other
information are entered, the system asks a series of questions. Each question
depends on the answer to the previous question. Usually only a few questions
are asked. They consider the patient’s demographics and medical history. Some
may also consider vital signs, input from physiological sensors, and laboratory
values. The systems provide the differential diagnoses and management plans.
They assist the physician in considering the full range of differential
diagnoses and avoid missing a potential diagnosis. These systems have become
practical tools for clinical diagnosis. They are most useful in pre-primary
care setting. They pre-diagnose, triage, and can connect the patient to the
appropriate medical facilities, providers, support services, and resources.
They can assist call centers and emergency departments in triaging. They are
most useful in diagnosing conditions that are not common.
Automated systems serve
only in an advisory role. No doctor-patient relationship is established. They
cannot make management decisions or order prescriptions. Many systems provide
the option to pass information to in-house or outside physicians who manage the
patient, usually for a fee or as part of a prepaid plan.
In response to the
pandemic, US Centers for Medicare and Medicaid Services initiated “Hospitals
without Walls” allowing care outside hospitals, in locations such as designated
hotels. In November 2020, it expanded to “Acute Hospital Care at Home [2].”
These regulatory changes along with advances in technology have greatly
increased the utilization of telehealth. Organizations such as Mayo Clinic and
Kaiser Permanente are utilizing it to provide at home acute care previously
provided in inpatient settings. Frequent detailed interaction between the
patient and the health system, facilitated by the PDHA is critical for this
purpose.
Most of the systems
participate in chronic care. They can prompt the user to collect clinical data
on an ongoing basis. Data can be acquired from sensors and monitors. Data can
be analyzed, transmitted, and shared with all the stakeholders. The systems can
prompt the patients to follow therapeutic recommendations such as taking their
medications. The systems can record the responses. They have become important
tools in providing longitudinal care.
During the pandemic,
much of the clinician provided mental health care is via PDHAs. This is likely
to continue in the future. Many fully automated systems are also available to
interact with clients.
PDHAs are playing an
important role in preventive care and health maintenance. Smart phone
applications that track physical activity are utilized widely. PDHAs can
provide individualized instructions on life-style choices. PDHAs are also
suitable for providing information about alternative and integrative medicine.
Many systems have their own libraries and can also be used to access outside
libraries.
Some systems have a
clinical educator mode which combines a differential diagnosis generator with
case histories. It is useful for teaching clinical reasoning and medical diagnosis
to clinicians. Links to medical literature relevant to the diagnoses are
provided.
Information in the
electronic medical records (EMR) is usually entered by clinicians and
facilities such as laboratories and is deemed reliable. Patients can usually
access some of the information in the EMR without being able to alter it. Many
entries in PDHAs are made by patients and caregivers. PDHAs may contain more
information than EMRs, especially during the period the patient is in the usual
state of health. Clinicians can extract reliable and relevant information from
the PDHAs and include it in the EMR. Some PDHAs provide an application
programming interface to facilitate communication with EMRs and other systems.
Usually, these systems
are proprietary. Interoperability and communication with other PDHAs and EMRs
are limited. They are generally not regulated by public bodies. They have not
been adequately validated by independent studies. Accuracy figures are usually
not available [3]. Because they are frequently updated, it is difficult to
evaluate them with studies that can take months. The results provided by any of
these systems can be incorrect and should be verified by a clinician [4].
When used for triaging,
these systems tend to be risk averse and are likely to triage the patient to
get unnecessary further care. When used for determining differential diagnoses,
their utility is limited if they provide a long list of possible conditions.
Hundreds of systems are
available. A few of the prominent ones are listed below.
Amwell.com: It connects
72 thousand providers to 80 million members. It assists in all stages of care.
teladochealth.com: It is used in the US and worldwide. It provides
comprehensive services to 52 million members. It includes Livongo which focuses
on chronic disease management.
ada.com:
This German system is used in many countries by over 11 million users. It uses
text chat on a mobile device.
buoyhealth.com: Developed
through the Harvard Innovation Laboratory, the company’s algorithm was
initially trained on data from 18 thousand medical papers, 5 million patients,
and 1,700 health conditions. The chatbot provides diagnoses and actionable
steps. It also has a database of health information.
isabelhealthcare.com: It can be integrated with electronic medical
record. Its paid clinical educator mode for physicians combines differential
diagnosis generator with case histories [3]. It is useful for learning clinical
reasoning and medical diagnosis.
amazon.care: It is available to employees of Amazon and other
companies. There are plans to substantially expand the system.
symptomate.com: It allows text chat or voice input through a
digital voice assistant. Multilingual support is available. It is one of the
few systems that can work on a phone as a voice-based application. More than
four million persons have used it.
medicallyhome.com: It facilitates provision of acute care in
extra-hospital settings. Mayo Clinic and Kaiser Permanente are utilizing it.
A rapidly increasing
number of AI aided PDHAs are already being utilized by millions of persons.
They are an integral part of telehealth. These systems are improving rapidly
and are bound to play an ever-increasing role in perioperative and other
medical care.