Page 23 - AIH-1-1

Artificial Intelligence in Health NLP in EHR

these records exhibit distinguishable parts. Notable diverse assessments in different emergency departments,
among these are the identity of the patient, the reason delayed arrival of specialist of day/on-call, late arrival
for the visit, the history and background of the patient, of investigation reports, poor patient response to initial
prevailing symptoms, assessment and treatment timelines, emergency department management, document processing,
specific points of documentation, treatment outcomes, the investigations toward personal and economic constraints
discharge letter, and the authorship of the record. of patients, and waiting for vacant beds. In a study by
[11]
During the doctor’s visit, the physician initially listens Bukhari , contributors to delays encompass laboratory
to the patient and subsequently conducts an assessment time, admission to observation, admission to trauma,
to identify and evaluate symptoms, working toward the duration of physical response to the final decision,
exclusion of potential symptoms. This process culminates consultation time, critical care management patients,
in a diagnosis specifying the name of the disease and door-to-final decision time, radiology time, non-critical
indicating the potential anatomical location of the disease. care management patients, triage cases, admission to
As patients may visit the same doctor multiple times or resuscitation room, doctor-to-consultation time, doctor-
multiple doctors, they acquire multiple prescriptions, to-radiology time, doctor-to-laboratory time, and door-
posing challenges in paper-based management. To to-doctor time.
address this, implementing a unified system for document Baker and Melby discussed the importance of
[12]
management becomes imperative . The paper record file communication with unconscious patients, highlighting
[1]
grows increasingly voluminous with each patient visit, that the patient’s level of consciousness, the extent of
presenting difficulties in analyzing paper-based patient physical care required, and the presence of relatives can
records using computational linguistic methods. This influence effective communication. In any emergency
challenge arises from the need for scanning and optical situation, crucial information is often sought, including
character recognition to extract information from paper- the patient’s name, place of stay or origin, phone and fax
based records. numbers, date of birth, blood type, social security number,

In addition, reliance on paper-based records health insurance details (both individual and group),
contributes to long waiting queues for registration, personal physician(s), emergency contacts, existing
impeding the ability to provide timely solutions. These conditions and disabilities, current medications, assistance
drawbacks of paper-based records prompted the invention requirements, allergy susceptibility, immunization dates,
of electronic health records (EHRs). Compared to their and communication/equipment/other needs. However, a
paper-based counterparts, EHR offer various advantages, challenge arises when communicating with unconscious
including remote access to patient information, enhanced patients and capturing this information. To address this
revenues, and improved communication among issue, the implementation of EHR accessible to authorized
practitioners . Deciphering someone’s handwriting can personnel becomes imperative.
[2]
sometimes require great imagination, but EHR alleviates A manual search through EHR has been found to
this challenge, enabling physicians to provide more increase the cognitive load on the user . The utilization
[13]
knowledgeable advice even during off-site or non-regular of fragmented information requires high cognitive
hours. A systematic review investigated errors related reasoning. To address this navigation problem, it is crucial
[3]
to handwritten prescriptions, covering studies published to minimize the actions required to retrieve intended
from 1985 to 2008. These studies reported an error rate of information. Thus, an information system is needed to
7% (interquartile range: 2 – 14) per medication order [4-7] , reduce unnecessary cognitive load on working memory,
with a comparable error rate ranging from 10.7 to 14.7%. thereby liberating cognitive resources for higher reasoning
Medical history has been shown to identify 70 – 90% of among clinician users. According to the Progress on
[8]
diagnoses . Without paper-based records, only 10.9% of Implementing and Using Electronic Health Record Systems
patients can remember their current medications. These – Developments in Organization for Economic Cooperation
factors underscore the need for implementing EHR. and Development (OECD) Countries as of 2021 , Australia,
[14]
According to a report from the World Health Belgium, Canada, the Czech Republic, Germany, Hungary,
Organization , the number of patients who die before Italy, Korea, Lithuania, Mexico, Slovenia, Switzerland,
[9]
reaching a hospital in low-income countries is twice that and the United States currently do not employ artificial
in high-income countries. Furthermore, on reaching intelligence (AI) for the processing and analysis of EHR
hospitals, various factors contribute to treatment delays. data. In eight countries – Costa Rica, Finland, Denmark,
Factors identified by Mohammad and Tashkandy Israel, Luxembourg, Portugal, Turkey, and the Netherlands
[10]
include multiple consultations, critical care management, – AI is utilized for automated alerts, messages, and actions

Volume 1 Issue 1 (2024) 17 https://doi.org/10.36922/aih.2147