Table 4 Tips for identifying clinical problems for AI solutions

1. Understand Healthcare Challenges

a. Can begin with your organisation’s value statements, then ponder on the current challenges and pain points within healthcare services. These could include issues related to

i. diagnosis accuracy,

ii. treatment effectiveness,

iii. patient outcomes,

iv. workflow inefficiencies,

v. resource allocation, or

vi. healthcare accessibility

b. May conduct medical audits to ascertain the relevant aspects or potential areas, or engage colleagues in services or those in the administration to explore and prioritise issues to be resolved

c. Identify potential use cases where AI can have a significant impact on improving patient outcomes, enhancing clinical decision-making, optimising healthcare delivery or reducing healthcare costs

d. Prioritise use cases based on factors such as clinical relevance, feasibility, scalability, and potential for positive impact

e. Literature review could be conducted to further understand related or prevalent problems, emerging trends, and possible interventions to improve the issues in healthcare delivery

2. Analyse Healthcare Data

a. Analyse available healthcare data, including electronic health records, medical imaging data, genomic data, wearable device data, and healthcare claims data. Look for patterns, trends, and anomalies that may indicate areas of concern or opportunities for intervention

3. Evaluate Feasibility and Resources

a. Assess the feasibility of implementing AI solutions for identified problems within the current healthcare ecosystem. Consider factors such as data availability, technical infrastructure, expertise needed, funding resources, and organisational readiness

4. Regulatory and Ethical Considerations

a. Consider regulatory requirements, ethical considerations, privacy concerns, and data security issues when identifying AI-driven solutions for healthcare problems. Ensure compliance with relevant regulations such as data protection regulations

Descriptive (Assistive AI):

An assistive AI algorithm that provides descriptive insights into patient demographics and healthcare utilisation patterns within a hospital system. This algorithm analyses historical data from electronic health records to generate reports and visualisations depicting patient demographics, admission rates, length of stay, and common diagnoses. Clinicians and hospital administrators can use these insights to better understand patient populations, allocate resources effectively, and optimise healthcare delivery processes

Diagnostic (Assistive AI):

An assistive AI algorithm for medical image analysis that assists radiologists in diagnosing breast cancer from mammography images. This algorithm uses DL techniques to analyse mammography images and detect suspicious lesions or abnormalities indicative of breast cancer. Radiologists can review the algorithm’s findings alongside their own assessments to improve diagnostic accuracy and reduce the risk of false positives or false negatives

Predictive (Autonomous AI):

An autonomous AI algorithm for predicting patient readmissions to the hospital within 30 days of discharge. This algorithm leverages ML algorithms trained on historical patient data, including demographics, medical history, diagnosis codes, and previous hospitalisation records. By analysing these data points, the algorithm generates personalised risk scores for each patient, indicating the likelihood of readmission. Healthcare providers can use these predictive insights to intervene proactively and provide targeted interventions to high-risk patients such as care coordination, medication management or follow-up appointments to prevent readmissions and improve patient outcomes

Prescriptive (Autonomous AI):

An autonomous AI algorithm for personalised treatment recommendation in oncology. This algorithm analyses genomic data, tumour characteristics, treatment history, and clinical outcomes from a large database of cancer patients. Based on this analysis, the algorithm generates personalised treatment plans tailored to each patient's unique profile, including chemotherapy regimens, targeted therapies or/and immunotherapies. Oncologists can use these prescriptive recommendations to make informed treatment decisions and optimise patient outcomes while minimising the risk of adverse effects or treatment resistance