Using Artificial Intelligence in Health Economics and Outcomes Research

Artificial Intelligence (AI): is the simulation of human intelligence activities by robots/machines/ technologies, especially computer systems

Health Economics and Outcomes Research HEOR: Read this HEOR post

AI can play a crucial role to strengthen healthcare systems from disease prediction to patient access.

Let’s explore the application of AI in HEOR specifically. AI-related analysis (such as machine learning [ML], deep learning, neural networks, etc) can help to improve disease prevention and diagnosis, which will help indirectly reduce R & D burden and costs, improve patient as well as clinical team outcomes, reduce costs, and impact population health. AI have the potential to change many facets of patient care within the provider, payer, and pharmaceutical industries.

• Population Health Management
  • AI when applied to electronic medical records, can predict the duration of hospitalization for patients and the probability of their readmission based on predictive models.
  • Predictive ML algorithms, probabilistic causal models, and unsupervised algorithms enable the prediction of the geo-location and time of the next outbreak. Similarly, whether the patient has access to medical care in a specific geo-location or not.
  • ML Approaches such as extreme gradient boosting can be widely used to identify the risk of particular diseases in populations using Big Data. E.g.: Detect patients with unrecognized mental health needs; Estimate risk of a post-surgical complication
  • Unknow barriers/risk factors to care such as socioeconomic status, the region can be identified using AI-enabled models to improve a patient’s outcome, quality of care.
  • This can help policymakers and health leaders to tailor interventions to the populations they serve by learning relationships between social determinants of health and population health challenges

• Drug effectiveness
  • During an outbreak of COVID-19 times, there is a race to develop drugs to control it. However, the drug needs to pass efficacy/effectiveness testing. AI can accelerate the prediction of drug effectiveness.  
  • Data from electronic health records and post-marketing reports of real-world use can be used to understand drug effectiveness. This approach has been used by multiple pharma companies.
  • AI can utilize big data to assess the effectiveness of drugs already in use among the different patient populations.
  • For improving patient-level outcomes, there is a need to understand drug efficacy in a real-world setting.
    
• Medication Adherence
  • A patient’s adherence to medication means the patient taking medicine as prescribed by the doctor. AI can use patient history to understand if the patient will be non-adherent. E.g., patients with Alzheimer’s will forget to take medication, patients with multiple issues may simply not take medication due to ‘pill burden’. AI can detect this beforehand and provide reminders via texts or apps to take medication.
• • As HEOR scientists, we can have an intervention to improve adherence as well generate evidence about cost reduction due to improved adherence.

• Adverse event detection
  • Adverse events are one of the top ten causes of death and disability worldwide and reducing adverse events with help to improve patient outcomes.
  • AI-based natural language processing (NLP) approach helps to identify and extract text from the vast field of social media, literature articles, clinician’s charts, etc. without requiring a human interface.  This monitors and identifies adverse events ahead of time, saving cost as well as improving patient care.
  • AI methods such as machine learning (ML) can be used to identify adverse events. ML algorithm can be used on the existing database of reported adverse events and understand known and unknown adverse events associated with medicines. The algorithm can help with predictions of adverse events even before prescribing to new patients.
  • HEOR teams can use this information in value messaging of drugs and related market access activities.

• Patient enrollment and retention for clinical trials
  • Patient experience in adverse events using AI will directly help with patient retention in trials. NLP  systems help to analyze  patient  records  to determine  if the  patients  met  given eligibility criteria  for clinical  trials
  • Mobile apps, wearable technology can be used to capture data, which further can be used to develop a patient profile.
  • Personalized patient monitoring can be implemented at the site with a reduction in human involvement. This will enlighten clinicians or site investigators to enhance patient engagement.
  • Patient-reported outcomes (PRO) development can be facilitated. ePROs will provide ease of access and reduce patient dropout from PRO studies.

Challenges that need to address while applying AI in HEOR:

• Data privacy  
• Ethical consideration
• Regulatory policies
• Data Availability
• Data Quality
• Infrastructure and systems requirements