Healthcare often seems complex and disintegrated for patients. Everyone involved in the system has serious concerns about legacy technology, including patients, nurses, administrators, and patients. Healthcare is expected to be more patient-centric, and more than half of the consumers said that the industry is more focused on itself.
Do you think that physicians and nurses are satisfied with the current system? The Physician burnout rate is continuously increasing, and the pandemic has added to the deepened healthcare issues. The number of administrative tasks, after-hours workload, and lack of personal control over the workloads often contribute to physician burnout. The burden of EHR on physicians is continuously increasing. On average, doctors spend 16 minutes on EHR per patient, while the time allotted is around 15 minutes per appointment. During the pandemic, patient-originated messages related to EHR increased, the average number of messages received per day almost doubled. Physicians are responsible for direct communications to their inbox, which may add to burnout in the future. Given the issues faced by all the stakeholders, AI has a significant role in healthcare. Fragmented care is no longer relevant, and patient-centric care is evolving. More than 50% of healthcare experts believe that AI will be ubiquitous in the industry by 2023. Our industry thought leader, Khalil Sheikh, shared his insights about reimagining Patient experience with data and AI. Let’s go through them in detail:
Why do you think that AI will be the future of healthcare?
Khalil Sheikh: The transformation of healthcare with AI has just begun. The changes have crept in to rethink patient experience, clinicians’ modus operandi, and the pharmaceutical industry operations. Broadly speaking, applications of AI in healthcare can be categorized to:
- Patient-centric AI
- Clinician-centric AI and
- Operational and administrative AI
AI in diagnosis and medical imaging has taken a big leap in recent years. Radiologists use AI technologies to screen millions of images to identify abnormalities and patterns. AI usage is not just to reduce the risk of misdiagnosis but also to reduce human efforts for more critical tasks. As we speak to experts in the industry, we see that around 30% of hospitals and imaging centers are using some form of AI in diagnosis and medical imaging, and about 60% are looking forward to investing in AI.
Most of us experienced the quick rollout of the Covid-19 vaccine. And AI was one of the driving forces for vaccine availability for a brief time. Pfizer leveraged AI to find signals in millions of data points generated from a 44,000 people study during the clinical trials.
The application of AIcould be everywhere from simple to complex tasks – symptom analyzer to medical records review, drug design, processing radiology images, advising treatment plans, and talking to patients. AI at large will help increase convenience, reduce costs and errors and drive access to the new value-based care model.
What are the emerging challenges in the wake of the pandemic and opportunities in healthcare that AI/ML/NLP can address?
Khalil Sheikh: Like every industry, the pandemic has increased workforce challenges in healthcare. And access to care for any other conditions barring covid was challenging. Virtual care/telemedicine resolved a few challenges and improved some care challenges.
But understanding the underlying healthcare condition still remains a challenge in most healthcare practices. Data regarding patients’ medical history or social determinants of health is still tough to find when in need. During the first quarter of 2020, data related to underlying health conditions were only available for 5.8% of the patients hospitalized with covid during Q1 2020. Most of us think that the challenge was resolved with EMR/HER, and there is more transparency. Yet the pandemic has highlighted the gaps in the existing technology.
AI is not about analyzing the early warning signs and providing treatment for future conditions; it can address the gaps in the current healthcare system. Simple examples like matching the medical records with high precision were overlooked. With the pandemic, experts now have the visibility into the gaps than ever before.
Read More: Infographics
How do you think AI/ML/RPA can be integrated in existing healthcare workflows?
Khalil Sheikh: As we speak about the wider AI adoption for different scenarios, AI needs to work efficiently within existing workflows to transform the patient experience, improve diagnosis, and optimize operations. You need to have deep healthcare knowledge and a complete understanding of the workflows to do this. For example, if you want to implement an AI tool, you need to understand the clinical benefits, security measures, and data management roadblocks. Physicians are already overloaded with clinical decisions and patient care. You can find more than 250 vendors in the medical imaging and radiology space alone.
A trusted service provider can neutralize your challenges to identify the right tools as per your existing landscape. Also, there are multiple approaches and algorithms for a single problem. You cannot wait long to see the results that can work for you. Another significant challenge in healthcare is the lack of strategic direction in implementing AI and the rising costs resulting from this. The service providers now understand multiple tools that can realistically integrate into the existing workflows.
Any specific Healthcare-centric AI implementation and its benefits that you wish to share with our readers?
Khalil Sheikh: As we talk about personalization in every industry, let us adopt the same to healthcare. Personalization in healthcare begins with access to self-service, continuous communication about health conditions, care, and well-being options. AI has a role everywhere – conversational AI for self-service, NLP, and text mining for communications.
In addition to these historical data and disease pattern analysis helps physicians to match the patients with treatment options that can be most effective for them.
Personalized or precision medicine shifts the care emphasis from being reactive to proactive. The tailor-made treatment plan also reduces the trial-and-error prescriptions, increases patient engagement, and provides better care for desired outcomes.
Is there a methodology to drive success when adopting the AI framework in the healthcare industry
Khalil Sheikh: AI systems continue to evolve based on the data and the challenges identified in the healthcare industry. But for successful implementation of any AI system, we follow a six-step methodology:
- Identify business objectives and create AI project goals – Every healthcare organization generates billions of data points. Our experts understand your business objectives to assess the data landscape and define the data and AI project goals. We also analyze the existing data landscape to determine the future state and the insights needed for the business objectives.
- Understand and visualize the critical aspects of data – Though the data collection and storage processes look simple for everyone, it is the first step to focusing on data quality. We explore the data in detail and verify the data quality to ease further processing.
- Data preparation – As all of us know, most of the data teams’ time is usually spent on data preparation for the AI model. Based on our past learnings and history, we try to leverage the best frameworks for data preparation and establish the ground truth of information.
- AI/ML models – There are numerous models for the same underlying challenge, but not all provide the same results. We analyze various data science techniques to derive the best AI model and leverage our domain expertise to uncover the true potential. You cannot simply plug an AI solution into the healthcare ecosystem, it must integrate into the clinical workflow effectively. Interoperability and workflow knowledge are keys to operationalizing any AI model.
- Evaluation of the models – Our evaluation framework is simple to understand and complex to cover all the intricacies in the models. We evaluate all the five parameters – data appropriateness, model testing, interoperability, security, version control, and governance for the data model. Most importantly, our experts constantly strive for more precision and accuracy from the underlying AI model.
- Deployment – We deploy the machine learning models based on the usage scenario – web service, batch prediction, or embedded models. Our experts perform periodic model evaluation and maintenance to keep you up to date with the changing ecosystem.