top of page

Shaping the future of healthtech with equity at its heart: event wrap-up

Shaping the future of healthtech with equity at its heart: event wrap-up by OHT Fellow, Charly Massey


“Equity: the practice of giving to others what is their due” 



The planning for this Refresh London OHT event started a while ago and it turned out to be amazing. The night was full of excitement. PWC, our wonderful partners, played the role of super-Host (thank you!) and the evening was filled with genuine connections and open conversations. The chat with our panel of true industry experts not only focused on the important theme of equity, but also provided a special opportunity for us to explore each other's stories. The main thing I took away from the panel session was the powerful impact of understanding one another—revealing motivations, exploring backgrounds, and appreciating the unique qualities that make each person special.


This understanding of interpersonal relationships emerged as a crucial element in promoting the inclusive spirit that is necessary for equitable health tech. There were so many wonderful pieces of advice and valuable insights generously shared by our three panelists, all of which I've tried to present in a user-friendly, easily scrollable format for maximum accessibility. See I listened!


A heartfelt thank you to everyone who attended, and please help us spread the love by sharing this blogpost with anyone you believe would find it helpful.


Let’s jump in to our 6 tips for shaping health tech with equity at its heart...


1. Equitable career advice: 

All of our panellists had followed a different path to what they had planned. Understanding the multitude of possibilities that their training and experience prepared them for, transformed their career’s. 😊 Anna studied to become a bioinformatician, but when she realized this opened up opportunities beyond the lab, a fire was lit! 

It can be tough and frustrating to try to achieve things on your own, so collaborating with the right people is crucial for project success, especially your own personal one. Don’t limit networks to computer systems. Build alliances with like-minded and useful people with whom you can open doors

The culture and diversity of an organisation are important in finding project champions and supportive friends who will help you design with optimism. Pay close attention to this when you're looking for opportunities, and although it can be frustrating to be the only female or feel other in a meeting, we are all part of the change. Take your place.

In real life, outside of work, we engage and connect with many diverse individuals. Remember that you are not just a persona. To rock at your job, don't limit yourself with a narrow mindset or role. Embrace your own humanity.


2. Equitable Data [Collection]: 

The quality of a model depends on the quality of its training data. Equity, being a subjective concept, is influenced by various factors such as ethnicity, gender, and individual expression. In the field of health tech, subjectivity also exists. We cannot assume that shared information, like symptoms, is consistently captured, constant, or comparable.

Communicating using familiar norms and cultures can greatly enhance the involvement of healthcare users in data collection. By providing clear explanations of the benefits of participating in research trials, we can motivate them to contribute valuable and up-to-date information. This, in turn, enriches the quality and diversity of information available for machine intelligence.

Deciding if the information is good enough to make decisions is an important thing to think about. In some cases, the data collected may be obtained too late. For example, in domestic violence cases, other public agents and services might already have noted important signs before someone seeks NHS primary healthcare. However, these signs are usually not digitised like primary healthcare data and aren't accessible. Technologies like speech-to-text conversion are now including historically excluded information and transforming unstructured data into structured formats.

However, there is still work to be done to improve the reliability of data collection methods. For example, PPG readers in watches are affected by skin colour.

Gathering data from multiple sources enhances comprehension. However, model infrastructures that rely on insecure data stored in various locations with scattered data flows are vulnerable to malicious attacks.

Also, it's incredibly important to take into account social biases in datasets, both inherent and at intersections. This is crucial because an individual may display both typical and atypical characteristics, but these biases are not always considered.


3. Equitable Model [Development]: 

AI should be treated as any other medical tool and undergo similar safety testing and meet the same standards as other software.

The issue of AI bias is often oversimplified by the public. It is not just about training the model to be representative of the population. It is also important to ensure accurate data capture and desired model performance.

Aligning human and machine intelligence to determine and achieve desired performance goals is crucial.

Reliability for all intended uses should be ensured during deployment.

Consider the model's performance in both lab settings and real-world scenarios. How does it adapt to diverse real-world data over time? Additionally, having an "off-switch" to safely decommission and replace unsafe or inaccurate models is important.

For instance, in a recent Mass Fail testing of two Electronic Prescribing Systems (EPRs) that included GP prescribing tools, it was discovered that the apparent preference for generic or brand selection was influenced by the position of the drugs in a dropdown menu list. This had significant funding implications for the healthcare systems using the two different tools.


4. Equitable Deployment  [User]: 

It's important to know that the average reading level of patients who use the NHS is only 9 years old. Shockingly, 43% of them don't fully understand the healthcare system.

LLM and NL models can now effectively work in over 200 languages with error rates below 1%, which presents a great opportunity to address the inequality between the 65% of patients in NHS England who do not speak English as their first language and rate their health as good or very good, and the 90% of native English speakers who say the same. The readmission rates for non-native English speakers are higher at 25%.

Language isn't the only important factor; the tone and style also make a big impact on engagement. ChatGTP has been used to simplify paediatric asthma care plans so that a 9-year-old can easily understand them.To really connect with these young patients, it also presents the same information within a fun Winnie the Pooh story, 

Using widely native communication channels like SMS and listening to people's voices, and considering the context of their decision-making are all important factors to consider. It's crucial to design for the user, keeping their needs in mind instead of just focusing on the service itself.

It is essential to prioritise what matters most to the end users. Instead of solely focusing on addressing alcoholism within hospital settings.  Providing necessary support and treatment in a familiar environment can yield better results.So engaging with individuals in places like pubs, where they feel comfortable, can be more effective.


5. Equitable Organisational considerations: 

It is concerning to see the lack of data fluency among decision-makers in the IT field. This concern led to the development of OpenSafely.

We still need specific use cases (PPIs), and we are considering ways to address this issue. Creating test bed environments can be challenging, but they offer significant value. Sam Roberts at NICE is working hard to make test beds accessible to potential suppliers for the NHS. This is particularly difficult for smaller organizations, which often lack the resources to conduct evidentiary tests for HRA1 and other requirements. Building trust, promoting interoperability, and acknowledging the value of time and people are essential.

NHS trusts are open to simplified and secure arrangements that revolve around clear and compelling commercial offerings. This marks the beginning of a promising journey, striking a balance between the benefits of ICS, shared care records, low latency, accurate information, and centralised controls.


6. Equitable Societal: 

Predictions are not decisions. 

Social determinants of health play a significant role in shaping individuals' well-being. Policy decisions, such as access to benefits, can have a profound impact on health outcomes. While statistics and figures provide valuable insights, it is equally important to ask ourselves, "What are the underlying problems we are truly trying to address?" By understanding the needs and preferences of the audience, we can provide more effective solutions.

In India, a hospital offered free cataract operations, along with a free week of post-operative recovery. However, there were no takers. The primary reason for this was the financial burden. The individuals in need of the operation were usually the primary breadwinners, and the costs associated with travelling to the hospital and losing a week's worth of earnings proved to be significant obstacles. These obstacles outweighed the potential benefits of improved eye health.



We had a blast putting together this event—care to join us? We'd be thrilled to have you as part of our community!





Comments


bottom of page