• Written by Azhar Aliyar, Director

A ward manager checks the patient census and bed allocation for the day, trying to balance demand and capacity. This is a familiar scenario in healthcare providers, where the pressure to deliver high-quality care while managing resources efficiently is constant. Behind the scenes, the data that supports these decisions is being generated and stored in various systems, from electronic health records to logistics and supply chain management. However, for an AI system like AsscherAi to provide accurate and helpful answers to queries about this data, certain foundations must be in place.

Challenges in healthcare data

Healthcare providers face unique challenges that make it harder to achieve enterprise data readiness for AI. Scheduling and utilisation of resources, such as operating theatres and staff, are complex and subject to frequent changes. Supply consumption variability is another issue, as the usage of medical supplies and equipment can fluctuate greatly depending on the type of patients being treated and the procedures being performed. Additionally, the documentation burden in healthcare is high, with a vast amount of data being generated and stored in various formats, from clinical notes to imaging reports.

These challenges can lead to data inconsistencies, inaccuracies, and gaps, which can negatively impact the performance of an AI system. For instance, if the data on patient admissions and discharges is not accurately captured, the AI system may not be able to provide reliable answers to queries about bed occupancy and resource allocation.

Field-level meaning and documentation

One common misconception is that column names in a database are sufficient documentation for the data they contain. However, this is not the case, as field-level meaning is critical to understanding the context and relevance of the data. For example, a column named "length_of_stay" may seem straightforward, but it may not be clear whether it refers to the total duration of a patient's stay, including pre-admission and post-discharge periods, or only the time spent in the hospital.

Most people get wrong the idea that data is self-explanatory and that column names are enough to understand the meaning of the data. They are wrong, as data is often nuanced and requires careful consideration of the context in which it was generated. A clear understanding of field-level meaning is essential to ensure that the data is accurately interpreted and used by the AI system.

Systems that get left out of scope

In the process of implementing an AI system, some systems may get left out of scope, either due to technical limitations or resource constraints. This can lead to partial answers and incomplete insights, as the AI system may not have access to all the relevant data. For instance, if the logistics system is not integrated with the AI platform, the system may not be able to provide accurate answers to queries about supply chain management.

This can be particularly problematic in healthcare, where decisions are often made based on a comprehensive understanding of the patient's condition, medical history, and treatment options. If the AI system is not able to access all the relevant data, it may not be able to provide reliable and accurate answers to clinical queries. To avoid this, it is essential to carefully consider which systems to include in the scope of the AI implementation and to ensure that all relevant data is integrated and accessible.

Gaps in the data

Stockouts, backfills, and other gaps in the data can also teach a model the wrong lesson. For example, if there is a stockout of a particular medical supply, the data may show a sudden drop in usage, which could be misinterpreted by the AI system as a decrease in demand. Similarly, backfills, which occur when data is added retroactively, can create inconsistencies and inaccuracies in the data. These gaps and inconsistencies can negatively impact the performance of the AI system and lead to incorrect insights and decisions.

To mitigate this, it is essential to carefully monitor the data for gaps and inconsistencies and to implement processes to address these issues. This may involve working with clinicians and other stakeholders to understand the context of the data and to identify potential issues. By doing so, healthcare providers can ensure that their AI system is providing accurate and reliable answers to queries, which can inform better decision-making and improve patient outcomes. For more information on how AsscherAi can support your organisation's data needs, visit our website at www.asscher.ai.

Limitations of AsscherAi

While AsscherAi can provide valuable insights and answers to queries about an organisation's data, it is not a replacement for human judgment and expertise. There are also situations where AsscherAi may not be the best choice, such as when the data is highly sensitive or when the queries require a deep understanding of clinical context. In these cases, it may be more appropriate to use other tools or methods, such as manual analysis or expert consultation. If you are unsure about whether AsscherAi is right for your organisation, or if you have questions about how to implement the system, please contact us to discuss your specific needs and requirements.