MIRROR, MIRROR: Healthcare data quality and the elusive digital twin

The volume of data used to describe patient conditions and care is ever-growing. At the same time, the analytics employed to mine this data are becoming increasingly sophisticated, informing everything from the prediction of care needs to more precise medicine. Given this wealth of information, it should be easy to paint a robust picture of a patient’s health – to create and leverage a precise “digital twin” in the medical record. After all, if companies like Amazon and Google have real-time insight into everything we browse and buy, surely a physician should be able to quickly access the relevant data in a patient’s medical history. But somehow, this isn’t the case. Not yet. That’s because data, while abundant, is often complex, unstructured, and inconsistent. And analytics and insights based on this flawed data is, by nature, unreliable. As a result, these “twins” we seek become a distorted reflection of the patient’s past and current state – a fragmented and flawed compilation of codes, test results, clinical notes, and other data. So, what’s standing in the way of every patient having a high-quality digital twin? A lot.?

SILOED DATA Individual patient data can be documented and stored1 in multiple places throughout the electronic health record (EHR), such as the problem list, clinical notes, imaging reports, and medication lists. On a grander scale, that unique data – as a whole or in pieces – can also live in a variety of EHR instances, in data warehouses, clinical data registries, and with other organizations that use or exchange clinical data. However, in order to depict an accurate picture of the patient, this data needs to flow readily from place to place. It must be accessible when and where it is required but silos often keep data compartmentalized and disjointed.?

For example, medication data is typically stored in its own tab or file within the EHR, separate from other relevant patient information. Without the ability to integrate this data into the larger patient picture – or even connect it to the problem list – the story told by the digital twin is incomplete. Knowing the prescribed dosage of a medication or that a medication has even been prescribed is essential to safe and timely patient care. But all too often, data silos stand in the way.

VARIATIONS IN TERMINOLOGY When documenting patient conditions in the EHR, clinicians use the words they’re most comfortable with, whether it’s short forms, acronyms, slang, or more formal phrases. For example, one physician may say essential hypertension, another primary hypertension, while yet another may simply write HTN. This practice introduces a great deal of variability into the patient record if these synonyms aren’t harmonized to a single, standard term – which is then precisely mapped to standard codes for billing and reporting activities. In the absence of a universal term, diagnosis data can be misinterpreted, omitted, or lost in translation as it moves from one provider to the next, or through data lakes and health information exchanges. While this may result in lost reimbursement for a provider, the patient’s situation is more dire. An incomplete digital twin is a flawed representation of a patient’s health and using that data can contribute to sub-optimal care – or the absence of care altogether

LACK OF SPECIFICITY In order to create an accurate digital twin, the data captured in the patient record must contain the appropriate amount of detail. However, patient data often lacks this level of specificity, or it becomes inconsistent as it’s transmitted from one system or location to another. This data quality problem has implications at both the individual and population levels. In the case of lab data, tests like lipid panels or complete blood counts (CBC) tell an important part of the patient story. But if, for example, LOINC? codes are used in some settings and local codes are used in others, data shared between providers may not convey the appropriate specificity – or be understood at all. In addition, organizations such as health information exchanges and clinical data repositories often rely on aggregated lab data to complete work for population health initiatives, medical research, and public health reporting. If what’s gathered at the individual “twin” level lacks consistency and detail, those errors and omissions impact downstream efforts as well.?

INCONSISTENT DATA STRUCTURE If data is to be effectively exchanged and shared, it has to mean the same thing to each party involved. To reach this mutual understanding, there must be an agreed-upon structure to patient data, similar to the need for a common terminology. In some cases, standard coding systems like SNOMED CT? provide this structure.2 But as medicine, technology, and our understanding of the factors that impact health evolve, new types of data are being developed – sometimes without consensus or universal adoption across the healthcare community. Social determinants of health (SDOH), such as access to transportation, healthy foods, and education, provide a useful illustration. While SDOH are widely recognized as factors that influence health and well-being, there is a lack of alignment on the screening tools and requirements for gathering and consistently recording this information. Without a consensus on how to structure this data to ensure it is accessible, meaningful, and interoperable – there will be significant gaps in the digital twin, rendering the patient story incomplete.?

VARIED DATA ACCESSIBILITY Patient data is recorded and stored in a variety of formats, each of which may be appropriate for the condition, test, or specialty involved. For instance, diagnosis data is frequently documented with ICD-10-CM codes, while surgeries use CPT?. This information is readily incorporated into the patient record, but other types of data aren’t as easily ingested or integrated into the EHR, which can cause gaps in the patient narrative. Here, genomics provides a helpful example. Healthcare is still in the early days of gathering and using genomics data, and while it can be valuable to help guide more precise and proactive care, this information is often contained in pdfs – a file format that is not readily searchable in the EHR.3 This obstacle to accessing highly personalized genomics data means that unstructured information is essentially “trapped” in pdfs, leaving insights untapped and unable to inform a robust and detailed digital twin.?

VOLUME OVER VALUE Challenges with file formats, accessibility, and structure aside, the sheer volume of data being created has become an obstacle in itself. In 2018, approximately 30% of the world’s data volume was generated by the healthcare industry, and by 2025 the compound annual growth rate for this data will reach 36%. 4 The result is an overabundance of information and too few mechanisms to make sense of what’s there. This poses a significant problem for clinicians who are charged with deciphering what is meaningful in this sea of data in order to provide optimal care. But for the vast majority of patients who have no medical training, understanding what is relevant and reflective of their health status can be nearly impossible. The issue has come into even sharper focus in recent years with the growing use of wearables. Data derived from smart devices like watches, gloves, and vests, can provide a steady stream of data, contributing – in real-time – to a patient’s multi-faceted twin. But even if it can be ingested into the EHR, which is yet another challenge, 5 this information requires compilation and interpretation6 before it can be useful for patients and providers. Without this important step, wearables data simply adds to the digital noise.??

The above exploration of obstacles to creating a complete and “healthy” digital twin is far from exhaustive. And the types of data used to illustrate these challenges are just the tip of the iceberg. However, understanding and addressing the healthcare data quality issues we face today will be critical if we are to tackle the problems that undoubtedly loom ahead. This is true not only in the aggregate, but for each person – each patient – who deserves the best possible care based on the most complete and accurate digital health record.?

To learn how IMO solutions can help your organization create and leverage clinical data that is complete, consistent, and accurate, visit imohealth.com/imo-portfolio

Visit this link for more details https://assets.asccommunications.com/whitepapers/imo-wp-january-2023.pdf?utm_campaign=IMO_WP_January_2023&utm_medium=email&_hsmi=242313818&_hsenc=p2ANqtz-8yOy5QPEsJPUOx-uQN3PYfMuwZIZA_OaE5x42iNN4bLgHmq1CkOHT9xcXw3oDuly1hhZ8kjs8V4Jd7NbNjsKdz3zY3Hw&utm_content=242313818&utm_source=hs_automation

#digital #data #health #healthcare #quality #google #like #medical #analytics #amazon #medicine