This course introduces health data, from the perspective of data analysts.
Health data connects complex health care systems. An understanding of health data is fundamental to health analytics.
Through this course, you will
gain a highly valuable skill in the healthcare sector
understand how health data records information about each patient and medical encounter
learn features of health data that enable you to perform more insightful analyses
be able to communicate more effectively with clinical and analytic colleagues
be empowered to improve care processes and make a difference to many people’s health and lives
The 4 sections we will cover
Where health data come from: 5 main sources including health insurance claims and EHR
What health data look like: Structured vs Unstructured data
Features of health data: Hierarchical structures, Disease etiology, chronology, supply vs demand
Issues of health data: Gaps, Errors, and how to practically deal with these
Where Health Data Come From
Welcome to Health Data 101. Thank you for choosing my course!
Health insurance claims data are a major source of health care data. As long as the the medical services are paid by health insurers, you will see the claims. This is a relative well structured and complete source, although clinical detail can be lacking at times.
Electronic health records are another major source of health data. EHR systems enable medical professionals to record information about patient visits. Clinical richness is the major advantage of EHRs, but accuracy can vary.
Research reports contain great health data, usually of a more scientific nature. When designed well, these can offer great insights. But be aware of sample size, biases of and any financial influence on the study.
Public health organizations contribute enormously to health data, especially for prevalence and incidence rates. Learn where to find these and know the limitations of each.
Wearables can track numerous measures including heart rate, blood pressure. They hold great promise for health care, but some challenges remain.
What Health Data Look Like
The most pervasive type of data is structured data. These general have (row/column) tabular form with clear linkages across different tables. We illustrate examples of diagnoses, procedures, drugs and laboratory tests here.
Unstructured data such as clinical notes and medical images, hold great clinical content. These can be analyzed using different set of tools than structured data.
Features of Health Data
Health data often have hierarchical structures. These structures represent layers of intelligence, that enable faster learning, analyses, and reduces impact of small sample size issues.
ICD10 diagnosis codes have very clear hierarchical structure. ICD9 also do, but to a lesser extent.
Drug codes, NDCs, are bewilderingly numerous. WHO created a ATC classification that has a clear hierarchical structure, and enables easier, faster and more complete analyses.
Procedure codes, reflecting the complexities of medical interventions, exist in a variety of forms, including HCPCS/CPT, ICD10PCS, ICD9CM. These also exhibit a hierarchical structure.
Laboratory test results are codified using LOINC codes. The analysis of unstructured data also relies on creating hierarchy and structure from the unstructured forms.
Disease Etiology, Chrology concepts enables you to do more insightful and more impactful analyses. Demand drives supply in healthcare.
Issues of Health Data
Gaps in data exist, where certain elements of data is completely missing from your datasets. e.g. no drug fill data in EHRs or no data across different medical institutions.
Examples of errors in health data and how to correct these. Pragmatism is key as perfect data does not exist.
A few practical considerations of health data, e.g. accuracy of source, extent of automation.
Congrats on finishing the course. Feel free to let me know your questions.