Read about data visualization, querying, sharing, and stewardship. Find out how to solve these actual data analytics challenges.
Challenges such as data collection, data cleaning, storage, and security were the «main characters» in the article «Healthcare Data Analytics: Challenges and Solutions. Part 1». Other data analytics challenges need to be solved.
Visualization
When a doctor provides medical care to a patient, It’s important for doctors to correctly interpret the data from screenings, tests, and investigations. Overlapping or cramped text, convoluted flow charts, and low-quality graphics can cause the provider to misinterpret or ignore important data about the patient’s condition. It can lead to medical errors. The World Health Organization (WHO) states that harm caused to patients during health care is the fourteenth leading cause of illness and death worldwide.
How to Address It?
When data is presented in easy-to-understand visuals, healthcare professionals can find trends, correlations, and patterns overlooked in the raw data. Typically, data visualization occurs in forms such as charts, graphs, pictures, and diagrams. Python tools are a good choice if a healthcare project involves large data loads and data visualization. A Belitsoft expert Dmitry Baraishuk explains that Python is popular because many libraries are available for it.
Developers can use Python libraries for almost any data visualization needs. For example, the Matplotlib library and its wrapper Seaborn are useful for plotting graphs. Seaborn allows access to several Matplotlib methods with less code. Another feature of Seaborn is its modern and aesthetically pleasing color palettes and styles for charts. Developers can customize Seaborn settings only if they know Matplotlib. Another library option is Plotly. Developers can create interactive and contour graphs, 3D diagrams, and dendrograms with its help. Healthcare technology organizations don’t need to specifically hire Python developers in-house to «pump up» data visualization. They can employ top-notch IT partners to outsource Python development services. Thus, the organization can reduce internal costs.
Stewardship
The amount of data in healthcare grows exponentially. Each US state has standards for how long a healthcare provider must retain records. For example, in California, adult patient records are retained for a minimum of seven years. Healthcare organizations must retain any records for six years related to the Health Insurance Portability and Accountability Act (HIPAA). In litigation, a healthcare provider must retain medical records for at least ten years.
Besides legal purposes, data retention may also have research purposes. A healthcare organization may also reuse data to conduct a comparative analysis of the company’s performance. In this context, ongoing data management and curation becomes an important task.
How to Address It?
A successful data governance plan requires the development of accurate, complete, and updated metadata. It’s like the labels on boxes that tell a customer what is in the box. In healthcare, metadata helps find relevant data about studies and patients. It allows analysts and researchers to understand who created the data, when, for what purpose, and how it has been used. Metadata helps analysts replicate previous queries accurately. It’s crucial for benchmarking and scientific studies. Metadata prevents the creation of «data dumpsters» where info is unstructured and poorly organized. A data steward can develop and curate meaningful metadata for a healthcare organization. They ensure all metadata elements have standard formats.
Querying
Every healthcare organization stores and processes great amounts of medical data. Business intelligence and data analytics tools help providers track morbidity, patterns, and trends in specific population groups. In addition, data analysis allows the provider to organize targeted interventions, improve healthcare delivery systems, and reduce health disparities. Web-Based Data Query Systems (WDQS) is a common method to distribute health data to population groups. But there is a problem with compatibility.
Each department of the healthcare organization can form its database with local storage. The data from one department may not be compatible in format with the data of another. For example, parts of a patient’s data set are stored in different formats or reside in multiple walled-off systems. It’s difficult for a provider to create a complete portrait of an individual patient’s health because of it.
How to Address It?
The format incompatibility can be solved by medical coding systems: LOINC, SNOMED-CT, and ICD. Structured Query Language (SQL) can help with relational databases and large data sets. Programmers use it readily to set up and run analytical queries and to create data integration scripts. An important nuance: if a developer uses SQL, they must be sure that the data they have is complete, accurate, and standardized.
Sharing
Healthcare organizations need to share their population health data. It allows them to provide value-based care to their populations. But if health data sets can be heterogeneous even within an organization, what about data across organizations? Doctors need info to follow up with patients and develop strategies to improve overall outcomes.
How to Address It?
The healthcare industry uses strategies and tools such as application programming interfaces (APIs) and Fast Healthcare Interoperability Resource (FHIR). They facilitate integration across diverse healthcare platforms and the uniform exchange of data.
API connectivity is a requirement for a provider to participate in the Medicare/Medicaid program from the Centers for Medicare & Medicaid Services (CMS). For example, a patient is discharged from the hospital. The API retrieves the Health Level 7 (HL7) message from the electronic health record (EHR) and sends it to post-acute care providers at their direct addresses. FHIR standardizes the way healthcare organizations exchange patient info, even if it’s exchanged or stored locally in another way.
To End Off
Different data is updated differently. Patient vital signs may update every minute or less. Data such as marital status and home address may change several times in a person’s lifetime. Healthcare organizations should regularly monitor their data assets to ensure that data doesn’t become outdated. Vendors should understand which data sets need to be updated manually and which can be automated. It’s also important to perform updates in a way that doesn’t compromise the integrity and quality of the data. Healthcare organizations should ensure that updates don’t create unnecessary duplicate records. If data is duplicated, its reliability is questionable.
