Visualizing, Interpreting, and Disposing of Healthcare Analytics Data

VISUALIZATION

The University of North Carolina (UNC) at Chapel Hill describes data visualization as a method for “exploring complex patterns or large quantities of data that cannot be easily perceived by looking at a table of numbers or reading paragraphs of text. The goal of data visualization is to communicate information more clearly, and it does so by employing our innate ability to recognize visual patterns in our environment.”

Data visualization can be exploratory or explanatory, according to Johns Hopkins University. Exploratory data visualization occurs when stakeholders are looking for trends or patterns within a dataset that may be worth exploring further, while explanatory, or communicative, visualization is used to present the findings of an analysis to an audience.

Data visualization in healthcare is growing more important as healthcare continues to generate more data daily. Many healthcare organizations are turning to advanced technologies like artificial intelligence (AI) and machine learning (ML) to analyze these massive amounts of information and identify important trends or patterns.

However, experts at Northern Kentucky University indicate that for these tools to be useful, the analytics outputs they generate must be presented as data visualizations to communicate insights to a broader audience effectively.

The American Health Information Management Association (AHIMA) calls data visualization an “essential skill” for health information management professionals, noting that visualizations can help most people better understand and recognize key data trends more effectively than they would when presented with raw data.

Data visualization uses a variety of techniques: infographics, tables, line and area charts, bar graphs, pie charts, scatter plots, histograms, heat maps, big data dashboards, and more. For healthcare, though, major players in the industry often create data visualization tools that researchers and policymakers can use to render healthcare trends.

For example, the Agency for Healthcare Research and Quality (AHRQ) has leveraged its wealth of data to build data visualizations for health insurance coverage, emergency department visits, COVID-19 hospitalizations, and other trends across the United States. To meet users ' needs, these can be customized according to income, geographic location, age, and other factors.

The Institute for Health Metrics and Evaluation (IHME) at the University of Washington also offers a host of data tools and interactive data visualizations to help capture global health trends such as pandemic recovery, disease burden, antimicrobial resistance, and health financing.

Research published last year in the Journal of Medical Internet Research demonstrated that interactive visualization may be particularly valuable in healthcare.

The study’s authors conducted a scoping review of interactive visualization applications in population health and health services research from January 2005 to March 2019, finding that these tools have been deployed for epidemiologic surveillance for infectious disease, health service monitoring and quality, resource planning, and studying medication use patterns.

Despite their potential, healthcare data visualization tools face several challenges, such as bias and trustworthiness.

To help address these issues, organizations like AHIMA have developed data visualization best practices while researchers continue to study how these tools can be further refined.

Experts writing in Diagnostics earlier this year described a framework to help improve the trustworthiness of interactive data visualization tools in healthcare using a medical fuzzy expert system.

Hennepin Healthcare, a health system in Minneapolis, utilized data visualization tools to help refine and enhance its health equity initiative for COVID-19 vaccine distribution, which helped flag at-risk patients and better target populations for outreach.

Once data has been visualized, it can be interpreted.

INTERPRETATION

Data interpretation involves making sense of the information analyzed and presented in the previous phases of an analytics project.

The interpretation step is crucial, as it helps reveal the meaning or insights generated within the context of the analysis, allowing stakeholders to use the information for decision-making. For this reason, data visualization and interpretation are often presented together, or the terms are used interchangeably.

However, because the interpretation of the data can have such a significant impact on decision-making or be used to guide future analytics projects, there are multiple considerations stakeholders must be aware of at this point in the data lifecycle.

The University of Edinburgh posits that “no data analysis occurs with pure objectivity,” highlighting ethical challenges at every step of the analytics and interpretation processes.

These challenges pop up at each “choice point” in a project, from deciding on the analytics method to interpreting the findings. Conscious and unconscious biases can slip in at any of these points and can impact what is “seen” in the data and how that is communicated.

Awareness of potential ethical dilemmas is the first step to tackling these problems. Considerations like when and how data should be used in a particular context or how to make a culturally sensitive interpretation of the data can help flag potential ethical pitfalls, but the University of Edinburgh further indicates that overarching considerations based on place, people, principles, and precedent should guide stakeholders.

Place-based considerations involve thinking about the context in which the data may have an impact, as data considered harmless in one community or country may be highly sensitive in another. Stakeholders should also consider the potential immediate, short-term, and long-term impact of a project’s findings.

People-based considerations take into account who may benefit or be harmed by the data. Prioritizing careful, culturally respectful, and accurate data interpretation can help mitigate harm.

Further, data analytics and interpretation should be guided by principles of honesty and respect. Doing so can help balance reliable, unbiased data analysis with presenting the data so that the insights can be understood in a way that benefits as many people as possible.

Precedent-based considerations help establish the best analytics strategy for a particular type of data and identify obstacles to analysis or interpretation. Looking to peer-reviewed studies that have analyzed similar data can help in this situation, but stakeholders should also be aware that their analysis and interpretation may also become part of the precedent available in the future.

The Centers for Disease Control and Prevention (CDC) also outlined some tips and best practices for stakeholders as they pursue data analysis, synthesis, and interpretation.

After the findings derived from the data have been successfully visualized and interpreted, those data can be disposed of appropriately.

DISPOSAL

Data disposal is the final, key phase of a healthcare analytics project.

Depending on the types of data used, stakeholders on a particular project may have to comply with state and federal guidelines for the retention and disposal of those data.

The Health Insurance Portability and Accountability Act of 1996 (HIPAA) mandated national standards to protect patient health information from disclosure, and the HIPAA Privacy Rule established how protected health information (PHI) may be used and disclosed by covered entities subject to the rule.

Healthcare providers and payers are covered entities who must comply with the HIPAA Privacy Rule but are also the organizations likely to undertake health-related analytics initiatives.

To balance the need for patient privacy with the potential for medical breakthroughs driven by the analysis of patient data, covered entities should create appropriate data retention and destruction schedules.

AHIMA provides guidance for how stakeholders across all healthcare settings can establish these schedules, but it is important to note that any plans must comply with HIPAA’s retention requirements.

Data retention policies are designed to “specify how long data must be retained to meet regulatory and/or organizational needs, and what should be done to the data after retention requirements have been met,” according to one AHIMA brief on healthcare data governance. But “organizations may choose to delete/destroy or archive data once retention requirements have been met.”

These data must be destroyed properly; however, the protocols differ depending on whether the information is in a physical or electronic format.

To properly destroy physical data, such as paper medical records, covered entities must implement safeguards to avoid incidental or prohibited uses and disclosures of PHI. A covered entity's employees must also be trained on any PHI disposal policy.

The safeguards a covered entity must implement are flexible, but it is recommended that each organization assess its individual circumstances, such as unique risks to patient privacy or the type of PHI being destroyed, when choosing how to best dispose of these data.

With that in mind, unsecured dumpsters are not recommended for PHI disposal, but locked dumpsters only accessible to authorized individuals are an option. Similarly, covered entities cannot dispose of physical PHI in trash cans or recycling bins accessible to the public.

Another option involves hiring a vendor to shred, burn, pulp, or pulverize physical PHI and dispose of it in a landfill, as long as the covered entity and the vendor have signed a business associate agreement (BAA) to maintain HIPAA compliance.

Properly destroying electronic PHI requires attention to many of these same considerations, but there are key differences stakeholders should note.

It is recommended that healthcare organizations follow the National Institute of Standards and Technology (NIST) Special Publication 800-88, Guidelines for Media Sanitization, and recommendations from the US Department of Health and Human Services (HHS).

HHS suggests that PHI stored on electronic media should be exposed to a strong magnetic field to disrupt the recorded magnetic domains, destroyed physically, or overwritten with non-sensitive media using hardware or software.

Once data are effectively destroyed, the data cycle is complete.

From here, stakeholders can repeat the cycle on a project that advances work done in a previous one, or they can move on to a separate project to support their organization’s goals in another area.