No room for failure: Delivering a successful data project
Subscriber Benefit
As a subscriber you can listen to articles at work, in the car, or while you work out. Subscribe Now
The amount of health care data generated grows exponentially each day. Approximately 30% of the world’s data volume is generated by health care and that percentage is expected to grow to 36% in the next two years. There isn’t room for failure when it comes to handling health care data, so it’s essential to understand what to do with the volume of information, how to safeguard and process the intelligence and ultimately how to extrapolate real-time insights from complex datasets to uncover risk signals, which can help improve patient care.
The first thing to understand is that having the right data is important, but it’s always going to have limitations. A staggering statistic from Gartner indicates that through 2022, only 20% of analytic insights will deliver business outcomes. Therefore, four key factors health systems and laboratories need to consider as they analyze their testing data are how it is integrated, cleaned, standardized and evaluated.
Integrating fragmented data
Health care data often comes from systems that are designed to collect the information necessary for reimbursement, not operational or patient care improvement, so there are likely to be gaps in the information that can be gleaned from it. For example, medical data may be stored in specialized formats and systems such as the LIS, billing, and HIS. Putting everything together into a single, HIPAA-compliant platform provides seamless visibility and can present a more complete picture of lab operations. Patient matching also plays an important role when combining datasets from different sources, such as laboratory testing data and medical claims, as one needs to be able to match patients between disparate systems.
Cleaning and standardizing data
Inputting missing values, standardizing formats, creating new categories and eliminating duplicates in patient records are all important steps when building models from a more manageable, performant dataset. Standardizing the format of zip code and state data points, for example, is important in public and population health use cases that involve patient geography, such as evaluating the risk of Monkeypox spread or tracking if a certain zip code correlates to higher incidence of diabetes.
Additionally this process can be utilized to standardize test names and abbreviations in order to compare test codes, analytes, and units of measure to determine whether tests with similar names are in fact referring to the same test. This standardization can be a valuable resource for determining if a reference lab is meeting its performance benchmarks or for monitoring trends over time.
Evaluating data for bias
Systems that leverage Machine Learning (ML) and Artificial Intelligence (AI) models are powerful tools, but unfortunately big data sets can be biased. For example, laboratory testing data is often skewed by gender, with women overrepresented when compared to the U.S. Census population average. Knowing that fact allows data scientists to adjust data models to reflect patient gender ratios that align with U.S. population numbers.
Devising ways to ingest health care data from multiple sources is an important first step, but if health tech vendors don’t have a solid process in place using predictive and prescriptive analytics to organize the data, how are they going to help providers make sound decisions and conclusions?
Realizing the value that can be gleaned from rapidly collecting, normalizing, enriching and leveraging health care data to connect with hidden patient stories and implement adjustments, will ultimately drive impact for healthcare systems and help solve real-world problems.
There is no room for failure or wasting such a valuable resource to unlock actionable, real-time insights and increase efficiency when patients’ lives may be on the line.