Laboratories generate massive amounts of data. However, data is meaningless without context, awareness of operational constraints, or simply accounting for human variability and errors. As laboratory managers, if we seek data-based decision making, we must understand how to decipher data, apply findings, and communicate action steps in a meaningful way. Understanding the four types of data analytics, how to contextualize data, and derive actionable insight statements are the key ingredients to making data-driven decisions.
The four types of data analytics
Data analytics is the process of leveraging raw data to answer a question, find insights into operational workflows, or observe trends.2, 3 The four major types of data analytics are descriptive, diagnostic, predictive, and prescriptive.4 As laboratory data sophistication will be unique to your lab, this article will focus on descriptive, diagnostic, and predictive aspects of raw data. Aspects of these raw data forms can often be managed manually and are accessible strategies for most managers.
There is no one-size-fits-all approach when drilling down into an observation and filtering the critical pieces of data needed to better understand what is occurring. You may choose to frame a structured inquiry around one or all types of data. It is advantageous, however, to proactively identify the questions you are trying to answer by the data instead of making generalized, wide data requests. Also, you should draft your intent and next steps. This safeguards that the data will not be underutilized once accessed. Specificity might seem tedious, but it will ensure you get raw data that has meaningful potential.
|Type of Data Analytics||Sample Data Inquiry Statement|
|Descriptive||Did first shift see an increase of Heparin contamination?|
|Diagnostic||Did Heparin contamination occur on one unit more than another?|
|Predictive||Does an increase in Heparin contamination spike at specific points in time (such as annual new nurse hiring)?|
Cleaning and contextualizing data
Raw data could be referred to as source data. It’s a direct, unprocessed export. This means it will need to be cleaned before it can resemble something actionable. Many organizations utilize data visualizations, such as dashboards, with descriptive charts or graphs. It’s important to remember that the sleek colors and flashy graphics are populated by raw data. Cleaning raw data and placing it in the right context are crucial steps to ensure that all visualizations are unaffected by junk or misplaced data entries. Clean data should not include inaccuracy, manual entry errors, be outdated, or contain incomplete documentation.5 Unfortunately, in the laboratory, we are often faced with process variability that leads to junk data creation.
While data can be scrubbed until it shines, context is what places data in the right position to generate actionable insights.
Those guidelines also hold true for setting parameters. Consider turn-around time monitors. While a simple example, depending on which point in the process you wish to include, the metric may not be represented if the original parameters excluded your data interest. For example, if you consider turn-around as from the time of sample collection, but the dashboard is measuring from the time of receipt in the lab, this would be an inaccurate representation.
It should never be assumed that a data analyst, who may or may not have a laboratory background, understands the context of your data, or classifies your data outputs in the same manner you would. Analytics are programmed to work in the way they are intended; junk or poorly contextualized data can mislead leaders into poorly informed outcomes.5 Thus, access to raw data is important because it allows leaders to trace back decisions and ascertain whether original data cleaning and context were approached accurately.
While data can be scrubbed until it shines, context is what places data in the right position to generate actionable insights. As leaders, it’s important to consider any environmental factors, operational workflows, or process deviations that may be impacting your data. Further, recognizing where convenience has replaced or impaired standardization is critical. For example, staff may choose to type comments instead of utilizing standard, pre-built ones. In some cases, it may be necessary to fix an operational process or procedural guidance prior to utilizing any data outputs. Your first data set may highlight staff selection variation requiring a standardizing clarification prior to any data-driven changes being made.
Developing actionable insight statements
With clean, contextualized data in hand, you’re ready to develop an actionable insight statement and share your findings with staff and stakeholders. An insight statement summarizes what you discovered during data analysis and how your discovery provides understanding into why something is happening.4 Additionally, to be actionable, interventions and initiatives that are generated by the data outcomes should be included. You can even incorporate your expected measurable outcomes from the data-driven approach and how you plan on monitoring success. Preparing actionable insight statements is a great exercise to get ready for the change management discussions with your staff and institution.
Leveraging actionable insight statements as an elevator-pitch for change is only the first step. Long-lasting transformation and sustainable data-driven culture depend on focused change management and monitoring efforts.6 As a leader, you must support, inform, and help individuals move toward proposed data-driven changes. For any new process, training and communication with internal and external personnel will be essential. You may need to discuss a go-live date for standardized efforts or drop-dead date for old procedures. You may also need to secure leadership alliances prior to making changes or create business readiness documentation. Likewise, you should develop improvement plans according to new data reviews at scheduled intervals.
While prioritizing staff observations can prove difficult in the quagmire of leadership tasks, having an open door affords leaders a valuable perspective on the laboratory’s performance. If you decide to move toward data-driven decision making, it is important to make clean data the expectation, seek standardization where you can find it, and create accessible visualizations once you know your data is clean and standards are in place. Also, proactively mapping out actionable insight statements, change management strategies, and monitoring steps will enable you to create sustainable change in your laboratory and institution.
- Mayhew, R. “Purpose of Workplace Observation”. CHRON. July 2020.
- Andreasiodmok. “Human Centered policy? Blending ‘big data’ and ‘thick data’ in national policy”. Gov.UK. January 17, 2020.
- Bay Atlantic University. “Characteristics of Big Data: Types, & Examples”. June 1, 2020.
- Cote, Catherine. “4 Types of Data Analytics to Improve Decision Making”. Harvard Business School. October 2021.
- Tableau. “Dirty Data is costing you: How to solve common data preparation issues”. Accessed January 2023.
- Miller, K. “5 Critical Steps in the Change Management Process”. Harvard Business School Online. March 19, 2020.