Seeing Mike Trout sign a 12-year $426.5M contract with the Los Angeles Angels reminded me of the 1996 movie Jerry Maquire. Jerry, a sports agent played by Tom Cruise, represents a fictional Arizona Cardinals wide receiver Rod Tidwell, played by Cuba Gooding, Jr. In one poignant scene, Rod chastises Jerry in an effort to motivate him to secure a higher-value contract for Rod’s services to the Cardinals. Through the repetitive mantra “show me the money, Jerry,” Rod makes it clear to his agent that only the money value of the contract matters.
So what does a movie about a sports agent have to do with creating an analytics culture among your clinical teams? Although your teams will not be screaming expletives at you to “show them the money,” clinicians will be just as animated asking you to show them the data.
Data credibility stands as the most important factor in gaining a physician’s trust in analytics. One single error in your data brings its validity into question, and casts a dark shadow over the remaining data and any analytics, patterns, or conclusions drawn from it.
Physicians are scientists who pride themselves on their ability to think through problems critically. Having been trained in statistics and research methodology, physicians quickly search for flaws in any data sources. These flaws then lead to skepticism of the results and resistance to any profiles or suggested actions to be taken in response to the data.
When presenting analytics to physicians a few simple rules need to be followed:
- Carefully curate data sources and make sure they are clean, accurate, and appropriately represent the metrics presented in the delivered analytics. Most healthcare data sources are dirty, and require removal of duplicates, significant cleansing, and ontology normalization. Organizations must put in the time to get this absolutely correct.
- Ensure that basic statistical principles, such as sample size, are considered when presenting results. Failure to deliver data that is statistically valid is probably the biggest mistake organizations make when presenting data to physicians. Reaching conclusions on small sample size is a big no-no, yet I see it all the time.
For example, a physician with a HbA1c quality metric rate of 80% tested is not necessarily 30 percentage points better than one with a 50% rate. If the physician with an 80% score had a sample size of five, then four of five patients received HbA1c testing. For the physician with a score of 50%, two of four patients received HbA1c testing. Do we really think there is a meaningful difference between these two physicians? Is it not possible that the physician at 50% could order one more HbA1c on a fifth patient and therefore raise his quality metric to 80% (4 of 5)?
- Deliver timely, actionable results so that physicians can take steps to improve care delivery. Providing analytics on activities too far in the past offers limited understanding of current activities, and therefore, little useful information on which to act upon. In addition, results that are related to activities out of the physician’s control (e.g. wait times in the ED due to staffing levels) may be nice to know but provide no actionable information.
To effectively engage physicians with data, organizations must develop an analytics culture that respects the methods physicians rely upon to treat their patients. This includes data accuracy, respect for critical thinking, and adherence to honest statistical methodology. Doing all of this is hard work, but the potential upside—improved quality, safety, and efficiencies in care—makes it a worthy goal.
To learn more about how your healthcare organization can use a modern visual analytics platform to achieve new levels of performance and patient satisfaction, visit our healthcare analytics page.