Big Data Can Reshape Lab Reference Values in the Era of Personalized Medicine

As might be expected, there are substantial differences in “normal” laboratory values determined by healthy outpatients and critically ill patients in the intensive care unit (ICU). However, even among ICU patients, there are significant differences in the distribution of laboratory values for patients with the best and worst outcomes, according to a study published Nov. 9 in JAMA Network Open. The authors say that big data enables stratification of laboratory reference ranges by populations, clinical context, and even outcomes, which can lead to a fundamental shift in interpretation of laboratory results.

“The concept of one reference interval for all might need to be redefined,” write the authors led by Patrick Tyler, M.D., from Beth Israel Deaconess Medical Center (Boston, Mass.). “While our simple model represents only a static snapshot of a dynamic situation, this effort represents the first attempt to generate a personalized, data-driven strategy for laboratory value interpretation based on clinical context and outcomes.”

Tyler and colleagues used data from the Medical Information Mart for Intensive Care database (Jan. 1, 2001, through Oct. 31, 2012), which collects patient data for all ICUs at a large tertiary medical center in Boston. The researchers identified 38,605 ICU patients over the study period (56.6 percent male; mean age, 74.5 years) and compared the hospital’s laboratory reference interval with one generated from data from patients in the ICU. Further, associations between laboratory values of ICU patients and outcomes (including mortality and length of stay) were evaluated.

Analysis focused on clinically relevant laboratory values, including minimum for albumin, ionized calcium, hemoglobin, and platelets; maximum for lactate; and both minimum and maximum for bicarbonate, blood urea nitrogen, creatinine, calcium, magnesium, phosphate, potassium, sodium, glucose, and white blood cell count. Albumin was ordered for 41 percent of patients, free (ionized) calcium for 50 percent, and serum lactate for 60 percent, while all other laboratory tests analyzed were ordered for more than 80 percent of patients.

The researchers found that hospital laboratory reference intervals were significantly different for ICU patients for all laboratory tests analyzed. The probability distributions of laboratory test values between the best and worst outcomes in ICU patients were all also significantly different from each other, with most laboratory tests having less than 0.8 overlap with the reference interval and about half of them having less than 0.5 overlap.

“By using data from ICU populations with high-resolution features and various outcomes, including both long term (e.g., survival) and intermediary (e.g., occurrence of arrhythmia), we can more accurately reference a patient against an outcome of interest, rather than with respect to a healthy population,” write the authors. “This enhancement will, with further research … help us better understand how different laboratory values should be interpreted. Using a fixed reference interval for critically ill patients may not be the most effective strategy; rather, probabilistic interpretation of laboratory values may be more valuable in guiding treatment decisions and prognostication.”

The authors say they realize that modifications of ranges would add complexity to the process of establishing and calibrating normal ranges, but in an era of big data and precision medicine, it “makes sense” to use available data “to increase the precision of what we do clinically.”

Takeaway: Big data may reshape the way laboratory reference values are calculated and interpreted by incorporating clinical context and outcomes to the definition of normal.