Is Big Data Clinically Significant?

A restaurant patron doesn’t leave a tip as he storms out of the trashy diner that Google “recommended” to him. This is an example of big data trying to predict what kind of restaurants that customer might enjoy, and missing the mark entirely. Nevertheless, the consequences of this blunder are mild, maybe resulting in an upset stomach or a dented wallet.  However, if a clinical decision support tool uses big data to supplement a clinician’s diagnoses for a patient, a misinterpretation of that data could have devastating results.

The term big data refers to the analysis of large sets of data that cannot be quantified by traditional methods. Where the data comes from does not matter as much as how it’s interpreted, which is the main hurdle when it comes to big data in healthcare.1 Attempting to make a sensible diagnosis from different drugs, diseases, allergies, genes, etc. is hard enough; when you take into consideration each patient’s unique biological makeup and medical history, it becomes nearly impossible for a computer system to tailor an accurate, personalized response.

Regardless of these barriers, big data is already applicable in our healthcare system. For each methodically controlled clinical drug trial, there are thousands of situational cases that are overlooked. Yet those cases are exactly the kind of circumstances big data thrives in. While most clinical trials focus on controlling test subjects and scenarios to produce the most accurate information possible, it cannot always be translated into clinical effectiveness since the real world can never recreate the perfect environments clinical trials were carried out in.2 However, a clever analysis of all the patients taking a particular drug may result in a plethora of useful information.

For example, let’s say we want to examine all patients taking tramadol, a painkiller commonly referred to as Ultram, and its efficacy over time on patients that have taken different painkillers in the past. Using big data, we can gather sets of data in the millions in a relatively short period of time, instead of the hundreds or thousands that may be examined during a clinical trial. With that data, we can then group the different painkillers taken by patients, how age and weight may contribute to efficacy, and how a patient’s medical history could challenge the expected results. With this, you can tell a computer program to analyze this data and come up with clinically significant results to help similar patients take the most effective painkiller.

Obviously, there are pitfalls to this thought experiment. A major roadblock that is distinct to healthcare IT is the lack of shared information among different EHRs and clinical support tools. This absence of data availability is the bane of big data since big data doesn’t only get its strength from an abundance of information, but also from the uniqueness of said information.3 Another hindrance important to all fields that use big data is data accuracy. While patients provide some information, clinicians rarely get the full picture, and that only hurts the big data movement.

Is big data clinically significant? Yes, but careful consideration needs to be taken when implementing it. The most basic analysis of big data could prove essential to healthcare informatics, but without a system in place to validate those results, we cannot totally trust the output of big data in healthcare IT.


  1. Raghupathi W, Raghupathi V. Big data analytics in healthcare: promise and potential. Health Inf Sci Sys. 2014; 2:3. doi: 10.1186/2047-2501-2-3
  2. Zhang Z. Big data and clinical research: perspective from a clinician. J Thorac Dis. 2014; 6:1659-64. doi: 10.3978/j.issn.2072-1439.2014.12.12
  3. Kaufman H. Big Data Analytics in Healthcare – How Laboratories Can Play a Leading Role. Quest Diagnostics. Published July 6, 2106. Accessed September 6, 2016.