First becoming a thing in the 1990s, big data continues to alter everything healthcare. To date, most big data consulting and software investments in healthcare come from the insurers’ side. Public and private payers drive the shift to value-based care by introducing quality measures to their provider networks and testing them with claims and service delivery analytics.
These measures then catch on and create certain benchmarks for performance evaluation among providers when it comes to dealing with the factors that typically hinder healthcare processes: fraud, unregulated costs, poor asset management, and more.
With its potential to transform care delivery, big data stands at the core of the technological evolution. Yet, in its turn it demands ever more advanced systems able to generate, store, process, and analyze clinical information. In this article, we will explore the current use of big data in healthcare across the entire care cycle, touching upon the most recent technological know-hows.
The following are the three major big data mechanisms that make the concept of ‘medical big data’ possible:
This mechanism uses historical data, such as diagnoses, lab results, previous treatment plans, and medication prescriptions, to build models predicting future outcomes. Such models can handle classification problems (whether a patient will develop a certain condition or not) and regression problems (the total cost of treatment for the provider).
This mechanism transforms chaotic electronic health records (EHRs) into meaningful insights. The phenotyping algorithm uses raw patient data, such as demographics, lab tests, clinical records, medication, and treatment plans, and converts this data into medical concepts, or phenotypes. This kind of data is used to support clinical billing and genome studies.
This mechanism is an efficient alternative to randomized clinical trials that require a controlled environment and population segmenting with only one item tested at a time, which is very expensive and time-consuming. Patient similarity algorithms cluster patients who share similar characteristics, and this way intelligently supports randomized clinical trials.
Big data is challenging enough to implement, particularly in such a siloed industry as healthcare. However, it brings many benefits to patients and healthcare facilities.
The following are the major big data use cases that are changing the industry right now, establishing new standards in provider-patient relationships.
We now live longer. It is a good enough reason for healthcare institutions to adapt their treatment models. As many conditions develop over long periods of time, doctors are looking for a proactive approach that will save lives, costs, and time.
Big data supports such a preemptive approach. Big data systems aggregate patient data and offer multiple prevention options based on picked-up warning signs. For example, smartphones and smart wearables, such as Jawbone, Fitbit, and Samsung Gear Fit, help users track basic vitals, including blood pressure, weight, temperature, and blood glucose.
A range of smart devices and apps support both healthy individuals, who want to track their wellbeing, and patients with chronic conditions, who need to balance their health status and prevent complications.
Businesses jump on this opportunity to improve population health. Alphabet’s life sciences company Verily aims at facilitating health management efforts by tracking biometric values of 10,000 volunteers and thus composing the baseline of human health.
The same population health improvement goal drives the Pittsburgh Health Data Alliance activities, which aim at aggregating data from multiple sources like medical and insurance records, genetic data, wearable sensors, and social media. This data is used by clinicians to create cohesive patient profiles and offer customized healthcare packages.
Speaking of heart disease prevention, Apple and Stanford Medicine partnered up for the Apple Heart Study. The study’s goal is to elicit and analyze irregular heart rhythms and help patients avoid more serious cardiac issues, such as heart failure and stroke.
When it comes to diagnosing, people are eager to get all available information about their possible conditions, treatment options, and anticipated outcomes. This wave of self-awareness that emerged with the first wearable sensors is now expanding to genome sequencing.
Patients turn to various genetic test kits, such as 23andMe and Color Genomics, to get on the informed side of their specific genetic susceptibilities. To get the result, they need to spit into a special tube, register this saliva collection tube via its barcode, and mail the specimen back to the lab. Currently, genetic test results can gauge the risks of developing certain hereditary disorders and chronic diseases, including Alzheimer’s and cancer.
However, patients are not the only ones who hunt for valuable health information. Providers collect huge amounts of data, but it is scattered across PCP’s offices, hospitals, and clinics. The issue with connecting this data to that from wearable sensors stays poignant today, and is unlikely to be easily solved in the upcoming years.
The good news is, there are already some successful cases of handling multi-source big data in healthcare. The Johns Hopkins Hospital in Baltimore runs an AI-powered data command center that pulls data from dozens of streams in real time, accessing EHR, emergency dispatch service updates, and lab results, while simultaneously checking the availability of hospital beds.
Applying human-trained algorithms to the acquired data, the center makes immediate decisions on triaging patients and transferring them into a proper facility, also notifying the care team on site. This approach is highly successful with medical emergencies where every minute is critical.
In his interview with Fortune, Jeff Terry, who was in charge of overseeing the Hopkins command center project at GE Healthcare, shared that “there’s been about a 60% increase in the ability to accept complex cancer patients, an over 25% reduction in emergency room boarding, and a 60% reduction in operating room holds with the command center since its launch in February 2016.” Since then, GE launched a number of such NASA-style hospital command centers across the US. As Terry claims, the centers should be yielding a roughly 4 to 1 ROI.
Another example of using big data in diagnostics is AI-powered software capable of recognizing patterns in patient data (for example, IBM Watson). Currently, such systems use machine learning to identify and localize malignancies in test results.
Big data analysis in medical imaging, for example, can relieve radiologists from examining each image individually and save them a good amount of time. Carestream, a medical imaging system, uses algorithms to read and analyze millions of images to spot specific patterns that can be identified as alarming.
“The more data is available, the easier and quicker it becomes to diagnose an ailment,” believes Ali Bagheri, VP of Global Operations at Cybernet Manufacturing. He continues, “Imagine a patient with a rare disease. There isn’t always going to be a specialist in their immediate area. Digital image pathology can be used to share test images with experts around the world for better diagnosis.”
As the technology evolves, it will hopefully bring in a wider range of early diagnosis possibilities for more conditions, allowing physicians to achieve better patient health outcomes.
When a patient’s journey comes to the treatment stage, big data analytics can substantially expand health specialists’ clinical expertise. Thanks to deep learning algorithms, providers can access the bulk of existing medical research on a particular condition or medication.
As a result, providers can personalize treatment plans for patients and lower the risk of side effects or exacerbations, which can be particularly alleviating for cancer and chronic conditions.
Big data also helps to fight medication non-adherence, costing about $100-300 billion a year because of the complications developing when patients fail to follow prescribed treatment.
Express Scripts, a Missouri-based pharmacy benefit management company, has created a custom algorithm that identifies how likely a patient is to follow prescriptions, based on 300 factors. The factors include basic demographic data, income levels, zip codes, and behavioral data.
Express Scripts states that their proprietary algorithm assigns risk scores to patients with a 94% accuracy. The company is then able to reach out to high-risk patients via various channels. Express Scripts reports reducing the non-adherence rate by 37% and thus saving more than $180 million for their clients.
Blue Cross Blue Shield, in their turn, teamed up with Fuzzy Logix, the company specializing in insurance and pharmacy data management. The partners aim at preventing opioid abuse cases. For this, they came up with 742 risk factors and developed an algorithm to sift through patients’ data and identify those at risk of misusing opioids. While this can’t eliminate opioid abuse fully, this can at least help reduce the number of affected patients.
A patient’s rehabilitation after an acute episode doesn’t end when they leave the hospital. That’s why providers have the 30-day readmission rate as one of the crucial benchmarks for clinical performance evaluation.
Big data combined with deep learning can point to patients in need of follow-ups and long-term care. Clinicians can then provide them with a better understanding of their current health status and help them avoid complications and emergencies.
Smartphones and wearable sensors are the prevailing tools used to handle patient data at this care point. For example, GPS-enabled inhalers for asthma patients can record information about a patient’s physical location, sleep patterns and physical activity to recognize if they might be feeling unwell.
In case of abnormal parameters, the inhaler will alert the authorized care team members and the family about possible anxiety episodes.
In another example, apps for diabetic patients can provide daily nutrition and activity suggestions, offer therapeutic education, and record blood glucose levels to keep those involved in the care cycle aware of the patient’s diabetes management progress.
Historically, healthcare has been prone to high levels of fraud and inaccurate claims. The common sources of fraud, waste, and abuse in healthcare are unintentional human mistakes, resulting in incorrect billing, inefficient diagnostics leading to wasteful and duplicate tests, and false claims. Collectively, this affects the whole industry, since only a slight share of losses can be recovered.
Fraud detection remains mostly a manual process, which is rather inefficient because of the massive amounts of data involved. However, with the technological advances such as EHR, big data becomes a tool that can be leveraged to curb healthcare fraud. It helps analyze a large number of claims at no time and identify suspicious patterns in data, signaling potential fraud or abuse.
For example, the most common types of insurance manipulations are billing for:
As perpetrators can’t typically get the whole claim filing cycle right, some key facts get misrepresented. Repeated misrepresentations form patterns, which are then picked up by big data analysis tools, compared to approved claims, flagged as potential fraud or error, and sent for review to experts.
While fraud detection with big data can save huge amounts of money to the healthcare industry, big data has more to offer in terms of cost-saving.
For example, by analyzing data about admission and staff allocation, hospitals can save costs by eliminating the problem of over- or understaffing, as well as by cutting time that patients have to spend at health facilities.
Hospitals also use big data to prevent unnecessary yet costly and time-consuming hospital visits. This benefit of big data is especially efficient when leveraged by a collective of hospitals sharing patient records. This way, hospital staff will be aware of:
With such data at hand, doctors won’t request repetitive tests and make patients visit the same specialists over and over again.
With time, patients are gradually evolving into consumers. Yes, ‘care consumers’ sounds a bit strange right now, but the entire healthcare ecosystem is moving toward consumerism.
One of the indicators is EHR vendors putting their systems into the cloud and embellishing them with CRM-like functionality, allowing caregivers to create comprehensive patient profiles and keep up multichannel communication.
In the patient-centric healthcare, big data brings forth an array of technologies that put control over the care cycle into patients’ hands. Some providers embrace the changes and start putting extra effort into adopting emerging technologies in order to gain strategic competitive advantages. Others stay on the cautious side, waiting until long-term trends prove to be more than fads. Although only time will tell what the winning approach is, you can’t go wrong with using big data technologies for more informed and connected care delivery.