As of now, healthcare costs seem to be even more effective than apples in keeping doctors away. In a recent report by CQC, 85% of surveyed patients admitted their concerns about both medical costs and care delivery quality. These numbers are even more frustrating concerning that more patients than ever had medical insurance in the U.S. in 2017.
Unfortunately, most pricing strategy factors are beyond the providers’ control. But healthcare organizations prioritize patient care needs and seek for the ways to overcome the insurance restrictions with the help of technology, precisely, big data analytics.
Big data paves the way to more accurate and streamlined diagnosis, personalized and evidence-based treatment options as well as assisted recovery outside the facility. Various techniques of processing and analyzing this data can help healthcare organizations trim costs in a way that will benefit patients’ wellbeing, for example, eliminating unneeded procedures and balancing therapy.
Whether providers are turning to third-party specialists for big data consulting and healthcare software development or referring to an in‑house team, they are achieving impressive results. Here is some real‑world evidence of big data analytics use in healthcare.
Standing at the core of care delivery and becoming the must for providers to adopt since HITECH introduction in 2009, EHR systems act as aggregators of big data. They store extensive information about each patient in the facility, including medical history, diagnoses, allergies, major surgeries, lab results, and demographics.
Predictive analytics algorithms can process these varied data points to define patients in need of special care, forecast readmissions, identify the possibility of extended LOS, support preliminary diagnoses, enable cost-effective treatment planning, and more.
The new deep learning tool created by Google effectively analyzed patient records drawn from EHRs in two hospitals, achieving high accuracy (up to 95%) in predicting diagnoses, medical events, and patient outcomes. This proprietary algorithm outperformed traditional predictive analytics techniques, which require the creation of customized datasets with particular variables and can’t deal with the mix of modalities, NLP, free-text, and structured data together.
Google’s model processed the information related to over 216 thousand hospitalizations, drilling into more than 46 billion structured and unstructured data points. The deep learning algorithm was able to accurately predict in-hospital mortality, 30-day readmission, prolonged LOS, and all diagnoses upon discharge.
According to North Shore-LIJ, 10% of their patients need special care or services, which accounts for 60–70% of their spending and resource utilization. Being an eager adopter of value-based care delivery approach, the health system decided to take preventive measures in order to meet their performance goals, optimize resource usage, and improve total care costs.
North Shore-LIJ started aggregating data from EHRs, pharmacy notes, and claims. The collected data sets are used for grouping patients according to their age scales, conditions, residence, and certain risk factors (blood pressure, cholesterol, behavior, nutrition, etc.).
The stratified groups are further processed to find specific patients with high risks to develop certain conditions or complications upon discharge. These patients are timely informed of the risks, advised to schedule an appointment, or even redirected to another PCP.
This systematic approach to the patients who might stay in the twilight zone outside the facility and don’t get more attention to their personal needs allowed North Shore-LIJ to get back on the track with resource utilization, cutting on unplanned admissions and critical cases. In particular, the system’s 30-day expenditures on the Total Joint care package were reduced by 4.2% within a year.
Sepsis can be challenging to detect early as its symptoms often are ambiguous. At the same time, only 50% of patients with septic shock receive proper treatment duly since every hour without the diagnosis increases the mortality rate by at least 7%.
Penn Medicine created a predictive algorithm to enable early sepsis detection. The model takes more than 500 clinical variables and 6 measurements from vitals and lab values to define whether a patient can develop severe sepsis. As a result, Penn Medicine can now detect 80% of severe sepsis cases in 30 hours since early symptoms have appeared.
Opioid abuse is a serious problem in the U.S., taking more lives each year. In 2017, 72 thousand people died from overdoses of prescription opioids, heroin, cocaine, fentanyl, and other synthetic drugs, as opposed to 50 thousand who died from the same reason in 2016.
Moreover, substance abuse usually signals that patients may struggle with mental disorders or social problems. This is why abusing patients need highly personalized interventions to ensure that they can recognize and be able to address their deeply rooted behavioral issues.
While the increasing number of accidental overdoses and deaths shows that the traditional methods of prevention prove themselves scarce, the state and clinical stakeholders join their efforts to introduce more tailored approaches to battling the opioid abuse crisis. Harnessing real-world data analytics is the first step.
The Health and Human Services Department plans to spread the emerging prescription-drug monitoring programs (PDMPs) across the states. These programs will analyze data gathered from hospitals, clinics, and pharmacies to identify the “doctor shopping” cases—patients that are seeing several health specialists and getting multiple prescriptions for the same medication.
The extracted findings will be injected into the nation-wide database of federal controlled substances so that physicians and pharmacists could pre-check their patients’ profiles, medication tracks, related conditions, and currently used drugs to exclude adverse effects or overdoses.
Some states, including Florida, Kentucky, and Ohio, already report positive results from PDMPs. According to the Centers for Disease Control and Prevention, Florida decreased the amounts of opioids prescribed in 80% of counties from 2010 to 2015. The state also reduced the number of oxycodone overdose deaths by 50%, as of 2012.
Utah-based Intermountain Healthcare sides with HHS on harnessing big data analytics for tackling the opioid abuse issue. Their goal is to achieve a 40% reduction in opioid prescriptions by the end of 2018. Processing their own EHR data in combination with external data sources, Intermountain Healthcare will be able to track the purchasing of Schedule II-V medications even outside their system or state and find even the most resourceful “doctor shoppers.”
Since the healthcare organization acknowledges root causes of substance abuse, they also plan to use the analyzed data to improve their pain management education for clinicians, so that they could help patients better and identify the early signs of patients falling into abusive behavior.
According to Blue Cross NC, 65% of ER visits aren’t that critical—practically, the health concerns of these patients can be addressed by PCPs and treated within the next 12 hours. However, these avoidable visits still happen, they still require attention from medical specialists, and they still entail Medicare penalties for high ER occupation rates.
Providers want to make sure they can help care consumers in the best way, not overloading their emergency departments with those who can get medical assistance on the outpatient basis. Big data promises to achieve this.
Minnesota Department of Health tapped into the state’s All-Payer Claims Database (APCD) and aggregated the information on ED utilization, hospital admissions, and preventable readmissions. They have found about 1.2 million preventable ER visits, amounting to $1.3 billion in care costs.
Among the most frequently reported conditions responsible for preventable emergency room visits were: back pain, respiratory infections, and abdominal pain. These conditions can be treated on the outpatient basis without additional costs arising from being admitted to the ER.
The department is now planning on the activities to improve emergency department utilization, including the measures for improving care coordination, support services, HIE, and medication management.
The multi-specialty group practice was concerned about the number of ER visits and decided to employ big data in an attempt to balance the care delivery costs and staff workload. They acquired a tool that enables analysis of all patient data, feeding the system’s algorithms new information monthly.
Columbia Medical Associates have found multiple care transition gaps leading to an increase in avoidable ER visits, including underreporting and poor information exchange with the hospitals. They addressed these gaps with staff training and education, improving the scheduling process, making the same-day follow-up calls, and even adding several health specialists to handle the urgent appointments only.
Navigating their way with the help of real-world evidence derived from big data, the group practice was able to improve patient satisfaction and deal with preventable emergency room visits. And yes, they achieved significant care cost savings—about $6.5 million, according to their estimations.
From buzzword to real-world evidence, big data analytics brings major benefits in healthcare, if applied right. The reason why this technique shows tremendous results is simple—healthcare is all about big data, and there’s plenty of room for improvement in various parts of the care delivery cycle and beyond.