A Walk Through the Brave New World of Healthcare Data Analytics

A Walk Through the Brave New World of Healthcare Data Analytics

With a stethoscope around the neck, a good flashlight, thermometer and genuine empathy for patients, you were all set to go as a doctor, just a few decades ago. The need for empathy remains, but the world of healthcare delivery has changed hugely since then.

Fast forward to today. Sci-tech has given us tools of remarkable capability. There is no corner of the human body, however small or remote, that cannot be imaged, measured, probed and altered in some fashion. We are armed today with devices that can even work upon the very stuff of life and creation: DNA. The words of the science fiction writer, Arthur C Clarke, come to mind: “Any sufficiently advanced technology is indistinguishable from magic.” What modern Medicine can deliver today is undoubtedly magical.

From Pieces, Into Bits — the Digital Transformation

In keeping with all other areas of human activity, the years have seen a shift in medical technology from analogue to digital. As an example, imaging studies are almost all captured in digital form. The old x-ray film is now obsolete. This change is convenient; reporting, viewing, archiving, transferring and analysing are all made much simpler with digital systems rather than physical.

The Data Tsunami

The power comes with a price. The data that pours in from any given patient is vast. Everyone dealing with healthcare delivery — users, caregivers, administrators or third-party payers — struggles with the effort of staying afloat in this deluge. Making sense of all this information is a task that can exceed the cognitive abilities of the smartest. It’s now an uphill task to stay up-to-date even in narrow specialities.

Looking at just one speciality, oncology (2005 – 2015):

  • 140 million patient encounters,
  • Generating 0.1 – 10 GB of data per patient (14 – 1400 TB overall)
  • 80% of which is unstructured.

An average hospital generates 665 TB of data, yearly. The quantity is doubling every two years.

A Triple Whammy

Three properties characterise the data deluge.

  1. Volume — as exemplified earlier.
  2. Velocity — the rate of accrual and change is estimated at 20 – 40% per year, meaning that the size of the data store doubles every other year.
  3. Variability — Captured data is stored in silos that can be difficult to penetrate. A large part of the problem is the lack of unified data storage structures and inter-operability. On top of all this, a substantial portion of the data resides as unstructured records, often descriptive text and narratives.

Tunnel Vision versus Big Picture

As a result, it is challenging for anyone — users, caregivers, administrators or payers— to get “the Big Picture”. Like the story of the blindfolded men encountering an elephant, each one interprets the whole through the narrow lens of what is immediately perceivable by the remaining senses.

New Wine, Old Bottles

Fortunately, the very same technology that has brought about the problem also can provide solutions. Data analytics is the hottest ticket in today’s information technology scene. Data mining, machine learning, deep learning and artificial intelligence offer us the means to make sense of this mass of bits and pieces in a way that individuals — even large teams of people — cannot.

Big data analysis has been a part of managing efficient businesses for some time now. We can extend the lessons learned from business into healthcare delivery to the great advantage of all —- users, caregivers, providers and payers.

Meanwhile, Behind the Scenes

While this flood has been building up, a paradigm shift is ongoing in the way quality healthcare delivery is assessed. For decades, the model of payment in healthcare was “fee-for-service”. Whether the outcome was good or bad, a service was appraised as being worth a specific sum of money and the amount released to the provider.

In recent decades, quality assessment has shifted from a physician-centric approach to a patient-centric one. The endpoint for satisfaction is an outcome that the patient feels is worth rewarding. Life and activities of daily living have been changed for the better (or at least, not worsened) by the transaction. The two points-of-view are, quite often, tangentially opposed.

A “pay-for-performance” model is slowly replacing fee-for-service.

This new model demands a panoramic view of service provision where the individual is compared against a population-based norm. The data keeps shape-shifting and has to be evaluated in real-time from the perspective of the 3 Vs: volume, velocity and variability. Humans can’t do this with ledgers or even spreadsheets. Big data analytics is the need of the day.

Promises to Keep

The intention to harness big data can be sincere, but the tools cannot be wished into existence. There is no magic wand. Starting from a well-organised assessment of needs, healthcare analytic systems have to be carefully designed and implemented. It’s all too common to take a “kitchen sink” approach to the exercise and end up with a product that no one likes.

The Winners

Healthcare data analytics will benefit four groups.

  1. Patients will receive the best quality of care.
  2. Professionals (caregivers) can deliver the best quality of care.
  3. Providers (hospital administrators) can assure users of getting the best quality of care.
  4. Payers (third-party agencies) can be confident in getting the best value for money.

Let’s take a walk through the garden of possibilities.


Data Sources

Where does this mass of information come from? Every aspect of healthcare delivery is today, a geyser of data.

The Patient Record

The patient record is the cornerstone of high-grade medical care. It’s where the process of data analysis begins.

The history and physical exam report is the core component. This document maps the patient’s current and past health status in great detail. Personal habits, past illnesses, family and social history, medications, and treatment plans enter the archive.

Other details are appended over time. They include:

  • lab summaries,
  • treatment plans,
  • procedure notes,
  • nursing notes,
  • medication records,
  • progress notes,
  • consultation requests.

Over time, the repository can become quite sizeable and bulky. Making meaning out of the record becomes a laborious and frustrating endeavour.

The Electronic Health Record (EHR)

The traditionally paper-based record is now captured digitally as an electronic health record (EHR).

EHRs have many advantages over paper.

  • They don’t need the vast spaces that physical records demand.
  • Multiple users can view them at the same time, from different points of the hospital.
  • They can be transmitted anywhere in the world.
    • Most of all, being digital, they lend themselves to data analytics.

The push for widespread EHR usage in recent years has led to the availability of an extensive database which, after analysis, may be repurposed as information packets directed at improving patient care.

There is data in plenty and, of concern, just as many standards for defining the record structure. Any given piece of data may be stored and coded in any number of fashions, often with the same package.

Other Hospital Data Sources

Modern healthcare delivery is a comprehensive, diverse, complex system, probably more so that any other activity in everyday use. Every element of this system pours in data which needs to be factored into the care of a patient.

Some typical sources include:

  1. Laboratory Information Management Systems (LIMS): The number of tests available for clinical use run in the hundreds. Starting from collection of samples from the patient to transporting them, processing them in highly sophisticated machines, reporting results and delivering reports back to patients and care providers, there are numerous points of data collection.
  2. Diagnostic Procedures: There is an equally large number of diagnostic procedures used today. Most of them are now capable of recording the entire transaction digitally. ECG, X-rays, scans, endoscopic tests, angiograms: every one of them can be piped into the data backbone of a hospital.
  3. Monitoring Equipment: The mandate for high standards of patient safety and outcome results in the need for intensely monitoring patients during their journey in a hospital. Multi-channel monitors, alarms, respiratory support devices and many more are data points.
  4. Wearable health devices are here to stay. Immense amounts of personal information are pouring in every day. We have access to perspectives of any persons’ health in a manner never imagined before.
  5. Pharmacy Management: Beginning with simple records of prescriptions, pharmacy systems offer an opportunity to keep track of the complex interactions between drugs and quality care.
  6. Scheduling Patient Flow: A hospital sees large movements of people in and out of the system: appointments have to be made, patients tracked during their journey from area to area, beds allocated to the satisfaction of the patient and the doctor. Computer-based systems coordinate these functions today. They are no longer hand done. Once again, tons of data.
  7. Radiofrequency identification (RFID) is increasingly integrated into healthcare to provide real-time management, tagging, and tracking of patients and staff
  8. Insurance Claims/ Billing: The entire process is now done online. The life of an institution hangs on the efficiency of financial management.
  9. Human Resources and Supply Chain Management—many healthcare organisations now use enterprise-level systems to manage the complexity of care in modern hospitals.

This list merely skims the surface of all that is available. Suffice it to say that modern healthcare delivery pivots around data management.

Meet the Data Scientist

We are now at a point in time where the 3 “Vs” of data which we talked about have to be tamed and converted into useful, actionable packets. The complexity of the task has led to the evolution of a distinct brand of information analyst: the data scientist.

They are highly skilled, specially trained, much-in-demand professionals who are a single-point resource for managing, analysis and interpreting Big Data. They have the capability of using tools that are in themselves complicated bits of engineering.


The Upside

I: Patients

Medical practice is, in its entirety, directed towards the welfare of patients. Let’s see how data analytics can improve what is delivered.

A: Chronic Disease Management

Patients diagnosed with chronic non-communicable diseases (NCD) consume a substantial portion of health services. A handful of specific conditions like diabetes, high blood pressure, heart disease and respiratory disorders account for a major share.

Cost-effective management of NCDs hinges on the ability of providers to pre-empt high-impact, high-cost complications which often occur in patients with these disorders.

NCDs, usually life-long afflictions, provide a wide window of opportunity for applying health care data analytics. The number of data points that need to be weighed and acted upon in each patient can overwhelm the cognitive capacity of the most well-informed, conscientious doctor.

Using smart devices, RFID-embedded machines and the universal availability of mobile telephony, patients can be closely monitored for specific target levels such as vital signs, oxygenation, blood sugar, glycosylated hemoglobin, blood pressure and many more. Detection of abnormal levels or worrisome trends permits early, evidence-based intervention which can slow down the rate of progression of many of these disorders.

Treatment can be personalised and tailor-made to fit the demands of each patient.

In a review of 49 studies of chronic disease management (Bhardwaj et al, 2018), big data analytics was beneficial in:

  • risk prediction,
  • diagnostic accuracy,
  • patient outcome improvement,
  • hospital readmission reduction,
  • treatment guidance and
  • cost reduction.

Population Health Management using predictive analyses has shifted the focus of Public Health from the traditional wait-and-watch approach to prediction and prevention.

B: Genomic Medicine

Genomic Medicine has changed the face of medical practice. Patient genotypes can provide pointers to the most effective drugs and treatment regimes, risk of complications and long-term outcomes.

The discipline is expanding at a breakneck pace. New information pours in every day. Genomic data has to be matched to the vast amounts of values observed for individual patients: a daunting task. The field is wide open for application of data analytics.

 

II: Professionals (Care Providers)

Although the doctor continues to be at the centre of healthcare delivery, modern medical practice is a collaborative effort involving many highly trained and certified providers: nursing professionals, pharmacists, physical therapists, social; workers to name a few.

Data analytics bears great promise for enhancing the skills of care providers.

A: Pre-empting acute/ critical events

As discussed earlier, predictive algorithms can point out and highlight patients with chronic disease who are at risk for crisis situations. Interventions can be made before a patient’s condition snowballs into an acute crisis requiring emergency department visits. Data analytics can identify such high-risk individuals early. Ongoing progress can be monitored, and customised care plans put in place.

B: Learning Health Systems

The information base of healthcare delivery is expanding and changing so rapidly that conventional learning tools like textbooks are obsolete almost from the time of publication. Medical information has to be far more dynamic and real-time.

Data analytics offers tools for designing and implementing “learning health systems”. Every patient visit is an opportunity to both learn and generate new knowledge. Knowledge bases can be looked up to provide the most current evidence. Recommendations can be matched to a patient’s specific data set. New patient information can be added to a global database and analysed on the fly.

Personalised Medicine is the mantra of the day.

C: Research

Data mining tools can pick up patterns that are not easily seen by humans. As data accrues, the analytic engine can keep sniffing out many gems of information and new knowledge. Some examples:

  • Risk assessment
  • Early detection
  • Epidemic detection
  • Potential cures
  • Quality of life improvement
  • Prevention strategies

The COVID 19 pandemic has shown us numerous instances of data analytics picking out potential treatment modalities.

 

III: Providers (administrators)

Hospital administrators are under constant pressure while performing the difficult balancing act between quality and cost. Despite its undeniable benefits to other business domains, healthcare has been slow, even reluctant, to adopt practices that are of proven value in business. The post-Covid years are sure to see notable changes in healthcare delivery methods. The role of data analytics will be crucial to survival and staying afloat in what promises to be a highly competitive arena.

Here are some critical areas where data analytics can find an application.

A: Key Performance Indicators (KPI)

Every process offered in healthcare has an outcome. Both process and outcome can be objectively assessed, tracked over time frames and outcomes compared against established norms or over changes in time within a given provider’s domain. This is a KPI.

Any number of KPIs are in use. Some common examples include the length of stay (LOS), 30-day readmission rates and healthcare-associated infection (HAI) rates.

The variables (process(es), factors) underlying each outcome can be complicated. Making associations between intervention and outcome can’t be done manually. Multi-factorial analysis of massive data requires data analytic tools.

KPIs can be keyed into performance dashboards (see below). Feedback to caregivers, when done in a sensitive, non-punitive manner, can be powerful tools for quality improvement.

KPIs can be used to set up “best practices” manuals that could be highly specific for a given institution.

Even with clear goals in mind and a manageable list of KPIs, the process gets very foggy when large volumes of data are involved. Enter, data analytics.

30-day readmission rates

30-day readmission rates are an important KPI of quality of care. Hasty discharges before adequate stabilisation of patients often result in readmissions within a few weeks.

These events lower patient satisfaction. Outcomes are often adverse. Hospital costs climb steeply. Payers often impose penalties on providers for this complication.

Data analytics give valuable insights into the mechanisms that could have been responsible for this event. Corrective measures and policy changes could be implemented.

B: Patient Traffic Flow Management

There is a constant movement of patients and personnel, both into and within a hospital. For long periods, this flow has been managed by personnel who acquire skills on the job, without any formal training in operations management. Considering the intricacies of patient movement in a modern hospital, data analytics can be handy for smooth service delivery.

Waiting time is a leading cause of patient dissatisfaction. Quite often, appointment times are delayed by long periods. Patients who need elective admission often simmer in lobbies till rooms are ready for occupation. The average waiting time in an emergency room is about 4 – 6 hours.

Radiofrequency identification (RFID) is a useful option for tagging and tracking patients and staff. Patient’s can be pinpointed with accuracy. The data is valuable for shaping patient flow in the care process. Once again, data analytics offers solutions for optimising and managing hot spots related to patient movement.

C: Billing and Finance

However competent the caregivers, efficient financial management is vital for organisations to stay afloat.

Key Performance Indicators (KPI) can be handy for finance managers. A variety of metrics are available from organisations like the {Healthcare Financial Management Association (HFMA)

Data analytics can provide up-to-the-minute assessments of the financial health of a hospital.

D: Human Error

Adverse events during healthcare delivery are commonly due to human error. Failure to note abnormal values, improper medication administration or misidentification of patients are all too common.

Data analytic systems can spot these events and issue warnings.

 

IV: Payers

Third-party payers usually make healthcare payments in modern practices. Be they governmental organisations or private insurers; they are always battling costs and seeking to get the most value for money.

Data analytics are vital tools for payers.

A: Comparative Analysis

Data analytics permit payers to survey the market for costs and effectiveness of specific disorders and interventions. They can be done both within an institution and between hospitals. Device and procedure costs can be compared.

Pricing data can be mapped against quality outcomes to identify the best quality, lowest cost providers. This data can be used to leverage prices with hospitals carried by the payers.

Once again, data analytics can provide detailed, up-to-date figures.

B: Fraud Prevention

Suspected fraudulent claims can be investigated with data analytics. Comparisons can be made for similar claims at other hospitals of known quality and integrity. Hard data can support rejections.


Dashboards and Displays

It’s not enough to capture and process data. Actionable information has to be displayed to users in a fashion that is easy to grasp. Anyone who has played video games will know that current-day computer graphics is more than up to the task.

The Old Way

The typical healthcare report is a static document delivered in a one-size-fits-all model. Revisions and updates are slow and time-consuming, often out-of-date at the time of printing.

The complexity of data available demands much more dynamic output.

Dynamic Displays

Look at the NYSE

Although nowhere as demanding, the rapidly moving and changing screens that we see on the floor of the NYSE and other financial centres, gives us an idea of how data can be displayed for the benefit of users.

Interactive, multi-coloured dashboards are available, showing data in easily-grasped formats. The data is updated in real-time or at least in short, frequent intervals.

Users can view critical metrics, trends, benchmarks and such.

Bells and Whistles

Complex data, when presented as easily-understood charts and tables, allow users to make confident decisions.


The Brave New World of Healthcare Data Analytics

Everywhere we turn, we keep seeing, reading or hearing about the rapidly expanding role of big data analysis and artificial intelligence. Computer power and software complexity have reached a point where hitherto fortressed domains are being breached. Recent reports of programmes generating sophisticated pieces of journalism that are hard to distinguish from human writing have induced a sense of fear in all professions. Robotisation revolutionised manufacturing. The automation wave is advancing relentlessly into white collared jobs and the service sector.

Healthcare has stayed defiantly refractory to the changes happening around it. This state can’t last for long. Major disruptions are in sight. Healthcare data analytics hold the promise for being a dominant force in bringing about a much-needed change in the area of healthcare delivery.