“Big data” is all around us. It powers the mass customization that allows Amazon to recommend new books for you. It is why Siri can talk to you on your phone. It provides the engine for the targeting of ads on Facebook. And, in health and human services, it is the fuel of decision support tools and population health management software (see Taking Decision Support From Concept To Practice and ‘Must Have’ Technologies For Cutting Edge Population Management).
But the use of big data doesn’t end there. I was fascinated with one new application of big data – provider billing audits. In the article, Analysis Of Big Data Leads To Big Arrests In Medicare Part D Fraud, federal authorities arrested 243 people in an alleged $712 million Medicare fraud. How did the Feds identify this fraud? In this case, the data was first compiled by the Office of the Inspector General (OIG) in the report, Ensuring The Integrity Of Medicare Part D, which helped the federal authorities identify “hotspots” that showed where over 700 physicians were billing for extremely high percentages of drugs with the potential for abuse. This led the authorities to investigate the pharmacies supplying those physicians, which further allowed for the identification of 400 pharmacies that were averaging 62 prescriptions per patient – or three-times the national average. This investigation led to the arrests.
This is an approach that the OIG will likely continue and expand, and has recommended that CMS do the same – as part of the above report, the OIG noted, “The availability and proactive use of data are essential to identify and address program vulnerabilities; identify providers with questionable billing; and meaningfully target program integrity resources to the areas of greatest vulnerability.” In addition, with the creation of the Chief Data Officer post in late 2014, CMS has signaled that they are fully committed to using big data in their standard Recovery Audit Program, noting on the website, “CMS is now routinely analyzing claims data in real time and applying predictive analytics to proactively identify fraud and abuse and track key metrics such as hospital readmissions.”
Finding patterns of billing problems is the likely short term approach – and likely to be fast. Fraudulent billings are an estimated 10 per cent of Medicare spending in the United States (see Financial Crimes Report to the Public), and in response the U.S. Department of Justice (DOJ) has embraced big data analytics to combat that fraud – it has been reported so far that for every dollar it spends using big data analysis, the DOJ has collected $8 for $4.3 billion in 2013 (see US healthcare: Big data diagnoses fraud). And, CMS announced last summer that it had identified or prevented $820 million in inappropriate payments over the past three years through its big data driven, Fraud Prevention System (see CMS Fraud Unit Uncovers $820 Million In Healthcare Scams In Past 3 Years) – this system uses predictive analytics to identify billing patterns in real time, and also reviews past patterns.
What does this mean for the field? The key for managers of provider organizations is to understand what your “data” says about your organization. And to know what patterns in your data may trigger further investigations and audits. By doing your own proactive practice pattern and compliance data analysis, you can get in front of any audits that are triggered by patterns in diagnosis, treatment, and billing. Using your own data, admittedly “little data” (see How To Make ‘Little Data’ Work For Your Organization and What To Do If You Don’t Have ‘Big Data’ – Making The Most Of ‘Little Data’), you head off problems and improve performance before they occur. And, all that is required is your current electronic health record (EHR) and billing data.
My colleague, OPEN MINDS Senior Associate Sharon Hicks remarked on this challenge, noting:
Many are preparing their internal compliance systems to account for the sea change that is caused by the wide use of health care data mining and analysis in government oversight. Especially for those who have questioned the methodology for extrapolation based on statistical sampling during audits, big data analytics is likely to change those models or even eliminate them. Most all of the Recovery Audit Contractors (RAC) for CMS are aggressively seeking new and innovative ways to find potential fraud, misuse, and/or abuse of CMS covered lives. The takeaway for providers is that they are especially vulnerable to methodology errors as these new models get implemented. Providers must have their own robust data systems if they are going to challenge the findings of a RAC.
Using your own data and your internal compliance or risk-management group, can you identify patterns, by clinician, by clinic, and by the entire organization that could signal a red flag to an external fraud, waste, or abuse auditor? Do you have any “unreasonable days”, like a clinician billing 15 hours in one day? Do you have routine triggers in place to keep claims from going out the door to payers, such as duplicate service or service for one level of care billed in conjunction with another, not allowed, level of care?
In the era of more data, it is inevitable that data will become more transparent – and not just to consumers. Payers and health plans will make greater use of the data to select the organizations with the “best” performance and identify those that aren’t optimal. The smart management teams will get ahead of this curve.