Volume 8  Number 4  September 21, 2011
Second Opinions

Deconstructing Databases I

The modern age has a false sense of security because of the great mass of data at its disposal. But the valid issue is the extent to which man knows how to form and master the material at his command.

Goethe, 1832

In the late 17th century, with increasing demand for long-term insurance coverage such as annuities, burial, life and other types of insurance, actuarial science became a formal branch of mathematics. By the time of the industrial revolution in the late 18th century, governments required increasingly advanced data collection in the form of the census and other measures to manage a 'crisis of control' created by increasingly interdependent societies. In the light of history, it is unsurprising that even in Goethe's day there was evidence of data overload, and other data difficulties, if not "data mining," subjects all too familiar in our own day.

While computers did not exist in 1832, it is clear that Goethe's insight prefigured the invention of the database defined in the dictionary as a "...usually large collection of data organized especially for rapid search and retrieval (as by a computer)"-my italics.

Medicine, Health, and Databases

We still live with Goethe's "false sense of security" despite the uncontrolled proliferation of databases in the medical world. Thanks to the CDC and its extensive statistical databases, in particular the National Center for Health Statistics, (NCHS) we have some of the richest, if limited, sources of clinical data in the world (see here, here and here). At the same time we sometimes underutilize this fountainhead of information, often preferring to slake our intellectual thirst on more capricious or fuzzy opinion/data collections, such as clinical guidelines, quality measures, clinical effectiveness data, and especially rigged drug trials. (vide ante)

Briefly, our problem is not the archipelago of medical databases instantaneously available with the touch of a mouse. Rather, it is the astounding lack of critical databases dealing with information we never bother to collect. As Shakespeare might have observed, "The fault, Dear Brutus, lies not in our stars (databases) but in ourselves."

A Few Examples of Missing Databases

If we are to "master the material at our command," we need look no farther than some of the statistics quoted in my last newsletter where I reference the fact that the number of CT scans performed in the United States has increased from three million per year in 1980 to 70 million today. From the time of publication of the first human MRI image in 1978 the number of MRI studies has grown for zero to 28 million per year in this country, 9% of the population, more than double the average in advanced countries. Yet, to my knowledge, we have designed no databases telling us the indications and results of all these studies. For example, how many millions of normal CT's and MRI's have resulted from this diagnostic infatuation?

Moreover, what fraction of any diagnostic studies result in actual benefit to the patient? What percentage of all imaging studies are indeed ill-advised or unjustified leading to Cascade Fiascos, the discovery of incidental findings without medical significance, often resulting in a series of extensive, sometimes risky diagnostic procedures ending in clinical catastrophes. In effect, we have virtually no statistics and therefore no databases relating the hundreds of millions of imaging, invasive, and other tests to the indications and outcomes of these diagnostic studies.

Implanted heart defibrillators cost around $25,000 each, but which patients will require or be benefited by having one remains uncertain. Just another missing database to consider. In 2004 Government and industry stuck a deal. Medicare agreed to expand approval rules for medical device use, nearly doubling the number of patients who qualified for one. The companies, in return, agreed to pay for a study to see which patients really benefited. But the companies reneged, backing out after putting in a measly $4 million, and the Government of the last Administration did not pick up the remaining tab.

As a result, researchers still cannot gather data to determine the types of patients who would most benefit from a defibrillator, and doctors keep implanting them in patients who may not benefit at all. Many of these patients have been harmed, even died, especially if the device has to be removed, a dangerous procedure. No one even knows whether one producer's model performs better than a competitor's. The problem is not limited to defibrillators. Consider cardiac stents. Again there is no systematic data, and therefore no database relating patient outcomes to the type of device, let alone consistent indications for their use.

Other areas of missing statistics and data continue to be publicized, though rarely on TV or newspaper front pages. Most physicians get this information via their specialty journals as opposed to generalists. For example, between 1996 and 2007, inpatient procedure rates among persons 4564 years of age doubled for total knee replacements from 12 to 26 per 10,000 population, and almost doubled from 7 to 12 per 10,000 population, for total hip replacements. Among persons 65 years of age and over, excision of intervertebral disc and spinal fusion procedure rates increased 67%, from 17 to 28 per 10,000 population, and total knee replacement procedures increased 60%, from 51 to 82 per 10,000 population. All we have up to now is the limited Maude database maintained by the FDA which consists merely of voluntary reports of adverse events caused by medical devices.

Why are so many more patients requiring hip and knee replacements as well as spinal operations? Have the indications for these procedures changed so radically in 10 years? What is the new complication rate? What were the outcomes of these procedures? Tens of thousands of artificial hips, knees, and now shoulders are implanted every year. Doctors placing these devices have little or no reliable comparative data on which brands last longest or work best in a given patient. Lord Cohen of Birkenhead, formerly a President of the British Medical Association, sometime physician to Queen Elizabeth, once remarked, "The desire of the physician to undertake an operation is not an indication for its performance." According to the McKinsey Global Institute, a consulting group, expenditures on implanted devices stand at about $17 billion annually in this country and are rising at a rate faster than the cost of drugs. Medical device expenditures, including related hospital costs are $78 billion annually.

Unlike other hospital products, implants are so-called physician preference items, meaning that doctors-not the hospitals-often choose which manufacturer's implant to use. It is a policy that invites corruption by doctors paid by companies to perform and publish studies and sabotages a hospital's ability to negotiate the best price. Profit margins on medical devices are also among the highest for any medical products, over 20% in the case of a defibrillator or an artificial hip. If you want to read more about the problem, do a Google search under "physician fraud on device purchases." The Affordable Health Care Act of March 2010 makes it easier for the government to prosecute under the federal anti-kickback statute, but if the well-heeled opposition has their way, they will ultimately kill the only major healthcare legislation passed in the last 45 years.

The only way to determine the types of patients who would most benefit from defibrillators, joint replacements, and other devices lies in requiring the maintenance of requisite databases on medical devices, standard in many European countries. Without follow-up on outcomes, medical opinion will remain divided. We still won't know how products compare with each other how many people really need a specific device, and which patients will be harmed.

More in store, stay tuned.

Martin F. Sturman, MD, FACP

Copyright 2011, Mathemedics, Inc.

EasyDiagnosis is an automated online service that analyses existing medical symptoms and predicts likely causes and conditions. Click here to find out more about this unique service and click here to try one of our modules for free.