Why bankers need to stick to banking, and keep their profound lack of knowledge of biomedicine and Medical Informatics to themselves

[Note: this post is very rich with hyperlinks. To fully understand the post, at least open the hyperlinks in a separate window and browse their material - ed.]

In April 2011 I referenced a 2007 comment about health IT ROI, by then-Congressional Budget Office (CBO) head Peter Orszag, in a post entitled "Medicare/Medicaid Cuts? Spend Money on Patients - Not Computer Experiments":

... More on purported cost savings - Peter Orszag, former head of the Congressional Budget Office, said the use of electronic health records, without a major change in health care delivery, "would not significantly reduce overall health care costs" in the agency's 2007 report on long-term health care spending. He also said that according to data from the report, the return on investment for EHR's "is not going to be as substantial as people think." The CBO concluded that predictions of cost savings from EHR's relied on "overly optimistic" assumptions and said much is unknown about the potential impact of health information technology. [That is, it is an experimental technology - ed.] Mass savings from health IT is an assertion that is both unproven and highly unlikely in my view.

Mr. Orszag is now Vice Chairman of Global Banking at Citigroup.



He seems to have changed his tune somewhat.



In an Aug. 16, 2011 piece in Bloomberg entitled "Health Care Prognosis Better With Digital Law: Peter Orszag" he opines:

Even with the all-too- depressing illustrations of political paralysis we’ve seen recently, government can still act to improve our lives. A good case in point: The U.S. health sector is rapidly digitizing, and federal legislation from early 2009 , passed well before the health-care reform act, is an important reason why.

He refers to the Health Information Technology for Economic and Clinical Health a.k.a. HITECH Act embedded in the 'economic stimulus' ARRA bill.

... Partners HealthCare has used its health IT to be more selective about which patients should have diagnostic imaging tests, such as MRIs and CT scans. The cost to Medicare for imaging tests nationwide roughly doubled from 2001 to 2009. And such tests are not only expensive but potentially dangerous. Frequently imaged patients face an increased risk of cancer because of exposure to excessive radiation. [That risk is not very large; on the other hand, insufficiently imaged patients are playing the "slot machine of life", which when they are unlucky leads to missed diagnoses, injury and death - ed.]



Doctors at Partners now order imaging scans through the computer system and are automatically queried about the patients’ characteristics. For each case, the software then provides an “appropriateness” score, reflecting evidence- based protocols [i.e., likely based on averages, not the individual patient's nuances - ed.] for the image requested. And in some cases, the program suggests an alternative to imaging.

I presume doctors' use of imaging is monitored based on the cybernetic "score." Orszag states:

Comparing doctors: the system is also used to compare doctors to one another, so they know if they use imaging tests more or less than their peers do.

I have yet to see robust (e.g., RCT-based) outcomes of this "regulator of care" function of health IT, wherein the physician is no longer the learned intermediary between patient and computer, but the computer instead is the cybernetic regulator of care between doctor and patient. I do know of a case, however, where cybernetic "regulation" of imaging seems to have helped kill an infant at another hospital (link).

From 2006 to 2009, imaging rates at Partners flattened, and in some specialties even started to decline, sometimes significantly. The number of outpatient images per patient, for example, fell 25 percent in that period, even after adjusting for patient characteristics such as age, ethnicity, gender, medical history and medications.



... The IT interventions appear to have been effective at reducing imaging rates across the board, including among the doctors who ordered the tests most. By 2009, that doctor at the 90th percentile ordered 20 images per 100 patients, a decline of almost 10. This one doctor’s net decrease in scans was larger than the total number of scans ordered by the doctor at the 10th percentile even in 2006. And the low-use doctor reduced his rate, too, by about two images per 100 patients.

I'm still not seeing comparisons of how patients fared under the cybernetic imaging-control regime.

More broadly, health IT is a necessary but not by itself sufficient step toward improving value in health care. A review of the health IT studies by the Congressional Budget Office, published in 2008, while I was the director of that agency, concluded that it “has the potential to significantly increase the efficiency of the health sector by helping providers manage information.” The CBO also found, however, that health IT couldn’t realize this potential without a supportive health-care delivery system that uses it aggressively. The most auspicious examples of IT use were in relatively integrated systems, such as Veterans Affairs, Partners HealthCare, Kaiser Permanente and Group Health Cooperative in Seattle.

Here's where Orszag goes off the irrational-exuberance rails.



He states the "potential" of HIT to "help providers manage information" is not reached because HIT is not widespread enough.



He omits the potential of the technology to cause providers to mismanage information. Such mismanagement turned a close relative of mine into roadkill, which apparently is OK on the road to HIT utopia.



He cites the VA, where highly specialized non-commercial HIT took decades to develop under relatively ideal conditions far removed from the traditional management information systems morass in the private sector (link), and Kaiser (with this history) and Group Health, who are not exactly prolific publishers about their mishaps and travails with HIT.



He omits the fact the FDA admits they know of HIT-related injury and deaths (link to FDA internal memo of Feb. 2010), do not know the true levels of it but that their data probably represents the "tip of the iceberg" (link), and that HIT is a medical device that should fall under the FD&C Act, but that is left unregulated (link) because it's a "political hot potato." (Thus, in effect, its subjects, patients, are experimental test subjects without having given informed consent.)



He omits the mission hostile nature of much commercial HIT (e.g., as demonstrated in the nine-part series here).



He omits a growing body of literature suggesting health IT's outcomes are neutral or negative, despite billions of dollars of expense (link to Reading List).



He omits acknowledged national HIT failures in the UK (link, link), and Australia and elsewhere (link), in healthcare systems far smaller, more government-controlled and less complex than ours, efforts largely led by and utilizing American companies and products.



He omits the apparently low quality and fundamental unreliability and unsuitability of today's commercial software (see examples at link, link).



He omits independent, renowned medical device testing organizations listing health IT as one of the top health risks of 2011 (link).



He omits the general futility of large-scale government IT initiatives, as in an article hidden in the Public Administration Review that contains near 150 references (link, PDF).



In effect, commercial health IT is nowhere near fit for national rollout in 2011.



I have two suggestions:



1. Bankers need to stick to banking, and keep their embarrassing lack of knowledge of biomedicine and Medical Informatics to themselves.



2. Mr. Orszag needs to spend some time reading the IT literature, without wearing blinders.



Finally, on a personal note, I will not be taking investment advice from Mr. Orszag. I would not trust it, often finding that those deficient in publicly-available knowledge in one domain are often lacking knowledge in others. That's simply my personal preference.



-- SS