Throwing computers at health care

Computerworld reports on an extensive new Harvard Medical School study, appearing in the American Journal of Medicine, that paints a stark and troubling picture of the essential worthlessness of many of the computer systems that hospitals have invested in over the last few years. The researchers, led by Harvard’s David Himmelstein, begin their report by sketching out the hype that now surrounds health care automation:

Enthusiasm for health information technology spans the political spectrum, from Barack Obama to Newt Gingrich. Congress is pouring $19 billion into it. Health reformers of many stripes see computerization as a painless solution to the most vexing health policy problems, allowing simultaneous quality improvement and cost reduction …

In 2005, one team of analysts projected annual savings of $77.8 billion, whereas another foresaw more than $81 billion in savings plus substantial health gains from the nationwide adoption of optimal computerization. Today, the federal government’s health information technology website states (without reference) that “Broad use of health IT will: improve health care quality; prevent medical errors; reduce health care costs; increase administrative efficiencies; decrease paperwork; and expand access to affordable care.”

As was true of business computing systems in general, at least until the early years of this decade, it’s been taken on faith that big IT investments will translate into performance gains: If you buy IT, the rewards will come. Never mind that, as the researchers note, no actual studies “have examined the cost and quality impacts of computerization at a diverse national sample of hospitals.”

Now, at last, we have such a study. The researchers combed through data on IT spending, administrative costs, and quality of care at 4,000 US hospitals for the years 2003 through 2007. Their analysis found no correlation between IT investment and cost savings or efficiency at hospitals and in fact found some evidence of a link between aggressive IT spending and higher administrative costs. There appeared to be a slight correlation between IT spending and care quality, in some areas, though even here the link was tenuous:

We found no evidence that computerization has lowered costs or streamlined administration. Although bivariate analyses found higher costs at more computerized hospitals, multivariate analyses found no association. For administrative costs, neither bivariate nor multivariate analyses showed a consistent relationship to computerization. Although computerized physician order entry was associated with lower administrative costs in some years on bivariate analysis, no such association remained after adjustment for confounders. Moreover, hospitals that increased their computerization more rapidly had larger increases in administrative costs. More encouragingly, greater use of information technology was associated with a consistent though small increase in quality scores.

We used a variety of analytic strategies to search for evidence that computerization might be cost-saving. In cross-sectional analyses, we examined whether more computerized hospitals had lower costs or more efficient administration in any of the 5 years. We also looked for lagged effects, that is, whether cost-savings might emerge after the implementation of computerized systems. We looked for subgroups of computer applications, as well as individual applications, that might result in savings. None of these hypotheses were borne out. Even the select group of hospitals at the cutting edge of computerization showed neither cost nor efficiency advantages. Our longitudinal analysis suggests that computerization may actually increase administrative costs, at least in the near term.

The modest quality advantages associated with computerization are difficult to interpret. The quality scores reflect processes of care rather than outcomes; more information technology may merely improve scores without actually improving care, for example, by facilitating documentation of allowable exceptions …

[A]s currently implemented, health information technology has a modest impact on process measures of quality, but no impact on administrative efficiency or overall costs. Predictions of cost-savings and efficiency improvements from the widespread adoption of computers are premature at best.

There is a widespread faith, beginning at the very top of our government, that pouring money into computerization will lead to big improvements in both the cost and quality of health care. As this study shows, those assumptions need to be questioned – or a whole lot of taxpayer money may go to waste. Information technology has great promise for health care, but simply dumping cash into traditional commercial systems and applications is unlikely to achieve that promise – and may backfire by increasing costs further.

13 thoughts on “Throwing computers at health care

  1. Mike Drips

    Having been both a computer consultant and a HIPAA consultant to hospitals, I can assure you that they are as f**ked up as any other enterprise out there. The only people that will benefit from government money on this scenario are the medical software companies who are already selling bloated, designed on a mainframe in the 60s, useless crap to hospitals.

    IT directors and IT staff at hospitals also are no smarter or more technically savvy than anyone else.

    Lower medical costs by selling more computerizations into hospitals? HA HA HA!

  2. Tom Lord

    The “IT for healthcare” push has its origins a few years back then the reform project really started. W.H.O. published its infamous piece about comparing national healthcare spending rates vs. outcomes. Many big firms were staring at their unfunded liabilities, especially for retirees, and realizing they needed lower healthcare costs. In and around the W.H.O. report various research concluded that the V.A. hospital system was the real exception: low costs and good outcomes among the best in the world. Some preliminary studies and a heaping helping of anecdotes resulted in a wide consensus that V.A. worked better mainly because (a) salaried doctors with outcome incentives rather than pay-per-procedure doctors; (b) integrated IT reducing the medical error rate and saving data entry costs. The avant guarde of the investment class was all over this space – elite conferences and everything.

    A few years back, in part in reaction to that motion, the software the VA system uses was made open source. The CDC jumped on board by going after real-time medical reporting as a hedge against pandemics (i.e., possibly make IT reform in health-care mandatory and dictate reporting features). Candidates advocating health care reform yet compliant with big money’s framing of the issues got bumps in campaign coffers. And voila, here we are today, that train pulling into the station.

    We parted ways with evidence-based health care reform long ago, in favor of angling for government-subsidized investment opportunities. That train pulling in is a gravy train, pure and simple, in the finest and most nauseating US tradition.

  3. finn


    This seems to be a much stronger claim than that posed in “Does IT Matter?”

    My understanding of the latter is that IT provided efficiency benefits BUT, because all your competitors got on the same IT treadmill, those efficiency benefits do not translate into competitive advantage (& thus profit).

    Reducing costs (in health care) would seem to be immune to this rebalancing because the efficiency gains really do translate to lower costs, without a profit margin to worry about.

    But the study says the exact opposite.

    Curious to know if you have a deeper analysis of this disjunct. Is it a matter of hospitals being “happy with crappy” (perhaps because they aren’t truly market-driven)? Is it due to the top-down nature of IT mandates in health care (all successful projects start out small, etc)?

  4. Tom Lord

    Finn, one thing that really “tight” IT brings to hospitals is a diminishment in the ability to cheat the system. Cheating the system is a big source of what efficiency there is in health care so if you lower the administrative costs but also, as a side effect, significantly reduce cheating – it’s a wash.

  5. Bertil

    My understanding (but I don’t think you mention that in your Does IT matter?) is that computer would bring savings, but only once the company is re-organised, or more often, when entire sectors are re-organised by newcomers; cases at hand: Google, Amazon wouldn’t have been possible, and they are extremely efficient business (in spite of having access to presumably the same IT as their competitors).

    Benefits in health will only come when hospitals and private practices are though-through with new organisations: Shared health data might bring new results (an area that will most likely be part of Web 3.0); Genetic-based drugs, coming from companies completely different then current drug-makers; Concierge medecine, etc. The important part about the VA case pointed out by Tom Lord is not the wages or the incentive, or the IT, but the set of all that that draws a completely different system — a system that was recently re-though through entirely to match tech of the day.

    My bet? A major player, or a type of player, will emerge, and doctors will hate it and need it as badly as newspapers hate and need Google News.

  6. Linuxguru1968

    If you walk into your typical hospital, you will notice A LOT of paper laying around! IT in health care is largely a boondoggle joke. When a large organization buys one of these “big” software packages, there is always the condition that “redundant” paper records be kept. So, the computer terminal for the system is pushed to the back of the ward and used by interns during slow hours to play solitaire.

    A lot of jobs and man hours in the US health-care are “paper pusher” jobs (biller and coders)which would be eliminated if a real paperless system would be implement. And, of course, doctors don’t like the idea of medical records being out of their control because they sometimes make mistakes and might have to change them( or loose them entirely if they file for bankruptcy!) – difficult to do if they are on secure servers that they don’t have access to!

    I know a nervous lady who works in the records department of a local large hospital; they are in the process of going paperless in patient records. There are terminals in all patent rooms for access to records and doctors don’t look at paper charts. She says that the hospital plans to reduce the staff for records from about twenty down to FOUR in the next several years. Obama’s idea of putting in funding for paperless automation in a bill designed to CREATE jobs just shows how stupid they are because it will ELIMINATE jobs not create them!

    This time,because of the recession, the attempt to go paperless might actually become a reality. In some sense a single health-care record system is just one step away from a single payer system which a lot of business and politicians favor anyway!

  7. finn

    Bertil, you’re looking forward to destructive innovation / creative destruction in the health care field. Given that health care is heavily regulated – imagine the investment required to bring a new genetic treatment to the market – I’m not sure I agree that this is a realistic possibility. I think we’re going to have to struggle toward a solution within our political system rather than hoping for a technocratic solution. This makes me sad – or at least frustrated – being a software engineer myself…

    Tom, I hear the faintest glimmers of hope in your first post. It sounds like there’s a good model out there; the difficulty is in implementing it in the face of captive elected officials and other political realities. I remember hearing about the VC outcomes & software awhile back but they dropped out of my forebrain. I wonder if 4,000 hospitals in this study included any of the VC archipelago, and how/whether they deviated from the mean? (yes, I realize i could probably suss this out by reading up on it all. sorry.)

  8. grizzly marmot

    This is no surprise to people working on the inside of these transitions. When the medical institution I work at decided to go all out on computerization, I actually sent the CEO a synopsis of “Does IT Matter”.

    I have to chuckle though at the start of the scholarly article – “Many believe that computerization will improve health care quality, reduce costs, and increase administrative efficiency”. Well, the people selling the computer systems NEVER make those claims. They claim that you COLLECT more money, not SAVE money, and indeed that is how they measure their success. As many know medical care does not follow the usual market forces. It is always better (financially) to do more, charge more and collect more. There is nothing like a computer to be certain all charges are charged. In addition, there are very few medical advances where the primary goal is “cost-effective” as opposed to better outcome.


    Since the systems in use are designed to collect money it is not a mystery that they do not improve outcomes.

  9. Gord Irish

    Are we simply expecting too much that IT would really reduce healthcare costs. Have similar studies to David Himmelstein’s been done that show IT reduces costs in regular business? The IT vendors have been telling us this is so for years and years but has it ever really been established?

  10. Linuxguru1968

    Automation will certainly reduce LABOR costs by eliminating jobs. However, since software vendors make most of their money in upgrades and maintance, those saving are likey not to make much of a difference; ending up just a transfer of debts from one account to the other with loss to the local tax base. Right now these systems are propriatory give-aways to the big political donnors. Adoption of completely open standard and open source applications would probably help – especially if competition was aimed at smaller local startup and companies.

  11. Zach Tumin

    There’s another doozy, also from Harvard. From my blog post on it (

    “The New York Times reported last week on a new study from Dr. Ashish K. Jha, at Harvard School of Public Health, and Catherine M. DesRoches of Massachusetts General Hospital. Looking at the results from 3000 hospitals, Jha and DesRoches assessed the impact on the quality of health care and outcomes from the move to electronic health records. The results? Let’s throw caution to the wind and call it “zero” –so far…”

  12. riskpundit

    Considering the study averages the benefits of health care IT over 4,000 hospitals, I can surely understand the results. However, what would be more interesting is comparing the top quartile to the bottom quartile.

    One of the great fallacies of health care is that it’s about the same all over. Not enough people have read Porter and Christensen.

    I would be a lot more interested in understanding Cleveland Clinic’s experience with IT than the average of 4,000 hospitals.

Comments are closed.