Our October 2009 What’s New in the Patient Safety World column “A Cautious View on CPOE” noted a new systematic review (Reckmann 2009) that concluded evidence showing a reduction in medication errors by CPOE, particularly serious medication errors, is scant. The few studies that did demonstrate positive results either had very focused outcomes (i.e. only certain types of medication errors in select populations were studied) or the study populations were very small or there were methodological problems. This month another study (van Doormaal 2009) adds to the uncertainty. This study, performed in the Netherlands, used times series analysis to get around some of the thorny methodological issues seen in before/after studies. It evaluated the effect of computerized physician order entry (CPOE), primarily medication order entry with rudimentary clinical decision support (CDSS), on both medication errors and preventable adverse drug events (those with harm). It showed that the percentage of orders with at least one medication error dropped from 55% pre-implementation to 17% post-implementation. The number of preventable adverse drug events also decreased post-implementation if one looked at just the conventional before/after design. However, there had been a significant downward trend prior to the CPOE/CDSS implementation. When the time series analysis was applied, that difference was no longer significant. That is, the number of preventable adverse drug events (those with harm) was not reduced after CPOE/CDSS.
Then this past week two other studies were released that conclude the cost savings from healthcare IT is negligible. One new study from researchers at the Harvard School of Public Health showed marginal differences in quality (as measured by CMS core measures) and costs (as measured by LOS) among hospitals with full-featured electronic medical records (EMR’s), partial EMR’s, and no EMR’s. The second study (Himmelstein 2009) looked at over 4000 hospitals and concluded that computerization was associated with modest improvement in process measures of quality but not with administrative or overall costs. Both studies, of course, cast doubt on the savings to the healthcare system projected to occur over the next decade.
So where does that leave us? Should we abandon our endeavors to use healthcare IT to improve quality and control costs? Of course not. However, we readily agree that we have yet to demonstrate that the theoretical benefits of healthcare IT actually translate into “hard” benefits, either in patient outcomes or efficiencies and costs. So what is wrong with the above studies? First of all, they probably looked at the wrong things. All hospitals have been working on improving the CMS core measures whether they were computerized or not. And the CMS core measures are primarily process measures, not true patient outcome measures. Similarly, all hospitals have worked hard to reduce LOS (length of stay) independently of their efforts to computerize. So it is no surprise that there should be little difference between the groups on these measures. The Himmelstein data on overall cost and administrative cost are more convincing. They offer 3 potential explanations for the lack of an impact of computerization on cost: (1) that the cost of computerization cancels out any efficiency savings or (2) the cost savings may come n the future or (3) that computerization actually has resulted in hospitals being able to increase their billings. Actually, there is probably a bit of truth in all three of those explanations.
What about the findings in the van Doormaal study? It may be that the degree of sophistication of the clinical decision support systems may have been inadequate to substantially reduce the number of medication errors leading to harm. The alerts generated were primarily related to dosing ranges, allergies, and some drug-drug interactions. They apparently did not include more complex clinical decision supports such as help with dosing based on renal function.
There are two common approaches to CPOE. One is to focus on reminding people to do those common evidence-based things you might forget to do (eg. DVT prophylaxis, remove or avoid altogether a Foley catheter, use beta blockers in post-MI patients, etc.). This is the approach many hospitals take when they first implement CPOE, alert systems, and standardized order sets. But there are other less hi-tech ways to do those things (patient safety rounds, clinical pharmacists rounding with clinical teams, paper order sets, etc.). The other approach is to focus on those things that go beyond the typical “computing” capability of the human brain. We know very few nephrologists, let alone non-nephrologists, who can appropriately dose medications just by looking at the GFR. On the other hand, computers are quite good at this. Similarly, most experienced clinicians are used to managing heparin and coumadin but if they have to use a less familiar anticoagulant drug like argotroban in a patient with heparin-induced thrombocytopenia they make errors. So developing computerized tools to help deal with these unfamiliar scenarios may be productive.
Perhaps lost in these papers is the fact that the biggest savings for our healthcare system in hospitals costs is actually in preventing hospitalization in the first place. The greatest potential, therefore, may be on the outpatient side, where computerization lags far behind the inpatient side. But we have also been unable to clearly demonstrate that those hypothetical savings from computerization in ambulatory care will actually translate to hard savings.
But the onus is still on us to show that computerization improves not just error rates but that it actually prevents those errors from reaching the patient and causing harm. Bedside medication verification (barcoding systems), a technology we love because it significantly reduces medication errors, still will not prevent a nurse from administering an inappropriate dose to a patient if that incorrect dose were originally ordered by a physician in error. So we desperately need to show that computerization actually prevents patient harm. In every other aspect of medicine today we demand evidence-based care so we should be no less demanding when it comes to IT as an intervention.
We still strongly believe that computerization will play a key role in both patient safety and improving the inefficiencies in our healthcare system. But make sure you don’t spend all our time and money developing a complex computerized solution where the classic paper checklist will produced the desired outcome.
Reckmann MH, Westbrook JI, Koh Y, Lo C, Day RO. Does
Computerized Provider Order Entry Reduce Prescribing Errors for Hospital
Inpatients? A Systematic Review.
J Am Med Inform Assoc 2009; 16: 613-623
van Doormaal JE, van den Bemt PMLA, Zaal RJ, et al. The Influence that Electronic Prescribing Has on Medication Errors and Preventable Adverse Drug Events: an Interrupted Time-series Study. J Am Med Inform Assoc 2009; 16: 816-825
Lohr S. Little Benefit Seen, So Far, in Electronic Patient Records.
New York Times November 15, 2009
Himmelstein DU, Wright A, Woolhandler S. Hospital Computing and the Costs and Quality of Care: A National Study.
The American Journal of Medicine (2009) xx, xxx