Do Computers in Hospitals Save Lives? Reduce Costs? Make the Food Better?

Well, a hot-off-the-press article in the Archives of Internal Medicine says maybe two out of three. Is this a landmark study? Some media reports suggest that's the case.

Given how starved adherents of healthcare technology are for any news that can justify Congressional largesse to the tune of $20 billion, this little gem of a manuscript may garner additional attention in the coming days. If you’re interested in knowing whether computers in hospitals really save lives, the Disease Management Care Blog is at your service with a summary. Another bonus is that you won’t have to rely on the mainstream media to get it wrong.

Dr. Amarasingham (Parkland in Texas), Ms. Plantinga (Bloomberg Public Health), Dr. Diener-West (Bloomberg Public Health), Dr. Gaskin (Hopkins) and Dr. Powe (Bloomberg Public Health) surveyed a variety of Texas hospitals’ physicians about their hospitals’ level of ‘automation.’ The survey they used was the validated Clinical Information Technology Assessment Tool (‘CITAT), which assesses four domains of ‘record keeping, test results, order entry and decision support.’ If 5 or more surveys were completed, each hospital was assigned an average score based on the physicians’ answers. The Texas Hospital Association then provided the hospitals’ data on mortality, costs, complications and length of stay for all patients as well as those with heart attack, chronic heart failure, open heart surgery and pneumonia. A total of 41 of 72 targeted hospitals had 5 or more surveys completed and could therefore be scored and included in this study.

Did high CITAT scores correlate with death rates, cost, complications or length of stay? Well, as the CITAT score increased in each of the four domains, adjusted odds ratios of death and complications for some conditions decreased. Costs and length of stay also decreased for some conditions. Many did not change.

The curmudgeonly DMCB says 'not bad.'

The insight here is that useful inpatient information technology – as defined by physicians, not technobabbly consultantspeak pseudoscience – is associated with impacts on death rates and costs. Importantly, there was a ‘dose response relationship’; as the CITAT increased, the impacts grew. Even better, the insight has greater credibility because it’s gained from real world hospitals, not disconnected academic medical centers authoring studies that are only read in other academic medical centers.

But the DMCB offers up some cautions:

Association is not the same as causality. It is possible that hospitals with the ability to invest in automation also have the ability to invest in nurses, maintain quality programs, attract the high caliber clinical/administrative leadership or leverage other unmeasured features that really account for the observed changes. The authors attempted to statistically control (neutralize) for hospital status, but this is never perfect (the same techniques were used to control for limitations in the same kinds of studies of estrogen in women, which were shown to be mistaken once a prospective randomized trial was done). The bottom line is that there is no guarantee that an install of this kind of IT in year 1 will lead to fewer deaths, decreased complications, lower cost and shorter length of stay in year 2. It might.

This is hardly a slam dunk panacea. The authors noted the swing in mortality rates was in the range of 0.5%. If you’re among those 5 in a thousand, that’s a lot. For the other 995, survivorship doesn’t change. However, those 995 are facing some other issues including never events, being adequately vaccinated, being disconnected from real doctors or being discharged safely. Swings in the amount of dollars numbered mostly in the low hundreds - when an inpatient stay costs thousands and the ambient national healthcare inflation trend rate eats hundred dollar bills for lunch. We don’t know how the savings profited the hospitals or the insurers and, what's more, we don’t know if any profit was greater than the cost of all this IT.

Sorry ye worshippers of the physician office-based electronic health record. This study tested elements of the EHR outside the physician office setting. It does not apply to your vision of a paperless physician office. You’ve still got work to do in terms of providing reasonable assurance that you really save money and reduce costs in that arena.

Multiple comparisons were performed, making the likelihood of statistical mischief greater. Many of the changes were statistically significant (seemed to be of a magnitude that were mathematically unlikely due to random chance) but barely so. To the authors’ credit, they attempted to statistically control for this also and recognized it as a limitation of their study.

The DMCB doubts a single study can answer the question, but this is an important addition to our knowledge base. Good work, authors. Finally, kudos to the Archives for making the manuscript readily available on-line.