An important paper was published online this week in the journal Pediatrics demonstrating for the first time that there was a decrease in hospital-wide mortality after the implementation of a commercial computerized physician order entry (CPOE) system. The research took place at Stanford at the Lucile Packard Children's Hospital. The investigators compared mortality before and after the implementation of an EHR from Cerner. Note that the adult hospital at Stanford uses Epic. The authors adjusted for differences in patient acuity using a case-mix index, and they also compared their mortality figures to those obtained from an administrative database of 42 non-profit tertiary pediatric hospitals called the Pediatric Health Information System (PHIS). The results were striking. There was a decrease of 20% in the mean monthly adjusted mortality rate after CPOE implementation. Also, when compared to data from the PHIS database for a comparable time period, the observed to expected (O:E) mortality ratios for Stanford were significantly lower than for the database (0.45 vs 0.72; p=0.0003). They conclude that CPOE implementation could have resulted in 36 fewer deaths in the time period studied. The investigators do point out that other interventions implemented at the same time might have contributed to the decrease in mortality. However, it seems likely that CPOE implementation itself was the driver.

The link can be found here. It appears that the PDF is available for free even to non-subscribers, probably for a short while since this is an online release. Here is a copy of the press release.

This is clearly an important study and adds to our knowledge in this area. It should be compared to the famous, or perhaps infamous, Han study from Children's Hospital of Pittsburgh in 2005 showing an increase in mortaility post implementation of a Cerner CPOE system. Much has been written both by critics and naysayers of EHR's and by EHR proponents in regards to Han's findings. Likely, this paper will draw similarly polarizing comments. However, if these data survive scrutiny, they clearly suggest that CPOE, when designed and implemented right, may really do what we always hoped - make care safer.