 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Joseph A. Carcillo, MD
Shekhar T. Venkataraman, MD
Robert S.B. Clark, MD
R. Scott Watson, MD, MPH
Hülya Bayir, MD
Richard A. Orr, MD
Departments of Critical Care Medicine and Pediatrics
University of Pittsburgh School of Medicine
Pittsburgh, PA 15213
Trung C. Nguyen, MD
Department of Pediatrics
Baylor College of Medicine
Houston, TX 77030
In Reply.—We appreciate the insightful comments of Longhurst et al, Jacobs et al, and Rosenbloom et al in response to our study findings. They all emphasize the critical importance of careful planning and preparation for successful computerized provider order entry (CPOE) implementation. We wholeheartedly agree. However, these authors also assert that our observations probably resulted from inadequate planning or foresight exercised by the CPOE project team at the Children's Hospital of Pittsburgh (CHP). Although we cannot, ourselves, refute this assertion, we note that members of CHP's CPOE project team seemed to have been well aware of the many potential pitfalls that can impede successful CPOE implementation.1 The report by Upperman et al,1 which described in great detail how CPOE was introduced at CHP, suggests that considerable thought and effort went into our institution's adoption of CPOE technology. This endeavor was rewarded, in fact, with a significant reduction in harmful adverse drug events.2 In this regard we ask, had we not performed our independent investigation specifically examining mortality outcomes, would anyone have questioned the adequacy of the preparations made at CHP? Clearly, debating the potential merits and shortcomings of various CPOE implementation strategies is important and necessary, but we submit that this debate must first proceed with a uniform definition of success. Because the mandate for widespread implementation of CPOE throughout all US hospitals stems from the promise that this technology will save patient lives, we further submit that the realization of this promise should serve as the "gold-standard" definition of success.
Longhurst et al, Jacobs et al, and Rosenbloom et al also seem to make a clear distinction between CPOE and the implementation process, implying that our observation resulted entirely from faulty implementation and not from CPOE technology, itself. Although we acknowledge this possibility in our article, we find it extremely difficult to separate one element from the other. A CPOE system does not operate in isolation, and its proper functioning requires its seamless integration into a strong and dynamic health information technology (HIT) infrastructure. In its broadest sense, implementation encompasses not only solving potential clinical workflow problems but also building this HIT infrastructure, resolving systems-integration glitches, and overcoming machine-human–interface obstacles. Although CPOE, itself, may not be to blame for undesired patient outcomes, a well-designed, well-programmed, userfriendly CPOE software architecture constructed "from the ground up" with a clear understanding of how clinicians think and operate in real-life situations can only serve to facilitate the implementation process. Conversely, a general, minimally modifiable, clinically awkward CPOE platform may require considerable creative efforts to "retrofit" this technology into operational form. Although CPOE, itself, may not be at fault for network problems related to ever-present threats from Internet worms and viruses, the vulnerability or resilience of a specific CPOE platform and its HIT infrastructure to properly function during periods of attack and after firewalls and other network security upgrades have been installed seem to be considerations intrinsic to the adoption of this technology.
Rosenbloom et al caution careful interpretation of our data because of inherent limitations of study design and raise questions regarding the statistical approach used. We have openly acknowledged and enumerated our study's limitations and share in their sentiment to cautiously interpret our data. We remain unclear, however, how our results might be invalid, because we have performed a well-accepted method of regression analysis. Still, to appease their criticism of our approach, we have subsequently performed a propensity score analysis as requested. Propensity scores were generated with CPOE as the dependent variable and with all of the variables listed in Table 1 of our article, except for Pediatric Risk of Mortality (PRISM) score, as independent variables. The propensity scores were then recorded into deciles, and a logistic-regression model was fitted using CPOE, decile group, and the interaction between CPOE and decile group as predictors of mortality. The interaction between CPOE and decile group was not found to be significant and was subsequently dropped from the model. Results then indicated that CPOE was significantly associated with increased odds of mortality (odds ratio [OR]: 2.420; 95% confidence interval [CI]: 1.508–3.883). The addition of PRISM score into this regression model continued to demonstrate increased odds of mortality (OR: 3.130; 95% CI: 1.848–5.302).
We applaud Jacob et al for continuing to monitor mortality outcomes (a practice we strongly support) after CPOE implementation at their institution. We find it very reassuring that they observed no increase in PICU mortality after using a different implementation strategy (and different CPOE platform) . The CPOE project team at Cincinnati Children's Hospital Medical Center should be commended for their diligence and their particular attention devoted to tackling the unique challenges of the PICU environment. On the other hand, if we are permitted to briefly play "devil's advocate," it seems that severity-of-illness–adjusted mortality rates did not substantially improve after CPOE implementation, either. As fellow pediatric intensivists, Jacobs et al should be well aware of the excitement after the publication of the adult clinical trials of drotrecogin alpha for severe sepsis that showed significant improvements in survival.3 In contrast, the pediatric clinical trials were terminated recently because similar survival benefits could not be demonstrated among children. Although drotrecogin alpha has received Food and Drug Administration approval as a treatment for adults with severe sepsis, this approved indication has not been extended to the pediatric population.
We wish to make it clear that despite our unexpected study results, we continue to believe that CPOE holds great promise as a tool to improve patient care and save lives, and we reject any proposal that calls to abandon this important technology. At the same time, the excellent points made by Longhurst et al, Jacobs et al, and Rosenbloom et al reinforce our impression that the very complexity of CPOE technology and its implementation requires a thorough, organized, systems approach to adequately address their many concerns and may be best resolved through carefully designed, multicenter trials supported by the National Institutes of Health and/or Food and Drug Administration. If the HIT community finds this suggestion unfeasible or unreasonable because of the unique nature of computer/software technology, another approach may be to examine how the aviation industry, long considered a model industry for maintaining high safety standards, has addressed the incorporation of these technologies into its field. We note that Federal Aviation Administration oversight ensures the mission-critical reliability of any new computer-related device or software throughout all US airport control towers. Alternatively, an independent self-governance HIT body analogous to the American Board of Pediatrics, for example, might provide another means to certify the proper deployment of CPOE.
REFERENCES
Related articles in Pediatrics:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2006 American Academy of Pediatrics. All rights reserved. |
||
|
||
