The Case

A 34-year-old woman with AIDS developed a fever and hypotension due to suspected pneumonia. Her past medical history included several AIDS-related complications, but a recent test showed that her viral load was undetectable on a drug regimen of stavudine, lamivudine, and Kaletra (a combination pill containing lopinavir and ritonavir). Given her critical condition, an infectious disease consultant recommended changing her stavudine, which is associated with lactic acidosis, to abacavir. The intern caring for the patient used a preprinted antiretroviral order template (paper form) to execute the medication orders, requesting a new agent, Trizivir, a combination pill containing abacavir, lamivudine, and zidovudine.

The following morning, a pharmacist noted that the patient’s revised orders called for continuation of stavudine, lamivudine, and Kaletra in addition to the new order for Trizivir. The patient was thus set to receive double doses of lamivudine and thymidine analogs, any of which could be terribly toxic in overdose. Apparently, the execution of orders via the template did not automatically cancel the other, free-form orders, a processing issue the intern failed to recognize. Fortunately, the pharmacist caught the error minutes before scheduled administration, and the patient suffered no adverse event, because only Trizivir was administered.

The Commentary

Medicating hospitalized patients is a task prone to errors that may result in serious injury.(1,2) Although less than 1% of them actually result in harm, adverse drug events (ADE) are still the most common cause of injury to patients in hospitals, ranging from discomfort to severe allergic reactions and death.(3) The Harvard Medical Practice Study (4) found 19% of all injuries to be attributable to drug complications. There is also significant associated cost of additional treatment and increased length of stay—estimated at around $2 billion annually.(5)

The majority of potential ADEs occur during ordering.(6) In most U.S. hospitals, clinicians write orders by hand or give verbal orders to staff, who then transcribe them and initiate the administration process. Handwritten orders are susceptible to interpretation errors due to poor legibility, inaccurate transcription, or omissions. For instance, medication orders may lack standard units of measure or frequency of dosing. Clinicians may also fail to add required corollary orders or to check for known allergies or drug interactions.

Pre-printed order forms and templates are in part intended to improve the completeness, accuracy, and legibility of orders. They also constrain the choice of medications, laboratory tests, and studies for specific clinical conditions and thus contribute to the standardization of care and promote adherence to institutional practice guidelines or clinical pathways.

Computerized physician order entry (CPOE) systems extend the effectiveness of paper forms by adding dynamic, real-time functionality and providing clinicians with immediate feedback on their actions. Algorithms running in the background may compare entered data with information in electronic medical records or clinical knowledge bases and trigger alerts suggesting medication dose or frequency adjustments (e.g., based on weight or renal function), give warnings about allergies or drug interactions, and display relevant laboratory results. Clinicians can then make corrections before the order is finalized and further steps are taken to administer the medication.

This near-overdose case was likely the outcome of several converging omissions and possible misconceptions. First, the intern gave orders appropriate for the given clinical state of the patient on the advice of an infectious disease consultant (substituting abacavir for stavudine, evidently by starting an abacavir-containing compound, Trizivir, and intending for the patient’s previous regimen to be discontinued) but apparently had inadequate knowledge of the administrative procedure that discontinues active orders. Individuals working within a complex system (e.g., the organizational and procedural structure of clinical ordering in a large hospital) may not have accurate conceptual understanding of the details, contingencies, and implications of all their actions as they propagate through the system. For example, clinicians may have a set of expectations about the processing and outcome of routine, free-form orders: for example, that superseded older orders will automatically be discontinued, that staff will inquire about clarifications for uncommon orders, and the like. This tacit knowledge (not explicit but implicitly acquired through the knowledge of the domain) may carry over to situations when the same activity (e.g., medication ordering) is completed by other means (e.g., by an order template) but may not be valid under these new circumstances. Switching from one method of task completion to another then requires extra cognitive effort to recall the appropriate set of processing rules, which opens the possibility for omissions and errors. This risk also intensifies during transitional periods when institutions are replacing one ordering system with another, which disturbs traditional patterns of conduct and forces modifications to already established practice routines.(7,8) Although templates enhance order completeness and compliance with institutional requirements, they may also engender erroneous assumptions about the situational context in which they are executed unless they contain explicit warnings about differences from routine free-form orders. An independent pharmacy review system needs to complement this process as a second-tier safety measure in a paper-based environment.

The second error contributing to the near overdose was the duplication of drug orders in a complex ordering pattern that involved compound chemical substances. The intern may have intended to substitute abacavir for stavudine in the original drug combination by ordering the compound Trizivir. Ordering abacavir not by itself but along with other drugs as a compound may have been a standard or usual procedure. The clinician would then have to remember to discontinue not only stavudine but also lamivudine. Moreover, he or she would have to realize that zidovudine, the third component in Trizivir, is in the class of thymidine analogs and would further increase potential toxicity and adjust the dosing accordingly. Although an expert physician may instantly recognize the danger of ordering this drug combination, the probability of an error due to momentary distraction, interruption, or a memory slip remains high. Increasing the odds of such an oversight is the large number of compound medications and their possible combinations.

The likelihood of similar errors could be lowered by implementing safeguards into the ordering process. Templates, unlike free-form orders, can contain printed information about components in compound drugs and remind clinicians to check current patient orders for duplicates. More effective safety measures, however, can be put into practice only in electronic devices such as CPOE systems. Automated algorithms alerting clinicians that a compound drug contains duplicates of currently active medications, for example, might have prevented both errors described in this case report.

There is research evidence (9), however, that order entry systems may create possibilities for error if users have incomplete conceptual knowledge of how a system processes and manipulates entered data. The complexity of advanced clinical information systems often poses a formidable challenge to novice users and may require extensive training to achieve desired levels of safety and performance.(10) Visual salience of critical information on ordering screens, for example, a common problem with many CPOE systems, needs to be addressed by software engineering that adheres to the principles of human-computer interaction.(11) Using CPOE systems is the safest way to order drugs.(12) However, failure to attend to usability issues related to human factors can facilitate errors and patient harm.(13,14) Providing software developers with a workable design paradigm that incorporates workflow considerations (15) and human-machine interaction issues is essential for allowing this technology to achieve its full potential as a medical tool.

Health care information technology (HIT) should be designed to correspond to human characteristics of reasoning, attention, and memory constraints. The introduction of new technology induces a change in user behavior as the user completes new tasks and follows different procedures. An optimal fit of HIT to its environment and to the needs of users is therefore of paramount importance for design and successful implementation in hospitals.

Take-Home Points

• Paper-based, handwritten clinical ordering is inherently prone to ambiguities and errors of omission, although it offers seemingly fast completion times.
• Order templates improve legibility and order completeness but fail to provide real-time decision support or to perform automated allergy and dosage checks.
• CPOE offers flexibility to support safe and complete ordering practices, but long learning times and user interface complexity may compromise its effectiveness.
• Transitioning between two modes of order completion (e.g., free-form and templates) necessitates extra cognitive effort from clinicians, which may negatively affect performance and potentially generate errors.
• Design of CPOE interfaces according to cognitive usability principles of human-computer interaction is essential for improving training times and error rates.

Jan Horsky, MA, MPhil Senior Doctoral Candidate, Department of Biomedical Informatics Columbia University, New York

Vimla L. Patel, PhD, DSc Professor of Biomedical Informatics Director of the Laboratory of Decision Making and Cognition Columbia University, New York

References

1. Bates DW, Boyle DL, Vander Vliet M, Schneider J, Leape LL. Relationship between medication errors and adverse drug events. J Gen Intern Med. 1995;10:199-205. [ go to PubMed ]

2. Bobb A, Gleason K, Husch M, Feinglass J, Yarnold PR, Noskin GA. The epidemiology of prescribing errors: the potential impact of computerized prescriber order entry. Arch Intern Med. 2004;164:785-792. [ go to PubMed ]

3. Classen DC, Pestotnik SL, Evans RS, Lloyd JF, Burke JP. Adverse drug events in hospitalized patients. Excess length of stay, extra costs, and attributable mortality. J Am Med Assoc. 1997;277:301-306. [ go to PubMed ]

4. Leape LL, Brennan TA, Laird NM, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med. 1991;324:377-384. [ go to PubMed ]

5. Bates DW, Spell N, Cullen DJ, et al. The costs of adverse drug events in hospitalized patients. Adverse Drug Events Prevention Study Group. JAMA. 1997;277:307-311. [ go to PubMed ]

6. Kaushal R, Bates DW, Landrigan CP, et al. Medication errors and adverse drug events in pediatric inpatients. JAMA. 2001;285:2114-2120. [ go to PubMed ]

7. Ash JS, Stavri PZ, Kuperman GJ. A consensus statement on considerations for a successful CPOE implementation. J Am Med Inform Assoc. 2003;10:229-234. [ go to PubMed ]

8. Massaro TA. Introducing physician order entry at a major academic medical center: I. Impact on organizational culture and behavior. Acad Med. 1993;68:20-25. [ go to PubMed ]

9. Horsky J, Kuperman GJ, Patel VL. Comprehensive analysis of a medication dosing error related to CPOE: a case report. J Am Med Inform Assoc. 2005;12:377-382. [ go to PubMed ]

10. Horsky J, Kaufman DR, Oppenheim MI, Patel VL. A framework for analyzing the cognitive complexity of computer-assisted clinical ordering. J Biomed Inform. 2003;36:4-22. [ go to PubMed ]

11. Patel VL, Arocha JF, Kaufman DR. A primer on aspects of cognition for medical informatics. J Am Med Inform Assoc. 2001;8:324-343. [ go to PubMed ]

12. Kuperman GJ, Teich JM, Gandhi TK, Bates DW. Patient safety and computerized medication ordering at Brigham and Women's Hospital. Jt Comm J Qual Improv. 2001;27:509-521. [ go to PubMed ]

13. Scanlon M. Computer physician order entry and the real world: we're only humans. Jt Comm J Qual Saf. 2004;30:342-346. [ go to PubMed ]

14. Koppel R, Metlay JP, Cohen A, et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA. 2005;293:1197-1203. [ go to PubMed ]

15. Malhotra S, Jordan D, Shortliffe EH, Patel VL. Workflow modeling in critical care: piecing together your own puzzle. In review.