Technology Enablers: Reducing Medication Errors Through e-Prescribing, Computerized Clinical Decision Support, and Drug Interaction Databases

US Pharm. 2013;38(8)(P&T suppl):8-10.

In health care delivery today, the various technologies employed have enabled clinicians to streamline and optimize the delivery process. These technologies, such as computerized clinical decision support in combination with electronic prescribing (e-prescribing), have great potential to improve patient safety by minimizing prescribing and medication dispensing errors.^1-5 However, as previously reviewed in U.S. Pharmacist’s TechRx column,¹ there are a number of operational challenges related to the implementation of computerized clinical decision support that must be overcome in order to fully realize the anticipated patient safety benefits.⁶ These challenges need to be further investigated and resolved for optimal best care delivery.

Electronic Health Records

Clinicians are increasingly using electronic health records (EHRs) to enter prescriptions via computer. Increased utilization of computerized medication order entry is being driven, in part, by the federal Meaningful Use program from the Centers for Medicare and Medicaid Services (CMS).⁷ This program includes incentives for the adoption and meaningful use of certified EHRs for eligible clinicians in both Medicare and Medicaid. E-prescribing is an integral component of the Meaningful Use program.

Once a prescription is entered into the prescribing software of the EHR, the prescription can either be printed or electronically transmitted. If the clinician elects to print the prescription, the information will be routed to a script printer in the office, and a paper prescription will be generated. Once the paper prescription is manually signed, it can either be handed to the patient or faxed or mailed to a pharmacy. In other instances, the clinician will elect to sign the prescription electronically and transmit the prescription directly to the patient’s pharmacy. After being electronically signed, the prescription is then transmitted electronically, usually via an intermediary, using a national data standard directly to the pharmacy computer system. The latter process constitutes e-prescribing.

Regardless of whether the prescriber elects to print or electronically transmit the prescription, the prescriber’s EHR can apply a series of edits to check for potential errors that could be harmful to the patient. The idea here is to catch and correct such errors before the prescription is printed or electronically transmitted. This process of routinely checking for potentially harmful errors is called computerized clinical decision support.

A number of clinical decision support tools can be built into the EHR software. These include drug-allergy and adverse drug reaction (ADR) edits, drug-drug interaction (DDI) edits, and drug-dosage, drug–patient condition, and drug-lab edits.

Drug-Allergy and ADR Edits

Almost all commercially available EHR systems allow clinicians to enter information about patient-specific allergies and ADRs. The nature of the reaction, if known, can be entered along with information about severity. For example, an entry for a penicillin-allergic patient might read: “Penicillin allergy (severe); pt developed rash, angioedema, and anaphylaxis.” Subsequent orders for penicillin or a potentially cross-reacting drug such as a cephalosporin will then trigger an alert. If the patient’s database is shared throughout an entire enterprise system, then any EHR in the system can trigger the alert. The patient’s allergy and ADR data can also be shared electronically with other clinicians and health care providers via health information exchange (HIE). There are standardized electronic formats for sharing such data, including the continuity of care document (CCD).⁸ Clinicians can override the edits if there is good cause. For example, if the patient has a serious infection and the clinician knows that the patient can tolerate the cephalosporin, then a note to that effect can be entered into the record.

Drug-Drug Interaction Edits and Alert Fatigue

DDI edits have emerged as an area of ongoing challenge. There are a number of commercially developed drug interaction databases that can be deployed on an EHR. Several examples include the National Drug Data File Plus from First DataBank, Medi-Span from Wolters Kluwer Health, and Cerner Multum.¹ These databases are very comprehensive and feature hundreds of potential edits. While some of these edits are high priority and relate to serious life-threatening interactions, many are low risk and of questionable clinical significance. Some EHR implementations display all edits with equal significance; hence, clinicians are presented with a stream of low-priority or irrelevant edits mixed in with occasional high-value edits. The consequences of this type of presentation are very serious because clinicians become overwhelmed and frustrated by the continual presentation of low-priority nuisance alerts. This pattern of edits interferes with clinician workflow and distracts clinicians, interrupting their thought patterns. This results in so-called alert fatigue.

A large multistate study of 2,000 clinicians found that approximately 90% of alerts were overridden.⁹ However, clinicians did accept 10% of them, implying that there is value in certain circumstances. However, alert fatigue can cause clinicians to miss high-value alerts that are buried in a stream of low-value noise. In some instances, the clinician or organization turns off the entire DDI software package, thereby defeating a major patient safety initiative.

Because alert fatigue threatens to jeopardize the entire concept of improving patient safety through DDI edits, the federal Office of the National Coordinator for Health Information Technology (ONCHIT) awarded a grant to the RAND Corporation and Harvard/Partners Healthcare in collaboration with the University of California, Los Angeles (UCLA), to study the problem and develop a solution.¹⁰ The approach taken by the study group was to identify a critical set of interactions that should be implemented universally. The group assembled multiple stakeholders and drew from a panel of experts that included knowledge-base vendors (Cerner Multum, First DataBank, Wolters Kluwer), academic experts, practicing physicians and pharmacists, and federal (FDA) and private (American Society of Health-System Pharmacist) agencies.

Thirty-one high-risk DDIs were reviewed and a final list of 15 interactions was adopted.¹⁰ The initial 31 interaction pairs encompassed 195 drug-drug pairs. These were categorized as 12 drug-drug pairs, 12 drug-class pairs, and 7 class-class pairs. The study group considers the final set of 15 interaction pairs to be a starter set that should be identified in all commercial products as high severity because they have serious potential for patient harm and are contraindicated for coadministration.¹⁰ The list might not represent all high-severity interactions, so additional research will be needed in this area.

The final list of 15 interactions contains agents and/or drug classes that should never be administered together. One example is selective serotonin reuptake inhibitors (SSRIs) and monoamine oxidase (MAO) inhibitors, because such a combination can result in the potentially fatal serotonin syndrome.¹¹ One can argue that edits like this, which could potentially prevent fatal interactions, should be hard stops on the EHR; in other words, the clinician cannot just automatically override the edit but must read a short monograph about the interaction and then provide a written justification as to why the override is being placed. This would prevent inadvertent dismissal of the edit due to alert fatigue. If such a procedure were followed, it is likely that these edits would almost never be overridden.

Deployment of the 15 interaction sets in EHRs as high risk, along with the elimination of clinically irrelevant edits, could greatly reduce the burden of alert fatigue; however, the actual commercial implementation of this approach has not been successfully accomplished due to legal issues, particularly concerns among database and EHR vendors about liability.¹² It appears that a more extensive effort will be needed to develop and maintain a comprehensive national database of all clinically-relevant and high-risk interactions that can be adopted by all EHR vendors. Concerns about liability will be mitigated only when a national database is developed, updated, and maintained on a regular basis, representing a complete set of all known high-risk and clinically-relevant interactions. A concerted effort would be required to remove all edits of low clinical value. Such a database would need continual updating as new information becomes available and new drugs enter the marketplace.

The methodology employed by the RAND study group could be the basis of such an effort.¹⁰ This includes broad-based stakeholder representation including practicing clinicians and academic experts and appropriate government, industry, and private organizations. However, the funding and exact methodology for moving this effort forward at the national level have not yet been identified.¹² Furthermore, it has not been determined whether the database should be maintained by a private entity or by a public agency such as the FDA.

Other approaches to alleviate alert fatigue have also been tested in the marketplace. One approach being tried by some database vendors, such as Wolters Kluwer, features customization at the organization or department level or by user type.¹ Improving the specificity of alerts by considering patient characteristics (e.g., renal function) is one approach to customization. Careful editing of the database to ensure that edits are consistent with best clinical practices can eliminate many clinically irrelevant edits. Individual end users can also customize their alerts by checking boxes to suppress a given alert for all patients or for a specific patient.

Some alerts may be of value for the patient, particularly those that relate to the timing or sequencing of drug administration. For example, calcium or iron ingestion may impair the absorption of levothyroxine, so such products should not be coadministered at exactly the same time with levothyroxine.¹ This type of information could be printed in an educational brochure for the patient, rather than appearing as an edit for the clinician.

Drug-Dosage Edits

Drug-dosage edits can be designed to be patient specific, so that information about drug indication, patient age, patient weight, body surface area, etc., can be incorporated into the alert logic. This can be particularly useful with pediatric populations, where dosing is often weight based.³

Drug–Patient Condition Edits

Edits that are specific for patient conditions will be useful for special patient populations. Examples include patients with major organ failure such as renal failure, congestive heart failure, or hepatic dysfunction, where drug dosages might need to be adjusted. The edits will help to ensure that appropriate dosage adjustments have been implemented. Another category would be women who are pregnant or breastfeeding, two situations where certain drugs should be avoided altogether.

In the future, genomic information could be included in the patient’s profile so that appropriate pharmaceutical choices could be made based on the patient’s genetic background. This step can facilitate the development of edits that are more specific and tailored for the patient.

Drug-Lab Edits

Edits can be developed that are responsive to the patient’s individual laboratory results. One important example would be edits that take into account the patient’s serum creatinine level or estimated glomerular filtration rate. Such edits could be effective both when medications are prescribed and when lab results are entered into the EHR. For example, a patient with diabetes might be appropriately prescribed metformin when the patient’s creatinine is 1.3. However, if the creatinine increases to 1.8 between office visits, the patient’s chart would be flagged for review when the lab results are received because now the metformin might need to be discontinued. Other lab results that could trigger a chart review might be liver function tests and potassium or thyroid levels. Such edits are likely to be highly patient specific and helpful to the clinician.

Conclusion

Today, technology is playing a pivotal role in the evolution of changing health care reform. If used prudently and in appropriate channels, it can empower the clinician to create and support best practices, which in turn supports the revitalization and recreation of health care practices now and in the years to come.

REFERENCES

1. Figge HL. Computerized clinical decision support and drug interaction databases. US Pharm. 2012;37(3):47-49.
2. Figge HL. What really is electronic prescribing? US Pharm. 2011;36(9):HS35-HS38.
3. Figge HL. E-prescribing and the impact on medication dispensing errors. US Pharm. 2012;37(7):31-33.
4. Kaushal R, Kern LM, Barron Y, et al. Electronic prescribing improves medication safety in community-based office practices. J Gen Int Med. 2010;25:530-536.
5. Moniz T, Seger A, Keohane C, et al. Addition of electronic prescription transmission to computerized prescriber order entry: effect on dispensing errors in community pharmacies. Am J Health Syst Pharm. 2011;68:158-168.
6. Sittig DF, Singh H. Electronic health records and national patient-safety goals. N Engl J Med. 2012;367:1854-1860.
7. Blumenthal D, Tavenner M. The “meaningful use” regulation for electronic health records. N Engl J Med. 2010;363:501-504.
8. Health Resources and Services Administration (HRSA). Health IT Adoption Toolbox. What is the CCD? www.hrsa.gov/healthit/toolbox/HealthITAdoptiontoolbox/PersonalHealthRecords/aboutccd.html. Accessed June 9, 2013.
9. Isaac T, Weissmann JS, Davis RB, et al. Overrides of medication alerts in ambulatory care. Arch Intern Med. 2009;169:305-311.
10. Phansalker S, Desai A, Bell D, et al. High-priority drug-drug interactions for use in electronic health records. J Am Med Inform Assoc. 2012;19:735-743.
11. Beasley CM, Masica DN, Heiligenstein JH, et al. Possible monoamine oxidase inhibitor-serotonin uptake interaction: fluoxitine clinical data preclinical findings. J Clin Psychopharmacol. 1993;13:312-320.
12. Darves B. Efforts seek to improve drug-drug interaction alert tools. iHealthBeat. April 15, 2013. www.ihealthbeat.org/features/2013/efforts-seek-to-improve-drugdrug-interaction-alert-tools.aspx. Accessed June 9, 2013.

To comment on this article, contact rdavidson@uspharmacist.com.