A former 24-week preemie was receiving post-op I.V. morphine for pain. Because of the patient's low weight, the physician chose a dilute morphine solution and ordered a continuous infusion of 60 mcg/hour. That evening, just before shift change, the patient's nurse needed to replace the morphine solution, as it was running low. However, the replacement solution wasn't yet up from the pharmacy. The RN used a more concentrated morphine solution stored on the floor after calculating an appropriate delivery rate based on the morphine concentration contained in this different solution.
The subsequent night shift nurse noticed occasional drops in the patient's O2 saturation, but was able to improve the situation by repositioning the patient. Around 8 a.m., however, the patient's O2 saturation dipped below 70%, chest movement with respiration was poor, and the patient was difficult to arouse. The nurse called the resident, who discontinued the morphine and administered naloxone. Fortunately, the patient responded to the naloxone. She became more alert and her O2 saturation increased. Later investigation revealed that the RN had made an error in calculating the dosage from the concentrated solution. As a result, the infant had received morphine at a rate of 600 mcg/hour, throughout the night—10 times the dose ordered.
Unfortunately, adverse events such as this are experienced far too often. This situation is underscored by The Institute of Medicine (IOM) report To Err Is Human , published in 1999.1
Due in part to the IOM report, the public has become increasingly aware of medical errors and their consequences. On the other hand, the IOM report has resulted in unprecedented attention toward patient safety. Moreover, there's mounting pressure from the public, from insurance companies, and from healthcare business groups for health professionals to recognize the reality that medical errors are far too frequent.2,3
Trigger tools
Over the last year, Cincinnati Children's Hospital Medical Center (CCHMC) has implemented several automated trigger processes, with some notable early successes. A trigger tool methodology for assessing ADE rates through random chart review has been in use at CCHMC since 2002. A “trigger” can be defined as an occurrence, prompt, or flag, found in review of a medical record, which can spark further investigation to determine the presence or absence of an adverse event. Trigger tools can help healthcare organizations identify adverse events and provide more consistent and accurate information than traditional error reporting systems.4
In contrast, by extending our trigger tool approach to the use of computerized techniques to provide “automated triggers,” we're able to address this shortcoming and target specific triggers of interest more effectively. These automated triggers enable us to identify a much higher proportion of adverse events than random chart review, thus yielding more information for revealing common causes.
Our experience supports the finding that a computerized approach can be both more accurate than voluntary reporting and less costly of time and money than manual chart review.5
An early aim of our Automated Safety Events (ASE) project has been to determine the extent to which the use of near real-time automated trigger methodology, along with corrective actions, will successfully reduce the frequency of the related adverse events.
Role of IT
Voluntary event reporting significantly underreports the true rate of adverse events, while randomized chart reviews tend to be both resource intensive and results-limited, in that the majority of charts reviewed will reveal no adverse events.6
The July 2006 IOM report Preventing Medication Errors specifically recommended that healthcare providers strive to “…create high-reliability organizations that constantly improve the safety and quality of medication use.”7 8
Methodology
• Project overview: To date, CCHMC has implemented four triggers, using this automated approach: 1) naloxone (for opiate over-sedation); 2) flumazenil (for benzodiazepine oversedation); 3) dextrose bolus (for insulin-related hypoglycemia); and 4) readmissions within 24 hours (for potential treatment errors or oversights). For each of our triggers, a list showing the occurrences is generated daily by computer. Each occurrence is then investigated via chart review, interviews with staff to determine whether there was an actual adverse event, and whether that event was preventable. In the case of preventable events, steps are then taken to implement corrective actions to ensure that events with similar causes don't happen in the future. The next few sections examine this process in more detail. The flowchart in Figure 1 depicts the decision-making process involved in identifying adverse events using our automated trigger approach.
• Identification: The case study described at the start of this article was identified with a naloxone trigger, and investigation confirmed it to be a preventable occurrence of opiate-related oversedation. However, not all occurrences of our automated triggers lead to actual adverse events. For example, in the case of naloxone, not all administrations are for opiate-related oversedation. Some are intended for diagnostic purposes. In those cases, the diagnostic use represents an exclusion criterion with regard to ADE identification. Different triggers will have different exclusion criteria. Accordingly, different triggers will have different degrees of success in being able to identify actual adverse events, as depicted in Figure 2.
Figure 2: Trigger Success in Identifying Adverse Events (as of May 08)
Information on events identified via the automated triggers is recorded in a relational database. In the case of confirmed adverse events, a “Safety Event Notification” that summarizes the event is automatically generated from the database and then sent to the appropriate staff (physicians, nurses, respiratory therapists, etc.). These personnel are asked to look over the event description, review their own records, and provide their perspective on what happened and why, what might have helped to prevent the event, and what actions might be taken to prevent future occurrence of similar events. The feedback from staff helps to determine whether an event was preventable, and this determination is also recorded in the database.
Unfortunately, deciding what's preventable can sometimes be rather subjective. Thus, it isn't always possible to reach a consensus on this question. Moreover, what seems nonpreventable today may prove preventable tomorrow. We prefer to err on the side of safety, however. If sufficient opinion supports the preventability of an event, we'll categorize it as such.
• Cause analysis and corrective actions: In the case of a preventable adverse event, we conduct a quick-strike meeting to examine its causes in detail, so that we can plan corrective actions. In line with the theory that medical errors are driven more by error-prone systems than by individuals, these quick-strike meetings are oriented toward identifying system-based causes.9 Figure 3 for team member descriptions.) Once causes are identified, we can devise interventions or process changes to address system vulnerabilities and shortcomings. The causes identified for our case study event are shown in Figure 4 , along with the interventions.
Figure 3: ASE Team Composition
Figure 4: Case Study Analysis Results
• Outcome/progress measure: For assessing overall success related to particular triggers, a measure that's seen growing use at CCHMC is that of “Days Between Events.” For relatively infrequent events, this measure can be more informative than measuring event rates, and it's this measure that serves as a principal measure of progress in reducing preventable events.
Results
Some kinds of events may have a large number of possible causes, which ultimately require a large number of interventions to have a noticeable impact on event rates. Others kinds of events may be more localized in both cause and intervention. Opiate-related overdose (detected by our naloxone trigger) is the kind of event that has many different causes. We're seeing some promising indications from our interventions, but it will take more time to confirm success or to reduce event rates effectively.
A substantial factor contributing to insulin-related hypoglycemia, on the other hand, has been successfully addressed by the single intervention of establishing a new protocol for nondiabetic patients on insulin. This involved the following steps:
defining the new protocol for use of insulin in nondiabetes-related hyperglycemia
placing a laminated “reminder” copy of the protocol on the front of each pertinent medical chart
empowering nurses to call the attending physician if the resident isn't following protocol.
Our view is that patients on insulin shouldn't generally need a dextrose bolus administration if the insulin is being managed properly. The new protocol is consistent with this view. Figure 5 illustrates the success of the new protocol in our Pediatric Intensive Care Unit (PICU). Success would have been even more pronounced, if not for failures to follow the protocol. We're working to strengthen adherence to this protocol.
Figure 5: Preventable Administrations of Dextrose Bolus to PICU Patients on Insulin
Prerequisites, benefits, and limitations
Intrinsic to implementing an automated trigger methodology is having a computerized provider order entry (CPOE) in place. This provides the principal basis for obtaining the daily lists of triggers. Even better would be having an integrated hospital information system that provides the ability to query triggers from multiple data sources. In fact, the implementation of such a system is currently underway at CCHMC. The effectiveness of our ASE team's efforts has also depended very much on a strong multidisciplinary team, since a variety of skills has been needed to address the many issues that have arisen. Both clinical and nonclinical skills have been essential. Sustaining the project in the long term requires the support of hospital senior leadership.10,11
The automated trigger approach has a number of recognized general advantages:
near real-time identification of adverse events, thereby facilitating timely investigation and correction efforts
a 100% capture of events, unlike the 10% to 20% typical of voluntary reporting
higher yield than random chart review in detecting adverse events, with a corresponding savings in time and effort; however, more important benefits emerged specific to our ASE teamwork and to our overall patient safety improvement efforts (Successful interventions for improving processes came about because of suboptimal clinical situations first identified by these automated triggers.)
more reliable safety processes, which should in turn decrease the number of adverse events12
insight into how best to prioritize the events needing further examination and to allot resources for improvement efforts
improved patient safety; moreover, analysis of events at a later time, across all departments, may reveal trends or areas of concern identifying a need for corrective actions on a much broader scope.13
There's occasional reluctance of staff to take the time to respond to our notifications and to participate in quick strike meetings. In addition, while acceptance of change is a growing part of our culture at CCHMC, resistance is still encountered at times. Unfortunately, perhaps the greatest overall limitation of the automated trigger process lies in the inability of hospitals to make use of it. A 2004 study found that the vast majority of hospitals didn't have CPOE.14 15
Expanded efforts
While CCHMC has fully implemented only four triggers to date, our longer term plan includes expansion to additional triggers. Our experiences with the automated trigger methodology have been highly positive, and we expect the beneficial impacts on patient safety to expand with our continued efforts in this area.
REFERENCES
1. Kohn L, Corrigan J, Donaldson M, eds.
To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 1999.
2. Anderson J. Information technology for detecting medication errors and adverse drug event.
Drug Safety. 2004;3(5):449–454.
3. Classen DC, Pestotnik SL, Evans RS, Burke JP. Computerized surveillance of adverse drug events in hospital patients.
Qual Safety Health Care. 2005;14:221–226.
4. Beyea S. Using trigger tools to enhance patient safety.
Assoc Oper Room Nurs. 2005;82(1):115–116.
5. Bates D, et al. Detecting adverse events using information technology.
J Am Med Inform Assoc. 2003;10(2):115–127.
6. Szekendi M, et al. Active surveillance using electronic triggers to detect adverse events in hospitalized patients.
Quality Safety Health Care. 2006;15:184–190.
7. Aspden P, Wolcott J, Bootman J, Cronenwett L, eds.
Preventing Medication Errors. Institute of Medicine of the National Academies. Washington, DC: The National Academies Press; 2007.
8. Girouard S, et al. Errors in children's health care. Role of partnerships. In: Second Annual Meeting of Child Health Services Researchers. June 27, 2000. Rockville, Md: Agency for Healthcare Research and Quality.
http://www.ahrq.gov/research/chsr2err.htm .
9. Leape L, Berwick D. Five years after
To Err is Human —what have we learned?
J Am Med Inform Assoc. 2005;293(19):2384–2390.
10. Bates D. Using information technology to screen for adverse drug events.
Am J Health-Syst Pharm . 2002;59:2317–2319.
11. Kloppenborg T, Tesch D, Manolis C, Heitkamp M. An empirical investigation of the sponsor's role in project initiation.
Project Manage J. 2006;37(3):16–25.
12. Luria J, Muething S, Schoettker P, Kotagal U. Reliability Science and Patient Safety.
Pediatr Clin North Am. 2006;53:1121–1133.
13. Leape L. Reporting of adverse events.
N Engl J Med. 2002;347(20):1633–1638.
14. Ash J, Gorman P, Seshadri V, Hersh W. Computerized physician order entry in U.S. hospitals: results of a 2002 survey.
J Am Med Inform Assoc. 2004;11(2):5–99.
15. Jha A, et al. Identifying adverse events: development of a computer-based monitor and comparison with chart review and stimulated voluntary report.
J Am Med Inform Assoc. 1998;5(3):305–314.
