Voluntary reporting has been the standard method for identifying adverse events in hospitals, yet its effectiveness at identifying a comprehensive array of adverse events has always been in question. The electronic health record (EHR) contains clinical data that can be systematically reviewed to identify adverse events and improve adverse event detection. Active use of an automated trigger tool that is embedded in an EHR can identify systematic issues with delivery of high-risk medications and is cost-effective and efficient. Further development of an automated adverse event detection protocol for pediatrics is needed to apply this approach systematically across pediatric institutions.
Key Points
- •
Children’s National Medical Center has benefited greatly from increased adverse event detection, in particular automated adverse event detection.
- •
This system has contributed to improved quality of care and cost-effectiveness for patients and families.
- •
The reduction in adverse events noted can be attributed to the targeted approach provided with the automated trigger tool.
- •
The current availability of triggers is being broadened with the development of pediatric specific triggers.
- •
Organizations can collaborate with their electronic health record contractor to design and build an automated trigger tool that best serves their needs and aids in quality improvement efforts.
Introduction
Hospitals strive to provide safe, efficient, and affordable care in a fast-paced environment and minimize adverse events. Knowledge surrounding adverse events in children remains inadequate. Approximately 70,000 children are harmed yearly because of adverse events, although more than half of these events may be preventable. All patients warrant an atmosphere of safety and comfort while receiving care in our hospitals and outpatient treatment centers. Thus, using innovative methods to detect and reduce or eliminate these adverse events is imperative.
Providing an optimal, safe environment means knowing the vulnerabilities that exist in a system. Only then can an organization begin to provide highly reliable care. Therefore the identification of adverse events or defects in a system is critically important. Creating a streamlined process to identify key trends in adverse events, leading to policy changes and a reduction in harm, is an achievable goal. What follows is a discussion regarding the state of adverse event detection methods to improve active surveillance of these events, based on reporting results from our own institution.
The standard method for identifying adverse events is most often surveillance of the voluntary reporting system of an institution. Because this approach has its flaws, many have attempted to augment detection in numerous ways. When compared with other more structured methods of adverse event detection, namely manual chart review as well as trigger tool methods, voluntary reporting consistently underidentifies events. Manual chart review can be resource-intensive because its primary method includes review of the medical record in its entirety. Because many adverse events and their associated triggers are rare, this detection method is often associated with a low yield. Trigger tools, whether manual or automated, have been one of the more popular methods. A trigger tool may be defined as data present in the electronic health record (EHR), or paper chart, that potentially identifies that an adverse event has occurred or could occur. Common examples of such data are documentation of reversal agents such as naloxone or flumazenil in the medical record. The administration of naloxone suggests that the patient has received too much opioid for their needs, an ideal trigger for further investigation. Similarly, the presence of laboratory values that are out of range of normal may be used as a trigger. Table 1 shows common detection methods used in many pediatric organizations.
| Detection Approach | Advantages | Disadvantages |
|---|---|---|
| Active surveillance: administrative data | Few resources needed Inexpensive Readily available | Limited identification Real-time clinical intervention not present Requires explicit documentation to ensure coding accuracy |
| Active surveillance: manual triggers | Substantial identification of events | Substantial resources needed Expensive Real-time clinical intervention is challenging Time-consuming chart reviewer training Real-time intervention limited by time report is generated |
| Active surveillance: electronic triggers | Substantial identification of events Real-time intervention present Time efficient | Substantial resources required Expensive Requires chart reviewer training Requires some technological sophistication |
| Active surveillance: technology reports | Inexpensive Few resources needed May detect more potential ADEs than other methods | Some identification of events Real-time clinical intervention is present, but depends on when report is generated Requires some technological sophistication |
| Active surveillance: direct observation | Real-time intervention is present Identifies more preventable events | Substantial resources required Limited identification of events Focus on medication administration events |
| Voluntary reporting | Inexpensive All types of events included Minimal resources required | Real-time intervention limited by notification of event Concentrations on ADE administration events Relies on staff member to submit event notification Requires nonpunitive environment and positive safety culture |
There are 2 standard methods of trigger-based adverse event detection system: manual and automated. Manual review is based on evaluation of randomly selected charts, usually a limited number of charts. Here, the reviewer searches for specific triggers in each of those medical records. By definition, this approach is based on a sampling method. By contrast, automated adverse event detection (AAED) is based on the identification of specific triggers in all EHRs for any given period. Rather than a reviewer searching for triggers in a sample of paper-based medical records, for AAED, algorithms are written to automatically identify triggers on a continuous basis. The benefits of this approach are evident. First, once the algorithms have been written, AAED is more time efficient, because each trigger is electronically forwarded to a reviewer. Second, because AAED allows for monitoring all EHRs, an organization likely obtains a more comprehensive understanding of adverse events, both the common and uncommon ones.
Introduction
Hospitals strive to provide safe, efficient, and affordable care in a fast-paced environment and minimize adverse events. Knowledge surrounding adverse events in children remains inadequate. Approximately 70,000 children are harmed yearly because of adverse events, although more than half of these events may be preventable. All patients warrant an atmosphere of safety and comfort while receiving care in our hospitals and outpatient treatment centers. Thus, using innovative methods to detect and reduce or eliminate these adverse events is imperative.
Providing an optimal, safe environment means knowing the vulnerabilities that exist in a system. Only then can an organization begin to provide highly reliable care. Therefore the identification of adverse events or defects in a system is critically important. Creating a streamlined process to identify key trends in adverse events, leading to policy changes and a reduction in harm, is an achievable goal. What follows is a discussion regarding the state of adverse event detection methods to improve active surveillance of these events, based on reporting results from our own institution.
The standard method for identifying adverse events is most often surveillance of the voluntary reporting system of an institution. Because this approach has its flaws, many have attempted to augment detection in numerous ways. When compared with other more structured methods of adverse event detection, namely manual chart review as well as trigger tool methods, voluntary reporting consistently underidentifies events. Manual chart review can be resource-intensive because its primary method includes review of the medical record in its entirety. Because many adverse events and their associated triggers are rare, this detection method is often associated with a low yield. Trigger tools, whether manual or automated, have been one of the more popular methods. A trigger tool may be defined as data present in the electronic health record (EHR), or paper chart, that potentially identifies that an adverse event has occurred or could occur. Common examples of such data are documentation of reversal agents such as naloxone or flumazenil in the medical record. The administration of naloxone suggests that the patient has received too much opioid for their needs, an ideal trigger for further investigation. Similarly, the presence of laboratory values that are out of range of normal may be used as a trigger. Table 1 shows common detection methods used in many pediatric organizations.
| Detection Approach | Advantages | Disadvantages |
|---|---|---|
| Active surveillance: administrative data | Few resources needed Inexpensive Readily available | Limited identification Real-time clinical intervention not present Requires explicit documentation to ensure coding accuracy |
| Active surveillance: manual triggers | Substantial identification of events | Substantial resources needed Expensive Real-time clinical intervention is challenging Time-consuming chart reviewer training Real-time intervention limited by time report is generated |
| Active surveillance: electronic triggers | Substantial identification of events Real-time intervention present Time efficient | Substantial resources required Expensive Requires chart reviewer training Requires some technological sophistication |
| Active surveillance: technology reports | Inexpensive Few resources needed May detect more potential ADEs than other methods | Some identification of events Real-time clinical intervention is present, but depends on when report is generated Requires some technological sophistication |
| Active surveillance: direct observation | Real-time intervention is present Identifies more preventable events | Substantial resources required Limited identification of events Focus on medication administration events |
| Voluntary reporting | Inexpensive All types of events included Minimal resources required | Real-time intervention limited by notification of event Concentrations on ADE administration events Relies on staff member to submit event notification Requires nonpunitive environment and positive safety culture |
There are 2 standard methods of trigger-based adverse event detection system: manual and automated. Manual review is based on evaluation of randomly selected charts, usually a limited number of charts. Here, the reviewer searches for specific triggers in each of those medical records. By definition, this approach is based on a sampling method. By contrast, automated adverse event detection (AAED) is based on the identification of specific triggers in all EHRs for any given period. Rather than a reviewer searching for triggers in a sample of paper-based medical records, for AAED, algorithms are written to automatically identify triggers on a continuous basis. The benefits of this approach are evident. First, once the algorithms have been written, AAED is more time efficient, because each trigger is electronically forwarded to a reviewer. Second, because AAED allows for monitoring all EHRs, an organization likely obtains a more comprehensive understanding of adverse events, both the common and uncommon ones.
Methods
Beginning in September, 2007, Children’s National Medical Center began using the AAED method using specifically designed pediatric triggers. Box 1 shows a complete list of the current trigger library, determined and maintained by a process using the following criteria:
- •
The presence of the trigger needs to be electronically identifiable
- •
The frequency of the trigger is manageable with current resources
- •
There must be a favorable positive predictive value of detecting an adverse event; the absolute positive predictive value may vary depending on the type of event(s) detected by the trigger (eg, low positive predictive values may be tolerated if the frequency of the trigger is relatively low yet the trigger identifies a high level of harm)
- •
Anticipated serious level of harm (≥level E on the National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP)) Index for Categorizing Medication Errors ( Table 2 )
