Adverse Drug Event Surveillance: Trigger Tools and Pharmacy-Based Screening

An adverse drug event (ADE) is harm to a patient that results from medication use — whether or not an error caused it, and whether or not anyone filed a report. Most hospitals still rely primarily on voluntary, self-reported incident data to surface these events, and the patient-safety literature has repeatedly shown that channel catches only a fraction of what actually occurs. Active surveillance — structured retrospective chart review using trigger tools, plus real-time pharmacy-based screening of orders and lab values — is a different detection method built specifically to close that gap, and it is the primary data source most functioning medication safety programs actually run on.

This guide describes the trigger-tool and pharmacy-based-screening methodology as documented in the patient-safety literature, including the IHI Global Trigger Tool framework and the NCC MERP Index for Categorizing Medication Errors. IHI’s own current tool pages could not be retrieved directly during this review (repeated 404s on the expected URLs); the methodology described below reflects the version of the tool widely taught and cited in peer-reviewed and AHRQ-hosted patient-safety literature. Confirm current trigger lists, licensing terms, and any revised methodology against IHI’s own published materials before adopting this as a formal program element.

What Counts as an ADE — and What This Guide Is Not

An ADE is broader than a medication error: it includes any harm from a drug’s use, whether the drug was dosed and administered correctly (an adverse drug reaction inherent to the medication) or whether an error — a wrong dose, a missed interaction check, an incorrect route — caused or contributed to the harm. A “potential ADE” or near miss, by contrast, is an error caught before it reached the patient or before it caused harm. Surveillance programs typically track all three categories separately, because they call for different responses: harm requires a severity-rated review, an error without harm is a process signal, and a near miss is a system-resilience signal.

This is hospital operational patient-safety surveillance — a quality and risk function run by pharmacy, nursing, and patient-safety staff to find and reduce harm inside the institution’s own patient population. It is a different activity from adverse-event reporting inside a clinical trial, which is governed by IRB oversight, 21 CFR 312, and ICH E6, and which reports to sponsors and regulators on a protocol-defined schedule rather than through hospital-internal chart sampling. If that is the reporting system you are looking for, see Pharmacovigilance in Clinical Research and the SUSAR definition instead.

Why Voluntary Reporting Undercounts ADEs

Voluntary incident-reporting systems depend on a clinician recognizing that an adverse event occurred, judging it worth reporting, and taking the time to file it — three separate points of attrition, each of which loses cases. Events that unfold gradually (a slow rise in creatinine from a nephrotoxic drug, a fall days after a sedating medication was started) are especially likely to go unreported, because no single clinician witnesses a clear causal moment. A frequently cited 2011 Health Affairs analysis using the IHI Global Trigger Tool method (Classen et al.) found hospitals’ actual adverse-event rates roughly ten times higher than what voluntary reporting and standard AHRQ patient-safety-indicator screening had been detecting at the same institutions — the gap wasn’t a difference in how harmful care was, only in how much of it the passive channel could see.

This is the case for running active surveillance alongside voluntary reporting rather than instead of it: voluntary reports still carry information active surveillance can’t reconstruct — the reporting clinician’s own judgment about what went wrong and why — but as a rate-tracking measure of how much medication-related harm the institution is actually producing, passive reporting alone is not a reliable denominator.

The Trigger Tool Method: Retrospective Chart Review

The trigger tool approach, developed under the IHI Global Trigger Tool framework, works by looking for indirect evidence that harm probably occurred, rather than waiting for someone to report it directly. A trained reviewer — typically a nurse — pulls a small, randomly selected sample of recently discharged records on a recurring cadence (commonly around 20 records, reviewed on a fixed schedule such as twice monthly) and scans each one for a defined list of “triggers”: specific orders, medications, or lab values that are strong circumstantial evidence an adverse event happened, even when nothing in the chart says so explicitly.

Common ADE-specific triggers include:

• Naloxone administration — suggestive of opioid-induced oversedation or respiratory depression.
• Flumazenil administration — suggestive of benzodiazepine-induced oversedation.
• Vitamin K administration, or an INR above the therapeutic range — suggestive of excessive anticoagulation.
• A glucose value below roughly 50 mg/dL — suggestive of insulin- or sulfonylurea-related hypoglycemia.
• An abrupt medication stop without a documented clinical reason — often a clinician quietly withdrawing a drug that caused a reaction rather than formally reporting it.
• Diphenhydramine administration outside an allergy history — suggestive of a drug reaction being treated as it happens.
• Clostridioides difficile toxin positive — often traceable to recent antibiotic exposure.
• Transfer to a higher level of care shortly after a medication change — a general-purpose trigger that catches harm the drug-specific triggers miss.

Finding a trigger does not by itself confirm an adverse event — it flags the record for a closer look. The reviewer reads the surrounding documentation to determine whether the trigger reflects a real ADE, and if so, a second reviewer (often a pharmacist or physician) confirms the finding and assigns a severity category using the NCC MERP Index for Categorizing Medication Errors, a nine-point scale (A through I) that separates events with no patient harm (categories A–D) from events that reached the patient and caused temporary harm, permanent harm, an intervention to sustain life, or death (categories E–I). Only categories E and above represent an actual ADE for surveillance-rate purposes; the lower categories are tracked separately as process and near-miss signals.

Because it is a retrospective sample, the trigger tool produces a rate — ADEs per 1,000 patient-days, or per 100 admissions, tracked over time — rather than catching every event as it happens or intervening on the individual patient it just found. That’s a deliberate tradeoff: a small, consistent sample reviewed the same way every cycle produces a trend line comparable across quarters and useful for showing whether an intervention actually moved the number, which a full census of every chart would be too resource-intensive to sustain.

Pharmacy-Based Screening: Concurrent, Not Retrospective

Pharmacy-based ADE screening runs the same trigger logic, but in near-real time and against individual patients rather than a monthly sample. Clinical decision support rules — built into the pharmacy information system or the EHR’s order-checking layer — watch for the same class of signals a trigger tool reviewer looks for after the fact: an order for naloxone or flumazenil, a critical lab result like an INR far above range or a glucose value in the hypoglycemic range, tied to a medication administered in the preceding hours. When a rule fires, it routes to a pharmacist’s work queue for same-shift review, rather than waiting for the next scheduled chart-sampling cycle.

The practical difference from the trigger tool matters for how a program uses each method: pharmacy-based screening is built to intervene on the specific patient it just flagged — hold the next dose, adjust a regimen, escalate to the prescriber — while the trigger tool’s retrospective sample is built to measure the institution’s overall ADE rate over time. A mature medication safety program runs both: concurrent screening for patient-level intervention, and periodic trigger-tool sampling for the program-level rate that shows whether systemic changes are working. Neither replaces voluntary reporting’s ability to capture a clinician’s own account of a near miss or a system failure — the three methods answer different questions, not the same question three ways.

Closing the Loop: From Detection to the Medication Safety Program

Detection is the input, not the output. A working medication safety program takes trigger-tool and pharmacy-screening findings through a defined path:

1. Aggregate the rate. Trigger-tool findings roll up into a quarterly ADE rate, tracked alongside pharmacy-screening intervention counts, and reported to the pharmacy and therapeutics (P&T) committee or a dedicated medication safety committee.
2. Route higher-severity events for root cause analysis. Any NCC MERP category F or above (harm requiring prolonged hospitalization, permanent harm, a life-sustaining intervention, or death) typically meets the threshold most hospitals use to trigger a formal review — often the same sentinel event or root-cause-analysis process used for other serious safety events, not a separate medication-specific track.
3. Turn recurring patterns into system-level fixes, not one-off counseling. A cluster of triggers pointing at the same drug class typically produces a concrete change: an order-set default, a hard-stop dose-range check, a designation on the hospital’s high-alert medication list, or — where look-alike naming is the root cause — adopting tall man lettering for that pair.
4. Re-measure. The same trigger-tool sample, run the same way on the same cadence, is what shows whether the rate actually moved after the fix — which is the reason the sampling method has to stay consistent cycle to cycle rather than being adjusted every time a new trigger seems more informative.

Trigger-tool surveillance is deliberately distinct from a Medication Use Evaluation (MUE): an MUE is a criteria-based retrospective review of how appropriately a specific drug or drug class was prescribed against a defined standard, not a harm-detection method — the two are often run by the same pharmacy department and feed the same P&T committee, but they answer different questions. An MUE asks whether use of a drug matched criteria; trigger-tool surveillance asks whether use of any drug caused harm. Programs commonly run both, and a trigger-tool finding pointing at inappropriate use of a specific agent is a reasonable trigger for opening an MUE on that agent.

Once a specific event is confirmed, causality assessment — whether the drug actually caused the reaction, versus the timing being coincidental — typically uses a structured instrument like the Naranjo algorithm, which is a separate step from the trigger-tool detection process described above; the trigger tool finds the case, the causality algorithm helps determine whether the drug was actually responsible before the case is coded and counted.

Frequently Asked Questions

What is the difference between a trigger tool and voluntary incident reporting?

Voluntary reporting depends on a clinician recognizing an event and choosing to file a report — it is passive and self-selected. A trigger tool actively searches a random sample of records for indirect evidence of harm regardless of whether anyone reported it, which is why trigger-tool sampling consistently finds substantially more adverse events than the same institution’s voluntary reports over the same period.

What is an ADE “trigger”?

A trigger is a specific order, medication, or lab value used as a clue that an adverse drug event may have occurred — for example, a naloxone order suggesting opioid oversedation, or a glucose value below roughly 50 mg/dL suggesting insulin-related hypoglycemia. Finding a trigger prompts a closer chart review; it is not itself proof that an ADE occurred.

How is ADE severity classified once one is confirmed?

Most programs use the NCC MERP Index for Categorizing Medication Errors, a nine-point scale (A–I) running from no error capacity through error without harm, error reaching the patient without harm, and then increasing degrees of actual harm up to death. Only the harm categories (roughly E and above) count as a confirmed ADE for rate-tracking purposes.

Is trigger-tool surveillance the same as reporting an adverse drug reaction to the FDA?

No. Trigger-tool surveillance is an internal hospital quality-and-safety measurement method. FDA adverse-event reporting is a separate, external regulatory channel — for the manufacturer- and facility-level postmarketing reporting obligations that channel involves, see 21 CFR 314.80. A confirmed ADE found through trigger-tool review may still warrant a separate FDA report; finding it through surveillance doesn’t substitute for filing one where required.

How often should a hospital run trigger-tool chart reviews?

There’s no single mandated frequency; the method is built around a small, consistent sample reviewed on a fixed recurring cadence rather than a full census. Programs commonly review a fixed number of randomly selected records on a set schedule (for example, twice a month) specifically so the resulting rate is comparable cycle to cycle — the exact sample size and interval should be set locally and then held constant, since changing them makes the trend line unreadable.

Related CASRAI Guides

Active surveillance is one input into the broader medication safety program alongside medication reconciliation at transitions of care, and downstream interventions it commonly drives include tall man lettering for look-alike drug names and formal sentinel event review for the highest-severity findings. For the analogous active-surveillance model in a different domain, see how hospitals run ongoing antimicrobial stewardship monitoring. For the broader hospital patient-safety context this surveillance function sits inside, see the patient safety pillar.