Section 1: Defining and Selecting Key Performance Indicators

• Number of medication orders verified today: 850.
• Number of allergy alerts that fired: 125.
• Current number of doses in the verification queue: 42.

• Pharmacist verification rate: 35 orders/hour (Target: 30 orders/hour).
• Allergy alert override rate: 88% (Target: < 90%).
• ADC stockout rate: 3.5% (Benchmark: < 2%).

• S: Highly specific – targets STATs, ED only, verification-to-admin.
• M: Measurable via timestamps in the EHR (order entry, verification, MAR admin). The metric is median time in minutes.
• A: A 5-minute (20%) reduction is challenging but likely achievable with process improvements.
• R: Directly relevant to patient safety and quality of care in the ED, a key hospital priority.
• T: Clear deadline – “by the end of Q3.”

• S: Extremely specific – focuses on a single, high-risk DDI, not all alerts.
• M: Measurable by querying the CDS database for the specific alert type and tracking the ratio of “override” actions to total firings.
• A: Achievable, as this specific interaction is almost always clinically significant, suggesting the current override rate is too high due to alert fatigue. A 25% reduction is a solid target.
• R: Highly relevant to preventing a well-known, potentially catastrophic ADE.
• T: Clear timeframe – “over the next 6 months.”

• S: Specific to a single nursing unit (M5) and defines “compliance” as both patient and med scan.
• M: Measurable via BCMA system logs. The metric is a percentage.
• A: Increasing from 92% to 98% is a significant stretch goal but achievable with targeted education and workflow analysis. 100% is not a realistic target.
• R: Directly relevant to the core patient safety function of BCMA – preventing wrong patient and wrong med errors.
• T: Time-bound with a 90-day target.

• S: Specific drug (IV acetaminophen) and specific intervention (therapeutic interchange).
• M: Measurable via purchasing data from the drug wholesaler (cost per vial, total vials purchased).
• A: Achievable, as IV acetaminophen is a common target for cost-saving initiatives with readily available, cheaper alternatives for many use cases.
• R: Directly relevant to the pharmacy department’s budget and the hospital’s financial health.
• T: Time-bound by the end of the fiscal year.

Measures the time from the initial order of a STAT medication to its administration to the patient. This KPI is a critical measure of emergency response capability.

Numerator: Sum of time elapsed (in minutes) for each qualifying event.

Denominator: Total number of qualifying STAT medication administration events.

Key Decision Points:

• Start Time: Order entry time? Or order verification time? “Order-to-Admin” is the gold standard, but “Verify-to-Admin” isolates pharmacy/nursing time. Be consistent.
• End Time: MAR documentation time.

Sources: EHR order tables (order timestamp), verification logs (verify timestamp), Medication Administration Record (MAR) (admin timestamp).

Pitfalls:

• Downtime: How are verbal STAT orders from a downtime event captured and timed? You must have a policy for excluding these.
• Nursing Documentation Lag: A nurse might give the med immediately but not document it for 10 minutes, artificially inflating the TAT. This is a major data integrity issue.
• Outliers: One single event where a patient was off the floor for a procedure can dramatically skew the “average” (mean) TAT. Using the median is statistically more robust and preferred.

Measures the adherence to the standard BCMA workflow, which is foundational to closed-loop medication safety. Can be broken down into patient scanning, med scanning, or an overall rate.

Numerator: Number of administrations where the required scan(s) were successfully performed.

Denominator: Total number of administrations requiring a barcode scan.

$$\text{Overall Compliance %} = \frac{(\text{Admins with Patient Scan AND Med Scan})}{(\text{Total Barcode-Required Admins})} \times 100$$

Sources: BCMA transaction logs within the EHR.

Pitfalls:

• Defining “Required”: Your denominator must exclude non-barcoded items (e.g., nebulizers, some ointments). This requires careful configuration of the medication records (NDC/formulary build).
• Workflow “Workarounds”: Are nurses scanning patient wristbands taped to the computer? Or scanning medication barcodes from a printed list instead of the unit-dose package? Your data won’t show this, requiring direct observation (Gemba walks) to validate.

Measures the percentage of critical alerts (e.g., severe DDI, anaphylactic allergy) that are bypassed by clinicians. This is a direct measure of alert fatigue and potential safety gaps.

Numerator: Number of high-severity alerts overridden by a clinician.

Denominator: Total number of high-severity alerts fired.

Must be broken down by specific alert type (e.g., Allergy vs DDI) and even by specific pairs (e.g., Warfarin-Bactrim) to be actionable.

Sources: Clinical Decision Support (CDS) system logs.

Pitfalls:

• Defining “High-Severity”: Your IT and P&T committee must clearly define and classify which alerts fall into this category. This is a critical governance step.
• Override Reasons: The quality of this KPI is massively enhanced if you can analyze the free-text or structured override reasons selected by the clinician. This tells you the “why” behind the override.
• Noise vs. Signal: A 95% override rate for a poorly designed, clinically irrelevant alert is expected. A 10% override rate for a critical anaphylaxis alert is a five-alarm fire. Context is everything.

The ultimate safety outcome metric. Measures the incidence of patient harm due to medication use that was considered preventable. This is a very complex, often manually abstracted KPI.

Numerator: Number of ADEs determined to be preventable (by clinical review).

Denominator: Patient-days (most common), Doses Administered, or Patients Discharged.

$$\text{Rate} = \frac{(\text{Number of Preventable ADEs})}{(\text{Total Patient Days})} \times 1000$$

Sources: Voluntary safety event reporting systems, chart review, trigger tools (e.g., searching the MAR for naloxone or diphenhydramine administrations to find potential opioid or allergy-related events).

Pitfalls:

• Underreporting: Voluntary reporting captures only a tiny fraction of actual events. This KPI is highly dependent on the organization’s safety culture.
• Subjectivity: Determining “preventability” requires expert clinical judgment and can be inconsistent between reviewers.
• Labor-Intensive: Reliable ADE data collection often requires manual chart review by pharmacists or nurses, which is expensive and time-consuming. This is why leading indicators (like alert overrides) are often used as surrogates.

Measures the average time an order sits in a pharmacist’s queue before it is verified. It’s a key indicator of pharmacy workload and potential delays in care.

Numerator: Sum of time elapsed (in minutes) from order entry to verification for all orders.

Denominator: Total number of orders verified.

This should be tracked by priority (STAT, ASAP, Routine) and time of day to identify bottlenecks.

Sources: EHR order and verification logs.

Pitfalls:

• Queue Segmentation: A single average is meaningless. You must be able to filter by the specific queue (e.g., Central Pharmacy, Oncology, Peds).
• “Stop the Clock” Events: If an order is “pended” while waiting for a pharmacist to call a provider for clarification, should that time be excluded? Your policy on this can significantly alter the data.

Measures the frequency with which a needed medication is not available in the ADC, forcing a nurse to call the pharmacy and causing a delay.

Numerator: Number of times a medication was requested from an ADC but was not available (stockout event).

Denominator: Total number of medication requests (removals) from ADCs.

$$\text{Stockout \%} = \frac{(\text{Number of Stockout Events})}{(\text{Total ADC Removals})} \times 100$$

Sources: ADC system server logs.

Pitfalls:

• Defining “Stockout”: Is it a true stockout if the drug is available in a different pocket of the same machine? The system configuration is key.
• “Ghost” Stockouts: A nurse might not find the drug and report a stockout, but it was actually there. This requires cross-referencing with inventory counts to validate.
• PAR Level Optimization: A low stockout rate isn’t always good if it’s achieved by setting PAR levels excessively high, leading to high inventory costs and expired medications. This KPI must be balanced with financial KPIs.

A normalized measure of drug cost that accounts for hospital volume and patient acuity. This is a high-level KPI used by finance and the C-suite.

Numerator: Total drug acquisition cost over a period (e.g., one month).

Denominator: Total Adjusted Patient Days for the same period.

(Adjusted Patient Days = Patient Days × (Gross Patient Revenue / Gross Inpatient Revenue) to account for outpatient activity).

Sources: Drug wholesaler purchasing records (numerator), hospital finance/admissions department (denominator).

Pitfalls:

• Acquisition Cost vs. WAC: You must use the true acquisition cost after all rebates and discounts (e.g., 340B pricing) to be accurate.
• Case Mix Index (CMI): A rising drug spend might be entirely appropriate if the hospital is treating sicker patients (a higher CMI). A more advanced version of this KPI is Drug Spend per CMI-Adjusted Patient Day.

Measures the percentage of medication orders that are for formulary agents, reflecting the effectiveness of P&T committee decisions and therapeutic interchange programs.

Numerator: Number of orders for formulary medications.

Denominator: Total number of medication orders.

Often more useful when measured for specific drug classes with high-cost non-formulary options (e.g., PPIs, statins, biologics).

Sources: EHR ordering data, Pharmacy formulary database.

Pitfalls:

• Formulary Status Build: This KPI is only as good as the formulary status field in your drug database. An inaccurate build will produce garbage data.
• “Justified” Non-Formulary Use: A low compliance rate isn’t always bad if the non-formulary use is clinically justified (e.g., allergies, failure of formulary agents). This requires the ability to audit the “why” behind non-formulary orders.

Measures the adoption of standardized order sets for common clinical scenarios (e.g., Sepsis, Community-Acquired Pneumonia).

Numerator: Number of qualifying patients with the specific order set initiated.

Denominator: Total number of patients with the qualifying diagnosis (e.g., ICD-10 code for CAP).

Sources: EHR order tables, patient diagnosis tables (from billing or problem list).

Pitfalls:

• Diagnosis Lag: The official billing diagnosis might not be assigned until discharge. Using admission diagnosis or problem list entries may be necessary for more real-time data, but can be less accurate.
• Order Set Modification: What if a provider uses the order set but deletes half of the recommended orders? A more advanced version of this KPI measures the “adherence rate” to the core components within the order set.

A cousin to the override rate, this measures the positive action. It’s the percentage of time a clinician accepts the guidance offered by a CDS alert.

Numerator: Number of times a specific CDS alert’s recommendation was accepted (e.g., provider changed the order based on a renal dosing suggestion).

Denominator: Total number of times that specific CDS alert fired.

Sources: CDS system logs, EHR order modification logs.

Pitfalls:

• Requires Sophisticated Logging: The system must be able to link the alert firing to the subsequent action (or inaction) by the user. This is not always a standard feature and may require custom build.
• The “Informational” Alert: For alerts that are purely informational and don’t suggest a specific action, this KPI is not applicable. It is most useful for alerts that recommend a specific change (e.g., “Change dose to X,” “Add lab order Y”).