Event Summary Group Algorithm

ControlCheck's automatic reconciliation and discrepancy identification rely on accurate determinations of event summaries. An event summary can be defined as a group of events that share a common patient, generic medication, and time frame (up to 12 hours apart). For the purposes of event grouping, all events given to the algorithm are assumed to have the same patient and the same generic medication.

The goal of event grouping is, given a group of events for the same patient and the same generic over an arbitrary period of time, to group or cluster the events into distinct event summaries that are close together and reflect the actual workflow of hospital employees. The degree to which two events are close to each other is primarily based on how close the events are to each other in time (i.e. when they occurred). The type and amount of medication for an event are also considered.

Definitions, Assumptions, and Basic Principles

Definitions

Event

• Types
  • Dispense-like
    • When a medication is taken or dispensed from some repository or store so that some or all of it can be given or administered to a patient
  • Admin-like
    • When a medication is given or administered to a patient
    • When a medication is wasted (i.e. thrown out) or returned to its repository or store
• Amount
  • The amount of medication in the event, in any units
  • Dispense-like events are negative
    • 5 mg dispense = -5 mg
  • Admin-like events are positive
    • 3 mL administration = +3 mL
• Time
  • The time the event occurred

Event Summary (or Bubble)

A grouping of any number of consecutive events.

• Value
  • The sum of the amounts of the events
• State
  • Closed/Frozen: when the value is zero
  • Open/Unfrozen: when the value is non-zero

Assumptions

• Events closer in time to each other are more likely to be in the same event summary. 
• Discrepancies are rare; most event summaries should be closed.

Basic Principles

• Examine groups of events that are close to each other before examining groups of events that are farther from each other.
• If at any point a group of events form a closed event summary, then all those events must be in the same summary at the end.
• Event summaries can only expand from both ends (events before the first event in the summary and after the last event) and absorb events, if and only if the expanded event summary remains closed.

Feature Enhancements to the Algorithm

Strict Event Type Rule

The strict event type rule enforces an additional step in the grouping algorithm that checks the type of event before grouping it with another. We group events together based on a linear timeline, which means that sometimes events that are closest together in time will be grouped together, even though they are "out of order" and the grouping would not make sense for a real-life case. ControlCheck has solved this nuance in the workflow by creating the Strict Event Type Rule, a hospital-specific setting, in minutes, that says:

No event summary closed by ControlCheck can start with an admin-like event if the distance between the initial admin-like event to the first dispense-like event is greater than the given strict event type cutoff value.

Put more simply, a closed-ControlCheck summary cannot begin with an admin-like event (admin, waste, or return) if the time distance between the first admin-like event in the summary and the first dispense in the summary are further apart in time than the strict event type cutoff value, in mins.

For example, let's say you have the event order: Dispense, Admin, Waste; and this order is repeated three times a day because the patient is receiving three scheduled doses a day. ControlCheck will receive all nine events and plot them on a linear timeline; the events closest together in time will be grouped together until they form a closed bubble (equal out to zero). If the first dispense is 30 mins apart in time from the first admin and waste, and the second dispense is only 10 mins away from the first admin and waste events, then the algorithm will think the events closest to each other go together. This then causes a domino effect on the rest of the day's events and you would end up with the following grouping results in ControlCheck for all nine events:

• Case 1 - Dispense1 (-2mg) = 2mg missing
• Case 2 - Admin1 (1mg), Waste1 (1mg), Dispense2 (-2mg) = 0mg missing but out of order sequence
• Case 3 - Admin2 (1mg, Waste2 (1mg), Dispense3 (-2mg) = 0mg missing but out of order sequence
• Case 4 - Admin3 (1mg), Waste3 (1mg) = 2mg missing

This grouping would be incorrect and cause a confusing and less than desirable user experience when auditing in ControlCheck. The strict event type rule allows ControlCheck to combat these common, but disruptive, factors in the workflow for users interacting with controlled substances.

Using our example above with the nine events for the same patient, med, and day, if strict event type rule was set to 10 mins, then that would mean that ControlCheck will not group those admins and dispenses out of order as we did in the first scenario. We instead would enforce a rule that would prevent the first waste and second dispense from ending up on the same case summary, as long as they were more than 10 minutes apart. We would then see the below grouping for the same nine events:

• Case 1 - Dispense1 (-2mg), Admin1 (1mg), Waste1 (1mg) = 0mg missing
• Case 2 - Dispense2 (-2mg), Admin2 (1mg), Waste2 (1mg) = 0mg missing
• Case 3 - Dispense3 (-2mg), Admin3 (1mg), Waste3 (1mg) = 0mg missing

The strict event type rule allows the ControlCheck grouping algorithm to make more informed decisions in cases like the above where events from separate cases happen close together in time. The purpose of the cutoff value, in minutes, is to prevent a caveat to this rule and allow the algorithm to account for slight nuances, such as back-charting. Back-charting has the potential to cause disruptions to correctly grouped event summaries. Let's say a nurse charts an admin at 0930 but she actually didn't dispense it until 950 and actually gave it at 1000. At the end of her shift, she documents the admin at

0930 which now creates a situation where the admin event will appear to happen before the dispense event when ControlCheck receives the data. Using the cutoff value, the algorithm will continue to group "out of order" sequences on the same case if their time difference is less than the cutoff value so that a back-charted admin only off by 5 mins is still closed by ControlCheck and does not appear as a false variance on the audit table. ControlCheck still has all of the documentation to close the case, the events are just out of order; but with the cutoff value, ControlCheck can make an assumption that they still should get grouped together and close out.

To learn more about the event grouping algorithm that ControlCheck uses, please feel free to reach out to our Help Desk, or send us an email to product-ControlCheck@bluesight.com.