The concept of “Inference” has been introduced into the drug serialization and ePedigree community.

What is ”inference” and how does it pertain to the compliance ?

“Inference” is a managed risk business process, not a regulatory process. A regulatory or enforcement body can not effectively legislate, monitor or regulate “inference.”

Let’s explore inference from a practical perspective. Inference is related to confidence.

I am going to put 48 bottles of serialized drugs in a case. Once I seal the case, I am going to infer the serial numbers are intact unless there is evidence the seal has been breached. That is a ’sound’ inference.

Let’s assume, I take the case of 48 bottles and I leave it on a shipping dock for a week, or I give it to a trucking company that takes 3 weeks to get the case 300 miles away. Inference may not be a good practice. There are too many points of vulnerability to confidently use inference.

Another scenario is that the original case has been opened at least once and ‘re-sealed.’ Not a good inference scenario.

Now, let’s move further down the supply chain. The original cases are opened at the wholesaler or retailer and they begin to building ‘totes’ with mixed product.  At this point, the inference is that all 48 bottles from the manufacturer are correctly serialized and as the drugs are inserted into the ‘mixed tote’ the wholesaler/retailer data captures the serial number on the item level.

This is called ‘picking.’

We know there are mistakes in picking.  Most companies have audits to catch mistakes.

The point that should be derived is that we have a pre-existing set of business rules in every organization that pre-defines financial risk associated with inventory shrink resulting from picking errors. Serialization does not change the risk formula, but serialization should reduce the errors.

Here is a picking scenario.

Today, an order from retailer location ‘111′ is for 4 bottles of item ‘123456.’ Modern distribution centers use some sort of data capture such as barcode (UPC) to ensure a high rate of accuracy by picking to the UPC rather than a name or a brand.

Now it’s not too practical to have picking of item level drugs to a specific serial number. That would require a human or a machine to sort through 48 bottles to find the unique serial number to pick, but it is practical to ensure the picking “system” contains the range of serial numbers for the specific drug that is acceptable to pick.

Once a case is opened for item level picking, the order picking system immediately creates a ‘data bucket’ of the 48 item level serial numbers and any quantity of pick must be validated against the range of 48 serial numbers.

There is basically no inference.

Now comes the ”tote.”

There is a tote with mixed product. As items are picked to the tote, it is presumed or inferred the product remains in the tote. Therefore there is an element of inference. Until the mixed tote is sealed there is always the possibility that an item can be removed or replaced, but this is relative to the security measures in place.

If this is an event that occurs at an unacceptable frequency, there various tools available to identify anomalies.  One tool is the broadening of CCTV capture stamped to time and location, as well as in-line scales to detect weight variations.

But there is no state of perfection where there isn’t some level of inference, but this inference is a business practice and risk event and not a legislative event.

In many respects the board oversteps it’s role when it entertains “inference.”

Assume the NTSB or the FAA agrees to some definition of inference of parts installed on an aircraft in association to an aircraft identifier.

Would you like to be on the aircraft that is missing a part when it left the factory because of an acceptable inference ? 

Now, if the airline removes a part, and doesn’t replace the part on the aircraft, that is a business practice and risk event. If the maintenance logs are not maintained accurately, that’s a business and risk event.

Below is the definition supplied by the California Board of Pharmacy:

“Inference” is a term used by some in the industry and by the board to refer to a vaied category of possible practices with regard to (temporarily) substituting aggregate-level (e.g., pallet, case, tote) tracking and validation for item-level tracking and validation, based on the limitations of, particularly, non-line-of-sight tracking technologies, and of distribution system partners.

The working definition of “inference” as it applies to serialization employed by the board is that it refers to the ability to “infer,” based on tracking and validation of a unique identifier attached to an aggregate package (e.g., pallet, case, tote) which has a hierarchical relationship with unique identifiers attached to each of the items/packages contained within the aggregate package – when appropriately combined with circumstances engendering confidence that an aggregate package identifier – that the items/packages within the aggregate package are what they are represented to be on the pedigree data, without individually validating (e.g., scanning) each item/package (immediately).

“Inference” does not appear in the pedigree law, and the the law does not presently permit the use of or reliance on “inference” in place of required item-level tracking or validation of pedigree data. However, the board has stated a willingness to hear input from industry regarding a perceived need for or possibility of “inference,” and has dedicated time during the E-Pedigree Workgroup meeting(s) to “inference.”

The board is exploring, and is interested in hearing from industry, whether and/or to what extent these are real or perceived issues to be addressed regarding “inference,” what solution(s) to these issues might exist or might be created, and whether these solutions would require some statutory change, promulgation of rules, or are a matter of enforcement discretion.