Nursing Home Antibiotic Stewardship: Build a Reviewable Data Packet

Start from the program’s question

CDC’s Core Elements for Nursing Homes, updated September 10, 2025, organizes stewardship around leadership, accountability, expertise, action, tracking, reporting and education. It recognizes that implementation varies with staffing and resources. Its linked companion checklist can support a baseline and later progress review.

Use those resources with the facility’s qualified clinical leadership. The packet below is an original evidence-design example; it does not recommend an antibiotic, dose, duration, diagnostic criterion or treatment change.

Keep program evidence and clinical evidence distinguishable

A file full of policy documents can demonstrate planned structure while omitting use. A utilization spreadsheet can demonstrate data preparation while omitting decisions. Index the different evidence so the committee can see which question each item answers.

A lower count can produce a higher rate

CDC’s checklist includes examples such as new antibiotic starts and days of therapy per 1,000 resident-days. The precise operational definitions still need qualified review; an order row is not automatically a new start or a day of therapy.

The following fictional calculation assumes the team has consistently defined and reconciled its local new-start events and resident-day denominator.

The raw count declines from 18 to 12. The rate rises from 20 to 30 because the denominator decreases more. Neither establishes better or worse prescribing. The calculation identifies why a committee should ask about population, definitions and clinical context before interpreting the chart.

Do not average site rates without checking their populations and weighting. Do not use people in the denominator of a rate labeled resident-days. Keep starts, orders, recorded administrations, days of therapy and residents distinct throughout the data path.

Reconcile an incomplete packet before interpreting it

Suppose period B omits an outside-pharmacy file, and imported corrections may have duplicated two order rows. The displayed rate is not yet a supported result. Mark the limitation, obtain the source and resolve the relationships before claiming the trend is final.

A useful discrepancy record identifies the missing source period, affected data fields, source owner and review disposition. A generic “data checked” checkbox cannot show whether the file was absent or deliberately excluded under a valid definition.

The packet should distinguish a current live view from the snapshot used at the meeting. If the omitted file arrives later, retain the original packet, explain the restatement and link the new reviewed result. A silent dashboard refresh can make a later decision appear to have been supported by evidence unavailable at the time.

Follow findings to the actual authorized decision

Create a separate record for a qualified reviewer’s finding. Preserve its source context, intended recipient and actual clinical response under the facility’s approved process. Acknowledgment can establish receipt without establishing a prescribing decision.

A decision may generate several follow-up tasks. Completing the physician response should not close a separate source-retrieval task, and an unanswered task should not erase an already documented decision. If a source changes later, make that new information visible for the appropriate professional review.

Keep individual review access limited to authorized participants. Regional leaders can often use a defined aggregate result and a secure reference to local evidence rather than receive resident-level details in every packet.

Use the packet to decide the next program action

The committee should be able to state what is known, what remains uncertain and what it is deciding. For the fictional period B example, the defensible immediate decision is to repair the source population and defer trend interpretation. That is a useful outcome of the meeting.

A software evaluation should ask whether staff can reproduce the calculation, identify the missing file and recover the evidence considered for a prior decision. It should also expose the work needed to gather outside sources. A claim of an automated stewardship dashboard does not answer those questions or establish clinical appropriateness.