Planning an audit and checking improvement

An audit compares what happened with a defined standard. In a clinical setting, qualified leaders must agree the relevant evidence-based standard and authorize the work. This course teaches a method using fictional administrative data; it does not authorize access to health records or changes to treatment.

Example: a team’s agreed procedure says every training request should receive an acknowledgement within two working days. An audit asks how often eligible requests met that standard during a defined period. A satisfaction survey asks a different question about experience. Both may help, but they are not interchangeable.

Practice: Write a fictional process, its standard and the question an audit would answer.

An audit can make a process gap visible, guide discussion and provide a baseline for later comparison. It does not automatically explain why a gap occurred, establish blame or prove a change caused an improvement. Involve people who use and deliver the process when interpreting results.

Example: 12 of 20 requests were acknowledged on time. The team first checks whether dates were captured consistently and whether the sample excluded weekends as intended. Only then does it explore causes such as an unclear owner or a missing notification. A percentage without context can mislead.

Practice: List two possible explanations for missing data that do not assume poor effort.

Use six practical stages: define the problem; agree criteria and standards; plan the authorized data collection; collect and analyze consistently; agree and test changes; then repeat measurement and decide the next action. Different organizations may name or group these stages differently.

Example: the team notices delayed acknowledgements, defines the two-working-day rule, selects a date range, measures performance, tests a named inbox owner and repeats the same measure. Stopping after publishing the first score leaves the improvement cycle unfinished. Record decisions so another person can understand how the result was produced.

Practice: Draw the six stages and identify who would authorize each part in a real project.

State the aim, setting, dates, eligible cases, exclusions, data source, collector, storage route and operational definition of success. Define a numerator for cases meeting the criterion and a denominator for eligible cases. Decide how missing data will be reported before collection. Sampling choices affect what conclusions can be drawn.

Example: include all 20 fictional requests received in one specified week; count those acknowledged within two working days. Do not quietly remove late requests to improve the result. For real clinical work, obtain local authorization and privacy advice before accessing data, and use the minimum necessary information.

Practice: Draft a plan for a fictional process with a numerator, denominator and missing-data rule.

In the worked example, 12 of 20 eligible requests met the standard: 12 ÷ 20 × 100 = 60%. If the agreed target is 90%, the difference is 30 percentage points. That is not the same as saying performance fell by 30 percent. Report the counts, period and missing data alongside the percentage.

If five records lack usable timestamps, show that limitation. Do not turn missing evidence into a confirmed success. A small or unrepresentative sample should not be described as the performance of every site or shift. Discuss the uncertainty before choosing a response.

Practice: Calculate 18/24 as a percentage and the percentage-point gap from a 90% target.

Choose a change related to a plausible cause, name an owner, agree a review date and measure possible unintended effects. A small authorized test can show whether an idea is workable before wider adoption. Healthcare Excellence Canada provides improvement resources to support structured testing and measurement.

Example: the team tests a designated inbox owner for one week, measures timely acknowledgement again and asks whether the new responsibility displaced other work. If 18 of 20 requests now meet the standard, the observed score is 90%; the team still considers sample differences and sustainability. Keep useful changes under review rather than declaring permanent success after one test.

Practice: Write a change, owner, repeat-measure date and one unintended effect to watch.