Step 2.2 – Define your sample

Outcome:  By completing Step 2.2, you will have a ‘sampling frame’ for your full implementation. This is separate to any sampling frame for pilots. It is the population of patients you would like to include in your regular surveys when you have fully implemented AHPEQS. This means deciding what types of patients will be eligible for the survey, what proportion of these patients you will invite to answer the survey, and the minimum number of completed responses you will require before you can draw valid conclusions from (or report) the data.

Things to consider

This page lists the items that need to be considered in Step 2.2  to define your sample.

Defining your eligible patient population

  • What types of patients would you like to send AHPEQS to? To ensure your AHPEQS results represent your patients and provide a good overview of your services, you should consider the different service areas within your organisation, and the demographics of your patients.
  • Are you interested in all patients discharged from the services you provide, or only a subset? If you are interested in a subset, what are the attributes of the patients you want to be eligible to complete the survey? Depending on your objectives for using AHPEQS, you could restrict your sample in various ways (or a combination of these), for example:
    • by age
    • by hospital, clinic, department or ward
    • by type of service received
    • by length of stay 
  • How will you automate the process of identifying your survey sample? Does your existing patient administration system allow you to filter patients by the attributes you are interested in?


You may decide to exclude some patients from your survey population on ethical or pragmatic grounds. These exclusions will depend on your organisation and the type of care it is responsible for, but you might need to consider how to treat patients who:

  • Are having treatment which means they have repeated visits within a short time period (for example, chemotherapy)
  • Are likely to be surveyed using the same or a similar survey by two entities (for example, mental health patients who are given the Your Experience of Service [YES] questionnaire; patients of privately owned public hospital services where a state government and private hospital group might both ask the questions)
  • Are mothers who have experienced a stillbirth
  • Are experiencing temporary or permanent loss of mental capacity 
  • Visit the emergency department but are not admitted to the hospital 
  • Have diagnosis codes and types of health service which were excluded from the Commission’s pilots (only if you want to be able to claim validity and reliability without further testing).

Sampling from your eligible patient population

Whether you want to know about the experiences of patients across all demographics and all services or only a subset of these, you still have to decide what proportion of the eligible patient group you would like to survey. The decision will be affected by your chosen mode of administration (which determines the cost of each completed response).

You can either ask for responses from every discharged patient in your chosen population or you can develop a sampling frame to determine how a subset of those patients can be chosen.

Sample stratification

If you are not intending to send the survey to all eligible patients (and have decided to take a sample) you now need to decide whether to stratify the sample. This means dividing your eligible patient population into mutually exclusive groups based on a variable of interest (for example, age, admission type, department of admission) and then sampling from each of these groups. This reduces the problem of random sampling from a population where you may entirely miss respondents with some attribute of the variable you are interested in.

Sample size calculation

If you are not intending to survey all eligible patients, you may wish to draw conclusions from your sample and assert that these apply to your whole eligible population within a margin of error (using confidence intervals). To do this, you will need to set a minimum number of completed responses to include in your analysis, to achieve statistical power.

Given that AHPEQS is primarily aimed at local quality and safety improvement at this stage, the results can still be meaningfully used at local level to improve the service's responsiveness to patients' views, to implement feedback loops to ensure patients receive the appropriate response to any issue raised, and to point to issues for further investigation.

Previous | Next