Measurement is about selecting what is to be measured, selecting and testing the
measure, collecting the data, validating the data, and using the data for improvement.
Validating the data is an important tool for understanding the quality of the quality data which is reliable, accurate, and defensible data that has been validated, for establishing the level of confidence decision makers can have in using data and in their implications for clinical practice.
An example of performance measurement is when an area for improvement in structure, process, and/or outcome is identified, new guidelines for patient care and safety are usually developed by the hospital using the data which had been selected, tested, collected, validated for patient care and safety improvement. This change process is normally managed by the hospital and include key stakeholders (e.g., clinicians) affected by the change.
An example of data validation when Health Information Management (HIM) / Medical Records (MR) practitioners who are generally specialised or experts is in disease coding may be involved, is when they provide advice in disease coding validation studies to determine staff training needs.
To ensure that a sample is valid when evaluating performance, it is critical to always determine an appropriate sample size ie. the number of subjects to choose, a procedure to ensure that your sample is representative of the population i.e the degree to which the subjects are similar to those in the intended use, and also determine the types of data to be used (administrative or clinical).
Well, you need to sample so as to try to get one that represents the population as closely as possible. This is because we rarely have enough time and money to look at the entire group of people that we are interested in (for example, the population of everyone attending a clinic at a particular hospital).
In trying to getting a valid sample, let us assume you had limited money, you cannot
study the target population as a whole. By all means do select a small sample size but when you choose a small sample size, there is always a higher risk of sampling error being present, for example when you could only choose only two patients out of the population of 30.
Unclear data definitions and inconsistent coding of data are reasons when data elements are found not to be the same. It is vital to have a list of codes with their definitions that you are going to be using throughout the collection of data. For example, if you are coding ward clerk as 1 and charge nurse as 2, it is important to ensure that you have used the same codes throughout the process of entering data into the dataset. In data validation, It is important to make corrective actions when inconsistent coding of data is found. However, if you do decide to change some data codes, it would be wise to note any changes as you progress.
The chart below characterises the process of data validation (by clicking on the chart below, it will open in a new tab of your current window, and by clicking on the image in this new tab, you can view a larger view of the chart).
