The PScore module of Statistics for Data Analysis implements the Propensity Score Matching technique , considered as an alternative to classic multivariate analyzes, when you need to analyze data from non-randomized studies .
In this video, we show how much this technique can help you by allowing you to explore the effectiveness of a treatment in a large number of patient subgroups, and obtain more representative data quickly and easily .
In general, we know that clinical trials are considered the gold standard, i.e. the most accurate diagnostic method for confirming a given doubt.
However, how often have you been unable to organise an experimental study, but have a lot of data available from observational studies?
It is often not possible to organise an experimental study for ethical reasons or because it is costly and time-consuming.
These are the cases where you can use PScore, the Premium Module of the Statistics for Data Analysis solution powered by SPSS.
The PScore, in fact, implements the Propensity Score Matching technique, which is based on data from real cases (health registers, surveys, medical records,...) obtaining results similar to those of randomised trials.
It is therefore very useful to explore the effectiveness of a treatment in a large number of patient subgroups and to achieve greater generalisability (compared to a trial) at a lower cost and time.
Statistics for Data Analysis, allows you to access the Propensity Score matching method easily, quickly and without the need for any specific programming language, with the PScore module, accessible and self-installing thanks to the LaunchBox.
After installation the Propensity Score technique is available in the menu Analyze/ Analyze Add-On/ Propensity Score Matching
After selecting the fields for the different functions and matching preferences, you can click on "Execute":
On the Source Editor you will find the new fields created by the Propensity Score matching process
At this point the PScore applied propensity score matching is a statistical technique in which a 'treated' case is matched with one or more 'control' cases based on the closeness/equality of the propensity score values.
This matching helps to recreate a "kind of random allocation" in observational studies, reducing the bias linked to the lack of randomisation.
In the Output environment, you will also find matching diagnostic metrics, including:
A contingency and summary table
The distribution of absolute standard differences
The distribution of standardised differences with and without matching
Propensity score distributions for treatment category combinations and matching for common support analysis
To summarise, there are only a few steps to apply the Propensity Score Matching technique and thanks to the PScore Module of Statistics for Data Analysis it is very easy to:
Identify the data (large sample size)
Define treatment/control and outcome
Open the PScore from the Menu Analize and select the covariates you are interested in
Estimate the Propensity Score using the dichotomous variable treatment as a dependent
Use the Propensity Score to 'match' groups.
Evaluate the success of 'matching' with the diagnostic methods that automatically appear in the Output.
Perform outcome analysis on the small sample size because it is adjusted by Propensity Score
For further information