November 1, 2023/ Last updated : August 9, 2024Chad HermanData Tables
DEPRIVATION
A deprivation record is only available for patients who have a valid postal code (recorded within the EMR and submitted to the central CPCSSN repository).
This table provides one measure of vulnerability indicator – the Pampalon index (link to Pampalon publication).
This is a neighborhood-level measure that uses the postal code to map to a dissemination area (DA), which is what the census data is linked to.
The Pampalon indicator provides a material index and a social index. Each index is a quintile. A patient who lives in a certain postal code has a probability of being in a certain quintile.
Since the mapping is not one-to-one between DAs and postal codes, it requires us to use probabilistic sampling and to consider the possibility of a single postal code mapping onto different SES levels (whether it be income quintiles or deprivation indices).
Within this table, you will see the probability distribution for each index (material and social) for each patient. There are several ways you can work with this:
1) Weighted average: For each patient take the weighted average. Essentially the probabilities within each quintile of the deprivation index will generate a number within the range of 1 to 5. This will transform the ordinal variable (1,2,3,4,5) onto a somewhat continuous scale (e.g., patient with 1.36 deprivation index).
2) Random assignment: Group patients according to their probability distribution and do a random assignment of patients to a quintile, by group. You will need to do the random assignment multiple times to get measures of variability. By creating multiple copies of the same data file through probabilistic sampling; and via multiple imputation methods (or its bootstrap analogues) should, on average, provide an estimate that closely resembles the underlying population-level quantity. Sampling variability will be high with small samples, and thus it can lead to estimation with greater variability. In contrast, sampling variability will be less with bigger samples.
