| Summary: | A research report submitted to the Faculty of Health Science in
partial fulfillment of the requirements for the degree of Master
of Science (MSc) in Epidemiology - Research Data Management
March, 2018. === Internal migration reconciliation involves the tracking of internal migrants to link their
places of origin and destination for each movement within a given Health and Demographic
Surveillance Area (HDSS) site. This involves several related activities adopted
to account for the time the person moves from one location to the other but, remaining
under the HDSS surveillance. This poses a major challenge for longitudinal studies particularly
Kintampo HDSS site, where manual data capture modality is still being used. For
most HDSS sites, all data related operations rely much on an individual unique personal
identifier which is issued to identify a resident member during registration. The identifier
enables the data system to keep track of resident members over an extended period of
time to enable an accurate estimation of the population under surveillance. Movement of
resident members within the geographical boundaries of the surveillance areas must be
tracked based on these personal identifiers. However, residents’ records may not be linked
to each other in the event of multiple movements when the personal identifier cannot be
recorded or is wrongly recorded. The effort in reconciling such cases of inconsistencies
resulting from internal migration involves the printing of mismatched records for field
supervisors to trace back to original locations to ascertain the identity of migrants. This
process is very expensive as a full-time field supervisor is required and the reconciliation
process is time demanding. In this project, we explored alternative automatic and
cheaper method of reconciling all categories of discrepancies that exist in the internal
migration datasets using the probabilistic record linking techniques.A theoretical foundation
for probabilistic record linkage technology was provided and the sequential order
was followed.The EM algorithm was used in the estimation of parameters. This research
report demonstrate clearly that the probabilistic record linkage frame work by Fellegi and
sunter is useful for HDSS internal migration reconciliation. The EM algorithm showed an
improved performance in terms of linked records compare to the probabilistic framework.
However, more work needs to done to explore other parameter estimation algorithm such
as the Frequency-based EM algorithm. Such results can be compared to the results in this
report. === LG2018
|
|---|
