Abstract
Background
A systematic review was conducted to examine the extent, range and nature of published research evidence using methods to attribute data reported at aggregate levels to small areas, taken from both the health sciences and geography literatures.
Methods
Four electronic bibliographic health databases were searched (MEDLINE, Embase, PsychINFO and CINAHL) and one human geography database (GEOBASE). We reviewed titles, abstracts and then full-text articles for their relevance to this review, based on pre-determined exclusion and inclusion criteria. All attribution methods were identified, reviewed in tables and assessed against a set of criteria to robustly compare their applicability to health data.
Results
Of 634 titles identified, 84 articles met the inclusion criteria. From these studies, we identified four broad categories of attribution methods: spatial interpolation, dasymetric mapping, regression methods and spatial microsimulation. Spatial interpolation and regression methods were the most utilised in the health science literature. Both groups of methods allow adjustments for the underlying demographic of the populations that are being disaggregated. In comparison, dasymetic mapping is the most utilised spatial attribution method in the geography literature. These methods did not adjust for the underlying demographic of the populations.
Conclusion
The type of spatial attribution method that should be applied to health data will depend on the health measures used. The prevalence of certain health conditions are much more dependent on the sociodemographic of a population than others. For the former, adjusting population distributions for underlying demographic factors is important. Where the prevalence of health conditions is more equally distributed across a population, we may wish to prioritise other criteria when selecting an attribution method.
A systematic review was conducted to examine the extent, range and nature of published research evidence using methods to attribute data reported at aggregate levels to small areas, taken from both the health sciences and geography literatures.
Methods
Four electronic bibliographic health databases were searched (MEDLINE, Embase, PsychINFO and CINAHL) and one human geography database (GEOBASE). We reviewed titles, abstracts and then full-text articles for their relevance to this review, based on pre-determined exclusion and inclusion criteria. All attribution methods were identified, reviewed in tables and assessed against a set of criteria to robustly compare their applicability to health data.
Results
Of 634 titles identified, 84 articles met the inclusion criteria. From these studies, we identified four broad categories of attribution methods: spatial interpolation, dasymetric mapping, regression methods and spatial microsimulation. Spatial interpolation and regression methods were the most utilised in the health science literature. Both groups of methods allow adjustments for the underlying demographic of the populations that are being disaggregated. In comparison, dasymetic mapping is the most utilised spatial attribution method in the geography literature. These methods did not adjust for the underlying demographic of the populations.
Conclusion
The type of spatial attribution method that should be applied to health data will depend on the health measures used. The prevalence of certain health conditions are much more dependent on the sociodemographic of a population than others. For the former, adjusting population distributions for underlying demographic factors is important. Where the prevalence of health conditions is more equally distributed across a population, we may wish to prioritise other criteria when selecting an attribution method.
| Original language | English |
|---|---|
| Article number | 117752 |
| Journal | Social Science & Medicine |
| Early online date | 23 Jan 2025 |
| DOIs | |
| Publication status | E-pub ahead of print - 23 Jan 2025 |