The Scaling up analysis included in AccessMod allows one to find the most suitable locations for health facilities to scale up an existing network. The scaling-up methodology uses multiple variables to produce a weighted composite map of suitability with an additive model approach.
The first step is to exclude from the analysis all the cells part of the exclusion zones. Once done, a physical accessibility analysis is performed on all the remaining cells presenting the highest value in the suitability distribution layer.
The site (cell) covering the largest population is the one selected for location of a new health facility. The type of facility placed on this site is then decided based on the content of the new health facility information table. Only one facility is therefore created at the location that maximizes both the suitability and population coverage. The process then goes to the next iteration to locate another new facility. The algorithm is highly dependent on the exclusion and suitability factors set by the user as well as the quality of the input data.
Exclusion areas
The exclusion areas are defined by the user by inputting any number of vector (points, polygons) or raster data sets. Typical exclusion zones can represent military areas, wetlands, private lands, etc. The user can also define a buffer (in kilometers) outside of the exclusion area that will be taken into consideration for the exclusion. This is especially useful when the user wants to avoid locating a new facility less than, say, 5km away of existing facilities (a constraint in Burkina Faso for example).
Suitability outside the exclusions areas
The way one defines suitability factors and one combines them has been carefully devised to enable maximum flexibility, so that the users can define a range of suitability models, from simple to very complex ones, in order to satisfy specific national/regional constraints.
The decision on the factors being included follows the consultation process that took place in summer 2015 when we gathered needs and ideas (from health staff from WHO and other organizations) on how to compute site suitability for the location of new health facilities.
Once the exclusion zone has been applied, non-excluded areas must be prioritized following the user-defined combination of suitability factors. This model can include any combination of the following factors to come up with a suitability index for each non-excluded cell in the study area:
- Sum of population within a used-defined radius (in kilometers) (for example, this allows us to target those areas with highest population)
- Euclidean distance from features of an input data set (e.g. roads, existing facilities) (for example, this allows us to prioritize new facilities close to roads or close to existing network of particular facilities such as those providing Emergency Obstetric and Neonatal care (EmONC))
- Travel time to/from features of an input data set (this takes into account the travel time, rather than the Euclidean distance, and might be more realistic depending on the data sets and aim of the scale up exercise)
- Generic priority map (this takes any raster data set as input and use its cell values as relative priorities. This gives a high degree of flexibility, as scaling up priority can come from any other external tool, such as one that would prioritize regions based on economic constraints)
For each of the suitability factors, the user must define whether higher values are more suitable or less suitable. By default, the relative weights among the model factors is equal, but the user can change these relative weights, giving more importance to one or several of the factors.
The additive model[1] then combines the factors as defined in part (3) of the equation below, and generates the raster suitability map.
The standardization procedure is taken from : http://support.esri.com/cn/knowledgebase/techarticles/detail/30961
Process
Once the model is set and validated, AccessMod computes each exclusion rule and creates a map of potential candidates. Only those candidates are considered in the next steps. The algorithm then processes in an iterative manner through the following steps:
- The algorithm rescales all factors defined in the suitability analysis, apply the user-defined relative weighting among factors, and combines the rescaled versions into one suitability map.
- The algorithm extracts all the locations presenting the highest suitability value and launches an accessibility analysis for each of them based on the maximum travel time set in the analysis settings.
- The resulting accessibility coverage (number of potential patients) for each site is then used to select the one site that presents the highest coverage. This site is where a new health facility will be located.
The accessibility coverage value for the selected site is compared against the input capacity table, so that the appropriate facility type, label and a maximum capacity are attributed to the candidate(s). This step can take a lot of time, depending on the number of locations to evaluate (i.e., if the suitability models is not discriminant enough, and that many locations have the same suitability).
- The location with the best coverage (i.e., where one new facility is found) is kept, and the polygon representing the catchment is stored.
- The next iteration is launched for the next facility location to be found.
End of the process
The process is finished when any one of the following user-defined limits is reached:
- The percentage of the total population covered by the scaled up facility network.
- The maximum computing limit set by the user.
- number of desired new facilities.
[1] Eastman R. (1999) Multi-criteria evaluation and GIS. Chapter 35 In: Longley PA, Goodchild MF, Maguire DJ, Rhind DW (eds) Geographical information systems, Wiley, New York. pp493-502.