5 Healthsites.io: The Global Healthsites Mapping Project
55
5.2 Healthsite.io Approach
5.2.1 Datasets Integration
Due to its open data license, its large number of entities, its worldwide coverage,
and its large community, OSM was chosen as the main dataset of the project. In
addition, several databases from trusted partners were chosen. All this data gathered
represented more than 150,000 health localities worldwide. In order to integrate
all these datasets with different structures and values, a data model, based on the
Entity–Attribute–Value model, and inspired by the OSM data model was chosen. The
OSM data model enables users to store information about anything. A FullTextSearch
index has been implemented to enable searching for textual data. However, this
flexibility of storing information represented a risk potentially hampering the goal of
having a curated database with easily accessible information. In order to address this
issue, 15 core attributes (Table 5.1) that are relevant for both the public and health
professionals were defined by the International Hospital Federation. For most of these
core attributes, defined values were set to enable comparisons. Indeed, an essential
question such as “What type of health facilities is it” can vary greatly according to
the national or organizational classification system.
Each time a record is created in Healthsites.io, a globally unique identifier is
created and assigned to that record. This is specific to the Healthsites.io database and
is used to provide a canonical point of reference for that record.
The upstream id is used to create a back reference to the original source data (e.g.,
from a national healthsites dataset).
5.2.2 Validation Process
In order to enhance the reliability of health facilities data, several validation processes
have been planned to be added to the project. There is an automatic verification
ranging from a simple: “email address should look like an email address” to more
complex which even rely on external services like: “check if the Locality address is
similar to results returned by external geocoding services” or “check if the telephone
number is correct by manually calling the number and verifying”.
There will also be a validation process based on user reputation (Fig. 5.1) to assess
the reliability of data through a Locality Validity Index (LVI). This reputation-based
process consists of four steps. The first step takes place during data integration.
Depending on source of the data, the LVI obtains a score ranging from 0 to 10; 0
being for data added by a new user, 10 for data from a community trusted user, and 5
if it is a batch of data coming from a Ministry of Health or healthcare organizations.
The reliability of the user is based on the monitoring of his activity on the platform
and the crosscheck of his activity by other users. Second, once the data is displayed
on the map of Healthsite.io, any user can complete missing attributes data and modify
55
5.2 Healthsite.io Approach
5.2.1 Datasets Integration
Due to its open data license, its large number of entities, its worldwide coverage,
and its large community, OSM was chosen as the main dataset of the project. In
addition, several databases from trusted partners were chosen. All this data gathered
represented more than 150,000 health localities worldwide. In order to integrate
all these datasets with different structures and values, a data model, based on the
Entity–Attribute–Value model, and inspired by the OSM data model was chosen. The
OSM data model enables users to store information about anything. A FullTextSearch
index has been implemented to enable searching for textual data. However, this
flexibility of storing information represented a risk potentially hampering the goal of
having a curated database with easily accessible information. In order to address this
issue, 15 core attributes (Table 5.1) that are relevant for both the public and health
professionals were defined by the International Hospital Federation. For most of these
core attributes, defined values were set to enable comparisons. Indeed, an essential
question such as “What type of health facilities is it” can vary greatly according to
the national or organizational classification system.
Each time a record is created in Healthsites.io, a globally unique identifier is
created and assigned to that record. This is specific to the Healthsites.io database and
is used to provide a canonical point of reference for that record.
The upstream id is used to create a back reference to the original source data (e.g.,
from a national healthsites dataset).
5.2.2 Validation Process
In order to enhance the reliability of health facilities data, several validation processes
have been planned to be added to the project. There is an automatic verification
ranging from a simple: “email address should look like an email address” to more
complex which even rely on external services like: “check if the Locality address is
similar to results returned by external geocoding services” or “check if the telephone
number is correct by manually calling the number and verifying”.
There will also be a validation process based on user reputation (Fig. 5.1) to assess
the reliability of data through a Locality Validity Index (LVI). This reputation-based
process consists of four steps. The first step takes place during data integration.
Depending on source of the data, the LVI obtains a score ranging from 0 to 10; 0
being for data added by a new user, 10 for data from a community trusted user, and 5
if it is a batch of data coming from a Ministry of Health or healthcare organizations.
The reliability of the user is based on the monitoring of his activity on the platform
and the crosscheck of his activity by other users. Second, once the data is displayed
on the map of Healthsite.io, any user can complete missing attributes data and modify