OpenStreetMap
Our Experience working with Health Facilities Import -India.
Posted by OSM_RMSI on 3 March 2020 in English.Hello everyone,
RMSI would like to share our experience and knowledge while working on the Health Facilities Import - India, which was initiated for the OpenGovernmentData - https://data.gov.in/ in April 2019.
Almost 68.86% Indian population lives in rural areas and the hospitals with emergency services and other small healthcare centers provide vital services in these areas. Which Often serve as the foundations of rural health care delivery systems. Long travel distances is one of the difficulties for residents of rural areas to access and seek health care services in these hospitals, as these hospitals normally serve multiple communities within a large region.
At RMSI, we’re on a mission to make the digital and physical worlds come together through technology driven solutions. When looking for solutions to the issues above, we noticed a lot of hospitals or healthcare centers didn’t even exist on most of the maps. OpenStreetMap as an open source editable map, it allows more affordable access to the residents living in the rural areas and digitizing healthcare centers in it will be a big helpful first step to resolve the problem. We believe strongly that the solution to most problems starts with getting the right data. India though has been devoid of meaningful healthcare data for a long time. Certainly, surveys and other rudimentary data collection through various data sources is important and do help, but they are normally time-consuming and less efficient with a lot of field trips, particularly while handling large datasets in large areas.
Hence we’re taking a stab at this by looking for our own data resources and making the insights available publicly so that we can spark and guide the conversation around what ails people in India and where healthcare data addition efforts to OSM Maps should be focused.
As being brought up at the beginning, long travel distance is definitely one of the major difficulties for residences in rural areas to access the health care system. And due to the increasing numbers of accidents and other emergency cases, although the government has provided many health care schemes along with toll free ambulance services, it’s still not enough to resolve the problem. There still exists many cases that couldn’t find nearest health services because of lacking information on the map. Thus RMSI conducts the import of government data we found with high accuracy, hoping to provide huge support to make more hospitals and health care service locations be available on the map.
The date resource searching wasn’t easy at all. There aren’t many available datasets that meet the standard of thoroughness and accuracy on hospital locations. After a long research, we finally found the Open Government Data (OGD) directories with a huge list of medical facility records which includes Hospitals, Clinics, Blood Banks, Health centers in India with geocode information. We were so happy to see the huge list of health facilities records in India and aimed to publish accessible government data to the people in need, through the OpenStreetMap(OSM).However, while verifying the quality of this data, we noticed the positional accuracy of coordinates were somewhat skeptical - some records are not located close to the same features on OSM, and some records are even falling out of the boundary of the country.
As the position accuracy of the records are not as expected, we have tried to contact the respective Open Government Data Chief Executive Officer for the latest datasets of hospitals and healthcare centers in India. But we were still told that the found published datasets are already the most up-to-date among other available ones.
The original quality of raw data didn’t properly fit the OSM import policy if we added them directly to OSM. Furthermore, it would be inappropriate to process the import with positional in-accuracy. Data pre-processing became a key factor for this import. Thus we decided to start processing the data in our locality (Hyderabad) with local knowledge applied, as a pilot trial. The team started using local knowledge on data cleanup for Hyderabad hospital records.
Other than inaccurate positional information, duplicated records and invalid values due to incomplete attributes became other challenges. We had to put lots of effort to pre-process all datasets by running a large scope cleanup tasks to remove duplicates and handle invalid data.
To import data to OSM, appropriate tagging is a common concern. But the global tagging schema for healthcare system on OSM isn’t thorough. More importantly, such global schema for health care service attribute does not fit cases in India.
As a team we have explored the existing OSM tags which are being used for the health facilities Import Ghana Health Facilities, Key:Healthcare etc.. and designed the tagging schema by considering the OSM compatibility tags for the OpenGovernmentData.
After applying and matching global OSM taggings to OGD attributes, we started handling the important and country-specific attributes. Example: Nursing homes - In india nursing homes provide the medical facilities but globally it is a “Old Age Home”.
Eventually we have defined necessary attributes required for the import and excluded few of the attributes which do not fit for the health facility data from the Open Government Data like Accreditation, Nodal_person Information, Misccellaneous_Facilities, Tariff_Range, etc… https://forum.openstreetmap.org/viewtopic.php?id=66300.
After finalizing all the attributes in the OGD and designing appropriate tagging schema, we then presented this newly designed tagging schema for India community in different platforms such as Telegram, OSM Forum, etc., to seek for suggestions and acceptance.
Incorporating advices from community members, we organized and updated country specific tagging to OSM wiki:
Respecting the OSM imports we have followed the defined import guidelines throughout our process from Prerequisite to Uploading.
After in-depth verification on compliance to ODbL of the OGD datasets, we began our journey to import the data to OSM, and built a OSM wiki page with detailed description of our imports process https://wiki.openstreetmap.org/wiki/India_Health_Facilities_Import which is being updated on regular basis.
When everything was set to go then we have announced our imports to the community along with the start date and detailed plans of the import process.
With full support on the import from the community, we began our first import in the local region – Hyderabad, by creating MapRoulette challenges to verify the position of each data point from our cleaned datasets. Please refer below figure 5 for the procedure.
After 15 days of effort, we have successfully identified the locations for the 477 hospitals and processed our first import to OSM. All these Hyderabad hospitals are live on OSM already.
With the positive output of hospital import in Hyderabad, we decided to continue the import for the other districts in Telangana state. We followed the same procedure and successfully imported 4533 health facilities out of 6799 records of Telangana using the information available on the OGD records - landmarks, street names, pincode, area name etc…
For the remaining 2266 record having position accuracy issues, we have decided to continue our approach of using local knowledge, but with a different mechanism – we sent our team to work on ground survey to verify locations of these dates.
We first designed and developed an in-house application with customized functionality to fit our survey method. We sorted all unsuccessful records district-wise, and distributed them to each squad to collect locations of the remaining OGD data in Telangana. With the help of our application we captured the locations of the missing facility more efficiently.
To locate the health facilities in rural areas based on the address is challenging as in India a standard address isn’t applied to residential locations in rural areas, instead, landmarks for local residents to recognize the location is more popularly used. Therefore, along with Health facilities important landmarks in the regions have been captured. These landmarks will help to identify the health facilities location for ease of navigation.
During the time when we conducted our ground survey, there were severe floods across Telangana state, which made our survey work more challenging. However, we still successfully located all the remaining 2,266 health facilities along with nearest landmarks in Telangana state. All these hospitals were added to OSM using the bulk import.
With all efforts above, we were able to add and update 6k+ of health facilities onto OpenStreetMap, including hospitals, blood banks, local clinics, nursing homes, sub centers, etc., benefiting 3.52 Cr. of people living in telangana.
This hugely increased our confidence to process other states in India. Working on other states in India, we now added 75% of the hospitals in India after verifying the government data with coordinates. For those of which the location accuracy needs further verification, we have started to collaborate with local OSM community users who can process them with their local knowledge.
To support and improve health facilities import at your local region few regional wise Maproulette Challanges has been created.
-
Chandigarh, Puducherry, Tripura https://maproulette.org/browse/challenges/9967
If you are interested in our progress, please keep watching the periodical updates on our import wiki. If you are interested in improving the health facilities in your local region please contact us at osm@rmsi.com so that we will share the respected maproulette for your region.
Discussion
Comment from impiaaa on 3 March 2020 at 20:14
This is awesome! Great job!
Comment from qeef on 4 March 2020 at 08:25
This is really cool.
Comment from kucai on 5 March 2020 at 03:39
Is subtown much different from suburb?