[Michael Edwards]

Thank you Becky for your post to the RHINO Forum. I think that Excel is probably the “Go-To” tool for most people that work with taking raw data to the stage of visualization. I find that I use Microsoft Access for almost all of my data management tasks.
You also mentioned 3 powerful tools for GIS, in which one aspect of GIS is the visualization of spatial data. RHIS data is so conducive to geographic presentations, since facilities have geo-coordinates and are located within districts, regions and countries. Of the 3 tools you mentioned, I’m a big fan of QGIS. This is because it is freely available and open source. I know that the 2 ESRI products you mentioned, ARCGIS and ARCGIS online are the “Cadillac/Mercedes Benz” tools of the power users, but I find the pricing of these products as deal-breakers. Now I can see asking a developer to pay for a developer’s license, but I’m not a fan of paying an annual fee to use the software. So the challenge remains, what do we use for our web-based geographic visualizations? Although I’m currently using the Google Maps API, I find it works great for putting points on a map, but thematic or chloropleth maps I find more of a challenge, and I’m still looking for something more user-friendly that is freely available and open source.
MikeE

[Michael Edwards]

Until recent advance of web-based routine information systems, dissemination of our RHIS analyses and data visualizations has been a challenge. Passing around Access databases and Excel files on discs or flash drives can’t always reach the public-at-large, or even easily reach the stake-holders that should/need access to the RHIS data for their decision-making needs.
Now, with web-based systems, we have powerful tools that can solve the RHIS data dissemination problem we had in the past. Of course there is still a problem where the Internet hasn’t yet reached, and you need a certain amount of bandwidth, but definitely progress has been made in many of the countries where we are working to strengthen health information systems.

My question is, should access to the RHIS dissemination tools, like decision support systems, data visualizers and reporting systems be password protected? I can understand the need for confidentiality with patient-based systems, EMRs and eRegistries, but what about our aggregate data? Should we have RHIS data publically available so that we can reach the ultimate stake-holders, the people in the community who use the health system? Is there a level of aggregation in the system, like the District Level that data can be publically available, but other levels, like maybe the Facility Level that we may restrict to those who provide the services?
I know the Demographic and Health Survey project has dealt extensively and appropriately with the issue of confidentiality, and thus can make data publically available at the aggregate level. What should be the policy with RHIS data?
I look forward to your opinions on this issue.
thanks, MikeE

[Michael Edwards]

Hi Theo,
The EDAP website needs a userid and password, and mine isn’t working anymore. 🙁
MikeE

[Michael Edwards]

Thank you Theo for your (as always) insightful comments.
Our most recent forum discussed the issues surrounding interoperability, and I’m sure that issues around data quality (accuraccy, completeness, and timeliness) should also be addressed again in the RHINO Forum series.
Concerning your request for how data visualization platforms(decision support systems) address quality issues, I would like to mention a couple of examples.

In Eritrea, the Decision Support System that is attached to their HMIS has Missing Reports as one of the available Indicators that can be graphed or mapped. So the number or percent of missing reports can be graphed monthly or annually, and as with any indicator, they can drill down from the national to sub-national (Regional or Zoba, Sub-Zoba, Health Facility) levels to see exactly when and where the problem exists.

Often times, systems only report the annual percentage of missing reports by district. I think that this analysis doesn’t fully expose the type of missing data quality problem that the district has. For example, one district could have 1 or 2 facilities that are totally non-compliant and don’t send any reports, and another could have numerous facilities that intermittently have a missing report. Both of these districts could show the same annual percentage of missing reports, but the intervention to resolve the problem would be different.

In the Niger SNIS (Système National d’Information Sanitaire), a special summary report on Missing Reports can be generated that gives an annual analysis of the missing reports by quarter (quarter is the reporting period). Thus the district managers and HMIS advisors can easily determine which facilities are missing reports and when they missed sending their reports.

Concerning data quality in general, in our recent MEASURE Evaluation Mid-Term All Staff meeting, it was announced that the most requested tool used by the project was the RDQA (Routine Data Quality Assessment) Tool. I think that this is a sign that routine data sources have taken on a more important role in determining health program effectiveness.
best wishes, MikeE

[Michael Edwards]

Thank you Amanda for your discussion of the user-centered design approach, and the link to your slide-deck.
It is also important to consider what information is needed and how it should be presented for the different audiences at each level of the system, and what are each of their different needs for the use of the data for decision making. How can our data visualizations help streamline the data analysis tasks so that the efforts can be more focused on the data use in a way that decisions lead to action.
MikeE

[Michael Edwards]

Thank you Sophia, Allison and Hayat for your posts about the various data visualization tool that are available and you have found useful. I’m sure that the most frequent tool in use today is Microsoft Excel.
Please check out this PDF file that lists many Data Visualization Resources and Tutorials that you may find useful:
http://hmn-tsp.net/DataVizRecommendedResources.pdf
http://hmn-tsp.net/DataVizRecommendedResources.pdf

Also, the resources that Allison mentioned and some others can be found here:
Website for diagrams: http://www.duarte.com/diagrammer/
http://www.duarte.com/diagrammer/
Website for data viz: http://datavizcatalogue.com/
http://datavizcatalogue.com/
Icons: https://thenounproject.com/
https://thenounproject.com/

best wishes, MikeE

[Michael Edwards]

Greetings RHINOs and welcome to the RHINO Forum on Data Visualization. I hope you were able to attend the webinar this morning that Amanda gave. I found it very informative, and I especially liked the brief history of data visualizations that she presented. If you didn’t see it, go to the link on the Forum Post announcing the webinar, and you can see the recording.

I would like to tell you about my long history of supporting Ministries of Health with the development of computerized data visualisation tools (Decision Support Systems or DSS) linked to national health information systems. My first system was in Niger where in 1994 we implemented a DOS-based DSS. We were working on the analysis of the Nation Health Information System data (SNIS, or Système National d’Information Sanitaire) for the annual report. We needed certain analyses, like time-trend graphs for key diseases, and a graph of the top 10 diseases at the national level, and we thought that if we needed these graphs for the national level, then they were also needed at the region, district and health facility level as well. So we developed a system that was flexible so that any indicator in the SNIS system could be graphed in some pre-defined graphic templates – time trend graphs, regional comparison histograms, and pie charts. The data was in dBase III, and we developed the DSS using Clipper and dGE (data Graphics Extender).

When everyone started using Windows, people would see our DOS-based system and say “That system does exactly what we want, but we want it in Windows”. So we started developing Microsoft Access-based Decision Support Systems with Active-X objects Graphics Server for the graphs and ESRI’s MapObjects LT for thematic maps. We chose these tools because they had royalty free distribution, so we could distribute our applications without having to make people pay for the application. They just needed Microsoft Office.

Now, people see our Access-based systems and they say “That system does exactly what we want, but we want it web-based and open source”. For these web-based Decision Support Systems, we are using the open-source programming languages PHP and Java Script, and we are using HighCharts for the graphs and the Google Maps API (Application Program Interface) for the maps. We are also working to develop links to both routine and non-routine data sources using existing APIs. For routine health information system data, the DHIS 2 application has an API that we are working with to connect with routine data, and the DHS (Demographic and Health Survey) also has an API that you can use to link to non-routine survey data.

Speaking of the DHIS 2 and DHS, both have built-in Decision Support tools that you can use. DHIS 2 has the Data Visualizer and DHS has the StatCompiler. We’d love to hear from those of you who have used these tools, and what your experiences are with the use of these tools.

Thanks, MikeE

[Michael Edwards]

I believe that data exchange between existing aggregate systems is not the piece of “cake” that Vikas suggests. If it was, then why don’t we see that it has been done successfully in many of the countries that have implemented aggregate systems? I just haven’t seen it yet. First is the issue of just mapping the facilities between systems. As Dave pointed out, he finds the situation with multiple “master” lists, and I’m sure reconciling these lists hasn’t been done primarily because doing so isn’t that easy, because you probably need to do this manually, as the facility names are probably not standardized, and there is no common unique identifier in use. Large countries have a lot of facilities. For instance, Nigeria has over 30,000 facilities, Tanzania has around 7,000 and Kenya around 10,000. Mapping this many facilities is no piece of cake. Even for a smaller country, there are issues with how facilities are named across different systems. In Haiti, after the 2010 earthquake having a Master Facility List proved indispensable during the rescue, and an international collaboration was brought together to quickly to provide this list. Those efforts are described in this article:
http://www.ghspjournal.org/content/2/3/357.full
http://www.ghspjournal.org/content/2/3/357.full

Besides the mapping facilities, you also need to address the issue of mapping fields and indicators. This is not always “a piece of cake” either. For example, when PEPFAR started using DATIM for entry of HIV/AIDS data, there were attempts to set up data exchange between existing systems and national systems. In DATIM, indicators for ART and Counseling and Testing have age and sex breakdowns that are often more detailed than those found in many national systems. DATIM had for Male and Females these age groups: <1, 1-4, 5-9, 10-14, 15-19, 20-24, 24-49, 50+. The national system had only these age groups for males and females: <1, 1-14, 15+. So it was impossible to use data exchange from the national system into DATIM without using calculated percentages, or in other words, making the data up.

[Michael Edwards]

I think that the development of interoperable aggregate systems (HMIS, LMIS, HRHIS….) should be the first priority for countries working towards a unified national health information system. To do this, countries need to establish a Master Facility List, with unique identifiers for not only health facilities, but also unique identifiers for the aggregate levels (regions, districts, communities) within the health system as well. This is not just a technical challenge, but also a management challenge as well. Strong leadership is required.

[Michael Edwards]

This editor seems to have not captured the links, so I will just copy them as text:

Kenya:
http://kmhfl.health.go.ke/#/home

Tanzania:
http://hfrportal.ehealth.go.tz/

Haiti Journal article:
http://www.ghspjournal.org/content/2/3/357.full

[Michael Edwards]

Dear Devan, Here are some MFL sites that have interesting search/analysis and mapping functionality, and a journal article about the Development of an MFL following the massive earthquake in Haiti:

Kenya MFL:

Tanzania MFL:

Development of MFL in Haiti following earthquake:

best wishes, MikeE

[Michael Edwards]

Interoperability for linking service statistics and HR databases, or any database, for that matter, has to start with a SHARED Master Facility List (MFL). So often systems are created, where they create a new unique identifier, or even don’t use a Unique ID number, but just have the facility name and what District it is in to go on. Trying to match facility records between databases is without a SHARED Unique Identifier is often very problematic.
For instance, in Nigeria, MEASURE Evaluation conducted a data collection survey in 2014 that collected facility name, State, LGA (District), type and ownership information for over 34,000 public and private health facilities. From another source, geo-coordinates (latitude/longitude) for over 24,000 health facilities had been collected, but there was no common identifier that could be used to easily link these 2 sources of data. Even the LGA names were problematic to match with over 60 LGA where the spelling of the name was different between the 2 data-sets. After resolving the LGA names, an attempt to match facilities between the 2 lists based on the name was attempted, but only about 1700 facilities had matching names. So now the task of individual matching of the rest of the sites is obviously a tedious and very time-consuming activity.
The moral of this story – establish a Unique Identifier for health Facilities that is SHARED among all stake-holders, and put a plan in place to manage the update of the Master Facility List.

MikeE

[Author]

Posts