dhis2 / who-data-quality-app Goto Github PK
View Code? Open in Web Editor NEWData Quality App by WHO
License: Other
Data Quality App by WHO
License: Other
If mapping a numerator to an indicator, only indicators with denominator = 1 should be options.
Dashbord 'Completeness' should display information for three years back, similar to the 'Consistency-time' tab.
Quarterly data on dashboard have limited value since only 4 periods are shown, should extended to e.g. 3 years/4 quarters.
The detail charts for one particular orgunits in the consistency module should show "shortest possible" period for a variable, not selected period only. E.g. if doing an analysis by year of a monthly data element, the detailed chart should show the months rather than years.
This is a place holder.
I am experimenting with methods for drilling down and reviewing facility-level trends in reporting completeness.
Currently, only English is supported. Extend support for other languages using standard DHIS2 i18n frameworks.
Add support in the external comparison to compare "raw" data e.g. from parallel systems, in addition to survey results.
Move descriptions/info on the meaning of different columns to the column headers, as is done in the outlier results. Applies to the numerator parameter settings for example.
This would be relevant for several countries including Liberia and Ethiopia
I have yet to see a country DHIS2 instance which has the data/data elements/indicators required to assess Domain 3 (consistency of routine estimates with survey estimates) and Domain 4 (External consistency of live birth estimates). The user manual should include some brief instructions on how to configure and populate the required indicators for this.
Currently, the sort icon does not indicate the sort order (a-z or z-a), this might be useful.
Functionality for combining various DQ metrics/outputs in reports, which can be accessed similarly to the annual report.
Only show data elements where the data element type is one which can be aggregated. For example, leave out text and date type data elements. Applies both to the in-depth analysis modules and the numerator/denominator mapping.
With the Analysis tools and the Data Export for Excel tool, when I select by year and 2016 I get data for the months from December 2015 to November 2016.
When hovering over dropout rate visualisations, show the name of the orgunit for each bar.
In the scatterplots, show the percentage each particular value represents.
When using Administration to configure the Tool:
a) Under "Numerator relations", with "Type" it is not self-evident what is meant by "Equal across orgunits" as opposed to "A = B". With the Liberia case, "Equal across orgunits" appears to place the reference line correctly (with a slope equal to the nationwide ratio) whereas "A=B" places the reference line at A=B. The user manual +/- the webpage itself should explain this.
b) Under "Numerator quality parameters", with "Compare orgunit consistency with" it is not self-evident what is meant by "Overall result" versus "Expected result". With the Liberia case, "Overall result" appears to correctly place the reference line. The user manual +/- the webpage itself should elaborate on this.
c) Under "Numerator quality parameters", what difference does it make whether "Expected trend" is set to "Increasing" versus "Decreasing" versus "Constant"? How can the graph in the bottom left of sections 2d be interpreted to decide whether the Expected trend is Increasing versus Decreasing versus Constant?
Expected trend.docx
In the annual report, the value to use for denominator when calculation % of orgunits with divergent score should be review. Currently only orgunits with at least on non-null completeness is included in the denominator (under the assumption that only those report on the given reporting form). This should probably be changed to include all orgunits.
Disable orgunit level/group selection when selected orgunit is a leaf.
When selecting custom date ranges for the analysis checks, monthly period type could be used as the default (with option of changing to dates/weeks when that is needed).
Ability to select a data value and open the corresponding data set report. Only works with orgunit "leaves".
Identifying outliers:
A more sophisticated method to identify outliers should be developed. Preferably the HB method (demonstrated at the DHIS 2 academy in August)
A graph in addition to tables to identify outliers
For a particular data element (operand), make a histogram with the typical reported values.
Dashboard can be slow to load, in particular the time consistency. To improved the (perceived) waiting time, fetch the quickest chart types first and display results as they are returned.
The version number of the app should be logged or displayed for example on the "About" page. This would make troubleshooting easier.
As it is, the report is too lengthy and users will find it hard to focus on the most important findings. As one example, I find the entire "Summary" section to be unhelpful. I like to skip it and jump to the detailed findings section.
After a particular Numerator has already been mapped, if you want to revise the mapping, you click on "edit".
The bottom half of the resulting dialogue box (Data mapping, Dataset for completeness, Variable for Completeness) is empty. It would be nice if the bottom half of this dialogue box showed the existing mapping so that you did not have to again search for each of the proper settings. This way it would be easier to re-set just one of 3 things being mapped or to switch the numerator from non-core to core.
A report or module for reviewing all DQ metrics related to one indicator (rather than by domain (dashboard) or group (annual report))
Upgrade to api version 2.26 and test with 2.28
Add in-depth completeness module. Could include the completeness-related outputs currently on the dashboard and in the annual report, but also additional functions.
Examples:
Review and improve handling of aggregationtype in the internal d2Data service.
Add a summary result when doing a DQ analysis. Could show e.g. distribution of outliers by
Certain indicator consistency checks might be relevant when looking at a single period (e.g. month), for example reported stock vs services. Others require a longer time period, such as dropout rates. If it was possible to specify this, the dashboard could show data for single months/quarters for these indicators.
When showing the visualisation for an orgunits and variable in the outlier module, it would be useful to include the trend in completeness as well.
Review the current handling of aggregationType in the d2data service.
With several country instances (e.g. Liberia) when reporting completeness is aggregated over multiple years (e.g. for the "Overall score" in section 1d of the Annual Report), the result is substantially greater than 100%. My guess is that DHIS2 administrators need guidance on how to define the aggregation for reporting completeness.
When loading a data element or indicator programatically to the d2select components, a group should be auto-selected as well.
Consistency - time tab
Better if the graph on the right has default level on district and not region. It would also be an advantage if the "guides lines" that distinguish between percentage deviations in scatter plots were removed. They give too much attention to small units.
Consistency - data tab
Is it correct to use the term dropout? It's not referring to same people (or is this a common way to use the term in DHIS2?)
Use dynamic graphs for "Consistency data"; In particular, histograms show little information when the y-axis is kept at 100% maximum at all times. By adjusting the maximum level on the y-axis of eg 5% over the highest observed percentage, the graph will be more intuitive. In that case, it is also important to point out that the y-axis is not 100%, but at X%.
Outliers tab
Consider dropping this tab. The outlier part does not work optimally, and much of the information here is better presented in the «consitency-time» tab. A good replacement can be to enter a "tables" tab. Where there is a table with the most important variables. Where the line starts at national level and where you can drill down.
Add info on the number and percentage of orgunits that are outside the given DQ threshold
"Dynamic" threshold for outliers according to the averagely reported value, similar to DHIS 1.4
Smaller deviations allowed for bigger facilities, and the smallest facilities potentially ignored.
Would have a similar effect to the HB-method proposed by SSB.
Change default on 'criteria' from 33% to 0% or remove it altogether.
We have not been able to identify the difference between "Totals" and "Details". If there is no difference, the option should be removed.
'District' as default on 'Disaggregation' under 'Orgunit'.
Add the opportunity to compare a specific month (eg April 2017) with the previous releases of April (April 2016, April 2015, April 2014, ..., April N) during 'Period'.
Some variables should not be aggregated on month to get the year. Eg. doctors, beds, etc. Is it possible to enter a functionality where you can get an average so that the numbers do not become x12? The error may be more in the main program?
Remove "dropout" / redefine "dropout" unless this is a known term in DHIS2. As we see it, "dropout" should refer to whether unique devices / people drop out of or continue a program (panel data)
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.