Homologous periods for indicators

bram_piot · 5 December 2013 07:17

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

···

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Jim_Grace1 · 5 December 2013 11:59

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)
You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.
You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

···

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Knut_Staring · 5 December 2013 13:53

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

···

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

olatitle · 6 December 2013 07:55

Would be great if we could somehow combine the logic and calculations of the surveillance rules with the analytics tables and make this data available in the analysis tools. Sounds like a good approach to try that in an app first.

Ola

···

Ola Hodne Titlestad (Mr)
HISP
Department of Informatics
University of Oslo

Mobile: +47 48069736
Home address: Eftasåsen 68, 0687 Oslo, Norway. Googlemaps link

On 5 December 2013 14:53, Knut Staring knutst@gmail.com wrote:

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Rodolfo_Melia · 6 December 2013 09:43

Hi - I just want to echo the need of more complex logic for doing indicators.

The way that we are addressing this at PSI is by creating ‘Calculated Data Elements’. That’s basically a Data Element that doesn’t have data entry, and instead we populate it via a SQL procedure that we run every night. A cron job currently does the job, but we are moving them into PgAgent. Our current use cases:

Copy the previous months ‘Closing Balance’ into the ‘Opening Balance’ of this month
Average distribution across the last 3 months (rather than the whole series)

Another similar need that we currently plan to address via iReports is giving improvement over time: variation (increase/ decrease)

Option a: Q2 this year vs Q2 last year
Option b: This Quarter (Q2 2013) vs Previous Quarter (Q1 2013)

In my mind, a potential implementation is to be able to define time ranges (default - all periods) when creating the numerator/ denominator of an indicator.

R

···

Rodolfo Meliá

Principal

+1 708 872 7636

+44 777 576 4090

Skype: rod.melia

On Fri, Dec 6, 2013 at 7:55 AM, Ola Hodne Titlestad olati@ifi.uio.no wrote:

Would be great if we could somehow combine the logic and calculations of the surveillance rules with the analytics tables and make this data available in the analysis tools. Sounds like a good approach to try that in an app first.

Ola

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Ola Hodne Titlestad (Mr)
HISP
Department of Informatics
University of Oslo

Mobile: +47 48069736
Home address: Eftasåsen 68, 0687 Oslo, Norway. Googlemaps link

On 5 December 2013 14:53, Knut Staring knutst@gmail.com wrote:

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

ismail · 8 December 2013 10:30

Hi,

We have pretty much similar requirement with Malaria Surveillance (weekly reporting) here in Tanzania where by threshold calculation to detect malaria outbreak is based on current reported value against three previous weeks values similar case like what Jim mentioned.

It will be best if we could this in either analytics / indicators or validation rules. As a temporary fix i implemented an external script which reads values through API and perform some calculations. Also HMTL Standard report which highlight weeks exceeding set threshold value could also do

Regards,

Ismail

···

On Fri, Dec 6, 2013 at 12:43 PM, Rodolfo Melia rmelia@knowming.com wrote:

Hi - I just want to echo the need of more complex logic for doing indicators.

The way that we are addressing this at PSI is by creating ‘Calculated Data Elements’. That’s basically a Data Element that doesn’t have data entry, and instead we populate it via a SQL procedure that we run every night. A cron job currently does the job, but we are moving them into PgAgent. Our current use cases:

Copy the previous months ‘Closing Balance’ into the ‘Opening Balance’ of this month

Average distribution across the last 3 months (rather than the whole series)

Another similar need that we currently plan to address via iReports is giving improvement over time: variation (increase/ decrease)

Option a: Q2 this year vs Q2 last year

Option b: This Quarter (Q2 2013) vs Previous Quarter (Q1 2013)

In my mind, a potential implementation is to be able to define time ranges (default - all periods) when creating the numerator/ denominator of an indicator.

R

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

–
Ismail Yusuf Koleleni,
Computer Science and Engineering Department
College of Information and Communication Technology (CoICT)
University of Dar es Salaam

+255713732238 / +255788732233

Rodolfo Meliá

Principal

+1 708 872 7636

+44 777 576 4090

Skype: rod.melia

On Fri, Dec 6, 2013 at 7:55 AM, Ola Hodne Titlestad olati@ifi.uio.no wrote:

Would be great if we could somehow combine the logic and calculations of the surveillance rules with the analytics tables and make this data available in the analysis tools. Sounds like a good approach to try that in an app first.

Ola

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Ola Hodne Titlestad (Mr)
HISP
Department of Informatics
University of Oslo

Mobile: +47 48069736
Home address: Eftasåsen 68, 0687 Oslo, Norway. Googlemaps link

On 5 December 2013 14:53, Knut Staring knutst@gmail.com wrote:

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Lars · 9 December 2013 18:51

Isma: I think you should explore the surveillance validation feature, this is exactly what it is designed for! Documentation here, some slides here. This way you can run these checks automatically, scheduled every night. Then when violations are detected, you can send alerts through messages to the right people. This is better from a monitoring perspective than having to manually generate reports.

Rodolfo: Thanks for the input on the analysis use-case. It makes sense. Would be good if you could write up the exact requirement in a blueprint.

regards,

Lars

···

On Sun, Dec 8, 2013 at 11:30 AM, Ismail Yusuf ismailkoleleni@gmail.com wrote:

Hi,

We have pretty much similar requirement with Malaria Surveillance (weekly reporting) here in Tanzania where by threshold calculation to detect malaria outbreak is based on current reported value against three previous weeks values similar case like what Jim mentioned.

It will be best if we could this in either analytics / indicators or validation rules. As a temporary fix i implemented an external script which reads values through API and perform some calculations. Also HMTL Standard report which highlight weeks exceeding set threshold value could also do

Regards,

Ismail

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

On Fri, Dec 6, 2013 at 12:43 PM, Rodolfo Melia rmelia@knowming.com wrote:

Hi - I just want to echo the need of more complex logic for doing indicators.

The way that we are addressing this at PSI is by creating ‘Calculated Data Elements’. That’s basically a Data Element that doesn’t have data entry, and instead we populate it via a SQL procedure that we run every night. A cron job currently does the job, but we are moving them into PgAgent. Our current use cases:

Copy the previous months ‘Closing Balance’ into the ‘Opening Balance’ of this month

Average distribution across the last 3 months (rather than the whole series)

Another similar need that we currently plan to address via iReports is giving improvement over time: variation (increase/ decrease)

Option a: Q2 this year vs Q2 last year

Option b: This Quarter (Q2 2013) vs Previous Quarter (Q1 2013)

In my mind, a potential implementation is to be able to define time ranges (default - all periods) when creating the numerator/ denominator of an indicator.

R

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

–

Ismail Yusuf Koleleni,
Computer Science and Engineering Department
College of Information and Communication Technology (CoICT)

University of Dar es Salaam

+255713732238 / +255788732233

Rodolfo Meliá

Principal

+1 708 872 7636

+44 777 576 4090

Skype: rod.melia

On Fri, Dec 6, 2013 at 7:55 AM, Ola Hodne Titlestad olati@ifi.uio.no wrote:

Would be great if we could somehow combine the logic and calculations of the surveillance rules with the analytics tables and make this data available in the analysis tools. Sounds like a good approach to try that in an app first.

Ola

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Ola Hodne Titlestad (Mr)
HISP
Department of Informatics
University of Oslo

Mobile: +47 48069736
Home address: Eftasåsen 68, 0687 Oslo, Norway. Googlemaps link

On 5 December 2013 14:53, Knut Staring knutst@gmail.com wrote:

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Thu, Dec 5, 2013 at 2:17 AM, Bram Piot bram.piot@gmail.com wrote:

Hi all,

Surprisingly, yesterday I had exactly the same request from one of our users in Uganda: they’d like to see monthly data for current and past year as separate lines on the same chart, very similar to Knut’s example. I can imagine that this would be rather complex to handle in the visualiser - maybe this needs a separate app?

bram piot

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

On 5 December 2013 10:08, Knut Staring knutst@gmail.com wrote:

Thanks Ola, it’s a good idea to have this also available as part of validation rules. However, my immediate need and what organizations dealing with seasonal epidemics such as WHO and CDC would like to see are charts with two lines:

The trend of cases and deaths for the current year as compared to the trend lines for a) the previous year or b) some kind of average for several previous years. See the attached example.

Knut

On Thursday, December 5, 2013, Ola Hodne Titlestad wrote:

Knut,

Your example looks more like the kind of functionality we have/could build into in surveillance type validation rules.

Ola

On 4 Dec 2013 22:07, “Knut Staring” knutst@gmail.com wrote:

We have the same requirement, in our case to compare this year’s weekly data with the average of the three previous years foe the corresponding periods (e.g. week 48), multiplied by 1.5. Perhaps this should be available in the indicator calculation? And/or an easy way to get it through the api. Could you start on a Blueprint for this?

des. 2013 20:46 skrev “Paulo Grácio” pgracio@criticalsoftware.com følgende:

Hi,

One of the challenges that I have at the moment is the creation of an indicator based on homologous period to have evolution. Any idea/suggestion?

For instance get Malaria deaths for period (2013) and Malaria deaths for period-1 (2012).

Regards

Paulo Grácio

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

http://dhis2.org

–

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Knut_Staring · 10 December 2013 18:15

I think part of the blueprint should be to extend the Web API to provide “All months this year”, “All weeks this year”, and “All days this year” in addition to the “Last 12 weeks” etc.

···

On Mon, Dec 9, 2013 at 7:51 PM, Lars Helge Øverland larshelge@gmail.com wrote:

Isma: I think you should explore the surveillance validation feature, this is exactly what it is designed for! Documentation here, some slides here. This way you can run these checks automatically, scheduled every night. Then when violations are detected, you can send alerts through messages to the right people. This is better from a monitoring perspective than having to manually generate reports.

Rodolfo: Thanks for the input on the analysis use-case. It makes sense. Would be good if you could write up the exact requirement in a blueprint.

regards,

Lars

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522

On Sun, Dec 8, 2013 at 11:30 AM, Ismail Yusuf ismailkoleleni@gmail.com wrote:

Hi,

We have pretty much similar requirement with Malaria Surveillance (weekly reporting) here in Tanzania where by threshold calculation to detect malaria outbreak is based on current reported value against three previous weeks values similar case like what Jim mentioned.

It will be best if we could this in either analytics / indicators or validation rules. As a temporary fix i implemented an external script which reads values through API and perform some calculations. Also HMTL Standard report which highlight weeks exceeding set threshold value could also do

Regards,

Ismail

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

On Fri, Dec 6, 2013 at 12:43 PM, Rodolfo Melia rmelia@knowming.com wrote:

Hi - I just want to echo the need of more complex logic for doing indicators.

The way that we are addressing this at PSI is by creating ‘Calculated Data Elements’. That’s basically a Data Element that doesn’t have data entry, and instead we populate it via a SQL procedure that we run every night. A cron job currently does the job, but we are moving them into PgAgent. Our current use cases:

Copy the previous months ‘Closing Balance’ into the ‘Opening Balance’ of this month

Average distribution across the last 3 months (rather than the whole series)

Another similar need that we currently plan to address via iReports is giving improvement over time: variation (increase/ decrease)

Option a: Q2 this year vs Q2 last year

Option b: This Quarter (Q2 2013) vs Previous Quarter (Q1 2013)

In my mind, a potential implementation is to be able to define time ranges (default - all periods) when creating the numerator/ denominator of an indicator.

R

Mailing list: https://launchpad.net/~dhis2-devs

Post to : dhis2-devs@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-devs

More help : https://help.launchpad.net/ListHelp

–

Ismail Yusuf Koleleni,
Computer Science and Engineering Department
College of Information and Communication Technology (CoICT)

University of Dar es Salaam

+255713732238 / +255788732233

Rodolfo Meliá

Principal

+1 708 872 7636

+44 777 576 4090

Skype: rod.melia

On Fri, Dec 6, 2013 at 7:55 AM, Ola Hodne Titlestad olati@ifi.uio.no wrote:

Would be great if we could somehow combine the logic and calculations of the surveillance rules with the analytics tables and make this data available in the analysis tools. Sounds like a good approach to try that in an app first.

Ola

Mailing list: https://launchpad.net/~dhis2-users

Post to : dhis2-users@lists.launchpad.net

Unsubscribe : https://launchpad.net/~dhis2-users

More help : https://help.launchpad.net/ListHelp

Ola Hodne Titlestad (Mr)
HISP
Department of Informatics
University of Oslo

Mobile: +47 48069736
Home address: Eftasåsen 68, 0687 Oslo, Norway. Googlemaps link

On 5 December 2013 14:53, Knut Staring knutst@gmail.com wrote:

Thanks a lot, Jim.

We are in fact already developing a powerful time series app for which this is the main missing thing, however we may have to change to a different javascript graphing library (because of the license).

Knut

On Thu, Dec 5, 2013 at 12:59 PM, Jim Grace jimgrace@gmail.com wrote:

Hi all,

By all means tell anyone requesting this type of analysis about the new surveillance-type validation rules. They do exactly this type of analysis, for example the calculation Knut mentioned, comparing this year’s weekly data with the average of the three previous years (e.g. week 48), multiplied by 1.5. Surveillance rule features also include:

You can average not only just over the same period in a number of previous years, but over a bounded set of weeks surrounding the week in question, or a combination of both (e.g. comparing week 48 this year with the average of weeks 47,48 and 49 in the previous two years, and week 47 this year.)

You can throw out a configurable number of high and/or low outliers from previous data before doing the average, so you’re not comparing the new data data against previous outbreaks and/or unusually low values.

You can analyze the data automatically as it comes in from data entry each day, for any data elements (singly or in a formula) in all organisation units at whatever levels you like, with customized formulas at each different organizational unit level (e.g. an alarming district-wide increase may be a lower percent than what could be random variation at individual sites.) And be notified automatically through DHIS messaging (by email if you like) of any outliers.

If they want the data for display, this would have to be added whether in the visualizer or a new app. But if they want the data for outbreak detection, by all means also tell them about the new surveillance rules.

Cheers,

Jim

–
Knut Staring

Dept. of Informatics, University of Oslo

+4791880522