how is the Annualized function of indicators working? I get some very
strange values in my pivot tables...
I have the following indicator:
ANC3 coverage, percent, annualized>
Numerator: ANC3 visits, in a Monthly data set, aggregation operator: sum
Denominator: Expected number of pregnancies, in a Yearly dataset,
aggregation operator: average
Bandajuma clinic reported 149 ANC3 visits in April, and have 220.4
Expected pregnant women per month in 2009, as I read it from the pivot
table, but have entered 451.5 for the year 2009. Is this right? In the
pivot table I get 401.5 % ANC 3 coverage, calculated like this:
numxfactor is 181283.3333
denominatorvalue is 451.5 (the same as expected number of pregnancies for
the year)
Indicator is annualized, level 4 (facilities)
I think there must be something wrong here, as I get percentages up to 500
for some months. Regardless of the catchment population not being correct,
how come a numxfactor turns out to have three decimals? How is this
indicator calculated? Maybe it has been defined wrong, or maybe there is
something funky in the code. Any comments?
how is the Annualized function of indicators working? I get some very
strange values in my pivot tables…
I have the following indicator:
ANC3 coverage, percent, annualized>
Numerator: ANC3 visits, in a Monthly data set, aggregation operator: sum
Denominator: Expected number of pregnancies, in a Yearly dataset,
aggregation operator: average
Bandajuma clinic reported 149 ANC3 visits in April, and have 220.4
Expected pregnant women per month in 2009, as I read it from the pivot
table, but have entered 451.5 for the year 2009. Is this right? In the
pivot table I get 401.5 % ANC 3 coverage, calculated like this:
What do you mean here? What is the correct value for the year of 2009?
451.5 or 12*220.4? Where does the monthly value of 220.4 in your pivot table come from?
how is the Annualized function of indicators working? I get some very
strange values in my pivot tables...
I have the following indicator:
ANC3 coverage, percent, annualized>
Numerator: ANC3 visits, in a Monthly data set, aggregation operator: sum
Denominator: Expected number of pregnancies, in a Yearly dataset,
aggregation operator: average
Bandajuma clinic reported 149 ANC3 visits in April, and have 220.4
Expected pregnant women per month in 2009, as I read it from the pivot
table, but have entered 451.5 for the year 2009. Is this right? In the
pivot table I get 401.5 % ANC 3 coverage, calculated like this:
What do you mean here? What is the correct value for the year of 2009?
451.5 or 12*220.4? Where does the monthly value of 220.4 in your pivot table
come from?
numxfactor is 181283.3333
denominatorvalue is 451.5 (the same as expected number of pregnancies for
the year)
Indicator is annualized, level 4 (facilities)
I think there must be something wrong here, as I get percentages up to
500
for some months. Regardless of the catchment population not being
correct,
how come a numxfactor turns out to have three decimals? How is this
indicator calculated? Maybe it has been defined wrong, or maybe there is
something funky in the code. Any comments?
If the annual denominator value is 451.5 and the monthly numerator value
is
149 then the annualised monthly value of 401.5 % is correct.
Num x Factor = 149 x 12 (approx.) x 100 = 178 000
Ind value is then 178 000 / 451.4 = 396 %
The annualisation factor in DHIS is based on days and not months, that is
why it is not exactly 12, but very close to 12 at least.
Ok, that explains the funky numxfactor value, which triggered my
suspicions. The data quality for ANC3 is anyway not very good, and in
addition ANC3 includes third and subsequent visits. So if one woman comes
for 6 visits during pregnancy, she contributes to one ANC1, one ANC2, and
three ANC3 visits during her pregnancy. In new forms to be implemented
octoberish, this has been addressed.
Your problem seems to be the dnominator data typed into DHIS.
Is the annual value really 451.5 or is that a montly estimate of the
annual
value?
The value entered for 2009 is 451.5, based on a catchment population of a
bit over 10 000 for this facility (calculated from the chiefdom catchment
population based on the relative service load of this facility within the
chiefdom). It might very well be wrong, since the service load is already
influenced by data quality and also the preferences of people to go there
(perceived quality of care for example, make people leave their "natural"
catchment area to go to another clinic). Anyway, it was the strange
numxfactor that put me off.
So annualized indicator means that a monthly numerator is factored by
around 12, based on days in the month, to match a yearly denominator? So
february data is factored by 12 point something, months with 30 days a
little over 12, while months with 31 days are factored by a little less
than 12?
Johan
···
Ola Hodne Titlestad |Technical Officer|
Health Metrics Network (HMN) | World Health Organization
Avenue Appia 20 |1211 Geneva 27, Switzerland | Email:
titlestado@who.int|Tel:
+41 788216897
Website: www.healthmetricsnetwork.org
how is the Annualized function of indicators working? I get some very
strange values in my pivot tables...
I have the following indicator:
ANC3 coverage, percent, annualized>
Numerator: ANC3 visits, in a Monthly data set, aggregation operator: sum
Denominator: Expected number of pregnancies, in a Yearly dataset,
aggregation operator: average
Bandajuma clinic reported 149 ANC3 visits in April, and have 220.4
Expected pregnant women per month in 2009, as I read it from the pivot
table, but have entered 451.5 for the year 2009. Is this right? In the
pivot table I get 401.5 % ANC 3 coverage, calculated like this:
numxfactor is 181283.3333
denominatorvalue is 451.5 (the same as expected number of pregnancies
for
the year)
Indicator is annualized, level 4 (facilities)
I think there must be something wrong here, as I get percentages up to
500
for some months. Regardless of the catchment population not being
correct,
how come a numxfactor turns out to have three decimals? How is this
indicator calculated? Maybe it has been defined wrong, or maybe there is
something funky in the code. Any comments?
If the annual denominator value is 451.5 and the monthly numerator value
is
149 then the annualised monthly value of 401.5 % is correct.
Num x Factor = 149 x 12 (approx.) x 100 = 178 000
Ind value is then 178 000 / 451.4 = 396 %
The annualisation factor in DHIS is based on days and not months, that is
why it is not exactly 12, but very close to 12 at least.
Ok, that explains the funky numxfactor value, which triggered my
suspicions. The data quality for ANC3 is anyway not very good, and in
addition ANC3 includes third and subsequent visits. So if one woman comes
for 6 visits during pregnancy, she contributes to one ANC1, one ANC2, and
three ANC3 visits during her pregnancy. In new forms to be implemented
octoberish, this has been addressed.
Your problem seems to be the dnominator data typed into DHIS.
Is the annual value really 451.5 or is that a montly estimate of the
annual
value?
The value entered for 2009 is 451.5, based on a catchment population of a
bit over 10 000 for this facility (calculated from the chiefdom catchment
population based on the relative service load of this facility within the
chiefdom). It might very well be wrong, since the service load is already
influenced by data quality and also the preferences of people to go there
(perceived quality of care for example, make people leave their “natural”
catchment area to go to another clinic). Anyway, it was the strange
numxfactor that put me off.
So annualized indicator means that a monthly numerator is factored by
around 12, based on days in the month, to match a yearly denominator? So
february data is factored by 12 point something, months with 30 days a
little over 12, while months with 31 days are factored by a little less
than 12?
Yes. The annualisation factor for monthly values is around 12 yes, depending as you say on the # of days of the actual month.
Annualisation factor for quarterly values is arounf 4, for weekly about 52. As you see the annualisation factor depends on the period type of the output data.
how is the Annualized function of indicators working? I get some very
strange values in my pivot tables...
I have the following indicator:
ANC3 coverage, percent, annualized>
Numerator: ANC3 visits, in a Monthly data set, aggregation operator:
sum
Denominator: Expected number of pregnancies, in a Yearly dataset,
aggregation operator: average
Bandajuma clinic reported 149 ANC3 visits in April, and have 220.4
Expected pregnant women per month in 2009, as I read it from the pivot
table, but have entered 451.5 for the year 2009. Is this right? In the
pivot table I get 401.5 % ANC 3 coverage, calculated like this:
What do you mean here? What is the correct value for the year of 2009?
451.5 or 12*220.4? Where does the monthly value of 220.4 in your pivot
table
come from?
Sorry, many similar sounding facility names. It should be 451.5 also shown
monthly. The annual value entered for 2009 (checked in DHIS) was 451.5,
and in the pivot table the same number is shown for each month. The
dataset where the DE is coming from is yearly, but the pivot table for raw
data also shows that same value for all months in the year.
Johan
···
2009/7/22 Ola Hodne Titlestad <olatitle@gmail.com>
2009/7/22 <johansa@ifi.uio.no>
numxfactor is 181283.3333
denominatorvalue is 451.5 (the same as expected number of pregnancies
for
the year)
Indicator is annualized, level 4 (facilities)
I think there must be something wrong here, as I get percentages up to
500
for some months. Regardless of the catchment population not being
correct,
how come a numxfactor turns out to have three decimals? How is this
indicator calculated? Maybe it has been defined wrong, or maybe there
is
something funky in the code. Any comments?
Yes, and that is correct. That is the point of using average aggregation operator, the value is the same at whatever time/month during that year. The use of annualisation for indicators helps to make the monthly indicator value more usable (multiplies the numerator by 12 and hence creates a monthly estimate of the annual denominator value).