Incidence trends analysis

[Mithila shrestja]

hello everyone, how could I find age standarised incidence rates in stata? Any help? Thank you

[Carlo Lazzaro]

Mithila:
this thread may give you some guidance: http://www.stata.com/statalist/archi.../msg00158.html

[Mithila shrestja]

Thank you so much Carlo for the thread.
I have not used stata much so I am not pretty clear on it.
I have data set in excel (glioma cases for various age groups from 1970-2013) for male population in one sheet and separately among female population. I have another separate excel file for male and female population during those years. have managed to merge the file and made glioma casesfor various age groups among male and male population according to the same age group during 1970-2013. i have calculated incidence rates for those age groups. But I do not know the standarized incidence rates and I got poisson regression as
poisson age04 dg_y, exposure(pop04) irr

Iteration 0: log likelihood = -95.658013
Iteration 1: log likelihood = -95.658013

Poisson regression Number of obs = 44
LR chi2(1) = 4.27
Prob > chi2 = 0.0387
Log likelihood = -95.658013 Pseudo R2 = 0.0219
age04 | IRR Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
dg_y | .9864721 .0065346 -2.06 0.040 .9737473 .9993631
_cons | 1.24e+07 1.64e+08 1.24 0.215 .0000756 2.04e+18
ln(pop04) | 1 (exposure)

How can I interpret this one?
Thank you.

[Carlo Lazzaro]

Mithila:
I would read your result as follows: the IRR for dg_y is 0.99 times smaller than for non_dg_y (and significantly so).
What does it mean in your research field, I do not know.
As an aside, things would be easier for any reader if you would:
- post what you typed and what Stata gave you back via Code delimiters (please, see FAQ #12);
- explain what the abbreaviations of your variables stay for.
Thanks.

[Mithila shrestja]

Hello Carlo, Yeah I will explain my data sets. Above explanations give no meaning.My data set looks like this in one sheet. dg_y is the diagnosis years and then corresponding number of glioma cases for various age groups. Years are upto 2013 and age group up to 85.

dg_y	age0-4	age5-9	age10-14	age15-19
1970	5	6	3	3
1971	1	4	4	1
1972	3	5	1	5
1973	2	4	3	3
1974	1	3	2	4

similarly, I have total population in the same format like:

dg_y	pop0-4	pop5-9	pop10-14	pop15-19
1970	173171	194431	202971	215689
1971	167275	192445	199232	215563
1972	160946	190605	197232	212305
1973	154914	187338	197906	208330
1974	153908	181760	196997	205224
1975	155364	175672	195205	201797

now, i would like to calculate poisson regression by code
poisson age04 dg_y, exposure(pop04) irr
I am sure if it is correct.
Stata gave me this result:
Iteration 0: log likelihood = -95.658013
Iteration 1: log likelihood = -95.658013

Poisson regression Number of obs = 44
LR chi2(1) = 4.27
Prob > chi2 = 0.0387
Log likelihood = -95.658013 Pseudo R2 = 0.0219
------------------------------------------------------------------------------
age04 | IRR Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
dg_y | .9864721 .0065346 -2.06 0.040 .9737473 .9993631
_cons | 1.24e+07 1.64e+08 1.24 0.215 .0000756 2.04e+18
ln(pop04) | 1 (exposure)

And I did not know how to interpret this result if I have done poisson regression correctly.
Thank you Carlo.

[Mithila shrestja]

more to add how can I calculate the overall incidence of gliomas by stata code?

Thank you.

[Carlo Lazzaro]

Mithila:
thanks for posting further details.
From what I can get your -poisson- is focused on glioma cases occurred during 1970-2013 in patient aged 0-4 years.
You have a single predictor -dg_y-, the year of diagnosis.
Hence the interpretation is that the IRR for glioma cases shows a statistical significative negative time-dependence (e.g.: in 1971 is 0.99 times smaller than in 1971, and so on). I do not know what this result implies (a change in the epidemiology of this disease?), nor if you considered all the predictors that other researchers in your research field plug in their model (maybe the type of health care facility where the diagnosis has been made would be an interesting as a predictor).

[Mithila shrestja]

I am exploring the incidence of brain tumor cases to see whether cases rise with increasing use of mobile phone use.
But when I draw line graphs, there is a sharp increase for 1985 and some how fluctuations for other years. It is so confusing
Carlo, is it possible to calculate overall incidence for all those years?
Thank you Carlo for your suggestions.

[Carlo Lazzaro]

Mithila:
if I'm correct in understanding what you're after,
I would take a look at -egen-, -rowtotal- command to sum up both glioma cases and population across years. Then I would re-run -poisson- with the overall data.
As an aside, please post attachments in Stata format only.

[Mithila shrestja]

Hello Carlo, I did not understand why do I need to sum up cases and population across years?
Actually I would like to get approximate Poisson method to analyze incidence trends in terms of years and ages.

[Carlo Lazzaro]

Mithila:
if yuour goal is only to calculate the overall Incidence for all those years, you may want something along the following lines:

egen Overall_cases_glioma=rowtotal( age*)
egen Overall_population=rowtotal( pop*)
poisson Overall_cases_glioma dg_y, exposure( Overall_population ) irr

Otherwise, if you're interested in the Overall incidence across age groups for each year, you may want to consider tweaking he previous code a bit:

egen Overall_cases_glioma=rowtotal( age*)
egen Overall_population=rowtotal( pop*)
poisson Overall_cases_glioma i.dg_y, exposure( Overall_population ) irr

[Mithila shrestja]

Many thanks Carlo. I want this one
egen Overall_cases_glioma=rowtotal( age*)

. egen Overall_population=rowtotal( pop*)

. poisson Overall_cases_glioma dg_y, exposure( Overall_population ) irr

Iteration 0: log likelihood = -182.88921
Iteration 1: log likelihood = -182.88921

Poisson regression Number of obs = 44
LR chi2(1) = 224.22
Prob > chi2 = 0.0000
Log likelihood = -182.88921 Pseudo R2 = 0.3800

--------------------------------------------------------------------------------------
Overall_cases_glioma | IRR Std. Err. z P>|z| [95% Conf. Interval]
---------------------+----------------------------------------------------------------
dg_y | 1.015905 .0010792 14.85 0.000 1.013792 1.018023
_cons | 1.16e-18 2.45e-18 -19.49 0.000 1.82e-20 7.37e-17
ln(Overall_popula~n) | 1 (exposure)
--------------------------------------------------------------------------------------

[Mithila shrestja]

Hello again,
Now I am trying to find average annual percentage change in incidence.Is it possible in jointpoint regression analysis only or in stata as well?
I have IRR for different years as you can see below:


poisson Overall_cases_glioma i.dg_y, exposure( Overall_population ) irr

Iteration 0: log likelihood = -146.62387
Iteration 1: log likelihood = -146.61141
Iteration 2: log likelihood = -146.61141

Poisson regression Number of obs = 44
LR chi2(43) = 296.78
Prob > chi2 = 0.0000
Log likelihood = -146.61141 Pseudo R2 = 0.5030

--------------------------------------------------------------------------------------
Overall_cases_glioma | IRR Std. Err. z P>|z| [95% Conf. Interval]
---------------------+----------------------------------------------------------------
dg_y |
1971 | .9369266 .1607511 -0.38 0.704 .6693643 1.311441
1972 | 1.099885 .1810847 0.58 0.563 .7965366 1.518759
1973 | 1.009403 .1694313 0.06 0.956 .7264203 1.402624
1974 | 1.046064 .1738451 0.27 0.786 .7552611 1.448836
1975 | 1.014466 .1697051 0.09 0.932 .7308763 1.408091
1976 | 1.234501 .1972177 1.32 0.187 .9026255 1.688399
1977 | .925588 .1581946 -0.45 0.651 .66212 1.293894

How could I do this in stata?Any help?

Thank you in advance.

[Carlo Lazzaro]

Mithila.
I would take a look at -margins-.