Trend analysis

[Sohil Manvar]

Hello everyone
I am using Nationwide Inpatient Sample for my research. I am looking for a trend in certain condition or procedure for example, trends of AFib in United states from 2003-20013. I am able to create trends to generate below mentioned baseline table with codes like

svy: tab age year, row col
svy: tab sex year, row col
svy: tab race year, col row
svy: tab diabetes year, col row
svy: tab smoking year, col row

This is the stata out put of above codes

svy: tab newage year, col row
(running tabulate on estimation sample)

Number of strata = 226 Number of obs = 45143
Number of PSUs = 6574 Population size = 213491.48
Design df = 6348

----------------------------------------------------------------------------------------------
| Calendar year
newage | 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Total
----------+-----------------------------------------------------------------------------------
1 | .0789 .0695 .0907 .094 .0812 .1035 .109 .0962 .098 .0861 .0929 1
| .0235 .0202 .0265 .0253 .021 .0254 .0277 .0252 .0245 .0231 .0237 .0242
|
2 | .0883 .0755 .0806 .089 .0952 .0962 .0932 .1 .0967 .0901 .0952 1
| .0891 .0742 .0796 .081 .0833 .0797 .08 .0885 .0817 .0817 .0822 .0819
|
3 | .0724 .0744 .0761 .0846 .0909 .0993 .0947 .102 .1016 .0988 .1051 1
| .2094 .2097 .2155 .2209 .228 .2359 .233 .2587 .246 .2571 .2601 .2348
|
4 | .0836 .0889 .0838 .0894 .096 .0967 .0956 .0882 .0958 .0891 .093 1
| .3729 .3858 .3657 .3599 .371 .3543 .363 .3451 .3576 .3573 .3546 .362
|
5 | .0834 .0871 .0873 .0947 .0935 .1014 .0952 .088 .0947 .0853 .0893 1
| .305 .3101 .3128 .3128 .2967 .3046 .2963 .2825 .2902 .2807 .2795 .297
|
Total | .0812 .0834 .0829 .0899 .0936 .0988 .0954 .0926 .097 .0903 .0949 1
| 1 1 1 1 1 1 1 1 1 1 1 1
----------------------------------------------------------------------------------------------
Key: row proportions
column proportions

Pearson:
Uncorrected chi2(40) = 115.0889
Design-based F(34.14, 2.2e+05)= 2.3005 P = 0.0000

Note: strata with single sampling unit centered at overall mean.

.
. svy: tab female year, col row
(running tabulate on estimation sample)

Number of strata = 226 Number of obs = 45143
Number of PSUs = 6574 Population size = 213491.48
Design df = 6348

----------------------------------------------------------------------------------------------
Indicator| Calendar year
of sex | 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Total
----------+-----------------------------------------------------------------------------------
0 | .0773 .0788 .0828 .0874 .0935 .0986 .0984 .0952 .0976 .0929 .0974 1
| .4534 .4497 .475 .4626 .4753 .4748 .491 .4896 .4792 .4896 .4885 .4759
|
1 | .0847 .0876 .0831 .0922 .0937 .099 .0927 .0901 .0964 .0879 .0926 1
| .5466 .5503 .525 .5374 .5247 .5252 .509 .5104 .5208 .5104 .5115 .5241
|
Total | .0812 .0834 .0829 .0899 .0936 .0988 .0954 .0926 .097 .0903 .0949 1
1 1 1 1 1 1 1 1 1 1 1 1
----------------------------------------------------------------------------------------------
Key: row proportions
column proportions

Pearson:
Uncorrected chi2(10) = 33.8803


This is the outcome I get by using those codes which i have converted in a table to get better and easy explanation.

Table 2 Baseline characteristics of study population, from

Variable	2003	2004	2005	2006	2007	2008	2009	2010	2011	P-value for trends
Overall	95062	11370	127207	143262	124210	119134	116941	94268	86457	<0.001
Age(%)
<65	36.8	35.7	36.4	38.5	37.8	38.4	37.6	39.2	38.8
>=65	63.2	64.3	63.6	61.5	62.2	61.6	62.4	60.8	61.2
Sex (%)
Male	76.0	75.1	74.3	73.0	72.3	72.1	71.2	70.9	70.3	<0.001
Female	24.1	24.9	25.7	27.0	27.7	27.9	28.8	29.1	29.7	<0.001
Race (%)
White	56.7	57.9	56.1	57.5	56.1	59.9	60.7	62.8	63.7	<0.001
Black	6.4	7.5	6.8	8.7	10.3	10.3	11.0	13.7	14.4	<0.001
Hispanics	4.5	3.9	5.1	5.1	5.9	4.9	6.1	6.0	6.7	<0.001
Others	2.7	2.5	3.2	2.6	4.0	4.9	4.9	4.2	4.6
Missing	29.7	28.2	28.7	26.1	23.6	19.9	17.3	13.3	10.6	<0.001

But how can I create Table 2, where there is a trend of specific variable for example here is a mortality trend among all population.

Table 2: In-hospital mortality (>60yrs of age) for Aortic Valve Disorders in US
Year	2000	2001	2002	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	Relative Change	P-value for Trend
Overall	4.53%	4.5%	3.86%	4.64%	4.29%	3.77%	3.74%	3.62%	4.32%	3.2%	3.34%	3.41%	3.46%	-30.92%	<0.001
Age(years)
60-69	2.56%	2.59%	2.03%	3.14%	2.73%	1.99%	2.32%	2.55%	2.37%	1.76%	2.3%	1.7%	2.26%	-13.27%	<0.001
70-79	4.03%	4.43%	3.51%	4.24%	4.31%	3.58%	3.63%	3.02%	3.98%	3.08%	3.33%	3.22%	3.19%	-26.33%	<0.001
>80	6.65%	5.86%	5.62%	6.17%	5.32%	5.23%	4.87%	4.97%	5.85%	4.29%	4%	4.59%	4.35%	-52.87%	<0.001
Male	3.83%	3.81%	3.34%	4.28%	3.67%	3.5%	3.33%	3.33%	3.91%	3.18%	2.94%	3.16%	3.41%	-12.32%	<0.001
Female	5.33%	5.28%	4.47%	5.06%	5.03%	4.11%	4.23%	4%	4.83%	3.22%	3.85%	3.73%	3.53%	-50.99%	<0.001
Non-hispanic Whites	4.45%	4.61%	3.85%	4.34%	4.22%	3.84%	3.74%	3.44%	4.4%	3.15%	3.3%	3.42%	3.39%	-31.27%	<0.001
Others	4.5%	5.72%	4.08%	4.24%	5.32%	2.98%	3.29%	3.76%	4.17%	2.4%	3.42%	3.78%	3.67%	-22.62%	<0.001

In above table first row showed percentage of mortality among all cases diagnosed with aortic valve disease in that specific year and subsequent rows showed mortality in different age group, sex, gender and so forth.

So I don't know what kind of code i can use to generate this second table and to get relative change.

Thanks.

[Phil Bromiley]

I think this was previously posted in precisely the same format. You'll increase your chances of a helpful answer if you follow the FAQ on asking questions.

The problem is partially that the question is not clear, and partially that you're asking for what may be a number of different things at once.

Don't waste our time and your time posting the same thing repeatedly.

[Sohil Manvar]

I think these one was more detailed post then previous one. But there is some limitation in stata so not everyone know how to get second tables. In SAS as long as I know it's much easier.
Thanks for your reply.

[Friedrich Huebler]

Phil asked you here to post an excerpt from your data and to explain how you expect it to be transformed. It is easier to help if you give us some data to experiment with. I would also like to reiterate his advice to read the FAQ.