topic Re: Pre Post Count data in Statistical Procedures

Pre Post Count data

chuie — Thu, 07 Sep 2017 16:50:02 GMT

Hi All,

I have quick question on how to analyze pre and post count data to know the effectiveness of school tele health program on ED (emergency Department) volumne reduction.

For example, We have opened a tele health clinic on X number of school.I would like to comapre if tele health clinic have any impact on ED redcution.

My data would be like this:

Schoolname beforeED count(2013) At intervention ED count(2014) After ED count(2015) After ED count(2016)

ABC 60 70 30 35

DEF 150 56 28 13

XYZ 86 56 64 25

If anyone could give me proper step by step instruction , I would really apapricate

Thanks

Re: Pre Post Count data

Reeza — Thu, 07 Sep 2017 17:12:18 GMT

1. Graph the data so you can see the differences.

2. Look at the paired t-test option as a basic starter

http://documentation.sas.com/?docsetId=statug&docsetVersion=14.2&docsetTarget=statug_ttest_examples03.htm&locale=en

Re: Pre Post Count data

chuie — Thu, 07 Sep 2017 17:26:11 GMT

Thank You Mam for promp reply. I forgot to add one variable called school tele health yes/no so that we could comapre the rates with those school who have tele heallth clinc vs not.

school Tele clinic (yes/no) pre ED count at intervention ED count after ED count (2015) after ED count(2016)

A YES 50 60 13 30

B NO 151 122 155 160

would like to answer:

1. how to know school tele health have reduced the ED volumn at 1. first year(2015) 2. second year (2016)

2. magnitude of reduction like ED volumn were reduced by x % at schools who have school tele health clinic comapred to those who do not have school tele health clinic

I would really appricate if you could answer those questions

Thank you so much in advance:)

Re: Pre Post Count data

Miracle — Thu, 07 Sep 2017 18:32:11 GMT

Hi Chuie,

I would reshape your current wide data to long data as such

school	teleclinic	time	count
A	Yes	0	50
A	Yes	1	60
A	Yes	2	13
A	Yes	3	30
B	No	0	151
B	No	1	122
B	No	2	155
B	No	3	16

where time=0=pre, ..., time=3=after ED count(2016) and model by Poisson Regression. Your outcome variable will be count. The independent variables will be teleclinic, time and teleclinic*time and specify them as categorical variables. The interaction term will test if the effect of teleclinic on your outcome variable differs at each time point. If overdispersion exists, you will have to try to model by other distributions i.e. Negative Binomial Regression, generalised poisson regression. Please refer SAS online documentation or Google for further information. HTH.

Re: Pre Post Count data

chuie — Thu, 07 Sep 2017 19:49:39 GMT

Thank YOU Wong.. This is very helpful I will do the anlysis and will post the sas code and output.

Please validate if I am on right tract.

Thank you again 🙂

Re: Pre Post Count data

chuie — Fri, 29 Sep 2017 22:25:25 GMT

Wong,

what test should I perform inoder to test the effectiveness of the clinic like ED count reduced by X% from the baseline

Please chare your thoughts