simultaneous causality correction?


Based on a data set I have to answer whether the use of tele-diagnosis and tele-evaluation by health professionals in Andalusia has a positive impact in their perception of the technology usefulness.

I have the variables:

dependent variable:
useful -how useful is tele-diagnosis and tele-evaluation in the medical practice and the possible answers are \Very Useful"=4, \Useful"= 3, \A bit useful"=2, and \Useless"=1

My explanatory variable of interest is:
Use- answer to \How often do you use tele-diagnosis and tele-evaluation in the medical practice. The answers are \Very often"= 5, \Often"= 4, \Rarely"=3, \Never"= 2, and \Not available"=1

My problem is simultaneous causality. I want to estimate the causal effect of use on useful, but there is also a causal effect from useful to use.

I know I can use instrumenal variables to correct for the bias.

But I don't seem to have any IVs I can use.

Here is the rest of the explanatory variables I can put into the model:



The rest of the variables are statements where the respondent exress the degree of agreement with the statements... Possible answers are \Totally agree"= 4, \Agree"=3, \Disagree"=2, and \Totally disagree"=1

B1: Electronic health record systems are not suciently adapted to the professionals
ˆ B2: `There are incompatibilities between the use of applications and the working processes'
ˆ B3: `There are infrastructure and connectivity limitations'
ˆ B4: `Lack of agreement regarding the data that must be stored'
ˆ B5: `It is dicult to integrate the applications I already use with new applications'
ˆ B6: `I am worried about the security and condentiality of data'
ˆ B7: `Lack of funding for ICT (information and communication technologies) integration'
ˆ B8: `Lack of economic incentives for the use of these applications'
ˆ B9: `Lack of knowledge and skills'
ˆ B10: `Lack of training '
ˆ B11: `Its use takes too much time during my clinical practice

Any suggestions to how I correct for the endogeneity??
Last edited:


Less is more. Stay pure. Stay poor.
Can you better define what you are talking about when it comes to "simultaneous causality".

Are you talking about whether they are both explaining the same thing in the dependent variable. If so, you can look at factor analysis and cronbach alpha to better understand their relationship.

Also if you put both into a general(ized) linear regression model, you can look at their collinearity.

If this is happening you may opt to select the better one for use.
Hey again.
Sorry if I did not express myself clearly. I will try once more.

I have to estimate the causal effect use have on usefull. That is estimate how using tele-diagnosis and tele-evaluation affects the usefull variable.

But as one can say that use affects usefull, one can also claim that usefull affects the use - that is how usefull you find this tele-diagnosis can affect the use of it.

Therefore I claim that I have a problem with simultaneous causality. My dependent variable useful also affects my independent variable use.

My question is, how to eliminate this bias. I know that in general I have to use instrumental variables, but based on the control variables I have, I don't see that as an opportunity.

I only have the variables listed earlier in the thread to use. Any suggestions how to forumlate a proper model, that will not be biased?
Last edited: