Correlation between paired samples and an independent variable?

[erik.jar]

Hi everybody!

This is from a study in medicine trying to find whether good consultation can improve patients well-being.

In the study patients got to complete three exact same questionnaires regarding health prior to, directly after and 2 weeks after their visit to the doctor. In the questionnaire directly after the visit they also completed a questionnaire rating the consultation/patient centredness.

Our hypothesis is that a good consultation will improve health/well-being.

Our data is ordinal and is not normally distributed. Therefore we used Friedman's test to find if there was a change in our variable measuring health which we found was significant. We then went on to do a Wilcoxon which also found significant change.

What we want to do now is to find whether or not these results correlate with the questionnaire rating the consultation/patient centredness. We have done a spearman's correlation but this wasn't significant.

So what I'm wondering is if there's any other test you can do to find a correlation between these samples?

Thank you very much!

 

[hlsmith]

How was centredness define and data formatted? How was health/well-being defined and formatted? Side question, were you able to randomized who got centredness care?

 

[erik.jar]

Patient centeredness or PCC was measured using a questionnaire with 9 lickert scale questions covering the most important aspects of PCC. Well being was measured using a other lickert scale questionnaire. We didn't randomize patients to different groups. The reason behind this is that the consultation is dynamic and even the most experienced doctor will have consultations with high and low PCC. Our plan was to split the sample in two groups, one with low and one with high patient centeredness and then analyze for correlations with change in well-being. Hope this answered your questions! Thank you.

 

[hlsmith]

Did your outcome "Likert" questionnaire have a total sum score, so you just want to compare two group's overall questionnaire score for patients?

Is there a set cutoff for converting PCC into low and high (dichotomizing)?