Statistical analysis of patient test results comparing self-testing and clinic management

In late 2008, the Community Anticoagulation Therapy (CAT) Clinic in Cedar Rapids, IA began a patient self-testing initiative for patients using anticoagulants. Previous studies had shown that other countries have had greater patient compliance and test results when switching to self-testing.

To conduct a similar study, home monitors were sent to eligible patients, and training was provided to patients on how to take proper measurements. The patients were also shown how to enter their INR results at home, so the CAT Clinic could continue to monitor and manage the patients.

In July of 2008, the CAT Clinic began to notice that the results from the patients who were self-testing seemed to be improved from their results prior to self-testing. The clinic wanted to statistically quantify the data, to see if their observations were valid, or just wishful thinking.

We helped modify the reporting features within INR Pro to make the data more accessible for this type of analysis, and provided statistical assistance to the clinic as described below.

In order to accurately measure the change, without mixing in patient differences, the data needs to be compared on a patient by patient basis, and rolled up to an overall summary. 

3 different statistical analyses were performed:

• 2 Proportions test comparing self-testing patients (69.8% in-range) vs their results prior to self-testing using lab/clinic testing (63.7% in-range).
  • Results showed a statistical difference of 6%, but that doesn’t appear to be a very practical difference.
• Paired T-test comparing normalized INR test results comparing self-testing patients (0.01) vs their normalized results prior to self-testing using lab/clinic testing (-0.08)
  • Results show not enough evidence to say there is a statistical improvement to the normalized INR value after self-testing, but we believe there will a statistical significance with more samples.
• Paired T-test comparing INR standard deviations comparing self-testing patients (0.57) vs their variation results prior to self-testing using lab/clinic testing (0.74)
  • Results showed a statistical change in the standard deviation of the normalized INR values after self-testing, therefore, the patient is more consistent after self-testing.

Therefore, there might be some evidence of self-testing effectiveness. However, even if there were no impact, there are significant cost and time savings for the patient and the healthcare professionals when self-testing is performed.

The details and all the statistical analysis of the study are described in the article below: https://www.inrpro.com/article.asp?id=22