Index of Non-Adherence

From Clinfowiki
Jump to: navigation, search

Computational analysis of non-adherence and non-attendance using the text of narrative physician notes in the electronic medical record.

Turchin A, Kolatkar NS, Pendergrass ML, Kohane IS. Medical Informatics and the Internet in Medicine 2007 Jun;32(2):93-102.


Question: Does the non-adherence of patients to physician recommendations create poor clinical outcomes?

Method/Data Sources: The authors developed process and software that allowed them analyze a large number of physician's free text notes for patient non-adherence. They compared that to patient Emergency Room visits to determine the outcome for non-adherence.

  • Data:
    • Clinical Outcomes is measured by subsequent Emergency Room visits adjusted for a number of factors.
    • Created software to quantitatively measure the compliance based on physician’s free text notes.
    • Limited records to those that show possible chronic disease.
      • (Mainly Hypertension and Diabetes Mellitus.)
    • Limited chronic records to those that had been seen regularly.
      • Those with more than a base amount of cancer were excluded as their compliance was not expected to be measured in visits to the ER.
    • Identified singular physician notes and the sentences within.
    • Identified 14 phrases that indicate non-adherence.
  • Method:
    • Index of Non-Adherence is the ratio of non-adherence sentences over the total number of sentences.
    • Other Indices were also created.
    • Various statistical methods (positive predictive rates) were used to extrapolate manual verification.

Main Results:

  • Created software and processes that can measure and monitor patient non-adherence as well as outcome via physician's notes.
  • Found that those patients who were most lax in following the physician’s recommendations visited the emergency department twice as many times as those that had the highest adherence to the physician’s recommendations.

Conclusion: Created software and process that allows for determination of patient non-adherence over a large number of free text records. Software and process verified several ways and the results made sense. The Authors plan to continue developing the software and processes.

Comments: Interesting approach to large scale research over free text data. As the author's mention, this can be used to identify those patients that may need more help with compliance. This reviewer wonders if it can also be used to determine which physician's need extra help with their patients.

--John Norris 22:13, 20 February 2008 (CST)