What is it about?
What is adherence? Adherence is how well someone's medicine use matches up with how their doctor told them to take it. Adherence has a big effect on how well the treatment works at keeping symptoms under control, and reducing the risk of asthma attacks. Having bad adherence might mean that the doctor feels the need to give them stronger medicine, or to prescribe them more medicines, when if they were taking their medicine properly, they would have their symptoms under much better control. For people who research medicines, being able to estimate how good someone’s adherence is means that we can spot people who are in danger of getting more ill (like having an asthma attack). Also, it gives us a better idea of how well certain medicines actually work in different groups, like in adults compared to elderly people. How can we estimate adherence? There are many ways of estimating adherence. We can ask people to keep medicine diaries, to track when they take their medicine. We can use a special machine, like a little computer on an inhaler, which records when it is used. We can take samples from people, like hair or blood, to look at the amount of medicine we can find in them. We can also look at the number of days between prescriptions being written by the doctor. This last option can only tell if medicine is being picked up from the pharmacy, and not if it’s actually being taken. However, we can estimate adherence over a huge number of people, without them needing to do anything. How can we estimate adherence by looking at prescriptions? Researchers before me have come up with lots of different ways to estimate adherence from prescription data, but no one really knows which method is best, and when. In this study, we compared many previously described methods of estimating adherence, in a Scottish prescription dataset, specifically focussing on asthma: a common, long-term condition with high rates of non-adherence. We used the estimates calculated from the different approaches people have described, and came up with some guidelines on the best way to estimate adherence, that could be applied to any disease. Our Results In our data, there were over 1.6 million asthma controller (to be taken every day, not just to relieve symptoms) prescriptions for 91,332 people with asthma, between January 2009 and March 2017. 16.7% of people had only a single inhaler during their follow-up (on average this was over 7 years!). In years in which someone got at least one inhaler, an average of 3 inhalers were prescribed, with most people having between 2 and 6. How do we choose how to estimate adherence from prescription data? We found seven key features of any research study which affected which method of estimating adherence would be best. These were: how long a time you are estimating adherence over (a few months, or a decade, for example), whether you need to also look out for patients using too much medicine, whether you are looking at adherence in a whole population (like average by age groups) or looking to make individual risk predictions, whether you can tell if a patient has been told to stop or change their treatment from the data you have, whether we assume a patient has run out early if they get a new inhaler earlier than expected, how far back the data goes in time, and whether we expect a recent change which we want to look at (like after someone changed inhaler, for example).
Featured Image
Photo by Sahej Brar on Unsplash
Why is it important?
Estimating medication adherence from prescriptions needs to be thought about carefully, as using the wrong methods may mean your estimate isn’t very good. We hope our guidelines are useful to other researchers, and that together we can improve the quality of adherence estimation across medical research, for better patient care.
Read the Original
This page is a summary of: Estimating Medication Adherence from Electronic Health Records: Comparing Methods for Mining and Processing Asthma Treatment Prescriptions, November 2022, Research Square,
DOI: 10.21203/rs.3.rs-2033577/v1.
You can read the full text:
Contributors
The following have contributed to this page