Novel Methods of Measuring Adherence Patterns Reveal Adherence Phenotypes with Distinct Asthma Outcomes

Area of Science:

• Pharmacology and Therapeutics
• Respiratory Medicine
• Data Science in Healthcare

Background:

• Poor adherence to asthma controller medications is linked to suboptimal symptom control and increased exacerbation risk.
• Traditional adherence metrics (mean proportion of doses) may not fully capture complex medication-taking behaviors.
• Novel metrics analyzing day-to-day variability could offer deeper insights into adherence patterns and clinical outcomes.

Purpose of the Study:

• To evaluate if new time- and dose-based adherence variability metrics provide additional value beyond mean adherence.
• To identify distinct adherence patterns associated with asthma symptom control (Asthma Control Test [ACT] score) and exacerbation risk.
• To analyze electronically recorded medication data from a 6-month randomized trial on inhaler reminders.

Main Methods:

• Calculated adherence metrics from the first two months of a 6-month study, including mean adherence, time adherence area under the curve (T-AUC), entropy, and standard deviation.
• Utilized factor analysis to identify dominant adherence metrics, followed by hierarchical clustering to define patient groups.
• Compared identified clusters based on symptom control (ACT scores) from months 2-6 and exacerbation risk over the entire study period.

Main Results:

• Two factors, primarily T-AUC and entropy, explained over 65% of adherence variance.
• Two patient clusters emerged: high time adherence (n=75) with better T-AUC, and low time adherence (n=23).
• The high time adherence cluster demonstrated less decline in ACT scores and better 6-month symptom control compared to the low time adherence cluster (P=0.012 and P=0.034, respectively).

Conclusions:

• Novel adherence metrics, particularly low time adherence, are associated with an increased risk of declining asthma symptom control.
• Adherence patterns may possess a predictive 'memory' for future clinical status.
• Further validation of these novel adherence metrics in larger datasets is warranted.