Purpose:
To evaluate different methods for identifying glaucoma patients and calculating adherence.
Methods:
We conducted a retrospective review of administrative and clinical data captured in the Southern California region of Kaiser Permanente. Five different operational definitions of glaucoma and three methods of calculating adherence were tested. Glaucoma definitions varied by ICD-9 diagnosis codes, specialist visits, and prescription fills. Incident glaucoma was defined as the first date each of the criteria was met. Adherence definitions varied in their calculations of days supplied. Differences in the characterizations of patients (demographics and utilization) and adherence patterns were evaluated.
Results:
Counts of incident patients between 2005 and 2009 ranged from 10,784 (requiring an ICD-9 of 365.10 or 365.11 from an ophthalmologist) to 60,538 (allowing any ICD-9 for 365.X). Using the days supplied field in the pharmacy claim reduced the medication possession ratio (MPR) while the Rylander (2008) method resulted in the highest estimates of adherence. With the exception of "any ICD-9 for 365.X" all definitions of glaucoma and adherence produced the same basic pattern of bimodal medical adherence by year 2 (Figure 1). Initial year adherence is not bimodal for definitions of glaucoma requiring minimal adherence for inclusion into the population.
Conclusions:
Retrospective analyses of chronic conditions and adherence can result in dramatically different results on the basis of operational definitions of the condition and measures of adherence. We found the basic pattern of adherence to topical ocular hypotensive agents was consistent across most definitions of glaucoma and adherence. The bimodal nature of the distribution indicates measures such as "average compliance" in a population are not meaningful.
Keywords: clinical (human) or epidemiologic studies: prevalence/incidence • clinical (human) or epidemiologic studies: biostatistics/epidemiology methodology