Identifying temporal patterns of adherence to antidepressants, bisphosphonates and statins, associated patient factors and healthcare cost outcomes

Low medication adherence (MA) for chronic conditions requiring long-term care is related to increased disease prevalence, premature death and healthcare spending. Traditionally, MA has been summarised as a single percentage value representing the amount of medication taken relative to the total amount prescribed during a specified period, such as the proportion of days covered (PDC). An alternative method that categorises temporal patterns of MA using group-based trajectory modelling (GBTM) has recently been explored internationally as an improved approach. We apply GBTM to examine MA to three medication types – antidepressants, bisphosphonates and statins – and its associated factors and healthcare cost outcomes, for those who participated in the Sax Institute’s 45 and Up Study, currently the largest ongoing health and ageing study in Australia on 267,153 participants aged at least 45 across New South Wales. The three medication types were chosen because the prevalence of each associated condition, depression, osteoporosis and cardiovascular disease, respectively, is high and imposes a large burden on society.

Of the three papers comprising the thesis, Paper 1 is a systematic literature review to assess the impact of MA to the three medication types on healthcare resource utilisation and healthcare cost (HRU), and compare these impacts across the medications. We found that nonadherence is generally positively associated with HRU for all medications and most prominently for bisphosphonates, although the impacts vary across medication types and HRU components. Paper 2 uses GBTM to categorise the types of MA to the three medication types based on temporal patterns of medication use, examines characteristics and associated patient factors of each type, and determines whether the categorisation can provide information and insights additional to those using the conventional measure, PDC. The GBTM analysis identified consistently high adherence, gradually decreasing adherence and early discontinuation for all three medication types, and additionally, moderate adherence followed by higher adherence, consistently low adherence, and discontinuation after seven to eight months for bisphosphonates and statins. Factors associated with the identified MA types were different across medication types. Although broadly consistent, GBTM generally provided more nuanced understanding of the factors associated with each type of nonadherence than PDC. Paper 3 identifies the associations between nonadherence to the three medication types defined using GBTM and healthcare costs, and compares these associations to those identified using PDC to understand comparative benefits of using GBTM. The associations varied across healthcare cost components and medication types including increased inpatient costs for antidepressant users with declining adherence and decreased osteoporotic inpatient costs for bisphosphonate users with declining or 6  consistently low adherence, compared to those showing consistently high adherence. The use of GBTM revealed some associations not identified when the PDC was used.

This research is the first to undertake several analyses including: provision of a comprehensive summary of the impacts of MA on HRU across the three medication types using a systematic literature review; application of GBTM to measure bisphosphonate adherence and to measure antidepressant adherence for Australians; and measurement of healthcare cost outcomes of nonadherence to antidepressants, bisphosphonates and statins using GBTM. Measuring MA based on temporal patterns can allow clinicians, policy-makers and researchers to differentiate between alternative types of nonadherence, better identify patients at risk of nonadherence based on their characteristics, and better understand the impact of nonadherence on HRU. Such improved knowledge of MA can help tailor and initiate more targeted MA interventions for maximum impact and assist in more efficient allocation of healthcare resources.