2. Methods

2.1 Study design and data source

We carried out a repeated cross-sectional study over five years (2012-2016) to assess trends in polypharmacy and the prevalence of PIM in patients with polypharmacy in the Netherlands. The data were obtained from the University of Groningen community pharmacy database IADB.nl. This database contains anonymized information of dispensed prescriptions, including the Anatomical Therapeutic Chemical code (ATC code) of the drug, the date of dispensing, the number of units dispensed, and the prescribed daily dose. The IADB.nl does not include information on medical conditions or on medication dispensed during hospitalization. The IADB.nl has been validated and used for many studies in the pharmacoepidemiology field [13,14]. Ethical approval is not required for studies using anonymous data in The Netherlands.

2.3 Study population

The study population consists of all adults known in the IADB.nl database, aged 45 years or older in 1 January of each study year and receiving at least two prescriptions for medication used for diabetes treatment (ATC codes: [A10A, insulin and analogues] and/or A10B [blood glucose lowering drugs, excluding insulin]) in the calendar year, and with follow-up data for at least 12 months.

2.4 Polypharmacy Definition

Polypharmacy was defined as the chronic use of at least five drugs concurrently [15]. Chronic use was defined as a drug being dispensed for at least 90 days or at least three times in a four-month period between 1 September and 31 December in the study year. Given the definition of the Dutch guideline [15], all drugs intended for topical use, contrast media, radiopharmaceuticals, surgical dressings, and general nutrients (ATC codes belonging to D, V, Y, Z), as well as records with invalid ATC codes were excluded from the calculation of polypharmacy prevalence. To determine the number of different drugs for chronic use, the third level of the ATC code (ATC3), which describes pharmacological subgroups, was used following the Dutch guideline on polypharmacy [15]. This implies that dispensing of different substances within the same pharmacological subgroup, e.g. beta-blocking agents, are counted as one chronic drug.

2.5 Potentially inappropriate medication (PIM) Criteria

We used the American Geriatrics Society (AGS) Beers 2015 criteria for assessment of PIM for patients of 65 years and older [16], and the PRescribing Optimally in Middle-aged People’s Treatments (PROMPT) criteria for patients 45-64 years old [17]. We included only those criteria for the assessment of a PIM that can be applied in a pharmacy database without clinical information. Beers criteria consist of a list of therapeutic classes and individual drug substances which should be avoided in people aged 65 and older in most circumstances or with some specified medical conditions [16]. The PROMPT criteria were in part derived from the Beers criteria but in addition include criteria related to drug-drug interactions [17]. Since such drug-drug interactions are not included in Beers, we excluded them from our PIM assessment. In total, 14 criteria from Beers and nine criteria from PROMPT were excluded because: (a) they could not be assessed with information available in the IADB database, (b) the drugs were not available on the Dutch market during the study period, or (c) they concerned drug-drug interactions. This resulted in a total of 15 PIMs for therapeutic classes and 15 PIMs for individual drug substances from Beers and PROMPT criteria that were used in our study (Figure 1). There were 24 PIMs from Beers applied for patients of 65 years and older, and 13 PIMs from PROMPT applied for patients between 45 and 64 years of age. In addition, there were nine PIMs from PROMPT that were not part of the Beers criteria, which we considered also of relevance for older patients. For comparison, these nine PIMs were also applied for patients of 65 years and older (Appendix Table 1).

2.6 Outcome

The primary outcome was polypharmacy prevalence per calendar year, stratified by age (45-54 years, 55-64 years and ≥65 years). Secondary outcomes were the prevalence of any PIM and the prevalence of the individual PIMs in polypharmacy patients. For the outcome of any PIM, patients were categorized as having no PIM or at least one PIM, using the 24 PIM criteria from Beers, and the 13 PIM criteria from PROMPT.

2.7 Analysis

The prevalence of polypharmacy and PIM were assessed for each year and expressed as percentages with 95% Confidence Intervals (CI). Additionally, absolute differences in these prevalences over the study period are presented. The prevalence of polypharmacy was calculated by dividing the number of people with diabetes on polypharmacy (numerator) by the total number of diabetes patients in each year (denominator). The prevalence of PIM was calculated by dividing the number of people with diabetes and polypharmacy who had at least one PIM by the total number of people with diabetes and polypharmacy. Linear-by-linear association with Chi-square analysis and regression analysis were used to test for significant linear or quadratic trends of the prevalence rates. Statistical significance was evaluated at the 2-sided α=0.05 level. All statistical analyses were carried out using SPSS software (version 23.0; IBM, Armonk, NY, USA).