One‐year persistence of potentially inappropriate medication use in older adults: A population‐based study

Abstract

Aims

To assess the 1‐year persistence of potentially inappropriate medication (PIM) use and identify associated factors in community‐dwelling older adults in Quebec, Canada.

Methods

A population‐based cohort study was conducted using the Quebec Integrated Chronic Disease Surveillance System. Individuals insured by the public drug plan and aged ≥66 years who initiated a PIM between 1 April 2014 and 31 March 2015 were followed‐up for 1 year. PIMs were identified using the 2015 Beers criteria. One‐year persistence of PIM use was defined as continuous treatment with any PIM, without interruption for more than 60 days between prescriptions refills. Poisson regression models were performed to identify factors associated with 1‐year persistence of any PIM.

Results

In total, 25.1% of PIM initiators were persistent at 1 year. In non‐persistent individuals, the median time to PIM discontinuation was 31 days (interquartile range 21‐92). Individuals were more persistent at 1 year with antipsychotics (43.9%), long‐duration sulphonylureas (40.2%), antiarrhythmics/immediate‐release nifedipine (36.5%) and proton pump inhibitors (36.0%). Factors significantly associated with persistence were an increased age, being a man and having a high number of medications and chronic diseases, especially dementia, diabetes and cardiovascular diseases.

Conclusions

One‐quarter of community‐dwelling older adults are continuously exposed to PIMs. To optimize medication prescribing in the older population, further interventions are needed to limit the use of PIMs most likely to be continued, especially in individuals most at risk of being persistent and also particularly vulnerable to adverse events.

What is already known about this subject

• The use of potentially inappropriate medications (PIMs) is common among older people living in the community.

• Despite several studies on the prevalence or the trend in PIM use, no study has yet determined the persistence of PIM use, that is, the continuous use without refill interruption by the same individuals.

What this study adds

• One‐quarter of older adults initiating at least one PIM persist after 1 year.

• Interventions are needed to limit the use of PIMs in individuals most at risk of being persistent who are vulnerable to adverse events (men, age > 75 years, those with polypharmacy and multimorbidity including some chronic diseases).

1. INTRODUCTION

Medication regimens remain complex and challenging in older people. Older adults may indeed receive multiple medications due to the presence of many chronic conditions. They often receive potentially inappropriate medications (PIMs), defined as medications whose risks outweigh expected benefits and/or medications for which a safer effective alternative exists for the same indication.1 Many studies conducted on the prevalence of PIM use in different settings2, 3, 4, 5 have revealed a significant use of these medications.

However, these studies only show instantaneous pictures of PIM use. They neither provide information on the evolution of PIM use over time nor allow determination of whether PIMs are used sporadically or chronically. Some longitudinal6, 7, 8, 9 or repeated cross‐sectional studies10, 11, 12, 13, 14, 15 have examined the trends in PIM use over time, which have shown mixed results. The prevalence of PIM use increased from 39.7% to 45.6% between 2010 and 2015 in Ireland11 while the proportion of PIM users decreased from 50.2% to 47.2% in 2011‐2016 in Canada10 and from 26.4% to 23.1% in 2010‐2016 in Germany.6 On the contrary, PIM use was constant in the UK and the Netherlands, where proportions of PIM users ranged from 38.7% in 2003‐2004 to 38.4% in 2011‐2012,13 and 34.7% in 2007 to 34.4% in 2014,9 respectively. Other studies have investigated chronic use of PIMs in community‐dwelling older adults using different definitions of such chronic use (e.g. presence of one, two or three prescriptions of at least one PIM during a period, with or without a defined number of days' supply). The proportion of chronic users was thus 33.7% in 2016 in Canada,10 17.4% in 2011‐2012 in the UK,13 13% in 2010 in Germany16 and 53% in 2013‐2015 in Taiwan.17 In a recent Finnish study conducted among community‐dwelling older adults, PIM initiators used at least one PIM during a cumulative average duration of 2.5 years over a 12‐year follow‐up.18 These studies may overestimate actual chronic PIM exposure as the treatment interruptions have not been considered. In other words, it is not known whether older adults use PIMs for prolonged periods of time with brief interruptions in treatment or, conversely, for a rather short period of time. Understanding the behaviour of PIM use in older adults is important as this will influence how exposure to PIMs can be prevented.

To the best of our knowledge, no study has yet accurately evaluated whether and for how long PIMs are used continuously, i.e. without refill interruption, by the same individuals. In the context of adherence to drug treatment, persistence is defined as the continuous use of medications for the intended duration of the treatment, especially for chronic diseases.19 This appears to be the most relevant method for estimating the continuous use of PIMs by the same individuals. However, it is the opposite of the commonly sought goal of adherence to treatment. Since PIMs are medications that should be avoided in the older population, the goal is that individuals do not persist with PIM use and that an interruption in PIM is detected when it occurs.

The present study thus aims to (a) determine the 1‐year persistence of PIM use and (b) investigate factors associated with persistence of PIM use in community‐dwelling adults aged 66 years and older, in the province of Quebec, Canada.

2. METHODS

2.1. Study design and data sources

A retrospective population‐based cohort study was performed in older adults with chronic diseases or potentially at risk of developing one chronic disease using the Quebec Integrated Chronic Disease Surveillance System (QICDSS). The QICDSS, managed by the Institut National de Santé Publique du Québec (INSPQ) and developed for the surveillance of the most prevalent chronic diseases in Quebec, links patient level records of five health administrative databases (demographic records, hospital discharge database, vital statistics death registry, physician claims database and the Quebec Health Insurance Board [RAMQ] drug plan database).20 Due to the criteria used to identify individuals to be included in the QICDSS and the high prevalence of chronic diseases in the older population in Quebec, the QICDSS includes about >99% of the population aged ≥65 years for medical data and also 90% for drug claims since older adults aged ≥65 years are automatically covered by the public dug plan.20

2.2. Study population

Individuals aged 66 years or older as of 1 April 2014 who were alive and continuously insured by the public drug plan between 1 April 2013 and 31 March 2015 were eligible for inclusion. From this population, we identified new users of PIMs (i.e. who initiated a PIM during the 2014 fiscal year [1 April 2014‐31 March 2015] and who had not used any PIM in the previous 365 days). The date of PIM initiation (the date of first drug claim) constituted the index date. In order to measure persistence, we excluded PIM initiators who died during the 2015 fiscal year (1 April 2015‐31 March 2016), who stayed in hospital for more 90 days between 1 April 2014 and 31 March 2016 and for whom a 1‐year follow‐up after the index date was not available in the public drug plan.

2.3. Identification of PIMs

PIMs consisted of drugs that should be avoided generally in older adults according to the American Geriatrics Society 2015 version of the Beers criteria (Table 2).21 The list used for the study was adapted to the Canadian market and some drugs for which clinical data were required to identify them as PIMs were excluded since these data were not available in the QICDSS (Appendix Table A1). Medications were classified based on non‐proprietary name, content and form codes. In accordance with the Beers criteria, we took into consideration particularities such as duration of use, concomitant use of medications and the presence of certain chronic diseases in determining the inappropriate status of certain medications. Complete details of the study methodology for identifying PIMs were presented in an earlier study.22

Table 2.

PIMs initiated at the index date by older Quebecers according to the 2015 Beers criteria, by sex

PIM class	Total (n = 75,844)		Women (n = 42,575)		Men (n = 33,269)
	n	%	n	%	n	%
Benzodiazepines	25,678	33.9	15,887	37.3	9791	29.4
Short‐ and intermediate‐acting	21,962	29.0	13,676	32.1	8,286	24.9
Long‐acting	3,852	5.1	2,283	5.4	1,569	4.2
PPIs (>56 days)	25,520	33.7	14,002	32.9	11,518	34.6
Skeletal muscle relaxants	8,255	10.9	4,196	9.9	4,059	12.2
Antipsychotics	4,598	6.1	2,591	6.1	2,007	6.0
First‐generation	275	0.4	159	0.4	116	0.4
Second‐generation	4,414	5.8	2,487	5.8	1,927	5.8
First‐generation antihistamines	4,156	5.5	2,246	5.3	1,910	5.7
Tricyclic antidepressants and paroxetine	3,151	4.2	1,962	4.6	1,189	3.6
Long‐duration sulphonylureas	2,181	2.9	792	1.9	1,389	4.2
Central alpha‐blockers	1,385	1.8	930	2.2	455	1.4
NSAIDsa	1,297	1.7	404	1.0	893	2.7
Antiarrhythmics and immediate‐release nifedipineb	723	1.0	280	0.7	443	1.3
Peripheral alpha‐1 blockers	694	0.9	267	0.6	427	1.3
Other analgesicsc	359	0.5	199	0.5	160	0.5
Antispasmodics	207	0.3	112	0.3	95	0.3
Hormonesd	125	0.2	102	0.3	23	0.07
Desmopressine	105	0.1	37	0.09	328	0.2
Antiparkinsonian agents	43	0.06	12	0.03	31	0.09
Antithrombotics	5	0.01	< 5	<0.01	< 5	<0.01
Barbiturates	5	0.01	< 5	<0.01	< 5	<0.01
Non‐benzodiazepine hypnotics	< 5	<0.01	< 5	<0.01	< 5	<0.01

Abbreviations: NSAIDs, non‐steroidal anti‐inflammatory drugs; PIMs, potentially inappropriate medications; PPIs, proton pump inhibitors.

^aChronic use of NSAIDs (>90 days), except for indomethacin and ketorolac, which were inappropriate when used once.

^bAntiarrhythmics include disopyramide, dronedarone, digoxin >0.125 mg/d and amiodarone.

^cOther analgesics include pentazocine and meperidine.

^dHormones include desiccated thyroid, estrogens with or without progestins and megestrol.

2.4. Persistence of PIM use

Persistence was defined as a continuous treatment with at least one PIM, starting at the date of initiation (index date) up to a year of follow‐up. Individuals were therefore considered persistent with PIMs if there was no period of interruption between two consecutives claims of any PIM for a year. Hence, to be considered persistent, the period between two consecutive PIMs claims had to be less than or equal to the duration of treatment (number of days' supply on the last claim), plus a duration of a permissible gap of 60 days (Appendix Figure A1). The discontinuation date was defined as the date of the last claim of a PIM, plus the treatment duration of the last PIM. Switches between different PIMs were not considered as treatment discontinuation, as we measured the 1‐year persistence with any PIM. When an additional PIM dispensing occurred before the date of the initial PIM discontinuation, a new period of treatment was defined from the date of the new dispensing (overlapping days supplied between consecutive drugs claims were not considered). Hospital stays were considered for the calculation of persistence as drug use during hospitalization was not captured in administrative databases. Since hospitalized patients only take drugs dispensed by the hospital, treatments received before hospitalization are not taken and are therefore stored by the patients; we thus assumed that there was probably not discontinuity in these treatments neither during hospitalization nor after discharge from hospital. Consequently, if the hospitalization occurred during the coverage period of a prescription, the number of days' supply overlapping the hospitalization was deferred to the end of the hospitalization. Overall, most prescriptions for chronic treatments are dispensed for 30 days in Quebec. However, drugs included in the Beers criteria also encompass drugs for acute treatments and/or to be taken as needed. For this purpose, a 60‐day permissible gap was chosen to capture medication intake as needed and repeated acute treatments.

In addition to 1‐year persistence with any PIMs, we measured and described the 1‐year persistence of the most initiated PIM classes during the 2014 fiscal year in subcohorts analyses. Individuals who initiated a specific PIM during the 2014 fiscal year were followed‐up for 1 year after the specific index date (date of the first specific PIM dispensing in the 2014 fiscal year).

2.5. Covariates

Baseline characteristics of patients, such as age, sex, urban‐rural residence, socioeconomic status and presence of chronic diseases, were determined on 1 April 2014. Rural (<10,000) or urban (≥10,000) residence was categorized according to a geographic area associated with the postal code of residence. The material and social deprivation index is an ecological substitute of the socioeconomic status developed by the INSPQ and is defined in quintiles.23 We included chronic diseases monitored by the QICDSS, which all have validated case definitions: hypertension, cardiovascular diseases (ischemic heart disease, stroke and heart failure), diabetes, respiratory diseases (chronic obstructive pulmonary disease and asthma), osteoporosis, mental disorders (anxio‐depressive disorders and schizophrenia), and Alzheimer's disease and related dementia.20 In addition, the diagnoses of atrial fibrillation (ICD‐9 427.3, ICD‐10‐CA I48) and bipolar disorders (ICD‐9 296, ICD‐10‐CA F30‐F31) gathered from the hospital discharge and physician claim databases which are required for the identification of some PIMs were added to the list of chronic diseases included in our study. The variable “number of chronic diseases” was calculated by the sum of the chronic diseases presented by each individual on 1 April 2014. The variable “number of different medications” was computed by the sum of the different non‐proprietary names received for each individual in the year before the first PIM prescription.

2.6. Statistical analysis

Descriptive analyses were conducted to describe patients' characteristics and the initiated PIMs at the index date. The Chi‐square test and Mann‐Whitney test were used to compare categorical and quantitative variables, respectively, with statistical significance set at P < .05.

The persistence of PIM use was estimated by calculating the number and proportion of patients who persisted with any PIM over a 1‐year period after the index date. The length of time (days) between index and discontinuation dates was also calculated. Kaplan‐Meier curves were used to describe the proportion of individuals who persisted with any PIM over a 1‐year period. Individual follow‐up was censured at the date of the first PIM discontinuation (failure to refill any PIM within an allowed period) or at 1 year (date of the end of follow‐up), whichever came first.

Similarly, we measured the 1‐year persistence with specific PIM classes by calculating the number and proportion of patients who persisted with a specific PIM for 1 year after the specific index date.

Poisson regression models with robust error variance estimator were conducted to explore the association between 1‐year persistence with at least one PIM as a binary dependent variable and patients' characteristics as independent variables. As the number of medications is an intermediate variable and may underestimate the association between chronic diseases and 1‐year persistence,24, 25 we built two models. In the first model, the variable number of medications (Table 4, model 1) was entered as the covariate while we removed this variable in the second model (Table 4, model 2). In two additional models, chronic diseases were entered as the type of chronic diseases with (Appendix Table A2, model 3) or without (Appendix Table A2, model 4) the variable number of medications. The results were reported as adjusted rate ratios (RR) with their 95% confidence intervals (95% CIs). Sensitivity analyses were performed to test the sensitivity of the permissible gap on the measure of persistence with any PIM and on the identification of associated factors. Permissible gaps of 15, 30, 45 and 90 days were tested. All the analyses were performed with SAS® software (SAS Institute, version 9.4, NC, USA).

Table 4.

Factors associated with 1‐year persistence with any PIM in older Quebecers

		Model 1d	Model 2d
	RR (95% CI)	RRa (95% CI)	RRa (95% CI)
Age, years a
66‐75	1.00	1.00	1.00
76‐85	1.32 (1.28‐1.35)	1.20 (1.17‐1.23)	1.21 (1.18‐1.25)
86 and older	1.82 (1.76‐1.89)	1.57 (1.51‐1.62)	1.59 (1.53‐1.65)
Sex
Men	1.00	1.00	1.00
Women	0.90 (0.87‐0.92)	0.87 (0.85‐0.90)	0.87 (0.85‐0.89)
Region a
Urban	1.00	1.00	1.00
Rural	1.05 (1.02‐1.09)	1.06 (1.03‐1.09)	1.07 (1.04‐1.10)
Material deprivation index a
First quintile, least deprived	1.00	1.00	1.00
Second quintile	1.15 (1.10‐1.21)	1.14 (1.09‐1.20)	1.15 (1.10‐1.20)
Third quintile	1.26 (1.16‐1.28)	1.19 (1.14‐1.24)	1.19 (1.14.‐1.25)
Fourth quintile	1.23 (1.17‐1.29)	1.19 (1.14‐1.24)	1.20 (1.15‐1.25)
Fifth quintile, most deprived	1.25 (1.19‐1.31)	1.19 (1.13‐1.25)	1.20 (1.15‐1.26)
Social deprivation index a
First quintile, least deprived	1.00	1.00	1.00
Second quintile	1.02 (0.97‐1.07)	1.00 (0.96‐1.05)	1.00 (0.96‐1.05)
Third quintile	1.03 (0.98‐1.08)	1.01 (0.97‐1.05)	1.01 (0.97‐1.05)
Fourth quintile	1.07 (1.02‐1.12)	1.04 (1.00‐1.09)	1.05 (1.00‐1.09)
Fifth quintile, most deprived	1.12 (1.07‐1.18)	1.10 (1.05‐1.15)	1.10 (1.06‐1.15)
Number of chronic diseases b
0	1.00	1.00	1.00
1	1.14 (1.09‐1.21)	1.06 (1.01‐1.11)	1.12 (1.07‐1.18)
2	1.29 (1.23‐1.36)	1.09 (1.04‐1.15)	1.23 (1.17‐1.29)
3	1.52 (1.43‐1.60)	1.17 (1.11‐1.24)	1.40 (1.33‐1.47)
4 or more	1.96 (1.85‐2.06)	1.34 (1.27‐1.41)	1.71 (1.63‐1.80)
Number of different medications c
0‐4	1.00	1.00
5‐9	1.27 (1.22‐1.32)	1.18 (1.13‐1.22)
10‐14	1.63 (1.56‐1.70)	1.39 (1.34‐1.50)
15 or more	2.06 (1.97‐2.16)	1.62 (1.55‐1.70)

Abbreviations: CI, confidence interval; PIM, potentially inappropriate medication; RR, crude rate ratio; RRa, adjusted rate ratio.

^aAt the beginning of the 2014 fiscal year (1 April 2014).

^bChronic diseases at the beginning of the 2014 fiscal year (1 April 2014): Alzheimer's disease and related dementia, cardiovascular diseases, diabetes, hypertension, mental disorders, osteoporosis, respiratory diseases.

^cMedications received for each individual 365 days before the index date (first PIM dispensed in the 2014 fiscal year [1 April 2014‐31 March 2015]).

^dPoisson regression models, including 1‐year persistence with at least one PIM as binary‐dependent variable. All covariates (age, sex, region, material and social deprivation index, number of chronic diseases and number of different medications) were entered as independent variables, except for model 2 (exclusion of the number of different medications).

2.7. Ethical approval

The use of QICDSS for surveillance purposes has been approved by the government bodies, the Public Health Ethics Committee and the Commission D'accès à L'information du Québec.

2.8. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY.

3. RESULTS

3.1. Characteristics of the study population

A total of 75,844 individuals were included to study persistence with PIM use (Figure 1). The mean age of the study population was 74.3 years (SD 6.8) and 56.1% were women (Table 1). The most prevalent diseases were hypertension (63.4%), cardiovascular diseases (35.8%) and osteoporosis (26.9%), and 15.5% of the population had at least four chronic diseases. A large proportion of individuals (86.6%) received five medications or more in the year before initiating a PIM and 46.7% received at least 10 medications. Almost all individuals (96.7%) initiated only one Proton pump inhibitors (PIMs) at the index date. The most frequently initiated PIM classes were benzodiazepines (33.9% of individuals), PPIs (33.7%), skeletal muscle relaxants (10.9%), antipsychotics (6.1%), first‐generation antihistamines (5.5%), tricyclic antidepressants and paroxetine (4.2%), and long‐duration sulphonylureas (2.9%) (Table 2).

Figure 1.

Flow chart of the study population

Table 1.

Characteristics of individuals who initiated a PIM in Quebec (Canada) during the 2014 fiscal year, according to 1‐year persistence of any PIM use

	Total (n = 75,844)	Persistent users (n = 19,051)	Non‐persistent users (n = 56,793)	P value
	n (%)	n (%)	n (%)
Age, years a				<.0001
Mean (SD)	74.3 (6.8)	75.2 (7.4)	73.9 (6.5)
Median (IQR)	73 (69‐79)	75 (69‐81)	72 (69‐78)
Age, years a				<.0001
66‐75	47,432 (62.5)	10,296 (54.0)	37,136 (65.4)
76‐85	22,587 (29.8)	6,453 (33.9)	16,134 (28.4)
86 and older	5,825 (7.7)	2,302 (12.1)	3,523 (6.2)
Sex				<.0001
Women	42,575 (56.1)	10,174 (53.4)	32,401 (57.1)
Men	33,269 (43.9)	8,877 (46.6)	24,392 (42.9)
Region a
Rural	16,689 (22.0)	4,363 (22.9)	12,326 (21.7)	.0148
Urban	59,028 (77.8)	14,641 (76.9)	44,387 (78.2)
Missing	127 (0.2)	47 (0.2)	80 (0.1)
Material deprivation index a				<.0001
First quintile, least deprived	13,484 (17.8)	2,821 (14.8)	10,660 (18.8)
Second quintile	13,354 (17.6)	3,224 (16.9)	10,130 (17.8)
Third quintile	13,918 (18.4)	3,540 (18.6)	10,378 (18.3)
Fourth quintile	14,849 (19.5)	3,820 (20.1)	11,029 (19.4)
Fifth quintile, most deprived	14,386 (19.0)	3,755 (19.7)	10,631 (18.7)
Missing	5,856 (7.7)	1,891 (9.9)	3,965 (7.0)
Social deprivation index a				<.0001
First quintile, least deprived	12,335 (16.3)	2,880 (15.1)	9,455 (16.6)
Second quintile	13,401 (17.7)	3,198 (16.8)	10,203 (18.0)
Third quintile	14,810 (19.5)	3,550 (18.7)	11,260 (19.8)
Fourth quintile	14,971 (19.7)	3,736 (19.6)	11,235 (19.8)
Fifth quintile, most deprived	14,471 (19.1)	3,796 (19.9)	10,675 (18.8)
Missing	5,856 (7.7)	1,891 (9.9)	3,965 (7.0)
Chronic diseases a
Hypertension	48,092 (63.4)	13,152 (69.0)	34,940 (61.5)	<.0001
Cardiovascular diseases	27,183 (35.8)	8,121 (42.6)	19,062 (33.6)	<.0001
Osteoporosis	20,372 (26.9)	5,305 (27.9)	11,495 (20.2)	<.0001
Respiratory diseases	17,371 (22.9)	4,869 (25.6)	15,503 (27.3)	<.0001
Diabetes	16,800 (22.2)	4,911 (25.8)	12,460 (21.9)	<.0001
Mental disorders	5,149 (6.8)	1,494 (7.8)	1,339 (2.4)	<.0001
Alzheimer's disease and related dementia	2,833 (3.7)	1,404 (7.4)	3,745 (6.6)	.0002
Number of chronic diseases a				<.0001
0	10,375 (13.7)	1,929 (10.1)	8,446 (14.9)
1	21,264 (28.0)	4,526 (23.8)	16,738 (29.5)
2	19,947 (26.3)	4,799 (25.2)	15,148 (26.7)
3	12,538 (16.5)	3,534 (18.5)	9,004 (15.8)
4 or more	11,720 (15.5)	4,263 (22.4)	7,457 (13.1)
Number of different medications b				<.0001
0‐4	10,183 (13.4)	3,698 (19.4)	16,363 (28.8)
5‐9	30,247 (39.9)	7,288 (38.3)	23,820 (42.0)
10‐14	21,560 (28.4)	4,998 (26.2)	11,610 (20.4)
15 or more	13,854 (18.3)	3,067 (16.1)	5,000 (8.8)
Number of different PIMs c				<.0001
1	73,274 (96.7)	17,807 (93.5)	55,467 (97.7)
2	1,977 (2.6)	932 (4.9)	1,045 (1.8)
3	485 (0.6)	255 (1.3)	230 (0.4)
4 or more	108 (0.1)	57 (0.3)	51 (0.1)

Abbreviation: PIMs, potentially inappropriate medications.

^aAt the beginning of the 2014 fiscal year (1 April 2014).

^bMedications received for each individual 365 days before the index date (first PIM dispensed in the 2014 fiscal year [1 April 2014‐31 March 2015]).

^cPIMs received for each individual at index date during the 2014 fiscal year.

3.2. Persistence of PIM use

Using a permissible gap of 60 days, 19,051 individuals (25.1%, 95% CI 24.8‐25.4) had a persistent use of at least one PIM at 1 year (Table 1 and Figure 2). Globally, 50% of older adults had stopped their PIM after 69 days (interquartile range [IQR] 31‐365) (Figure 2). In non‐persistent individuals specifically, the median time to first discontinuation was 31 days (IQR 21‐92).

Figure 2.

One‐year persistence with any PIM in older Quebecers who initiated a PIM during the 2014 fiscal year

The proportion of individuals with 1‐year persistence was higher for those who initiated antipsychotics (43.9%), long‐duration sulphonylureas (40.2%), antiarrhythmics (i.e. amiodarone, digoxin >0.125 mg/d, disopyramide and dronedarone)/immediate‐release nifedipine (36.5%) and PPIs (36.0%). Conversely, the proportion of persistent individuals was weaker for nonsteroidal anti‐inflammatory drugs (NSAIDs) (5.7%), first‐generation antihistamines (2.5%) and skeletal muscle relaxants (1.5%) (Table 3).

Table 3.

One‐year persistence with PIM classes initiated by older Quebecers during the 2014 fiscal year

PIM classes (number of PIM initiators)	n	%	95% CI
Any PIM (n = 75,844)	19,051	25.1	24.8‐25.4
Antipsychotics (n = 6,307)	2,769	43.9	42.7‐45.1
First‐generation (n = 551)	90	16.3	13.2‐19.4
Second‐generation (n = 6,061)	2,696	44.5	42.6‐46.4
Long‐duration sulphonylureas (n = 2,637)	1,059	40.2	37.2‐43.3
Antiarrhythmics and immediate‐release nifedipine (n = 889)a	324	36.5	33.3‐39.6
Proton pump inhibitors (>56 days) (n = 30,590)	11,014	36.0	35.1‐36.9
Tricyclic antidepressants and paroxetine (n = 4,477)	711	15.9	14.8‐17.0
Central alpha‐blockers (n = 1,834)	283	15.4	13.7‐17.1
Benzodiazepines (n = 30,256)	3,294	10.9	10.5‐11.3
Short‐ and intermediate‐acting (n = 26,209)	2,854	10.9	10.5‐11.3
Long‐acting (n = 5,021)	428	8.5	7.7‐9.3
NSAIDs (n = 1,585)b	91	5.7	4.6‐6.8
First‐generation antihistamines (n = 5,466)	136	2.5	2.1‐2.9
Skeletal muscle relaxants (n = 10,314)	156	1.5	1.3‐1.7

Abbreviations: CI, confidence interval; NSAIDs, non‐steroidal anti‐inflammatory drugs; PIM, potentially inappropriate medication.

^aAntiarrhythmics include disopyramide, dronedarone, digoxin >0.125 mg/d and amiodarone.

^bChronic use of NSAIDs (>90 days), except for indomethacin and ketorolac, which were inappropriate when used once.

3.3. Factors associated with 1‐year persistence of PIMs

The main factors significantly associated with 1‐year persistence with any PIM were an increased age (76‐85 years: adjusted RR 1.20, 95% CI 1.17‐1.23; 86 years and older: 1.57, 95% CI 1.51‐1.62) and having a high number of chronic diseases (RR range 1.06‐1.34) and medications (RR range 1.18‐1.62) (Table 4, model 1). Women were less likely to be persistent with PIM treatment than men RR (1.10, 95% CI 1.10–0.90). The risk of being persistent increased by 10% for the most socially deprived individuals (RR 1.10, 95% CI 1.05‐1.15) while there was no significant difference between the different quintiles of the material deprivation index (except for the first quintile, least deprived individuals). Regarding the type of chronic diseases, Alzheimer's disease and related dementia RR (1.81, 95% CI 1.74–1.89), diabetes RR (1.25, 95% CI 1.21–1.28) and cardiovascular diseases RR (1.14, 95% CI 1.11–1.17) were associated with the risk of being persistent with any PIM at 1‐year (Appendix Table A2, model 4).

3.4. Sensitivity analysis

The results of sensitivity analysis showed that the larger the gap, the higher the proportion of persistence with any and specific PIMs (Appendix Table A3). Factors associated with 1‐year persistence with any PIM remained the same regardless of the permissible gap but there were stronger associations when the permissible gap was shorter (Table 4 and Appendix Table A4).

4. DISCUSSION

One‐quarter of community‐dwelling older adults (25.1%) who initiated at least one PIM were persistent at 1 year, that is, they used a PIM on a continuous basis throughout the year. In other words, continuous exposure to PIMs does not appear to be as chronic in three out of four community‐dwelling older adults since half of them do not refill their initial PIM prescription for more than a 1‐month period during a 1‐year follow‐up. Antipsychotics, long‐duration sulphonylureas, antiarrhythmics/immediate‐release nifedipine and PPIs were PIM classes with the highest 1‐year persistence. Older people, men, multimorbid individuals, notably those with Alzheimer's disease and related dementia, diabetes and cardiovascular diseases, and those with a high number of different medications were the most at risk of being persistent with PIM treatment at 1 year.

Our results are difficult to compare with other studies because we used an original approach to estimate exposure to PIM without treatment interruption. Only one Finnish study suggested that exposure to PIM in community‐dwelling older adults was probably repeated over time, as the cumulative average duration to PIM exposure was of 2.5 years over a 12‐year follow‐up.18 However, this study did not indicate if the older adults use PIMs for prolonged periods with short treatment interruptions or vice versa. Our study demonstrates that a high proportion of older adults did not refill their initial PIM prescription for more than a 1‐month period during a 1‐year follow‐up. These results seem robust as if we extend the gap period to 3 months; only 3% of persistent older adults are identified in addition. Nevertheless, the continuous use of PIMs may potentially increase iatrogenic risk. This raises important clinical questions because the individuals most at risk of being persistent with PIMs (older patients, people with a large number of medications and chronic diseases) are particularly vulnerable to adverse drug events, which may lead to negative health outcomes. Indeed, PIM use has been associated with adverse events such as hospital admissions, increased mortality or higher costs.18, 26, 27 In addition, previous studies have shown a positive association between the number of PIMs and the occurrence of adverse events28, 29, 30 but no study has evaluated the relationship between duration of PIM exposure and risk of adverse events.

Our results show that 1‐year persistence varied by PIM classes. Persistence was higher for certain PIM classes more often used in the treatment of chronic diseases (e.g. long‐duration sulphonylureas or antiarrhythmics/immediate‐release nifedipine). In addition, 1‐year persistence with PPIs was high (36.0%), which is consistent with earlier studies supporting the widespread use of this class for much longer than indicated or without a clear indication.10, 11, 31, 32 This illustrates the need to review the prescription of PPIs given the risks associated with their long‐term use (pneumonia, Clostridium difficile infection, fractures).33, 34 Moreover, the persistence with potentially inappropriate antipsychotic treatment was significant in our study (43.9%). These results corroborate with the fact that antipsychotics are often prescribed for the treatment of behavioural symptoms in the older population, particularly in those with dementia.35, 36 One‐year persistence was lower for some PIM classes, mostly those used in the treatment of acute diseases or for an intermittent use (e.g. NSAIDs, first‐generation antihistamines or skeletal muscle relaxants). Although benzodiazepines are the most frequently initiated class in our study (33.9%) and most studies report a high prevalence of this class among older people worldwide,10, 37, 38, 39 the 1‐year persistence with benzodiazepines was weak (10.9%). This result compares with the proportions of 12.1% and 7.6% of older adults in Quebec and in Finland, respectively, who were chronic users of benzodiazepines (defined as the presence of at least two claims and 180 cumulative supply days over the year).10, 40 However, a higher proportion of chronic users (31.4%), defined as 120 days' supply or more per year, was observed in US community‐dwelling older adults (65‐80 years old) in 2008.41 Despite the low persistence with benzodiazepines found in our study, it would be necessary to ensure that the use of benzodiazepines was not replaced by other non‐appropriate sedatives such as zopiclone, a phenomenon that was observed elsewhere,42 or by sedative atypical antipsychotics (e.g. quetiapine), for which 1‐year persistence was not negligible in our study.

We identified several factors associated with persistence of PIM use, such as increased age, being a man, multimorbidity and polypharmacy. These factors were also reported in previous repeated cross‐sectional studies investigating the chronic use of PIMs (defined as the presence of at least a certain number of PIM claims in the year).13, 16 However, contrary to our results, women were more often at risk of using PIMs. In our study, the higher proportion of men who persist with PIMs at 1 year may be explained in part by the fact that men have initiated PIMs with higher persistence (e.g. long‐duration sulphonylureas, antiarrhythmics/immediate‐release nifedipine and PPIs). Among the chronic diseases studied, individuals with Alzheimer's disease and related dementia were 80% more at risk of persisting with PIMs at 1 year. This may be related to the significant 1‐year persistence with medications acting on the central nervous system (e.g. antipsychotics), which are commonly used for controlling behavioural disorders associated with these diseases.35, 43 Previous findings seem contradictory as to the association between dementia and PIM use. Indeed, prior studies found that PIM use is common in older adults with Alzheimer's disease and related dementia,43, 44 while other studies have observed the opposite.45, 46 Finally, in our study, the risk of being persistent with PIMs at 1 year increased by 25% and 14% for individuals with diabetes and cardiovascular diseases, respectively. This finding is consistent with the fact that these are two chronic diseases for which subjects are more likely to take their treatment on a regular basis and are therefore more likely to be persistent.

The sensitivity analyses showed variations in the measurement of persistence with any and specific PIMs depending on the different permissible gaps tested. The changes observed in measuring the persistence with any PIM may be partly due to the consideration of several specific PIMs with different treatment goals (for chronic or acute treatment). In addition, different behaviours were captured when the duration of the permissible gap was changed and may impact the 1‐year persistence with any PIM. For instance, with the 90‐day permissible gap, we probably selected more individuals who resumed their treatment after a discontinuation, whereas with a 15‐ or 30‐day permissible gap, we captured more individuals taking their treatments regularly. However, our results seem robust because there was little variation in the order of persistence with the different PIMs. For instance, regardless of the permissible gap, antipsychotics and long‐duration sulphonylureas remain the two classes of PIMs with the greatest persistence while first‐generation antihistamines and skeletal muscle relaxants are the two classes with less persistence. Moreover, there was no variation in the factors associated with 1‐year persistence using different permissible gaps. Thus, maintaining the choice of the 60‐day permissible gap for the main analyses was a good compromise to take into account both chronic treatments traditionally prescribed for 30 days but also to capture repeated acute treatments or medication intake as needed.

Given the higher risk of persistence among vulnerable older adults in our study, further interventions are needed to limit the use of PIMs. Deprescribing may be an optimal way to reduce existing PIMs in the older population.47 Indeed, deprescribing interventions are reported to be an effective and safe way to reduce the number of medications and thus inappropriate prescribing.48, 49 In addition, to be successful, deprescribing interventions require shared decision making between patients and prescribers.47 The importance of a good relationship between these two actors has been confirmed in previous studies where a significant proportion of older adults were willing to stop taking one or more medications if their physician said it was possible.50, 51, 52, 53, 54, 55, 56, 57 Moreover, patient‐centred educational interventions such as an increased awareness of potential harms associated with PIM use by healthcare providers were approaches for successful patient‐initiated deprescribing interventions.58

To the best of our knowledge, this is the first study to investigate persistence of PIM use, that is, the continuous use of PIMs without interruption of treatment during a 1‐year follow‐up in PIM initiators. Also, this present study was conducted from a population‐based cohort of community‐dwelling older adults using the QICDSS, which includes practically the entire elderly population insured by the public drug plan in the province of Quebec.

However, some limitations should be mentioned. First, drugs dispensed in hospitals stays, in long‐term care facilities and reimbursed by private assurance companies are not available in the QICDSS. Likewise, drugs not reimbursed by the public drug plan and dispensed over the counter were not captured in the RAMQ drug plan database. In addition, we have not included all the medications that should be avoided in older adults listed on Beers criteria since clinical data required to identify some PIMs are not available in the database. Thus, all the above‐mentioned elements may have underestimated the persistence with PIMs. Second, data on drugs were based on data reimbursement, and therefore we cannot say with certitude whether dispensed drugs were actually taken. Third, the Beers criteria provide a list of medications potentially inappropriate in older adults, but in some cases medications can be appropriate when considering patient characteristics and goals of care.

Although continuous exposure to PIMs does not seem to be chronic for three‐quarters of community‐dwelling older adults initiating a PIM, one‐quarter of those were persistently exposed to at least one PIM the following year. Further interventions are needed both to limit the initiation of PIMs and to deprescribe existing PIMs most likely to be continued. These interventions should notably target individuals most at risk of being persistent such as men, older people (>75 years) or those with a high number of medications and chronic diseases (e.g. dementia, diabetes and cardiovascular diseases), as they are particularly vulnerable to adverse drug reactions, which may lead to unfavourable health outcomes. Thus, our findings highlight the great need to conduct further studies on the impact of the persistence of PIMs in the older population.

COMPETING INTERESTS

C.S. has a research scholarship from the Fonds de recherche du Québec‐Santé. M.‐E.G. has a scholarship from the Fonds de recherche du Québec‐Santé in partnership with Unité Soutien from the Strategy for patient‐oriented research and has received a scholarship from the Canadian Institute for Health Research. B.R., M.S. and M.‐L.L. declare that they have no conflict of interest.

CONTRIBUTORS

B.R., C.S., M.S. and M.‐L.L. conceived and designed the study. B.R. performed the statistical analysis, analysed and interpreted the data, and drafted and critically revised the manuscript. M.S. acquired and interpreted the data, and critically revised the manuscript. M.‐E.G. interpreted the data and critically revised the manuscript. C.S. and M.‐L.L. provided supervision, interpreted the data and critically revised the manuscript.

APPENDIX A.

Figure A1.

Definition of persistence measurement

Table A1.

Medications from the 2015 Beers version included in this study

Category	Medication	Comments
Anticholinergics
First‐generation antihistamines	Brompheniramine
	Chlorpheniramine
	Cyproheptadine
	Dimenhydrinate
	Diphenhydramine (oral)
	Hydroxyzine
	Meclizine
	Promethazine
Antiparkinson agents	Benztropine (oral)
	Trihexyphenidyl
Antispasmodics	Atropine (excludes ophtalmic)
	Belladonna alkaloids
	Clinidium‐chlordiazepoxide
	Dicyclomine
	Scopolamine
Antithrombotics
	Dipyridamole oral short‐acting
	Ticlopidine
Cardiovascular
Peripheral alpha‐1 blockers	Doxazosin
	Prazosin
	Terazosin
Central alpha‐blockers	Clonidine
	Guanfacine
	Methyldopa
Antiarrhythmics	Disopyramide
	Dronedarone	Avoid for atrial fibrillation (ICD‐9: 4273, ICD‐10: I48) or heart failure (ICD‐9: 428, ICD‐10: I50)
	Digoxin >0.125 mg/d
	Amiodarone	Avoid for atrial fibrillation (ICD‐9: 4273, ICD‐10: I48) except heart failure (ICD‐9: 428, ICD‐10: I50)
	Nifedipine, immediate release
Central nervous system
Antidepressants, alone or in combination	Amitriptyline
	Amoxapine
	Clomipramine
	Desipramine
	Doxepin >6 mg/d
	Imipramine
	Nortriptyline
	Paroxetine
	Trimipramine
Antipsychotics, first‐generation (conventional)	Droperidol	Avoid, except for schizophrenia (ICD‐9: 295, ICD‐10: F20, F21, F23.2, F25) and bipolar disorders (ICD‐9: 296, ICD‐10: F30, F31) Drugs used as antiemetics during chemotherapy were not included (chlorpromazine, perphenazine, prochlorperazine)
	Fluphenazine
	Flupentixol
	Haloperidol
	Loxapine
	Methotrimeprazine
	Pimozide
	Thioproperazine
	Thiothixene
	Thioridazine
	Trifluoperazine
	Zuclopenthixol
Antipsychotics, second‐generation (atypical)	Aripiprazole	Avoid, except for schizophrenia (ICD‐9: 295, ICD‐10: F20, F21, F23.2, F25) and bipolar disorders (ICD‐9: 296, ICD‐10: F30, F31)
	Asenapine
	Clozapine
	Olanzapine
	Paliperidone
	Quetiapine
	Risperidone
	Ziprasidone
Barbiturates	Butalbital
	Phenobarbital
Benzodiazepines (short‐ and intermediate‐acting)	Alprazolam
	Bromazepam
	Lorazepam
	Oxazepam
	Temazepam
	Triazolam
Benzodiazepines (long‐acting)	Clorazepate
	Chlordiazepoxide (alone or in combination with clinidium)
	Clonazepam
	Diazepam
	Flurazepam
Nonbenzodiazepine and benzodiazepine receptor agonist hypnotics	Zaleplon
	Zolpidem
	Zopiclone
	Ergoloid mesylates
Endocrine
	Desiccated thyroid
	Estrogens with or without progestins	Avoid, except vaginal form
	Megestrol
Long‐duration sulphonylureas	Chlorpropamide
	Glyburide
Gastrointestinal
	Mineral oil (oral)
Proton pump inhibitors	Dexlansoprazole	Avoid for use >56 d unless with concomitant use of oral corticosteroids (dexamethasone, cortisone, fludrocortisone, hydrocortisone, methylprednisolone, prednisolone), oral non‐COX selective NSAIDs, COX‐2 selective agents (celecoxib, valdecoxib) and oral antiaggregant aspirin
	Esomeprazole
	Lansoprazole
	Omeprazole
	Pantoprazole
	Rabeprazole
Pain
	Meperidine
Non‐COX selective NSAIDs, oral	Aspirin >325 mg/d	Avoid for chronic use (>90 d) unless with concomitant use of proton‐pump inhibitors or misoprostol
	Diclofenac
	Diflunisal
	Etodolac
	Ibuprofen
	Ketoprofen
	Mefenamic acid
	Meloxicam
	Nabumetone
	Naproxen
	Piroxicam
	Sulindac
	Indomethacin
	Ketorolac, includes parenteral
	Pentazocine
Skeletal muscle relaxants	Carisoprodol
	Chlorzoxazone
	Cyclobenzaprine
	Methocarbamol
	Orphenadrine
Genitourinary
	Desmopressin

Abbreviations: COX, cyclooxygenase; NSAIDS, non‐steroidal anti‐inflammatory drugs.

Table A2.

Additional analyses on factors associated with 1‐year persistence with any PIM in older Quebecers

		Model 3c	Model 4c
	RR (95% CI)	RRa (95% CI)	RRa (95% CI)
Age, years a
66‐75	1.00	1.00	1.00
76‐85	1.32 (1.28‐1.35)	1.18 (1.15‐1.22)	1.20 (1.17‐1.24)
86 and older	1.82 (1.76‐1.89)	1.49 (1.43‐1.55)	1.52 (1.46‐1.58)
Sex
Men	1.00	1.00	1.00
Women	0.90 (0.87‐0.92)	0.90 (0.88‐0.93)	0.91 (0.89‐0.94)
Region a
Urban	1.00	1.00	1.00
Rural	1.05 (1.02‐1.10)	1.06 (1.03‐1.09)	1.07 (1.04‐1.10)
Material deprivation index a
First quintile, least deprived	1.00	1.00	1.00
Second quintile	1.15 (1.10‐1.21)	1.14 (1.09‐1.19)	1.15 (1.10‐1.20)
Third quintile	1.26 (1.16‐1.28)	1.19 (1.14‐1.24)	1.19 (1.14‐1.24)
Fourth quintile	1.23 (1.17‐1.29)	1.19 (1.14‐1.24)	1.19 (1.14‐1.25)
Fifth quintile, most deprived	1.25 (1.19‐1.31)	1.19 (1.14‐1.24)	1.20 (1.15‐1.25)
Social deprivation index a
First quintile, least deprived	1.00	1.00	1.00
Second quintile	1.02 (0.97‐1.07)	1.00 (0.96‐1.05)	1.00 (0.96‐1.05)
Third quintile	1.03 (0.98‐1.08)	1.01 (0.97‐1.06)	1.01 (0.97‐1.06)
Fourth quintile	1.07 (1.02‐1.12)	1.05 (1.00‐1.09)	1.05 (1.01‐1.09)
Fifth quintile, most deprived	1.12 (1.07‐1.18)	1.10 (1.05‐1.15)	1.10 (1.06‐1.15)
Chronic diseases a
Hypertension	1.29 (1.25‐1.33)	1.04 (1.01‐1.07)	1.11 (1.08‐1.15)
Cardiovascular diseases	1.33 (1.29‐1.37)	1.05 (1.02‐1.08)	1.14 (1.11‐1.17)
Osteoporosis	0.93 (0.90‐0.97)	0.90 (0.87‐0.92)	0.92 (0.90‐0.95)
Diabetes	1.36 (1.31‐1.40)	1.12 (1.09‐1.16)	1.25 (1.21‐1.28)
Respiratory diseases	1.17 (1.13‐1.21)	1.02 (0.99‐1.05)	1.09 (1.06‐1.12)
Mental disorders	1.09 (1.03‐1.15)	1.11 (1.06‐1.16)	1.12 (1.07‐1.17)
Alzheimer's disease and related disorders	2.19 (2.10‐2.31)	1.79 (1.71‐1.86)	1.81 (1.74‐1.89)
Number of different medications b
0‐4	1.00	1.00
5‐9	1.27 (1.22‐1.32)	1.18 (1.14‐1.22)
10‐14	1.63 (1.56‐1.70)	1.41 (1.36‐1.47)
15 or more	2.06 (1.97‐2.16)	1.67 (1.60‐1.75)

Abbreviation: PIM, potentially inappropriate medication.

^aAt the beginning of the 2014 fiscal year (1 April 2014).

^bMedications received for each individual 365 days before the index date (first PIM dispensed in the 2014 fiscal year [1 April 2014‐31 March 2015]).

^cPoisson regression models, including the 1‐year persistence with at least one PIM as binary dependent variable. All covariates (age, sex, region, material and social deprivation index, type of chronic diseases and number of different medications) were entered as independent variables, except for model 4 (exclusion of the number of different medications).

Table A3.

Sensitivity analyses considering different permissible gaps for 1‐year persistence with PIMs

PIMs (number of PIM initiators)	Persistence
	15‐d gap		30‐d gap		45‐d gap		60‐d gap		90‐d gap
	na (%)	95% CI	n (%)	95% CI	n (%)	95% CI	n (%)	95% CI	n (%)	95% CI
Any PIM (n = 75,844)	12,820 (16.9)	16.6‐17.2	15,791 (20.8)	20.5‐21.1	17,815 (23.5)	23.2‐23.8	19,051 (25.1)	24.8‐25.4	21,423 (28.2)	27.9‐28.5
Antipsychotics (n = 6,307)	2,160 (34.2)	33.0‐35.4	2,457 (39.0)	37.8‐40.2	2,659 (42.2)	41.0‐43.4	2,769 (43.9)	42.7‐45.1	2,910 (46.1)	44.9‐47.3
First‐generation (n = 551)	73 (13.3)	10.5‐16.1	79 (14.3)	11.4‐17.2	88 (16.0)	12.9‐19.1	90 (16.3)	13.2‐19.4	96 (17.4)	14.2‐20.6
Second‐generation (n = 6,061)	2,102 (34.7)	32.7‐36.7	2,391 (39.5)	37.4‐41.6	2,588 (42.7)	40.6‐44.8	2,696 (44.5)	42.6‐46.4	2,834 (46.8)	44.7‐48.9
Long‐duration sulphonylureas (n = 2,637)	697 (26.4)	24.7‐28.1	894 (33.9)	32.1‐35.7	998 (37.9)	36.0‐39.8	1,059 (40.2)	37.2‐43.3	1,117 (42.4)	38.3‐42.1
Cardiovascular medications (n = 889)	232 (26.1)	23.2‐29.0	295 (33.2)	30.1‐36.3	315 (35.4)	32.3‐38.6	324 (36.5)	33.3‐39.6	331 (37.2)	34.1‐40.4
Proton pump inhibitors (n = 30,590)	8,077 (26.4)	25.9‐26.9	9,647 (31.5)	31.0‐32.0	10,597 (34.6)	34.1‐35.1	11,014 (36.0)	35.1‐36.9	11,994 (39.2)	38.7‐39.7
Tricyclic antidepressants and paroxetine (n = 4,477)	462 (10.3)	9.4‐11.2	581 (13.0)	12.0‐14.0	659 (14.7)	13.7‐15.7	711 (15.9)	14.8‐17.0	782 (17.5)	16.4‐18.6
Central alpha‐blockers (n = 1,834)	189 (10.3)	8.9‐11.7	252 (13.7)	12.1‐15.3	273 (14.9)	13.3‐16.5	283 (15.4)	13.7‐17.1	308 (16.8)	15.1‐18.5
Benzodiazepines (n = 30,256)	1,567 (5.2)	4.9‐5.5	2,256 (7.5)	7.2‐7.8	2,850 (9.4)	9.1‐9.7	3,294 (10.9)	10.5‐11.3	4,034 (13.3)	12.9‐13.7
Short‐ and intermediate‐acting (n = 26,209)	1,390 (5.3)	5.0‐5.6	1,982 (7.6)	7.3‐7.9	2,480 (9.5)	9.1‐9.9	2,854 (10.9)	10.5‐11.3	3,485 (13.3)	12.9‐13.7
Long‐acting (n = 5,021)	183 (3.6)	3.1‐4.1	274 (5.5)	4.9‐6.1	359 (7.2)	6.5‐7.9	428 (8.5)	7.7‐9.3	524 (10.4)	9.6‐11.2
NSAIDs (n = 1,585)	50 (3.2)	2.3‐4.1	73 (4.6)	3.6‐5.6	89 (5.6)	4.5‐6.7	91 (5.7)	4.6‐6.8	106 (6.7)	5.5‐7.9
First‐generation antihistamines (n = 5,466)	55 (1.0)	0.7‐1.3	77 (1.4)	1.1‐1.7	102 (1.9)	1.5‐2.3	136 (2.5)	2.1‐2.9	188 (3.4)	2.9‐3.9
Skeletal muscle relaxants (n = 10,314)	79 (0.8)	0.6‐1.0	109 (1.1)	0.9‐1.3	137 (1.3)	1.1‐1.5	156 (1.5)	1.3‐1.7	203 (2.0)	1.7‐2.3

Abbreviations: NSAIDS, non‐steroidal anti‐inflammatory drugs; PIMs, potentially inappropriate medications.

^aNumber of persistent users.

Table A4.

Sensitivity analyses considering different permissible gaps for factors associated with 1‐year persistence with any PIM

	15‐d gap		30‐d gap		45‐d gap		90‐d gap
	Model 1d	Model 2d	Model 1d	Model 2d	Model 1d	Model 2d	Model 1d	Model 2d
	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)	RRa (95% CI)
Age, years a
66‐75	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
76‐85	1.33 (1.28‐1.38)	1.35 (1.31‐1.40)	1.24 (1.20‐1.28)	1.26 (1.22‐1.30)	1.21 (1.17‐1.24)	1.22 (1.19‐1.26)	1.17 (1.14‐1.20)	1.19 (1.16‐1.22)
86 and older	1.89 (1.81‐1.98)	1.93 (1.84‐2.02)	1.69 (1.62‐1.77)	1.72 (1.65‐1.79)	1.61 (1.55‐1.68)	1.64 (1.57‐1.70)	1.49 (1.44‐1.55)	1.51 (1.46‐1.57)
Sex
Men	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Women	0.85 (0.82‐0.88)	0.85 (0.82‐0.88)	0.86 (0.84‐0.88)	0.86 (0.84‐0.88)	0.87 (0.85‐0.89)	0.87 (0.85‐0.89)	0.89 (0.87‐0.91)	0.90 (0.88‐0.92)
Region a
Urban	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Rural	1.05 (1.01‐1.10)	1.07 (1.03‐1.11)	1.05 (1.02‐1.09)	1.07 (1.03‐1.10)	1.05 (1.02‐1.09)	1.07 (1.03‐1.10)	1.05 (1.02‐1.08)	1.06 (1.03‐1.09)
Material deprivation index a
First quintile, least deprived	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Second quintile	1.17 (1.11‐1.24)	1.18 (1.11‐1.25)	1.14 (1.09‐1.20)	1.15 (1.09‐1.21)	1.15 (1.10‐1.20)	1.15 (1.10‐1.21)	1.11 (1.07‐1.16)	1.12 (1.07‐1.16)
Third quintile	1.28 (1.21‐1.36)	1.29 (1.22‐1.36)	1.23 (1.17‐1.29)	1.24 (1.18‐1.30)	1.21 (1.16‐1.27)	1.22 (1.16‐1.27)	1.16 (1.12‐1.21)	1.17 (1.12‐1.22)
Fourth quintile	1.26 (1.19‐1.33)	1.28 (1.21‐1.35)	1.22 (1.17‐1.29)	1.24 (1.18‐1.30)	1.20 (1.14‐1.25)	1.21 (1.15‐1.26)	1.16 (1.12‐1.21)	1.17 (1.13‐1.22)
Fifth quintile, most deprived	1.29 (1.22‐1.36)	1.30 (1.23‐1.38)	1.23 (1.17‐1.29)	1.25 (1.18‐1.31)	1.21 (1.15‐1.26)	1.22 (1.17‐1.28)	1.18 (1.13‐1.23)	1.19 (1.14‐1.24)
Social deprivation index a
First quintile, least deprived	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
Second quintile	0.99 (0.94‐1.05)	0.99 (0.94‐1.05)	0.99 (0.94‐1.04)	0.99 (0.94‐1.04)	0.99 (0.95‐1.04)	0.99 (0.95‐1.04)	1.01 (0.97‐1.05)	1.01 (0.97‐1.05)
Third quintile	0.98 (0.92‐1.03)	0.98 (0.92‐1.03)	0.99 (0.95‐1.04)	0.99 (0.95‐1.04)	0.99 (0.95‐1.04)	0.99 (0.95‐1.04)	1.02 (0.98‐1.06)	1.02 (0.98‐1.06)
Fourth quintile	1.03 (0.97‐1.08)	1.03 (0.98‐1.10)	1.03 (0.98‐1.08)	1.03 (0.98‐1.08)	1.03 (0.99‐1.08)	1.04 (0.99‐1.08)	1.05 (1.01‐1.09)	1.05 (1.01‐1.09)
Fifth quintile, most deprived	1.11 (1.05‐1.17)	1.11 (1.05‐1.17)	1.10 (1.05‐1.15)	1.10 (1.05‐1.16)	1.09 (1.05‐1.14)	1.10 (1.05‐1.15)	1.09 (1.05‐1.13)	1.09 (1.05‐1.14)
Number of chronic diseases b
0	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
1	1.07 (1.00‐1.15)	1.15 (1.08‐1.23)	1.08 (1.02‐1.15)	1.15 (1.09‐1.22)	1.07 (1.02‐1.13)	1.14 (1.08‐1.19)	1.04 (0.99‐1.08)	1.09 (1.05‐1.14)
2	1.16 (1.09‐1.24)	1.35 (1.27‐1.44)	1.13 (1.06‐1.19)	1.29 (1.22‐1.36)	1.10 (1.04‐1.16)	1.25 (1.19‐1.31)	1.06 (1.01‐1.11)	1.19 (1.14‐1.24)
3	1.28 (1.19‐1.37)	1.60 (1.50‐1.71)	1.23 (1.16‐1.31)	1.50 (1.42‐1.59)	1.18 (1.12‐1.25)	1.42 (1.35‐1.50)	1.12 (1.07‐1.17)	1.32 (1.27‐1.39)
4 or more	1.52 (1.42‐1.63)	2.09 (1.96‐2.23)	1.44 (1.36‐1.54)	1.91 (1.81‐2.02)	1.37 (1.30‐1.45)	1.79 (1.70‐1.88)	1.26 (1.20‐1.33)	1.60 (1.53‐1.67)
Number of different medications c
0‐4	1.00		1.00		1.00		1.00
5‐9	1.22 (1.16‐1.27)		1.18 (1.14‐1.24)		1.19 (1.14‐1.24)		1.17 (1.13‐1.21)
10‐14	1.52 (1.44‐1.60)		1.45 (1.38‐1.52)		1.42 (1.36‐1.49)		1.38 (1.33‐1.43)
15 or more	1.85 (1.74‐1.96)		1.72 (1.63‐1.81)		1.67 (1.59‐1.75)		1.59 (1.52‐1.66)

Abbreviations: CI, confidence interval; PIM, potentially inappropriate medication; RR, crude rate ratio; RRa, adjusted rate ratio.

^aAt the beginning of the 2014 fiscal year (1 April 2014).

^bChronic diseases at the beginning of the 2014 fiscal year (1 April 2014): Alzheimer's disease and related dementia, cardiovascular diseases, diabetes, hypertension, mental disorders, osteoporosis, respiratory diseases.

^cMedications received for each individual 365 days before the index date (first PIM dispensed in the 2014 fiscal year [1 April 2014‐31 March 2015]).

^dPoisson regression models, including the 1‐year persistence with at least one PIM as a binary‐dependent variable. All covariates (age, sex, region, material and social deprivation index, number of chronic diseases and number of different medications) were entered as independent variables, except for model 2 (exclusion of the number of different medications).

Roux B, Sirois C, Simard M, Gagnon M‐E, Laroche M‐L. One‐year persistence of potentially inappropriate medication use in older adults: A population‐based study. Br J Clin Pharmacol. 2020;86 1062–1080. 10.1111/bcp.14214

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are not available due to policy restrictions.