Abstract
Objective
To analyze primary medication nonadherence across several prescription indications and test the predictors of drug nonadherence in an adult primary care population.
Design
Retrospective observational study using primary care provider prescriptions linked to pharmacy-based dispensing data from 2012 to 2014.
Setting
Manitoba.
Participants
Patients in the Manitoba Primary Care Research Network.
Main outcome measures
Prevalence of primary medication nonadherence by drug class. Multivariable logistic regression models were used to test the associations of patient demographic and clinical or provider characteristics with primary medication nonadherence. The C statistic was used to assess the models’ discriminative performance.
Results
A total of 91,660 unique prescriptions were assessed from a cohort of more than 200,000 patients. Primary medication nonadherence ranged from 13.7% (antidepressants) to 30.3% (antihypertensives). In conditions that typically present symptomatically (eg, infections, anxiety) nonadherence ranged from 13.7% to 17.5%. The range was 21.2% to 30.0% for medications related to asymptomatic conditions or those typically detected by screening. The discriminative performance of the models based on patient demographic, clinical, or provider characteristics was weak.
Conclusion
Primary medication nonadherence is common, occurring more often in asymptomatic conditions. The poor predictability of the models suggests that caution is required when considering characteristic-based interventions or prediction tools to improve primary medication nonadherence.
Résumé
Objectif
Analyser la non-adhésion primaire à la médication parmi plusieurs prescriptions indiquées et mettre à l’essai des facteurs de prédiction de la non-adhésion dans une population adulte en soins primaires.
Type d’étude
Une étude observationnelle rétrospective à l’aide des ordonnances prescrites par des professionnels de la santé reliés à une pharmacie qui a fourni des données de 2012 à 2014.
Contexte
Manitoba.
Participants
Des patients dans le Réseau de recherche en soins primaires du Manitoba.
Principaux paramètres à L’étude
La prévalence de la non-adhésion primaire selon la classe de médicaments. Des modèles de régression logistique à variables multiples ont servi pour mettre à l’essai les associations entre les renseignements démographiques des patients, les caractéristiques cliniques ou les caractéristiques propres aux professionnels, et la non-adhésion primaire à la médication. La statistique C a été utilisée pour évaluer le rendement discriminant du modèle.
Résultats
Un total de 91 660 ordonnances individuelles prescrites à plus de 200 000 patients ont été évaluées. La non-adhésion primaire à la médication variait de 13,7 % (antidépresseurs) à 30,3 % (antihypertenseurs). Pour les problèmes typiquement symptomatiques (p. ex. infections, anxiété), la non-adhésion variait de 13,7 à 17,5 %. Elle se situait entre 21,2 et 30,0 % pour les médicaments associés à des problèmes asymptomatiques ou à ceux habituellement détectés par dépistage. Le rendement discriminant des modèles fondés sur les données démographiques des patients, les caractéristiques cliniques ou les caractéristiques propres aux professionnels était faible.
Conclusion
La non-adhésion primaire à la médication est fréquente et se produit plus souvent dans le cas des problèmes asymptomatiques. La faible prévisibilité des modèles fait valoir qu’il faut user de prudence lorsque sont envisagées des interventions fondées sur des caractéristiques ou des outils de prédiction pour réduire la non-adhésion primaire à la médication.
Primary medication nonadherence—when a patient does not fill their initial prescription—is common.1 Several studies indicate that poor medication adherence increases adverse clinical events, morbidity, mortality, and overall health care costs.1-8 Nonadherence is a critical barrier to treatment success and a considerable challenge for health care providers.9 Being aware of the extent to which nonadherence occurs in specific conditions is the first step to addressing it.5,10
Primary medication nonadherence is distinct from secondary medication nonadherence, where medications are filled but not taken as prescribed.1 We focused on primary medication nonadherence because, until recently, many of the studies evaluating primary non-adherence to medication were limited by small sample sizes, suboptimal methodologies, or surrogate measures of adherence behaviour.5,8 The recent availability of electronic medical record (EMR) prescribing information linked to medication dispensation information available from provincial health administrative data offers a more complete picture of drug initiation and adherence behaviour, particularly in primary care where treatment is often initiated.1,5,8,11
Rates of primary nonadherence have been shown to vary between 2.4% and 30.7%.1 Several factors related to the patient, provider, social context, and medical condition have demonstrated associations with higher rates of nonadherence. However, the influence of specific characteristics and their relevance as predictors of primary medication nonadherence vary greatly.1,3,10,12-14 Additionally, much of the research on primary medication nonadherence has focused on specific chronic disease populations (eg, diabetes, hypertension) and, therefore, a more comprehensive approach is warranted.2,15-20 The objective of this study was to analyze primary nonadherence across several prescription indications and test the predictors of drug adherence behaviour in an adult primary care patient population in Manitoba.
METHODS
In Canada, health care delivery is a provincial and territorial responsibility; medical services are provided in hospitals and community settings via single-payer, publicly funded insurance. Provinces and territories must meet standards described in the Canada Health Act to receive federal funding. Administrative health records are stored and maintained by provincial and territorial government bodies. We performed a retrospective observational study with primary care provider prescriptions linked to pharmacy-based dispensations data from the Manitoba Population Research Data Repository at the Manitoba Centre for Health Policy (MCHP). The MCHP Population Research Data Repository contains population-based de-identified administrative records submitted to the Manitoba government to report health and social service use by residents of Manitoba and includes data on emergency department visits (Emergency Department Information System) and all prescriptions filled in Manitoba pharmacies (Drug Program Information Network [DPIN]).21-23
The MCHP data repository also holds clinical data from the Manitoba Primary Care Research Network (MaPCReN), which includes a database of de-identified primary care EMR data from community-based primary care practices. Primary care providers who have consented to participate in MaPCReN include family physicians, nurse practitioners, and community pediatricians. It is one of the networks participating in the Canadian Primary Care Sentinel Surveillance Network, which also has 10 other regional networks. Prior studies have demonstrated that patient populations within Canadian Primary Care Sentinel Surveillance Network and MaPCReN practices are representative in terms of disease prevalence and prescribing rates when compared with other national data sources.24,25 At the time of this study, MaPCReN included 44 primary care clinics representing 241 providers (including approximately 20% of the family physicians in Manitoba) caring for more than 200,000 patients aged 18 years or older.
Prescriptions written from April 1, 2012, to December 31, 2014, from the MaPCReN database were linked to Manitoba’s DPIN. The DPIN contains data for all prescriptions dispensed by community pharmacies in the province of Manitoba. We included medication classes indicated for common chronic and acute conditions often treated in primary care and where nonadherence may affect health outcomes.
A separate study cohort was produced for each medication class and indication pairing listed in Table 1. Table 1 also provides the corresponding World Health Organization’s Anatomical Therapeutic Chemical codes used to define each medication in the grouping.
Table 1.
Medications, typical indications, and corresponding Anatomic Therapeutic Chemical codes selected for the study
| MEDICATION CLASS | TYPICAL INDICATIONS | CODES |
|---|---|---|
| Antibiotics | Infection | J01, G04A |
| Antidepressants | Depression | N06A |
| Antihypertensives | Hypertension | C02, C03, C04, C05, C07, C08, C09 |
| Benzodiazepines | Anxiety | N05BA, N03AE |
| Bisphosphonates | Osteoporosis | M05B |
| Hypoglycemics | Diabetes | A10 |
| Lipid-lowering agents | Cardiovascular disease, hyperlipidemia | C10 |
The main outcome, primary medication nonadherence, was defined as the absence of a dispensing record (in DPIN) after a new prescription (in MaPCReN) within 90 days of the date the prescription was written. New prescriptions were defined as those not dispensed to the patient within the previous 365 days.
Multivariable logistic regression models were used to test the association of patient demographic and clinical or provider characteristics with primary medication nonadherence. The specific covariates for the models were calculated using standard definitions,22,23,26 and those included in the models are described in appendices available from CFPlus.* A mixed-effects multivariable logistic regression model, which included a random intercept for provider, was also fit to the data to assess the magnitude of clustering of patients among providers. The intraclass correlation was used to provide a measure of the clustering effect; it has a lower bound of 0 and an upper bound of 1, with higher values indicating a stronger clustering effect.
The C statistic was used to describe the discriminative performance of the models for each medication class. A model with a C statistic greater than 0.7 is considered to have reasonable discriminative performance, while a model with a C statistic greater than 0.8 is considered to have strong discriminative performance.27
All data management, programming, and analyses were conducted using SAS, version 9.4. This research was approved by the University of Manitoba Health Research Ethics Board and by the province’s Health Information Privacy Committee.
RESULTS
The demographic characteristics of the study cohorts are described in Table 2. Overall, there were 91,660 unique prescriptions from more than 200,000 active patients. The smallest number of prescriptions was for bisphosphonates (N=670) and the largest grouping was antibiotics, which included 37,402 prescriptions. Most prescriptions were provided to female patients in all medication class cohorts, with the exception of lipid-lowering agents. Most prescriptions were for patients in the younger age groupings (18 to 44, 45 to 64), with the exception of bisphosphonates, which were mostly started in the 65 to 74 age grouping. The distribution of prescriptions within the income quintiles was fairly even, falling within a range of 14.4% to 22.8% in each category.
Table 2.
Demographic characteristics by medication class
| CHARACTERISTIC | ANTIBIOTICS, n (%) (N=37,402)* | ANTIDEPRESSANTS, n (%) (N=14,257)* | ANTIHYPERTENSIVES, n (%) (N=18,609)* | BENZODIAZEPINES, n (%) (N=4448)* | BISPHOSPHONATES, n (%) (N=670)* | HYPOGLYCEMICS, n (%) (N=11,294)* | LIPID-LOWERING AGENTS, n (%) (N=4980)* |
|---|---|---|---|---|---|---|---|
| Sex | |||||||
| • Male | 11,623 (31.1) | 4596 (32.2) | 8463 (45.5) | 1414 (31.8) | 79 (11.8) | 4609 (40.8) | 2733 (54.9) |
| • Female | 25,779 (68.9) | 9661 (67.8) | 10,146 (54.5) | 3034 (68.2) | 591 (88.2) | 6685 (59.2) | 2247 (45.1) |
| Age group, y | |||||||
| • 18-44 | 14,980 (40.1) | 6815 (47.8) | 2981 (16.0) | 1681 (37.8) | 35 (5.2) | 3436 (30.4) | 394 (7.9) |
| • 45-64 | 13,421 (35.9) | 5179 (36.3) | 7912 (42.5) | 1774 (39.9) | 159 (23.7) | 4554 (40.3) | 2673 (53.7) |
| • 65-74 | 4634 (12.4) | 1200 (8.4) | 3964 (21.3) | 579 (13.0) | 203 (30.3) | 1745 (15.5) | 1304 (26.2) |
| • ≥75 | 4367 (11.7) | 1063 (7.5) | 3752 (20.2) | 414 (9.3) | 273 (40.7) | 1,559 (13.8) | 609 (12.2) |
| Income quintile | |||||||
| • Q1 (lowest) | 6911 (18.5) | 2710 (19.0) | 3309 (17.8) | 639 (14.4) | 135 (20.1) | 2045 (18.1) | 799 (16.0) |
| • Q2 | 7083 (18.9) | 2845 (20.0) | 3453 (18.5) | 812 (18.3) | 139 (20.7) | 2156 (19.1) | 969 (19.5) |
| • Q3 | 7274 (19.4) | 2835 (19.9) | 3777 (20.3) | 878 (19.7) | 130 (19.4) | 2116 (18.7) | 982 (19.7) |
| • Q4 | 8225 (22.0) | 2997 (21.0) | 4056 (21.8) | 1035 (23.3) | 122 (18.2) | 2549 (22.6) | 1136 (22.8) |
| • Q5 (highest) | 6604 (17.7) | 2451 (17.2) | 3425 (18.4) | 940 (21.1) | 123 (18.4) | 2062 (18.3) | 990 (19.9) |
| • Missing | 1305 (3.5) | 419 (2.9) | 589 (3.2) | 144 (3.2) | 21 (3.1) | 366 (3.2) | 104 (2.1) |
N represents the no. of prescriptions in each grouping.
The range of primary medication nonadherence was 13.7% (antidepressants) to 30.3% (antihypertensives). In conditions that typically present symptomatically, such as infections, depression, and anxiety, nonadherence ranged from 13.7% to 17.5%. Medications related to asymptomatic conditions or those typically detected by screening, such as hypertension, osteoporosis, and diabetes, demonstrated nonadherence rates from 21.2% to 30.3%. The exception to this trend was medications used for hyperlipidemia and cardiovascular disease (lipid-lowering agents), which had a primary nonadherence rate of 15.2%. Table 3 describes overall primary nonadherence by medication class, age group, and sex.
Table 3.
Primary nonadherence rates for the study cohorts by medication class, sex, and age group
| NONADHERENCE, n (%)* | |||||||
|---|---|---|---|---|---|---|---|
| VARIABLE | ANTIBIOTICS | ANTI-DEPRESSANTS | ANTI-HYPERTENSIVES | BENZODIAZEPINES | BISPHOSPHONATES | HYPOGLYCEMICS | LIPID-LOWERING AGENTS |
| Total | 6552 (17.5) | 1956 (13.7) | 5634 (30.3) | 694 (15.7) | 185 (27.6) | 881 (21.2) | 775 (15.2) |
| Sex | |||||||
| • Male | 1827 (4.9) | 642 (4.5) | 2607 (14.0) | 220 (5.0) | 27 (4.0) | 452 (10.9) | 408 (8.2) |
| • Female | 4725 (12.6) | 1314 (9.2) | 3027 (16.3) | 474 (10.7) | 158 (23.6) | 429 (10.3) | 347 (7.0) |
| Age group, y | |||||||
| • 18-44 | 3070 (8.2) | 959 (6.7) | 934 (5.0) | 232 (5.2) | 31 (4.6) | 168 (4.0) | 62 (1.2) |
| • 45-64 | 2204 (5.9) | 724 (5.1) | 2266 (12.2) | 293 (6.6) | 49 (7.3) | 436 (10.5) | 342 (6.9) |
| • 65-74 | 701 (1.9) | 152 (1.1) | 1244 (6.7) | 95 (2.1) | 42 (6.3) | 186 (4.5) | 224 (4.5) |
| • ≥75 | 577 (1.5) | 121 (0.9) | 1190 (6.4) | 74 (1.7) | 63 (9.4) | 91 (2.2) | 127 (2.6) |
N represents the no. of prescriptions in each grouping.
The intraclass correlation was very low for the multivariable mixed-effects logistic regression model that included a random intercept for provider, suggesting minimal clustering effects. Therefore, subsequent models fit to the data were based on the assumption of independence of observations.
The C statistic provided for each medication class cohort (Table 4)28 indicates how well patient demographic and clinical characteristics predicted primary medication nonadherence. In general, the models demonstrated weak discriminative performance. In addition, there was little consistency in the association of the model covariates with primary nonadherence for each medication class.
Table 4.
Primary nonadherence C statistics, ORs, and 95% CIs for patient demographic and clinical characteristics by medication class: Statistically significant differences are presented in boldface.
| MEDICATION CLASS | |||||||
|---|---|---|---|---|---|---|---|
| VARIABLE | ANTIBIOTICS (N=37,402)* | ANTIDEPRESSANTS (N=14,257)* | ANTI-HYPERTENSIVES (N=18,609)* | BENZODIAZEPINES (N=4448)* | BISPHOSPHONATES (N=670)* | HYPOGLYCEMICS (N=11,294)* | LIPID-LOWERING AGENTS (N=4980)* |
| C statistic | 0.59 | 0.56 | 0.57 | 0.61 | 0.69 | 0.62 | 0.59 |
| Sex, OR (95% CI) | |||||||
| • Male | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • Female |
0.87
(0.82-0.93) |
0.99 (0.89-1.10) |
1.08
(1.01-1.15) |
0.94 (0.78-1.12) |
0.94 (0.78-1.12) |
1.23
(1.14-1.33) |
0.99 (0.84-1.16) |
| Age group, y, OR (95% CI) | |||||||
| • 18-44 |
0.55
(0.49-0.61) |
0.78
(0.63-0.97) |
0.93 (0.83-1.04) |
1.19 (0.87-1.63) |
0.04
(0.01-1.12) |
1.29
(1.12-1.47) |
1.35 (0.96-1.91) |
| • 45-64 |
0.74
(0.67-0.82) |
0.79
(0.64-0.98) |
1.11
(1.02-1.22) |
1.08 (0.80-1.44) |
0.69 (0.43-1.12) |
1.38
(1.22-1.56) |
1.74
(1.38-2.21) |
| • 65-74 |
0.81
(0.72-0.91) |
0.89 (0.69-1.15) |
1.02 (0.92-1.12) |
1.14 (0.81-1.60) |
1.26 (0.49-2.01) |
1.18
(1.02-1.36) |
1.26 (0.98-1.61) |
| • ≥75 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| Income quintile, OR (95% CI) | |||||||
| • Q1 (lowest) |
0.85
(0.78-0.93) |
1.10 (0.94-1.30) |
1.23
(1.11-1.37) |
1.31 (0.98-1.75) |
1.55 (0.84-2.86) |
1.12 (0.99-1.28) |
0.88 (0.68-1.15) |
| • Q2 | 0.97 (0.89-1.06) |
1.00 (0.85-1.17) |
1.15
(1.04-1.27) |
0.87 (0.68-1.12) |
1.18 (0.67-2.11) |
1.11 (0.98-1.26) |
0.91 (0.71-1.17) |
| • Q3 | 1.04 (0.95-1.14) |
0.99 (0.84-1.15) |
1.32
(1.19-1.46) |
1.29 (0.99-1.67) |
1.36 (0.75-2.46) |
1.13 (1.00-1.29) |
0.98 (0.76-1.26) |
| • Q4 | 1.10 (1.00-1.20) |
1.05 (0.90-1.23) |
1.14
(1.03-1.25) |
1.20 (0.93-1.53) |
1.06 (0.58-1.90) |
1.02 (0.90-1.15) |
1.01 (0.79-1.29) |
| • Q5 (highest) |
Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • Missing |
1.22
(1.03-1.44) |
1.00 (0.74-1.35) |
1.05 (0.87-1.27) |
0.65
(0.42-0.99) |
1.26 (0.41-3.86) |
0.89 (0.71-1.12) |
1.94 (0.95-3.95) |
| No. of hospitalizations† OR (95% CI) | |||||||
| • 0 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • 1-2 | 0.99 (0.90-1.09) |
1.06 (0.90-1.26) |
1.02 (0.91-1.14) |
1.40
(1.01-1.93) |
1.58 (0.87-2.86) |
0.85
(0.74-0.98) |
0.96 (0.72-1.29) |
| • ≥3 | 1.03 (0.77-1.38) |
0.80 (0.51-1.27) |
1.03 (0.75-1.40) |
1.13 (0.44-2.89) |
0.33 (0.04-2.67) |
0.77 (0.52-1.13) |
0.65 (0.28-1.52) |
| No. of ambulatory visits/ OR (95% CI) | |||||||
| • 0 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • 1-2 | 1.04 (0.93-1.16) |
1.17 (0.95-1.43) |
0.96 (0.81-1.12) |
1.21 (0.84-1.76) |
1.12 (0.31-4.07) |
1.06 (0.88-1.27) |
1.31 (0.92-1.86) |
| • 3-9 | 1.09 (0.98-1.22) |
1.14 (0.94-1.39) |
0.88 (0.75-1.03) |
1.13 (0.91-1.85) |
1.03 (0.30-3.52) |
0.94 (0.79-1.12) |
1.34 (0.96-1.88) |
| • ≥10 | 1.13 (1.00-1.29) |
1.05 (0.84-1.32) |
0.89 (0.75-1.05) |
1.57
(1.04-2.38) |
1.11 (0.30-4.06) |
1.03 (0.85-1.25) |
1.05 (0.71-1.56) |
| No. of ED visits,† OR (95% CI) | |||||||
| • 0 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • 1-2 |
0.79
(0.72-0.87) |
1.01 (0.87-1.18) |
1.26
(1.12-1.41) |
1.09 (0.81-1.46) |
1.57 (0.86-2.86) |
1.09 (0.96-1.25) |
1.12 (0.84-1.50) |
| • ≥3 |
0.61
(0.51-0.72) |
0.76 (0.58-1.00) |
1.25 (0.97-1.60) |
0.93 (0.52-1.65) |
1.53 (0.51-4.56) |
0.99 (0.75-1.30) |
1.92 (0.85-4.34) |
| No. of prescriptions,† OR (95% CI) | |||||||
| • 0-2 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • 3-5 | 0.99 (0.92-1.07) |
1.22
(1.07-1.38) |
0.75
(0.69-0.82) |
1.10 (0.89-1.37) |
0.78 (0.45-1.34) |
0.96 (0.86-1.07) |
1.06 (0.86-1.31) |
| • ≥6 | 0.97 (0.89-1.06) |
1.42
(1.23-1.65) |
0.69
(0.63-0.76) |
1.21 (0.94-1.56) |
0.65 (0.36-1.15) |
0.91 (0.81-1.03) |
1.14 (0.90-1.45) |
| Charlson Comorbidity Index,28 OR (95% CI) | |||||||
| • 0 | Ref | Ref | Ref | Ref | Ref | Ref | Ref |
| • 1-2 |
1.20
(1.11-1.29) |
0.98 (0.86-1.12) |
1.13 (1.05-1.22) |
0.80
(0.65-1.00) |
1.20 (0.77-1.86) |
0.88
(0.79-0.97) |
0.91 (0.75-1.09) |
| • ≥3 | 1.07 (0.91-1.25) |
0.72
(0.54-0.95) |
0.99 (0.84-1.15) |
0.56
(0.34-0.91) |
1.03 (0.48-2.23) |
0.71
(0.58-0.88) |
0.76 (0.52-1.11) |
ED—emergency department, OR—odds ratio, Ref—reference category.
N represents the number of prescriptions in each grouping.
†Based on the 365 days prior to the index date.
DISCUSSION
While our results closely follow the upper range of primary medication nonadherence reported in other studies (prevalence estimates ranging from 13.7% to 30.3%),1,14 they still add important information for clinical practice. Medications related to chronic, typically asymptomatic conditions had rates of primary medication nonadherence that were 10 to 20 percentage points higher compared with symptomatic conditions (eg, depression, anxiety, infection). Our data demonstrate the intuitive notion that patients not experiencing symptoms may be less motivated to take medication to treat their conditions. Prescribers should consider this for conditions such as diabetes or hypertension and, particularly, for a new diagnosis or change in medication management. For symptomatic conditions, most patients filled their incident prescriptions. However, the literature suggests that this may be different for secondary adherence, where a prescription is filled but not taken as prescribed.6,29-34
Within the literature, a recent meta-analysis by Lemstra et al pooled 24 studies that considered 550,485 prescriptions and reported a combined primary non-adherence rate of 14.6% (95% CI 13.1% to 16.2%) when considering antihypertensives, lipid-lowering agents, hypoglycemics, and antidepressants.35 In the same paper, the nonadherence rate for lipid-lowering agents was 17.0% (95% CI 14.4% to 19.5%), which closely reflects the rate observed in our study.35 We found that lipid-lowering agents (eg, statin medication) had a relatively low nonadherence rate (15.2%), despite being used for asymptomatic conditions such as hyperlipidemia and cardiovascular disease. This may be owing to patients preferring medication over dietary changes36 or, as Casula et al suggested, a difference in counseling, patients’ motivation, or the perceived benefit of particular medications.37 While the underlying reasons for nonadherence are beyond the scope of our study, our results confirmed the variability of patient characteristics as predictors of primary medication non-adherence.2,13,38 Our models were unable to determine provider characteristics that predict primary medication nonadherence, reinforcing the concept of complex patient-provider interactions.
Our study did demonstrate patient demographic and clinical factors associated with primary medication nonadherence, including having pre-existing polypharmacy, frequently presenting to hospital, and having comorbidities. Still, the significant associations were not consistent across various indications. These findings diverge from the literature; Davidson et al found that an increasing number of medications was associated with improved medication adherence,12 and a review by Seal et al noted several studies where primary medication nonadherence was associated with comorbidity and poorer health status.13
However, since much of the existing literature has focused exclusively on a specific setting or single disease population, our study adds a more comprehensive view of drug initiation and adherence within primary care.
The results of our study share an important similarity with other medication nonadherence literature. Much of the literature suggests a compelling driver of the patient’s decision to fill their prescription is having insurance that covers the cost of the medication.1,39 In our models, several medication classes indicated an increased likelihood of adherence in the most-affluent income quintile; however, this finding was not consistent. Nonetheless, while drug coverage deductibles are lower in less-affluent income quintiles, these costs may still present a barrier for patients with limited financial means. Thus, the results of this study continue to reinforce the important role of the primary care provider in working with patients to consider sustainable drug management or alternative plans.14,40
Our findings suggest that providers may not be able to use particular attributes to predict which patients will or will not fill a new prescription, adding to the notion that primary medication nonadherence is complex and often influenced by varied and competing influences.2,38,41 Based on the poor predictability of our models, caution is required when considering characteristic-based interventions or prediction tools to improve primary medication nonadherence. Rather, an individual approach based on knowledge of the patient and their values is preferable in order to understand the influences and motivations associated with primary medication nonadherence and address the issue.
Limitations
This study had several key limitations. First, since we used medication dispensing information from administrative data linked only to EMR prescribing data of consenting providers, it may not be fully representative of the entire population, despite the large numbers of prescriptions assessed. Further, we were unable to include the cost of individual medication or copayment amounts in our analyses. We also cannot be certain of the exact indication for each prescription or the clinical circumstances related to the interaction that led to the prescription because a review of narrative text was outside the scope of this study. Finally, we did not study the causes of primary medication nonadherence.
Conclusion
Primary medication nonadherence is common, occurring more often with asymptomatic conditions. Overall, we were unable to identify a consistent pattern of association related to patient demographic, clinical, or provider characteristics and primary nonadherence, indicating the need for individual, patient-centred approaches to improving primary medication nonadherence.
Supplementary Material
Editor’s key points
▸ Primary medication nonadherence—when a patient does not fill their initial prescription—increases adverse clinical events, morbidity, mortality, and overall health care costs. Nonadherence is a critical barrier to treatment success and a challenge for health care providers. This study aimed to offer a robust picture of drug initiation and adherence in an adult primary care population, where treatment is often started.
▸ The authors analyzed primary nonadherence across several medication classes for chronic and acute conditions commonly treated in primary care, where nonadherence may affect health outcomes.
▸ This study was unable to identify a consistent pattern of association between primary nonadherence and patient demographic, clinical, or provider characteristics.
Points de repère du rédacteur
▸ La non-adhésion primaire à la médication, c’est-à-dire lorsqu’un patient ne fait pas remplir sa première ordonnance, augmente les événements cliniques indésirables, la morbidité, la mortalité et les coûts globaux en soins de santé. La non-adhésion est un obstacle majeur à la réussite du traitement et un défi pour les professionnels de la santé. Cette étude visait à brosser un tableau probant du démarrage d’une médication et de l’adhésion au médicament au sein d’une population adulte en soins primaires, un milieu où un traitement est souvent amorcé.
▸ Les auteurs ont analysé la non-adhésion primaire dans le cas de diverses classes de médicaments pour des problèmes chroniques et aigus couramment traités en soins primaires, lorsque la non-adhésion peut nuire aux résultats sur le plan de la santé.
▸ Cette étude n’a pas réussi à cerner un modèle uniforme d’associations entre la non-adhésion primaire et les données démographiques des patients et d’autres caractéristiques cliniques ou particulières aux professionnels.
Footnotes
Appendices outlining specific covariates included in the models are available from https://www.cfp.ca. Go to the full text of the article online and click on the CFPlus tab.
Contributors
All authors contributed to the concept and design of the study; data gathering, analysis, and interpretation; and preparing the manuscript for submission.
Competing interests
Dr Alexander G. Singer received a grant from the Canadian Institute for Military and Veteran Health Research with funding and in-kind support from IBM. The other authors have no competing interests to declare. The results and conclusions are those of the authors and no official endorsement by the Manitoba Centre for Health Policy, Manitoba Health, or other data providers, institutions, or funders is intended or should be inferred.
This article has been peer reviewed.
Cet article a fait l’objet d’une révision par des pairs.
References
- 1.Tamblyn R, Eguale T, Huang A, Winslade N, Doran P.. The Incidence and determinants of primary nonadherence with prescribed medication in primary care: a cohort study. Ann Intern Med 2014;160(7):441-50. [DOI] [PubMed] [Google Scholar]
- 2.Raebel MA, Ellis JL, Carroll NM, Bayliss EA, McGinnis B, Schroeder EB, et al. Characteristics of patients with primary non-adherence to medications for hypertension, diabetes, and lipid disorders. J Gen Intern Med 2012;27(1):57-64. Epub 2011 Aug 31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Feehan M, Morrison MA, Tak C, Morisky DE, DeAngelis MM, Munger MA.. Factors predicting self-reported medication low adherence in a large sample of adults in the US general population: a cross-sectional study. BMJ Open 2017;7(6):e014435. Erratum in: BMJ Open 2017;7(8):e014435corr1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Conn VS, Ruppar TM.. Medication adherence outcomes of 771 intervention trials: systematic review and meta-analysis. Prev Med 2017;99:269-76. Epub 2017 Mar 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Nieuwlaat R, Wilczynski N, Navarro T, Hobson N, Jeffery R, Keepanasseril A, et al. Interventions for enhancing medication adherence. Cochrane Database Syst Rev 2014;(11):CD000011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Adherence to long-term therapies: evidence for action. Geneva, Switz: World Health Organization; 2003. Available from: http://whqlibdoc.who.int/publications/2003/9241545992.pdf. Accessed 2022 Jun 20. [Google Scholar]
- 7.Iuga AO, McGuire MJ.. Adherence and health care costs. Risk Manag Healthc Policy 2014;7:35-44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Solomon MD, Majumdar SR.. Primary non-adherence of medications: lifting the veil on prescription-filling behaviors. J Gen Intern Med 2010;25(4):280-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Simpson SH, Eurich DT, Majumdar SR, Padwal RS, Tsuyuki RT, Varney J, et al. A metaanalysis of the association between adherence to drug therapy and mortality. BMJ 2006;333(7557):15. Epub 2006 Jun 21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dalvi V, Mekoth N.. Regimen difficulty and medication non-adherence and the interaction effects of gender and age. Hosp Top 2018;96(2):35-41. Epub 2017 Dec 8. [DOI] [PubMed] [Google Scholar]
- 11.Birtwhistle RV. Canadian Primary Care Sentinel Surveillance Network. A developing resource for family medicine and public health. Can Fam Physician 2011;57:1219-20 (Eng), e401-2 (Fr). [PMC free article] [PubMed] [Google Scholar]
- 12.Davidson E, Lam S, Sokn E.. Predictors of medication nonadherence from outpatient pharmacy data within a large, academic health system. J Pharm Pract 2019;32(2):175-8. Epub 2017 Dec 18. [DOI] [PubMed] [Google Scholar]
- 13.Seal F, Cave AJ, Atkinson LL.. Primary non-adherence of prescribed pharmaceutical treatments and interventions: an investigative review to improve quality in primary care. Qual Prim Care 2017;25(5):344-59. [Google Scholar]
- 14.Fischer MA, Choudhry NK, Brill G, Avorn J, Schneeweiss S, Hutchins D, et al. Trouble getting started: predictors of primary medication nonadherence. Am J Med 2011;124(11):1081.e9-22. [DOI] [PubMed] [Google Scholar]
- 15.Abrams EM, Singer AG, Lix L, Katz A, Yogendran M, Simons FER.. Adherence with epinephrine autoinjector prescriptions in primary care. Allergy Asthma Clin Immunol 2017;13:46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Brunton SA, Polonsky WH.. Hot topics in primary care: medication adherence in type 2 diabetes mellitus: real-world strategies for addressing a common problem. J Fam Pract 2017;66(4 Suppl):S46-51. [PubMed] [Google Scholar]
- 17.Goldstein KM, Zullig LL, Bastian LA, Bosworth HB.. Statin adherence: does gender matter? Curr Atheroscler Rep 2016;18(11):63. [DOI] [PubMed] [Google Scholar]
- 18.Reynolds K, Muntner P, Ceetham TC, Harrison TN, Morisky DE, Silverman S, et al. Primary non-adherence to bisphosphonates in an integrated healthcare setting. Osteoporos Int 2013;24(9):2509-17. Epub 2013 Apr 18. [DOI] [PubMed] [Google Scholar]
- 19.Mehat P, Atiquzzaman M, Esdaile JM, Aviña-Zubieta A, De Vera MA.. Medication nonadherence in systemic lupus erythematosus: a systematic review. Arthritis Care Res (Hoboken: ) 2017;69(11):1706-13. Epub 2017 Sep 21. [DOI] [PubMed] [Google Scholar]
- 20.Van der Zwaard BC, van Hout W, Hugtenburg JG, van der Horst HE, Elders PJM.. Adherence and persistence of patients using oral bone sparing drugs in primary care. Fam Pract 2017;34(5):525-31. [DOI] [PubMed] [Google Scholar]
- 21.Katz A, Enns J, Smith M, Burchill C, Turner K, Towns D.. Population data centre profile: the Manitoba Centre for Health Policy. Int J Popul Data Sci 2020;4(2):1131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Manitoba Centre for Health Policy. Concept dictionary and glossary for population based research. Term: Emergency Department Information System data. Winnipeg, MB: University of Manitoba; 2014. Available from: http://mchp-appserv.cpe.umanitoba.ca/viewDefinition.php?definitionID=104664. Accessed 2020 Mar 1. [Google Scholar]
- 23.Manitoba Centre for Health Policy. Concept dictionary and glossary for population based research. Term: Drug Program Information Network. Winnipeg, MB: University of Manitoba; 2010. Available from: http://mchp-appserv.cpe.umanitoba.ca/viewDefinition.php?definitionID=102596. Accessed 2020 Mar 1. [Google Scholar]
- 24.Williamson T, Green ME, Birtwhistle R, Khan S, Garies S, Wong ST, et al. Validating the 8 CPCSSN case definitions for chronic disease surveillance in a primary care database of electronic health records. Ann Fam Med 2014;12(4):367-72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Queenan JA, Williamson T, Khan S, Drummond N, Garies S, Morkem R, et al. Representativeness of patients and providers in the Canadian Primary Care Sentinel Surveillance Network: a cross-sectional study. CMAJ Open 2016;4(1):E28-32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Manitoba Centre for Health Policy. Concept dictionary and glossary for population based research. Winnipeg, MB: University of Manitoba; 2022. Available from: http://umanitoba.ca/faculties/health_sciences/medicine/units/chs/departmental_units/mchp/resources/concept_dictionary.html. Accessed 2020 Mar 1. [Google Scholar]
- 27.Hosmer DW, Lemeshow S.. Applied logistic regression. 2nd ed. New York, NY: John Wiley & Sons Inc; 2000. [Google Scholar]
- 28.Concept: Charlson Comorbidity Index. Winnipeg, MB: Manitoba Centre for Health Policy; 2021. Available from: http://mchp-appserv.cpe.umanitoba.ca/viewConcept.php?printer=Y&conceptID=1098. Accessed 2022 Jun 16. [Google Scholar]
- 29.Halaris A. A primary care focus on the diagnosis and treatment of major depressive disorder in adults. J Psychiatr Pract 2011;17(5):340-50. [DOI] [PubMed] [Google Scholar]
- 30.Lam RW. Challenges in the treatment of anxiety disorders: beyond guidelines. Int J Psychiatry Clin Pract 2006;10(Suppl 3):18-24. [DOI] [PubMed] [Google Scholar]
- 31.Olfson M, Marcus SC, Tedeschi M, Wan GJ.. Continuity of antidepressant treatment for adults with depression in the United States. Am J Psychiatry 2006;163(1):101-8. [DOI] [PubMed] [Google Scholar]
- 32.Shubber Z, Mills EJ, Nachega JB, Vreeman R, Freitas M, Bock P, et al. Patient-reported barriers to adherence to antiretroviral therapy: a systematic review and metaanalysis. PLoS Med 2016;13(11):e1002183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Corless IB, Hoyt AJ, Tyer-Viola L, Sefcik E, Kemppainen J, Holzemer WL, et al. 90-90-90-Plus: maintaining adherence to antiretroviral therapies. AIDS Patient Care STDS 2017;31(5):227-36. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Sweeney SM, Vanable PA.. The association of HIV-related stigma to HIV medication adherence: a systematic review and synthesis of the literature. AIDS Behav 2016;20(1):29-50. [DOI] [PubMed] [Google Scholar]
- 35.Lemstra M, Nwankwo C, Bird Y, Moraros J.. Primary nonadherence to chronic disease medications: a meta-analysis. Patient Prefer Adherence 2018;12:721-31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Harmsen CG, Støvring H, Jarbøl DE, Nexøe J, Gyrd-Hansen D, Nielsen JB, et al. Medication effectiveness may not be the major reason for accepting cardiovascular preventive medication: a population-based survey. BMC Med Inform Decis Mak 2012;12:89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Casula M, Tragni E, Catapano AL.. Adherence to lipid-lowering treatment: the patient perspective. Patient Prefer Adherence 2012;6:805-14. Epub 2012 Nov 8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Steiner JF, Ho PM, Beaty BL, Dickinson LM, Hanratty R, Zeng C, et al. Sociodemographic and clinical characteristics are not clinically useful predictors of refill adherence in patients with hypertension. Circ Cardiovasc Qual Outcomes 2009;2(5):451-7. Epub 2009 Aug 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Shrank WH, Hoang T, Ettner SL, Glassman PA, Nair K, DeLapp D, et al. The implications of choice: prescribing generic or preferred pharmaceuticals improves medication adherence for chronic conditions. Arch Intern Med 2006;166(3):332-7. [DOI] [PubMed] [Google Scholar]
- 40.Law MR, Cheng L, Dhalla IA, Heard D, Morgan SG.. The effect of cost on adherence to prescription medications in Canada. CMAJ 2012;184(3):297-302. Epub 2012 Jan 16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Chan DC, Shrank WH, Cutler D, Jan S, Fischer MA, Liu J, et al. Patient, physician, and payment predictors of statin adherence. Med Care 2010; 48(3):196-202. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.