Sage Journals: Discover world-class research

Abstract

Purpose:

Examine patterns of medication use, changes in medication patterns over time, and investigate factors associated with medication patterns among older Australian women with Atrial Fibrillation (AF).

Methods:

It is a retrospective analysis of the 1921-26 birth cohort of the Australian Longitudinal Study on Women’s Health (ALSWH), diagnosed with AF between 2000-2015 (N = 1206). Survey data of these women was linked with national registries for medications and death. Latent Transition Analysis (LTA) identified distinct patterns of medication use and transitions among these patterns for 3 consecutive years following AF diagnosis. LTA with co-variates determined the factors associated with latent status membership.

Results:

One-tenth (9.6%, 11.7%, 11.4%) of the study population did not receive any medication for AF in all 3 years following AF diagnosis and about 60% did not receive any medication for the prevention of thromboembolism. Among those who received medications, almost three-quarters (76.6%, 68.4%, 68.5%) received some kind of combination of medications. LTA indicated at least 6 different patterns of AF medications. These patterns had transition probabilities >85% for most of the latent statuses. All factors but diabetes mellitus among the CHA₂DS₂-VA scoring scheme were independently associated with latent status membership at the time of AF diagnosis.

Conclusions:

Evaluation of pharmacological treatment indicates that prevention of thromboembolism is inadequate among women with AF. There exists wide variations in medication patterns. However, once in a particular pattern, women are likely to continue the same medications long-term. This underscores the importance of initial assessment of patient profile and stroke risk score in determining the treatment for AF. Failure to assess risk makes women susceptible to devastating AF complications.

Keywords

atrial fibrillation aged 80 and over latent transition analysis medication patterns longitudinal studies stroke prevention

Introduction

Globally, Atrial Fibrillation (AF) is the most common cardiac arrhythmia. Prevalence of AF is age dependent and increases with advancing age.¹ In people younger than 40, AF prevalence is less than 1%,¹ rising up to 20-25% among those aged 80 years and above.^2
-4 With the changes in population demographics and improved longevity, AF is considered an emerging epidemic and a major public health challenge.⁵ Literature suggests that AF is independently associated with a number of health impacts, such as reduced quality of life,⁶ increased risk of embolic stroke,^1,7 and premature death.^1,7
-9

Treatment of AF involves careful assessment of risks and other comorbidities and involves two major decisions - rate or rhythm control, and initiation of anticoagulation therapy to prevent stroke. CHA₂DS₂-VASc (Congestive heart failure, Hypertension, Age 65-74, Diabetes, Stroke or TIA, Vascular disease, Age ≥ 75 years and Sex[female]) score is the most widely used scoring scheme to calculate the risk of stroke among AF patients.¹⁰ It is recommended that a score of one or more for men, and two or more for women indicates high risk of stroke and patients be treated with anticoagulants. However, the 2018 Australian guidelines,¹¹ recommend the use of the risk-prediction model CHA₂DS₂-VA (sexless version of CHA₂DS₂-VASc recommending same threshold for both genders) to guide decisions about anticoagulation treatment.¹¹ Literature, however, indicates that medication use does not correlate well with the stroke risk.¹² Rather, the use of anticoagulant decreases with increasing stroke risk especially among women and older patients.¹³

Women with AF are particularly at increased risk of AF complications such as stroke and death compared to men.^7,9 However, despite improvements in the overall trends of medication uptake, treatment among women is less aggressive and less effective¹⁴ leading to increased complications and consequences related to AF. Sabouret et al.¹⁵ confirms that women < 75 years were 17 times less likely to receive antithrombotic treatment compared to men. This is doubled to 34% in women ≥ 75 years of age. With age associated increased risks, there is a need to understand this potential for under treatment of older women with AF particularly for ones in their 80s and 90s.

To the best of our knowledge, there are no studies to date of the long term use of AF medication among octa- and nonagenarian women in Australia or elsewhere. This study therefore aims to explore the medication use for AF among older women. Specific objectives include understanding the medication use for AF among older Australian women, investigating the distinct patterns of AF medications over time, and examining factors associated with these identified medication patterns.

Methods

Study Population and Datasets

The study is based on the data from the oldest cohort (born 1921-26) of the Australian Longitudinal Study of Women’s Health (ALSWH, www.alswh.org.au). ALSWH is a prospective, longitudinal cohort study examining the socio-demographic and lifestyle factors over the life course which affect the physical and mental health of women and their use of health services. In 1996, this cohort of women participated in the first survey (N = 12432) and were regularly surveyed every 3 years thereafter until 2011, returning self-completed comprehensive paper questionnaires. Since 2011, due to the age of the participants, a shortened survey continues to be sent to participants every 6 months. Further details about the ALWSH are reported elsewhere.¹⁶

AF was determined using linked state-based hospital data from New South Wales (NSW), Queensland (QLD), Western Australia (WA), South Australia (SA) and the Australian Capital Territory (ACT), during the period from 2000 to 2015. Previously, 1827 women from this cohort were identified as having AF based on the hospital records, using ICD-10-AM (I48, I48.0, I48.1, I48.2 and I48.9) codes.¹⁷ Details of the AF identification process is reported in another paper.²

Eligibility for the Analysis

Among 1827 women previously identified with AF, 1206 women were eligible for the current analysis. Women were included in the analysis if they: i) consented for Pharmaceutical Benefit Scheme (PBS)¹⁸ data linkage, and ii) were alive for at least 30 months after first hospital reported diagnosis of AF to ensure assessment of medication use for at least 3 years post AF diagnosis (Year₁ [T₁], Year₂ [T₂], Year₃ [T₃]). The main reason for exclusion from the current analysis was death within 30 months of first AF diagnosis (n = 620 women), with only 1 woman who did not consent to PBS data linkage.

AF medication data for all eligible women was obtained through PBS data linked with ALSWH participants. The PBS is an administrative dataset initiated to monitor prescription medicines in Australia and is available for ALSWH participants from 2000-2018. Using Anatomical Therapeutic Chemical (ATC)¹⁹ codes, AF medications use was evaluated for each subsequent year following the first reported diagnosis of AF for 3 consecutive years. (Details of the AF medications and corresponding ATC codes are presented in Supplementary Table S1.) Yearly medication use for all identified ATC codes was considered positive if at least one PBS record indicated dispensing of that medication. All medications were initially classified into 6 categories: 1) Rate Control (RC); 2) Rhythm Control (RyC); 3) Vitamin K Inhibitors (VKI); 4) Platelet-aggregation Inhibitors (PAI); 5) Direct Thrombin Inhibitors; and 6) Factor Xa Inhibitors. Categories 5 and 6 were later merged for analysis due to low prevalence of these drugs in the study population and were labeled as Non-Vitamin K Antagonist Oral Anticoagulants (NOACs).

Medication Use and Time Period for Ascertainment

Medication use for each category was defined dichotomously (yes/no), and was calculated for 3 consecutive years. The 3 year time periods were calculated as 12 (T₁), 24 (T₂) and 36 (T₃) months from the date of first reported AF diagnosis in hospital data. Medication use was considered positive for a given category per year if a woman had at least one record of medication in that category during that year.

Statistical Analysis

Descriptive statistics were used to describe study participants under 4 domains: (i) demographics; (ii) health behaviors; (iii) access to health care services; and (iv) self-reported co-morbid conditions. Variables under these domains are described in supplementary Table S2. Data on these variables were obtained from ALSWH Survey 3 conducted in 2002, corresponding to the earliest available linked hospital data for most states. These were considered as baseline characteristics of the study population. Missing data were filled in from Survey 2 (1999), initially, and then from Survey 1 if necessary.

Prevalence of AF medication use

Prevalence of AF medication use was assessed for each year following AF diagnosis based on the drug categories and the total number of drug categories used by AF women simultaneously in a year.

Latent transition analysis (LTA)

Latent Transition analysis (LTA) was applied to identify underlying patterns of AF medication use among older women in Australia and to understand how they change over time for at least 3 years after AF diagnosis. LTA is an extension of Latent Class Analysis (LCA).²⁰ In addition to identifying mutually exclusive latent statuses, known as latent classes in LCA, LTA also allows the estimation of the incidence of transition among these latent statuses from one time point to the next, conditional on the latent status membership at the previous time point. The analysis was conducted for a range of latent statuses from 2 to 9. The iterative model building process was stopped when the fit indices showed no further improvement. Model selection was based on parsimony, model interpretability, and information criteria (Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC), for which lower scores represent better fitting models).

Following model selection, 3 sets of parameters were calculated for each latent status: latent status prevalence, gamma ( $γ$ ); item response probabilities, rho ( $ρ)$ ; and transition probabilities, Tau $(τ)$ ; from T₁ to T₂ and T₂ to T_3,. Item response probabilities were assumed to be constant over time and were used to assign labels to latent statuses. The assumption of measurement invariance of item response probabilities was statistically assessed. This was followed by LTA with covariate analysis. Individual characteristics from CHA₂DS₂-VA score were independently assessed as covariates at T₁. Whether each covariate was a significant predictor of latent status membership at T_1, was tested using the likelihood ratio test between the model with and the model without the covariate. Multivariable LTA with covariates was performed to understand what combination of CHA₂DS₂-VA factors determine the class memberships in the model.

All statistical analysis were performed using SAS 9.4 (SAS Institute, Cary, NC, USA).

Ethical Approval

The ALSWH project has ongoing ethical clearance from both the University of Newcastle (UoN) and University of Queensland’s (UQ) Human Research Ethics Committees (HREC) [HREC UoN no.: H-076-0795 & HREC UQ no. 2004000224]. This project has been approved by the Data Access Committee of ALSWH, registered as Project A685.

Results

Baseline Characteristics

Most women lived either in major cities or regional areas at baseline with only 2% of women living in remote Australia, and 42% were married or partnered at the time of survey with another 56% previously married or partnered (Table 1). More than three quarters of all women reported having excellent access to hospital (87%) and General Practitioners (GPs). A quarter of women reported difficulty managing on their available income. The 3 most common comorbidities included hypertension (HTN, 53%), cancer (52%, includes skin cancers) and arthritis (47%). Less than 5% of women reported being a current smoker, with a similar proportion reporting high risk alcohol consumption. More than half of the women had a BMI which was overweight or obese (54%) with 42% of women reporting a sedentary lifestyle (defined as 0-10 minutes of moderate physical activity per week). At the time of first reported AF diagnosis, all women were aged 75 years or older (range 75-94).

Table 1.

Baseline Characteristics of Older Australian Women (Born 1921-26) Who Were Identified With Atrial Fibrillation Between 2000 to 2015 (N = 1206).

Baseline^ characteristic	N	%
Sociodemographic
Residential area
Major cities	548	45.4
Regional areas	630	52.2
Remote areas	28	2.3
Marital status
Married or partnered	411	42.0
Widowed/separated/divorced	547	55.9
Single (never married)	21	2.1
Missing	227
Manage on available income
Easy always/not too bad	838	69.7
Difficult sometimes	260	21.6
Impossible	104	8.7
Missing	4
Health Behaviors
Body Mass Index (BMI, kg/m²)
Underweight (BMI < 18.5)	39	3.3
Healthy BMI (18.5 ≤ BMI < 25)	502	42.9
Overweight (25 ≤ BMI < 30)	422	36.1
Obese (BMI ≥ 30)	206	17.6
Missing	37
Exercise status
Nil-Sedentary	452	42.4
Low	278	26.1
Moderate	156	14.7
High	179	16.8
Missing	141
Alcohol status
Never drank	453	37.6
Low risk/rarely drinks	715	59.4
High risk/risky drinker	36	3.0
Missing	2
Smoking status
Never smoked	732	65.5
Ex-smoker	336	30.0
Current smoker	50	4.5
Missing	88
Access to health services
Access to hospital
Excellent, very good, good	917	87.2
Fair or poor	93	8.8
Don’t know	42	4.0
Missing
Access to GP
Excellent, very good, good	784	76.0
Fair or poor	143	13.9
Don’t know	105	10.2
Missing	174
Self-reported health conditions*
**Hypertension (HTN)	632	52.5
Missing	1
**Diabetes mellitus	120	10.0
Missing	1
Angina	123	12.6
Missing	233
Heart attack	44	4.5
Missing	233
Stroke	52	4.3
Missing	0
**^&Cancer	521	51.9
Missing	203
**Arthritis	525	47.0
Missing	88
**Osteoporosis	224	18.6
Missing	1
Asthma	135	11.2
Missing	1
**Depression	167	17.5
Missing	251
**Bronchitis/Emphysema (COPD)	78	6.5
Missing	2

^{^} Baseline values were obtained from Australian Longitudinal Study on Women’s Health 1921-26 cohort, Survey 3 (conducted in 2002). Missing values were filled in from Survey 2 and then Survey 1 where possible.

*Self-reported health conditions were recorded as yes or no. Numbers reported here are for women reported “yes” for that particular condition.

**Enduring conditions (once reported, they were considered to have that condition thereafter)

^& Includes all cancers, including skin cancer

Use of Medication

Despite strong recommendation for initiation of medication for prevention of thromboembolism among women ≥ 75 years with AF (using the CHA₂DS₂-VA or CHA₂DS₂-VASc scoring schemes), only 44% of women received any mediation for stroke prevention and a further ∼10% did not receive any medication for AF during the first year post-diagnosis (Table 2). RC medications were the most commonly used groups of medications, followed by PAI and VKI. The overall pattern of medication use was similar across all 3 years, though reduction was observed for all medications from T₁ to T₂ except the use of NOACs. NOACs were the least-used medications among the study population, but data suggests a slight increase in use of NOACs over the study time period. This is understandable as NOACs are relatively new drugs, approved for prevention of thromboembolism in 2013 in Australia.²¹ The comparison of medication use from T₂ to T₃ indicates similar prevalence of drug use for all categories.

Table 2.

Prevalence of Atrial Fibrillation Medications Among Older Australian Women (Born 1921-26, N = 1206) for 3 Consecutive Years Following First Diagnosis of Atrial Fibrillation.

	1st year after AF diagnosis		2nd year after AF diagnosis		3rd year after AF diagnosis
	n	(%)	n	(%)	n	(%)
Type of medication^a
Rate Control (RC)	870	72.1	829	68.7	827	68.6
Rhythm Control (RyC)	225	18.7	163	13.5	158	13.1
Vitamin K Inhibitor (VKI)	493	40.9	443	36.7	437	36.2
Platelet Aggregation Inhibitor (PAI)	574	47.6	468	38.8	468	38.8
NOACs^b	36	3.0	40	3.3	53	4.4
No medication	116	9.6	141	11.7	137	11.4
Number of different medications^c
1 drug only	255	23.4	337	31.6	337	31.5
Any combination of 2 drugs	604	55.4	595	55.9	602	56.3
Any combination of 3 drugs	190	17.4	116	10.9	118	11.0
Any combination of 4 medications	40	3.7	17	1.6	12	1.1
All 5 drugs	1	0.1	0	0	0	0

^a N = 1206; women may receive more than one type of medication; percentages do not add up to 100%

^b Non Vitamin K antagonist Oral Anticoagulants (includes Direct Thrombin Inhibitors and Factor Xa Inhibitors)

^c Restricted to women who received medications in that year (n = 1090, 1065 and 1069 for the 1st, 2nd and 3rd year respectively)

Among women who received any type of medication (Table 2), the majority received some combination of medications. The majority of women (55%) received a combination of 2 medications, this prevalence remained stable across all 3 years. The prevalence of women with one type of AF medication increased from 23% to 31% from T₂ to T₃. In the first year, 17% of women used 3 different types of AF medication although this decreased to 11% for the second and third years. Very few women used 4 types of medications; the prevalence decreased from 4% in the first year to 1% in the third year.

Latent Transition Analysis

Model selection for number of statuses

Latent transition analysis was performed to further understand the medication combinations and change in these medication patterns over time. Latent statuses were identified based on the observed use of AF medications, e.g. RC, RyC, PAI, VKI, and NOACs, in each of the 3 years following AF diagnosis. Based on model selection criteria of parsimony, model interpretability and the information criteria, a model with 6 latent statuses was selected as the optimal model (Supplementary Table S3). It indicates that AIC and BIC both declined through to the 6 status model. At the seventh status, the AIC continued to decline but the BIC rose. Thus, AIC points to a 7 status model and BIC points to a 6 status model. However, the 6 class model was chosen after careful analysis of latent class separation, as it was more parsimonious, while being conceptually interpretable and more clinically relevant.

Latent status prevalence

Following careful review of the item response probabilities (Table 3), labels were assigned to the latent statuses: 1) Low medications; 2) Rate Control only (RC only); 3) Platelet-aggregation Inhibitors only (PAI only); 4) Rate Control with Platelet-aggregation Inhibitors (RC with PAI); 5) Rate Control with Vitamin K inhibitors (RC with VKI); and 6) Rate, Rhythm and Platelet-aggregation Inhibitors (RC, RyC with PAI). For each time period, “RC with VKI” was the most prevalent latent status (29.5%, 27.4% & 27.8%) followed by “RC with PAI” (25.5%, 21.7% & 21.7%). Prevalence for latent status “low medications,” “RC only” and “RC, RyC with PAI” was similar ranging from 10-13%. The least prevalent latent status was “PAI only.” A small decrease was observed in the 2 most prevalent latent statuses (“RC with VKI” and “RC with PAI”) and slight increase in all others from T₁ to T₂. The prevalence of all latent statuses did not change much from T₂ to T₃.

Table 3.

Six Latent Status Model of AF Medications Use for 3 Consecutive Years of Older Australian Women Were Identified With Atrial Fibrillation Between 2000 to 2015 (N = 1206).

	Latent statuses
	Low medications	RC only	PAI only	RC with PAI	RC with VKI	RC, RyC & PAI
Latent status prevalence
Time 1 (1 year from first reported AF diagnosis)	13.8%	10.2%	7.6%	25.5%	29.5%	13.4%
Time 2 (2nd year of first reported AF diagnosis)	15.7%	15.1%	9.6%	21.7%	27.4%	10.5%
Time 3 (3rd year of first reported AF diagnosis)	14.0%	16.5%	9.7%	21.7%	27.8%	10.4%
Item response probabilities (“yes” response)^*
Rate control	0.018	0.956	0.022	0.982	0.972	0.527
Rhythm control	0.000	0.027	0.021	0.019	0.094	1.000
Vitamin K Inhibitors	0.323	0.000	0.018	0.069	0.998	0.293
Platelet Aggregation Inhibitors	0.048	0.047	0.838	0.962	0.142	0.596
NOACs	0.019	0.132	0.036	0.016	0.011	0.039

* Item response probabilities constrained equal across times

Item response probabilities ≥ .5 are in bold to facilitate interpretation

RC = Rate Control, PAI = Platelet Aggregation Inhibitor, VKI = Vitamin K Inhibitor, RyC = Rhythm Control.

Measurement invariance

Hypothesis testing for equality of item response probabilities across time points indicated that differences (either qualitative or quantitative) exist between the latent statuses across times (Supplementary Table S4). Item response probability patterns were visually assessed for all 3 time points and were found to be similar for all latent statuses except for minor difference in latent status 6 “RC, RyC with PAI” at T₃. Women in this latent status at T₃ were less likely to be using rate control compared to at T₁ and T₂. Moreover, the BIC favors the restricted model. For the reasons above and for the ease of interpretability, item response probabilities ( $ρ$ ) were assumed to be equal across time for this analysis. This ensured that the same set of latent statuses would be identified at T₁ to T₃. $ρ$ parameters were then assessed and evaluated to assign labels to the latent statuses. Supplementary Figure S5 shows the 6 different patterns of AF medications among the ALSWH cohort.

Transition probabilities

Table 4 shows transition probability matrix. It indicates that despite some movement between latent statuses, in general a woman’s most likely latent status at T₂ was her latent status membership at T_1. Women in “Low medication” at T₁ had 85% chance of remaining in “Low medication” at T₂. Women in all other latent statuses also showed similar high probabilities (>74%) of staying in the same latent status at T₂ except for women in “RC, RyC with PAI”. Women in this latent status had only 68% chance of remaining in the same latent class at T₂ with 11% likely to be in “RC with PAI,” 9% in “low medication,” 7% in “PAI only,” and 5% in “RC only” latent statuses at T₂. Thus, women belonging to almost all latent statuses had high probability of remaining in the same latent status at any time except for “RC, RyC with PAI.” This indicates that once prescribed, women are most likely to continue with the same medications for subsequent years.

Tables 4.

Transition Probabilities for 6 Class Model for AF Medication Patterns Among Older Australian Women Who Were Identified With Atrial Fibrillation Between 2000 to 2015 (N = 1206).

	Low medications	RC only	PAI only	RC with PAI	RC with VKI	RC, RyC with PAI
Probability of transitioning to……….….Time 2 latent statusConditional on Time 1 latent status
Low medications	0.85	0.03	0.00	0.02	0.05	0.05
RC only	0.06	0.88	0.00	0.07	0.04	0.00
PAI only	0.09	0.00	0.87	0.03	0.02	0.00
RC with PAI	0.01	0.13	0.08	0.74	0.01	0.02
RC with VKI	0.04	0.06	0.00	0.03	0.88	0.00
RC, RyC with PAI	0.09	0.05	0.07	0.11	0.00	0.68
Probability of transitioning to……….….Time 3 latent statusConditional on Time 2 latent status
Low medications	0.88	0.01	0.00	0.00	0.09	0.02
RC only	0.00	0.87	0.00	0.11	0.02	0.00
PAI only	0.00	0.02	0.89	0.07	0.00	0.02
RC with PAI	0.00	0.06	0.05	0.86	0.03	0.00
RC with VKI	0.00	0.06	0.00	0.00	0.93	0.00
RC, RyC with PAI	0.01	0.00	0.00	0.05	0.00	0.94

RC = Rate Control, PAI = Platelet-aggregation Inhibitor, VKI = Vitamin K Inhibitor, RyC = Rhythm Control.

LTA with covariates

We used individual factors from the CHA₂DS₂-VA score (Congestive heart failure, Hypertension, Age > 65-74 years, Diabetes mellitus, Stroke or TIA and Vascular disease) as covariates to predict latent status membership at T₁. All variables were dichotomous, with the exception of age. Age was taken to be “age at diagnosis,” and was a continuous variable which was standardized. The latent status “low medication” was considered as reference category for this analysis. Table 5 indicates that women with HTN were 2 times likely than women without HTN to be in all latent status at T₁ and 4 times more likely to receive “RC only” at T₁ compared to women with “low medication” category. Most striking was the contribution of history of stroke to the latent class membership. Previous history of stroke was less likely to be associated with all but one “RC with VKI” latent status. Only 14% women with the history of stroke were likely to receive “RC with VKI” rendering most women at high risk of recurrence of stroke. Table 5 also indicates that for every 1 unit increase in age at diagnosis women were twice as likely to be categorized in “RC only” compared to “low medications”.

Table 5.

Predictors of Membership (Expressed as Odds Ratios) in Time1 Latent Statuses of AF Medication Patterns Among Older Australian Women Who Were Identified With Atrial Fibrillation Between 2000 to 2015 (N = 1206).

Co-variate	Low medications	RC only	PAI only	RC with PAI	RC with VKI	RC, RyC with PAI
History of hypertension	ref	3.87	1.94	2.02	1.98	2.33
History of congestive heart failure	ref	1.70	1.17	1.91	2.75	2.28
History of diabetes mellitus	ref	1.37	2.09	1.94	1.99	1.53
History of stroke	ref	0.94	0.65	0.52	1.06	0.17
History of vascular disease	ref	1.55	2.39	2.59	1.14	2.64
Age	ref	2.21	1.15	1.18	1.07	1.09

RC = Rate Control, PAI = Platelet Aggregation Inhibitor, VKI = Vitamin K Inhibitor, RyC = Rhythm Control.

Odds ratios generated using Latent Transition Analysis including covariates.

Discussion

To the best of our knowledge, this is the first population based study into understanding the medication patterns among older women with AF, both globally and in Australia. Our study indicates there is low use of medications for prevention of thromboembolic events among older AF patients. Only 40% of women in our study population were either on Vitamin K inhibitors or NOACs suggesting that about 60% women with AF did not receive any anti-thrombolytic medication despite being at high risk of disabling condition like stroke. Our study did not explore the dosages of medications and therefore cannot confirm if the women receiving vitamin K inhibitors were using adequate dosages. Most previous studies have reported low prevalence of anticoagulation medication^22,23 among AF patients, similar to our findings, while patients who do receive medications are not anticoagulated adequately having sub therapeutic levels of INR.²⁴

Our study also indicates that there are wide variations with respect to management of AF. There were 6 distinct patterns of AF medication use ranging from “low medication” (14-16%) to two thirds of women using some combination of medications to one particular drug group preferred “RC only” (10-16%) or “PAI only” (8-10%). One possible reason for this wide variation could be the differences in health care providers. Literature confirms diverse preferences for medication prescriptions among different health care providers.²⁵ Cardiologists are more oriented toward prevention of thromboembolism and may be more confident prescribing anticoagulants compared to other health care providers.²⁵ Our study cannot identify healthcare provider prescribing the medication which could be the potential reason for the differences in the medication patterns. Despite individual preferences for treatment, all women in our study were ≥ 75 years of age and based on the treatment guidelines in Australia or elsewhere should have been treated with anticoagulants. In contrast, only 40% of women received either vitamin K inhibitor or NOACs suggesting inadequate treatment of AF among women, with resultant increased risk of stroke, disability leading to poor quality of life and high economic burden on individuals, families and health system.

About two thirds of the women were using a combination of 2 or 3 medications. For instance “RC with VKI” was the most common combination reported in this study (29.5%, 27.4%, 27.8%) followed by “RC with PAI” (25.5%, 21.7%, 21.7%). The next most prevalent category was the combination of 3 medications “RC, RyC with PAI” (13.4%, 10.5%, 10.4%) This combination was the least stable latent status with only 68% of the women having the probability of remaining in the same latent status. Mostly all other latent statuses were stable over all 3 years and women’s drug pattern in the subsequent year is dependent on the medication pattern in the previous year. This suggests that once medications are prescribed, it is unlikely for older women to be prescribed other medications. This underscores the importance of careful and diligent medication prescription at the time of diagnosis as older women are likely to continue with the same medication patterns in subsequent years and perhaps even longer.

Another interesting finding was that a history of diabetes does not play a role in the decision making for the medication prescription for AF among women. All other individual factors from CHA₂DS₂VA score were associated with latent status membership.

Strengths and weakness

This is the first population-based study to understand the medication pattern for AF among older women. This study did not investigate medication dosage, monitoring of coagulation status or the type of health care provider, reflecting limitations of the provided linked administrative medication data. We acknowledge that our study had a low prevalence of NOACs due to the time frame of the study. NOACs were approved around 2013 for use among AF patients for the prevention of thromboembolism and their uptake took another couple of years. We are unsure if the low prevalence of NOACs in our study is due to the delayed approval of the medication or because of a lack of studies about the efficacy of these medications among older age groups. This study included women who lived for at least 30 months after the first AF diagnosis report in the hospital data. Though this is considered as a selection bias to include women who were relatively less sick but it was performed to evaluate the longitudinal use of AF medication.

Conclusion

Despite the high risk of stroke among older women with AF, less than half are treated for the prevention of thromboembolism. This makes them at high risk of stroke with devastating consequences or disability further affecting their quality of life. More awareness and education for clinicians might further help in reducing this gap.

Supplemental Material

Supplemental Material, Supplementary_Material_for_Patterns_of_AF_Medications - Patterns of Medications for Atrial Fibrillation Among Older Women: Results From the Australian Longitudinal Study on Women’s Health

Supplemental Material, Supplementary_Material_for_Patterns_of_AF_Medications for Patterns of Medications for Atrial Fibrillation Among Older Women: Results From the Australian Longitudinal Study on Women’s Health by Shazia Shehzad Abbas, Tazeen Majeed, Balakrishnan R. Nair, Peta M. Forder, M. Biostatistics, Natasha Weaver and Julie E. Byles in Journal of Cardiovascular Pharmacology and Therapeutics

Footnotes

Authors’ Note

The corresponding author is a PhD candidate and has received “Higher Degree by Research” scholarship from the University of Newcastle. Data may be obtained from contacting Australian Longitudinal Study on Women’s Health () and are not publicly available.

Acknowledgments

The research on which this study is based was conducted as part of the Australian Longitudinal Study on Women’s Health by University of Queensland and the University of Newcastle. We are grateful to the Australian Government Department of Health for funding and to the women who provided the survey data. The authors acknowledge the assistance of the Data Linkage Unit at the Australian Institute of Health and Welfare (AIHW) for undertaking the data linkage to the National Death Index (NDI). The authors also acknowledge the following: Centre for Health Record Linkage (CHeReL), NSW Ministry of Health and ACT Health, for the NSW Admitted Patients Data Collection, and the ACT Admitted Patient Care Data Collection; Queensland Health, including the Statistical Services Branch, for the Qld Hospital Admitted Patient Data Collection; Department of Health Western Australia, including the Data Linkage Branch, for the WA Hospital Morbidity Data Collection; SA NT Datalink, and SA Health for the SA Public Hospital Separations Data Collection. The authors acknowledge the Department of Health and Medicare Australia for providing MBS and PBS data, and the Australian Institute of Health and Welfare (AIHW) as the integrating authority.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Shazia Shehzad Abbas

Supplemental Material

Supplemental material for this article is available online.

References

Ball

Carrington

McMurray

JJV

Stewart

. Atrial fibrillation: profile and burden of an evolving epidemic in the 21st century. Int J Cardiol. 2013;167(5):1807–1824.

Abbas

Majeed

Nair

Forder

Weaver

Byles

. Burden of atrial fibrillation and stroke risk among octa- and nonagenarian women in Australia. Ann Epidemiol (forthcoming). 2020;44:31–37.e2.

Briffa

Hung

Knuiman

, et al. Trends in incidence and prevalence of hospitalization for atrial fibrillation and associated mortality in Western Australia, 1995–2010. Int J Cardiol. 2016;208:19–25.

Ding

Fratiglioni

Johnell

Fastbom

Ljungdahl

Qiu

. Atrial fibrillation and use of antithrombotic medications in older people: a population-based study. Int J Cardiol. 2017;249:173–178.

Braunwald

. Cardiovascular medicine at the turn of the millennium: triumphs, concerns, and opportunities. N Engl J Med. 1997;337(19):1360–1369.

Thrall

Lane

Carroll

Lip

GYH

. Quality of life in patients with atrial fibrillation: a systematic review. Am J Med. 2006;119(5):448.e441–448.e419.

Stewart

Hart

Hole

McMurray

. A population-based study of the long-term risks associated with atrial fibrillation: 20-year follow-up of the Renfrew/Paisley study. Am J Med. 2002;113(5):359–364.

Piccini

Hammill

Sinner

, et al. Incidence and prevalence of atrial fibrillation and associated mortality among Medicare beneficiaries: 1993–2007. Circ: Cardiovasc Qual Outcomes. 2012;5(1):85–93.

Chugh

Havmoeller

Narayanan

, et al. Worldwide epidemiology of atrial fibrillation. Circulation. 2014;129(8):837–847.

10.

Kirchhof

Benussi

Kotecha

, et al. 2016 ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur Heart J. 2016;37(38):2893–2962.

11.

Brieger

Amerena

Attia

, et al. National Heart Foundation of Australia and the Cardiac Society of Australia and New Zealand: Australian clinical guidelines for the diagnosis and management of atrial fibrillation 2018. Heart Lung Circ. 2018;27(10):1209–1266.

12.

Hodgkinson

Taylor

Hobbs

. Treatment pathways for patients with atrial fibrillation. Int J Clin Pract. 2012;66(1):44–52.

13.

Piccini

DeWald

Hernandez

Curtis

. Pharmacotherapy in Medicare beneficiaries with atrial fibrillation. Heart Rhythm. 2012;9(9):1403–1408.

14.

Ikemura

Kohsaka

Kimura

, et al. Assessment of sex differences in the initial symptom burden, applied treatment strategy, and quality of life in Japanese patients with atrial fibrillation. JAMA Netw. 2019;2(3):e191145.

15.

Sabouret

Depret-Bixio

Cotte

F-E

Marie

Bedira

Blin

. Sex differences in stroke prevention in atrial fibrillation in French primary care. Results of the AFIGP (Atrial Fibrillation In General Practice) Database. Clin Res Cardiol. 2014;103(11):887–893.

16.

Lee

Dobson

Brown

, et al. Cohort profile: the Australian longitudinal study on women’s health. Int J Epidemiol. 2005;34(5):987–991.

17.

Australian Consortium for Classification Development. The International Statistical Classification of Disease and Related Health Problems, Tenth revision, Australian Modification (ICD-10-AM), Australian Classification of Health Interventions (ACHI) and Asutralian Coding Standards (ACS) - ICD-10-AM-AM/ACHI/ACS. 10th ed. Independent Housing Pricing Authority; 2017.

18.

Australian Longitudinal Study on Women’s Health. Data Linkage: Pharmaceutical Benefits Scheme (PBS). Published 2019. Accessed December 03, 2019. https://www.alswh.org.au/how-to-access-the-data/external-linked-datasets/pharmaceutical-benefits-scheme.

19.

WHO Collaborating Centre for Drug Statistics Methedology. ATC/DDD Index 2019 Web site. Updated December 13, 2018. Accessed December 03, 2019. https://www.whocc.no/atc_ddd_index/

20.

Collins

Lanza

. Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. Vol. 718. John Wiley & Sons; 2009.

21.

Australian Government Department of Health. The Pharmaceutical Benefits Scheme (PBS). Noval Oral Anticoagulamts: Predicted vs actual analysis Web site. Published 2016. Updated October 2016. Accessed January 23, 2020. http://www.pbs.gov.au/info/industry/listing/participants/public-release-docs/2016-06/noacs-non-valvular-atrial-fibrillation-june-2016

22.

Ang

Peterson

Friesen

Vial

. Review of antithrombotic drug usage in atrial fibrillation. J Clin Pharm Ther. 1998;23(2):97–106.

23.

Bajorek

Krass

Ogle

Duguid

Shenfield

. The impact of age on antithrombotic use in elderly patients with non-valvular atrial fibrillation. Australas J Ageing. 2002;21(1):36–41.

24.

Andrikopoulos

Pastromas

Mantas

, et al. Management of atrial fibrillation in Greece: the MANAGE-AF study. Hellenic J Cardiol. 2014;55(4):281–287.

25.

Fosbol

Holmes

Piccini

, et al. Provider specialty and atrial fibrillation treatment strategies in United States community practice: findings from the ORBIT-AF registry. J Am Heart Assoc. 2013;2(4):e000110.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.24 MB