Long-term use and cost-effectiveness of secondary prevention drugs for heart disease in Western Australian seniors (WAMACH): a study protocol

Effect of frailty on initiation of statins following incident acute coronary syndromes in patients aged ≥75 years

INTRODUCTION

Statin use for preventing recurrent acute coronary syndromes (ACS) is low in older people due to many clinical factors, including frailty. Using the recently developed hospital frailty risk score, which allows ascertainment of frailty from real-world data, we examined the association between frailty and initiation of statin treatment following incident ACS in patients aged ≥75 years. Our secondary aim was to determine whether non-initiation of statins was associated with more conservative treatment, defined as non-receipt of evidence-based medicines and/or coronary artery procedures.

METHODS

We used person-linked hospital administrative and Pharmaceutical Benefits Scheme data to identify incident ACS admissions between 2005 and 2008 in Western Australia and prescription medicine use, respectively. Outcomes were receipt of any statin, high-dose statin, beta-blockers, renin-angiotensin system inhibitors (RASI), antiplatelets and coronary artery procedures within six months of the incident ACS and were analysed using multivariable generalised linear regression models.

RESULTS

In 1,558 patients (52.4% female, mean age 82.6 years), initiation of any statin or high-dose statin decreased with increasing frailty. The adjusted risk ratios for any statin were 0.89 (95% CI: 0.82-0.97) and 0.67 (95% CI: 0.54-0.85) for the intermediate- and high-frailty categories compared with the low-frailty category, respectively. Compared with patients who received statins, those not receiving statins were less likely (p<0.001) to receive beta-blockers (80.8% vs 51.5%), RASI (86.9% vs 62.1%), antiplatelets (90.9% vs 65.1%) or a coronary artery procedure (65.9% vs 21.1%).

CONCLUSIONS

Increasing frailty is inversely associated with initiation of statins and generally leads to a more conservative approach to treatment of older patients with ACS.

Machine learning risk prediction model for acute coronary syndrome and death from use of non-steroidal anti-inflammatory drugs in administrative data

Our aim was to investigate the usefulness of machine learning approaches on linked administrative health data at the population level in predicting older patients’ one-year risk of acute coronary syndrome and death following the use of non-steroidal anti-inflammatory drugs (NSAIDs). Patients from a Western Australian cardiovascular population who were supplied with NSAIDs between 1 Jan 2003 and 31 Dec 2004 were identified from Pharmaceutical Benefits Scheme data. Comorbidities from linked hospital admissions data and medication history were inputs. Admissions for acute coronary syndrome or death within one year from the first supply date were outputs. Machine learning classification methods were used to build models to predict ACS and death. Model performance was measured by the area under the receiver operating characteristic curve (AUC-ROC), sensitivity and specificity. There were 68,889 patients in the NSAIDs cohort with mean age 76 years and 54% were female. 1882 patients were admitted for acute coronary syndrome and 5405 patients died within one year after their first supply of NSAIDs. The multi-layer neural network, gradient boosting machine and support vector machine were applied to build various classification models. The gradient boosting machine achieved the best performance with an average AUC-ROC of 0.72 predicting ACS and 0.84 predicting death. Machine learning models applied to linked administrative data can potentially improve adverse outcome risk prediction. Further investigation of additional data and approaches are required to improve the performance for adverse outcome risk prediction.

Cardioprotective medication adherence in Western Australians in the first year after myocardial infarction: restricted cubic spline analysis of adherence-outcome relationships

Adherence to cardioprotective medications following myocardial infarction (MI) is commonly assessed using a binary threshold of 80%. We investigated the relationship between medication adherence as a continuous measure and outcomes in MI survivors using restricted cubic splines (RCS). We identified all patients aged ≥65 years hospitalised for MI from 2003–2008 who survived one-year post-discharge (n = 5938). Adherence to statins, beta-blockers, renin angiotensin system inhibitors (RASI) and clopidogrel was calculated using proportion of days covered to one-year post-discharge (landmark date). Outcomes were 1-year all-cause death and major adverse cardiac events (MACE) after the landmark date. Adherence-outcome associations were estimated from RCS Cox regression models. RCS analyses indicated decreasing risk for both outcomes above 60% adherence for statins, RASI and clopidogrel, with each 10% increase in adherence associated with a 13.9%, 12.1% and 18.0% decrease respectively in adjusted risk of all-cause death (all p < 0.02). Similar results were observed for MACE (all p < 0.03). Beta-blockers had no effect on outcomes at any level of adherence. In MI survivors, increasing adherence to statins, RASI, and clopidogrel, but not beta blockers, is associated with a decreasing risk of death/MACE with no adherence threshold beyond 60%. Medication adherence should be considered as a continuous measure in outcomes analyses.

Warfarin Use and Mortality, Stroke, and Bleeding Outcomes in a Cohort of Elderly Patients with non-Valvular Atrial Fibrillation

Aims

To determine exposure to warfarin and the associated outcomes in a population of older patients with non-valvular atrial fibrillation (NVAF).

Methods

Cohort study of patients aged 65-89 years admitted to hospital July 2003-December 2008 with newly-diagnosed or pre-existing AF. Outcomes at three years among one-year survivors post-index admission (landmark date) were all-cause mortality, stroke/systemic thromboembolism (stroke/TE) and bleeding. Multivariate Cox models were used to identify factors associated with each outcome.

Results

AF was the principal diagnosis for 27.5% of 17,336 index AF admissions. Of 14,634 (84.4%) patients alive at one-year 1,384 (9.5%) died in the following year. Vascular disease (42%) was the most frequent cause of death.Warfarin use, prior to the index admission and/or the 1-year landmark, did not exceed 40%.Compared to non-exposure or discontinuation at the index admission, initiation or persistence with warfarin prior to the landmark date was associated with reduced risk for all-cause mortality, a statistically non-significant reduction in risk for stroke/TE, and an increased risk for bleeding. Higher CHA2DS2-VASc scores were associated with increased risk for each outcome.

Conclusions

In a population-based cohort of hospitalised NVAF patients, the initiation and persistent use of warfarin was associated with lower all-cause mortality risk to three years, although reduction in risk for stroke/TE did not reach statistical significance. The apparent under-use of warfarin in this older, high-risk cohort reinforces the opportunity for further reduction in stroke/TE with the uptake of non-vitamin K oral anti-coagulants (NOACs) among those not prescribed, or not persistent with, warfarin.

Feature selection and transformation by machine learning reduce variable numbers and improve prediction for heart failure readmission or death

Background

The prediction of readmission or death after a hospital discharge for heart failure (HF) remains a major challenge. Modern healthcare systems, electronic health records, and machine learning (ML) techniques allow us to mine data to select the most significant variables (allowing for reduction in the number of variables) without compromising the performance of models used for prediction of readmission and death. Moreover, ML methods based on transformation of variables may potentially further improve the performance.

Objective

To use ML techniques to determine the most relevant and also transform variables for the prediction of 30-day readmission or death in HF patients.

Methods

We identified all Western Australian patients aged 65 years and above admitted for HF between 2003–2008 in linked administrative data. We evaluated variables associated with HF readmission or death using standard statistical and ML based selection techniques. We also tested the new variables produced by transformation of the original variables. We developed multi-layer perceptron prediction models and compared their predictive performance using metrics such as Area Under the receiver operating characteristic Curve (AUC), sensitivity and specificity.

Results

Following hospital discharge, the proportion of 30-day readmissions or death was 23.7% in our cohort of 10,757 HF patients. The prediction model developed by us using a smaller set of variables (n = 8) had comparable performance (AUC 0.62) to the traditional model (n = 47, AUC 0.62). Transformation of the original 47 variables further improved (p<0.001) the performance of the predictive model (AUC 0.66).

Conclusions

A small set of variables selected using ML matched the performance of the model that used the full set of 47 variables for predicting 30-day readmission or death in HF patients. Model performance can be further significantly improved by transforming the original variables using ML methods.

Machine learning-based prediction of heart failure readmission or death: implications of choosing the right model and the right metrics

AIMS

Machine learning (ML) is widely believed to be able to learn complex hidden interactions from the data and has the potential in predicting events such as heart failure (HF) readmission and death. Recent studies have revealed conflicting results likely due to failure to take into account the class imbalance problem commonly seen with medical data. We developed a new ML approach to predict 30 day HF readmission or death and compared the performance of this model with other commonly used prediction models.

METHODS AND RESULTS

We identified all Western Australian patients aged above 65 years admitted for HF between 2003 and 2008 in the linked Hospital Morbidity Data Collection. Taking into consideration the class imbalance problem, we developed a multi-layer perceptron (MLP)-based approach to predict 30 day HF readmission or death and compared the predictive performances using the performance metrics, that is, area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (AUPRC), sensitivity and specificity with other ML and regression models. Out of the 10 757 patients with HF, 23.6% were readmitted or died within 30 days of hospital discharge. We observed an AUC of 0.55, 0.53, 0.58, and 0.54 while an AUPRC of 0.39, 0.38, 0.46, and 0.38 for weighted random forest, weighted decision trees, logistic regression, and weighted support vector machines models, respectively. The MLP-based approach produced the highest AUC (0.62) and AUPRC (0.46) with 48% sensitivity and 70% specificity.

CONCLUSIONS

We show that for the medical data with class imbalance, the proposed MLP-based approach is superior to other ML and regression techniques for the prediction of 30 day HF readmission or death.

Operationalization and validation of a novel method to calculate adherence to polypharmacy with refill data from the Australian pharmaceutical benefits scheme (PBS) database

Background

Electronic health care data contain rich information on medicine use from which adherence can be estimated. Various measures developed with medication claims data called for transparency of the equations used, predominantly because they may overestimate adherence, and even more when used with multiple medications. We aimed to operationalize a novel calculation of adherence with polypharmacy, the daily polypharmacy possession ratio (DPPR), and validate it against the common measure of adherence, the medication possession ratio (MPR) and a modified version (MPR_m).

Methods

We used linked health data from the Australian Pharmaceutical Benefits Scheme and Western Australian hospital morbidity dataset and mortality register. We identified a strict study cohort from 16,185 patients aged ≥65 years hospitalized for myocardial infarction in 2003–2008 in Western Australia as an illustrative example. We applied iterative exclusion criteria to standardize the dispensing histories according to previous literature. A SAS program was developed to calculate the adherence measures accounting for various drug parameters.

Results

The study cohort was 348 incident patients (mean age 74.6±6.8 years; 69% male) with an admission for myocardial infarction who had cardiovascular medications over a median of 727 days (range 74 to 3,798 days) prior to readmission. There were statins (96.8%), angiotensin converting enzyme inhibitors (88.8%), beta-blockers (85.6%), and angiotensin receptor blockers (13.2%) dispensed. As expected, observed adherence values were higher with mean MPR (median 89.2%; Q₁: 73.3%; Q_3: 104.6%) than mean MPR_m (median 82.8%; Q₁: 68.5%; Q_3: 95.9%). DPPR values were the most narrow (median 83.8%; Q₁: 70.9%; Q_3: 96.4%). Mean MPR and DPPR yielded very close possession values for 37.9% of the patients. Values were similar in patients with longer observation windows. When the traditional threshold of 80% was applied to mean MPR and DPPR values to signify the threshold for good adherence, 11.6% of patients were classified as good adherers with the mean MPR relative to the DPPR.

Conclusion

In the absence of transparent and standardized equations to calculate adherence to polypharmacy from refill databases, the novel DPPR algorithm represents a valid and robust method to estimate medication possession for multi-medication regimens.

Secondary preventive medication use in a prevalent population-based cohort of acute coronary syndrome survivors

AIM

Describe the dispensing patterns for guideline-recommended medications during 2008 in people with acute coronary syndrome (ACS) and how dispensing varies by gender and time since last ACS hospitalization.

METHOD

A descriptive cohort spanning 20 years of people alive post-ACS in 2008. We extracted all ACS hospitalizations and deaths in Western Australia (1989-2008), and all person-linked Pharmaceutical Benefits Scheme claims nationally for 2008. Participants were 23 642 men and women (36.8%), alive and aged 65-89 years in mid-2008 who were hospitalized for ACS between 1989 and 2008. Main outcome was the proportion of the study cohort (in 2008) dispensed guideline-recommended cardiovascular medications in that year. Adjusted odds ratios estimating the association between type (and number) of guideline-recommended medications and time since last ACS hospitalization.

RESULTS

Medications most commonly dispensed in 2008 were statins (79.6% of study cohort) and then angiotensin-converting enzyme inhibitors or angiotensin-receptor blockers (ACEi/ARBs) (71.1%), aspirin or clopidogrel (59.4%), and β-blockers (54.6%). Only 51.8% of the cohort was dispensed three or more of these drug types in 2008. Women with ACS were 18% less likely to be dispensed statins (adjusted odds ratio (OR)=0.82; 95% CI 0.76-0.88). Overall, for each incremental year since last ACS admission, there was an 8% increased odds (adjusted OR=1.08; 95% CI 1.07-1.08) of being dispensed fewer of the recommended drug regimen in 2008.

CONCLUSION

Longer time since last ACS admission was associated with dispensing fewer medications types and combinations in 2008. Interventions are warranted to improve dispensing long term and any apparent gender inequality in the drug class filled.

Long-term use of secondary prevention medications for heart failure in Western Australia: a protocol for a population-based cohort study

INTRODUCTION

Heart failure (HF) is a chronic, debilitating and progressive disease associated with high morbidity and mortality. Evidence-based medications (EBMs) are the cornerstone of management of patients with HF. In Australia, these EBMs are subsidised by the Commonwealth Government under the Pharmaceutical Benefits Scheme. Suboptimal dispensing and non-adherence to these EBMs have been observed in patients with HF. Our study will investigate trends in dispensing patterns, as well as adherence and persistence of EBMs for HF. We will also identify factors influencing these patterns and their impact on long-term clinical outcomes.

METHODS AND ANALYSIS

This whole population-based cohort study will use longitudinal data for people aged 65-84 years who were hospitalised for HF in Western Australia between 2003 and 2008. Linked state-wide and national data will provide patient-level information on medication dispensing, medical visits, hospitalisations and death. Drug dispensing trends will be described, drug adherence and persistence estimated and the association with all-cause/cardiovascular death and hospitalisations reported.

ETHICS AND DISSEMINATION

This project has received approvals from the Western Australian Department of Health Human Research Ethics Committee and the Western Australian Aboriginal Health Ethics Committee. Results will be published in relevant cardiology journals and presented at national and international conferences.