An observational study of clinical and health system factors associated with catheter ablation and early ablation treatment for atrial fibrillation in Australia

Data sources

This is a retrospective cohort study using linked administrative inpatient and mortality data from NSW, Australia. Data on AF hospitalizations were extracted from the NSW Admitted Patient Data Collection (APDC), which includes records of all inpatient separations (discharges, transfers, and deaths) from public and private hospitals in NSW. The APDC records up to 50 procedures, coded using the Australian Classification of Health Interventions (ACHI) [20] and up to 51 diagnoses, coded using the International Classification of Diseases, 10th Revision, Australian Modification (ICD-10-AM) Death records were extracted from the NSW Registry of Births, Deaths, and Marriages (RBDM) death registration file. Data linkage was performed by the NSW Ministry of Health Centre for Health Record Linkage, which reports rates of false positive and false negative links of 0.5% [21]. This study was granted ethical approval by the University of New South Wales, NSW Population and Health Services Research (HREC/18/CIPHS/56), Aboriginal Health and Medical Research Council of NSW (1503/19), and Australian Institute of Health and Welfare (EO2018/2/431) research ethics committees.

Study cohort

The cohort included patients with a primary diagnosis of AF or atrial flutter (ICD-10-AM codes: I48, I48.0, I48.1, I48.2, I48.91) between January 2009 and October 2017, with follow-up through October 2018. The index admission was identified as the first admission with a primary diagnosis of AF between January 2009 and October 2017, with no prior admission with an AF diagnosis in any diagnosis field or prior cardioversion in a three-year look-back period. Patients under 18 years at the time of the index admission were excluded. If a single hospital stay was made up of various episodes of care, we aggregated all diagnosis and procedure codes from these episodes, and for all other fields, we kept the information from the first of these related episodes.

The primary aim of this study was to explore factors associated with receiving CA as an interventional procedure following a diagnosis of AF. As we only have access to hospital administrative data, we excluded patients whose index AF admission included an ablation procedure, under the assumption that these were patients who had been previously diagnosed in outpatient care. We excluded patients who had a diagnosis of valvular heart disease (ICD10 code I05 - rheumatic mitral valve diseases, Q23 - congenital malformations of aortic and mitral valves) or mitral valve stenosis (I34 nonrheumatic mitral valve disorders), or a replacement of mitral valve procedure (coded as ACHI codes 38488-09, 38488-02, 38488-03, 38489-02) during the look-back period or the index admission.

Outcomes and Covariates

The primary outcome was CA for AF (ACHI codes 38290-01, 38287-02). We defined early ablation as CA administered within one year of the index AF admission. Medical history and comorbidities were determined by examining diagnosis codes (primary and otherwise) in the index AF admission and prior admissions in the three-year look-back period. Patient age and sex were obtained from the APDC record for the index AF admission. Rather than using all comorbidities from all available diagnosis codes in the cohort, we restricted selection to relevant comorbidities based on the expertise of two cardiologists (DB and RS). Prior diagnosis codes were used to calculate the AF stroke risk score (CHA₂DS₂-VASc [22]) when describing the characteristics of the cohort. Cardioversion was included as a covariate as it is a common procedure for AF.

Health system factors were included to assess associations between sociodemographic and health resource utilization factors with CA: year of index admission, patient payment status, and socio-economic status (SES) of their area of residence. Australia has a mixed public and private health system, and patients with private health insurance have access to shorter waiting times and greater choices in treating practitioners. Patient payment status was measured on the index admission and was categorized as public or private patients. Patients with a status of workers compensation, veteran affairs, and defense force were included with private patients.

We used patient’s Statistical Local Area of residence (Australian Statistical Geography Classification 2011 Boundaries) to measure SES using publicly available socio-economic indices of deprivation (the Index of Relative Socio-economic Advantage and Disadvantage) [23]. This index was stratified into quintiles (with quintile 1 indicating lowest relative SES and quintile 5 indicating highest SES).

Statistical analyses

We used Cox proportional hazard models for determining the risk factors associated with receiving catheter ablation (dependent variable). The time-to-event of catheter ablation was determined by calculating the number of months from the index admission until the earliest event among 1) the first occurrence of ablation, 2) the end of the time-at-risk window (31 October 2018), and 3) death. Patients with an index admission after 31 October 2017 were excluded to ensure all patients were followed for at least one year. Among patients who received an ablation, the risk factors associated with receiving early ablation (within the first year of index AF admission) vs late ablation was determined using logistic regression. The same covariates and adjustment strategy were used for both analyses. The base model included age and sex. The first adjusted model included clinical factors: cardioversion, congestive heart failure, hypertension, diabetes, stroke, myocardial infarction (MI), coronary artery disease, and major bleeding events. The second adjusted model included health system factors: private patient status (public patient status as reference), socio-economic index of advantage and disadvantage (quintile modeled as an ordinal variable), and year of index admission. The third adjusted model included both clinical and health system factors. The models’ goodness-of-fit was evaluated using the concordance index (for the Cox models) and the area under the receiver operating characteristic curve (AUC, for the logistic regression models) under 10-fold cross-validation. Univariate survival analysis using Kaplan-Meier was estimated using the entire dataset to calculate the probability of receiving ablation vs no ablation. For readability purposes (CA is a positive event), we use the wording “likelihood” when describing the hazard ratio results. We present odds ratios for the logistic regression results.

Baseline characteristics

Table 1 shows the baseline characteristics of the entire cohort at the first AF admission (N=46,764). The ratio of male to female patients with a diagnosis of AF was almost equal (51.8% males), but the ratio of males to females receiving ablation was 2:1 (64.1% males). The patients who received ablation were younger (median age 61) compared to those who did not receive ablation (median age 74) and were more likely to have been private patients in their index admission (69.2% private patients, compared to 45% in those who did not receive ablation). Those who received ablation had lower CHA₂DS₂-VASc risk scores, lower comorbidities in general, and more likely to have had a cardioversion. Patients from the two upper-income quartiles received more ablations.

Amongst patients that received ablation (see Appendix for baseline characteristics stratified by gender), females were older (median age 64) compared to men (median age 61), had higher CHA₂DS₂-VASc risk scores (mean score 2.0 vs 0.9), had lower rates of comorbidities (except diabetes, 21.9% of females vs 14.9% of males), were less likely to have had a cardioversion (29.6% vs 40.9%) or to have been private patients in their index admission (63.7% vs 72.3%). Men that received ablation had higher rates in the highest income quartile, whereas females had higher rates in the four lower-income quantiles.

Patients who received early ablation were slightly older (median age 63 vs median age 60), had higher rates of cardioversion and diabetes, and were more likely to have been private patients in their index admission, than patients who received a late ablation.

Characteristics Associated with Ablation

Figure 1 shows the Kaplan Meier survival curve representing the cumulative probability of receiving ablation after index admission, with about 5% of patients receiving CA and a quarter of them occurring within the first year. The hazard ratios (and corresponding p-values) from the time-to-ablation models are shown in Table 2.

• Download figure
• Open in new tab

Figure 1

Kaplan Meier survival curve showing the cumulative probability of receiving catheter ablation after index AF admission, with about 5% of patients receiving catheter ablation, a quarter of ablations occurring within the first year and half of the ablations occurring within the first three years.

Older age was associated with a lower likelihood of receiving CA (likelihood decreasing for each additional year of age), even after adjusting for clinical and health system factors. Female sex was associated with a lower likelihood of receiving CA in the base model, but not after adjusting for clinical and health system factors. Cardioversion during index admission was associated with a higher likelihood of receiving CA, even after adjusting for health system factors. A history of congestive heart failure, hypertension, diabetes, and MI were associated with a lower likelihood of receiving CA, even after adjusting for health system factors. A history of stroke, coronary artery disease, and major bleeding events were not associated with a lower or higher likelihood of receiving CA. All health system factors were associated with a higher likelihood of receiving CA even after adjustment for clinical factors, including non-public payment (private patient), as well as increasing SES of the area of residence. For the goodness-of-fit, the model with health system covariates had a higher concordance index compared to the model with only clinical factors (Table 2).

Characteristics Associated with Early Ablation

Table 3 displays the odds ratios associated with receiving early ablation. Cardioversion during index admission and history of diabetes were associated with a higher likelihood of early ablation, even after adjusting for health system covariates. Private patient status was associated with a higher likelihood of early ablation even after adjusting for clinical covariates. The model with the health system factors had a higher AUC than the model with clinical factors.

Main findings

Age was associated with a lower likelihood of receiving CA (likelihood decreasing for each additional year). Patients with a cardioversion during index admission were more likely to receive CA, consistent with prior work showing an association between the risk of ablation and the number of performed cardioversions [19]. This likely reflects a desire for a “rhythm control” strategy in these patients, based on clinical factors. Women in our cohort, regardless of receipt of ablation, received fewer cardioversions than men. Patients with a medical history of congestive heart failure, hypertension, and diabetes being less likely to receive CA is consistent with other studies [10], [17] and suggests that ablation is performed on relatively healthier patients (fewer comorbidities), who are at an earlier stage in the disease process and more likely to have a successful procedure. Health system factors associated with an increased likelihood of receiving CA included the year of index AF admission (risk increasing with year), private patients, and patients from areas of higher socioeconomic status. Patients with a cardioversion during index admission, a history of diabetes, and who were private patients (at their index admission) had higher odds of receiving early ablation.

Of all model adjustments for determining factors associated with CA and early ablation, the models with health system covariates best explained the data. This is alarming, as it shows health system factors being stronger predictors than clinical factors.

Impact of Health System Factors on Receiving Ablation

Our findings suggest the increased use of ablation as a procedural intervention for AF patients over time, consistent with studies in Australia [13] and the US [17]. The strong association of private patient status with ablation suggests that patients who enter the health system for their care of AF through the private system (or even through other forms of subsidized care) have faster access to ablation. These findings most likely reflect that private patients in Australia have increased access to elective procedures through their health insurance, including shorter waiting times and choice of treating practitioner. Our results also highlight that patients living in more advantaged areas are more likely to receive ablation even after accounting for insurance status, suggesting there may also be inequities in access to the procedure – either through capacity for affording out-of-pocket costs, or in the geographic location of services. The association between private patient status and early ablation highlights that public patients may first undergo alternative or lower-cost care, with ablation performed at a later stage. Whether this delayed access is resulting in different outcomes for public and private patients requires further investigation.

A study with a cohort of under 65 years of age in the US found insurance type to be a predictor of ablation [17] and a study using data from Swedish health registries showed that university education and income in the highest quartile were factors associated with undergoing ablation [18]. More recently, a study comparing surgical utilization in NSW and other countries showed that residents of lower-income neighborhoods had lower rates of surgery compared to residents of higher-income neighborhoods [24]. Our work similarly highlights disparities in receiving CA associated with patient status and the neighborhood index of social advantage and disadvantage, raising concerns on the equitable delivery of CA in Australia. However, administrative data is unlikely to capture the complexity of the risk-benefit analysis in individual patients, and there may be further socio-demographic and health system factors that influence clinical recommendation and patient utilization of ablation procedures.

Limitations

This study used linked administrative records from hospitalizations in NSW, Australia. These do not capture information about the use of antiarrhythmic medications and anticoagulants, family history, medical history, social and lifestyle factors, and any diagnosis or procedure from outpatient care. As such, the date of index AF admission is a proxy for date of AF diagnosis. Our analysis relied solely on coded hospital diagnoses, which are recorded only for conditions that significantly affect patient management during an episode of care. As such, our use of diagnoses as a proxy for incident or prevalent disease, the inferring of date of onset of AF, and the quality of hospital diagnosis coding varying widely (accuracy 51-98% across 32 studies) [25] are potential sources of bias in our analysis. However, this source of bias is lessened by our use of a large dataset, hospitalizations coming from public and private hospitals across the state of NSW. Finally, our survival model did not take into consideration competing interventions (open ablation) or death as a competing risk, which may have produced a slight overestimation in the results.