Geographic Clustering of Emergency ... - Wiley Online Library

ORIGINAL CONTRIBUTION

Geographic Clustering of Emergency Department Presentations for Atrial Fibrillation and Flutter in Alberta, Canada Rhonda J. Rosychuk, PhD, Hensley H. Mariathas, PhD, Michelle M. Graham, MD, Brian R. Holroyd, MD, MBA, and Brian H. Rowe, MD, MSc

Abstract Objectives: Atrial fibrillation and flutter (AFF) are the most common arrhythmias seen in the outpatient setting, and they affect more than 300,000 adult Canadians. The aims of this study were to examine temporal and geographic trends in emergency department (ED) presentations made by adults (age ≥ 35 years) for AFF in Alberta, Canada, from 1999 to 2011. Statistical disease cluster detection techniques were used to identify geographic areas with higher numbers of individuals presenting with AFF and higher numbers of ED presentations for AFF than expected by chance alone. Geographic clusters of individuals with stroke or heart failure follow-up within 365 days of ED presentations for AFF were also identified. Methods: All ED presentations for AFF made by individuals aged ≥35 years were extracted from Alberta’s Ambulatory Care Classification System. The Alberta Health Care Insurance Plan provided population counts and demographics for the patients presenting (age, sex, year, geographic unit). The Physician Claims File provided non-ED physician claims data after a patient’s ED presentation. Statistical analyses included numerical and graphical summaries, directly standardized rates, and statistical disease cluster detection tests. Results: During 12 years, there were 63,395 ED presentations for AFF made by 32,101 individuals. Standardized rates remained relatively stable over time, at about two per 1,000 for individuals presenting to the ED for AFF and about three per 1,000 for ED presentations for AFF. The northern and southeastern parts of the province were identified as clusters of individuals presenting for AFF, and ED presentations for AFF, and several of the areas demonstrated clusters in multiple years. Further, several of the geographic clusters were also identified as potential clusters for stroke or heart failure within 365 days after the ED presentations for AFF. Conclusions: This population-based study spanned 12 fiscal years and showed variations in the number of people presenting to EDs for AFF and the number of ED presentations for AFF over geography. The potential clusters identified may represent geographic areas with higher disease severity or a lower availability of non-ED health services. The clusters are not all likely to have occurred by chance, and further investigation and intervention could occur to reduce ED presentations for AFF.

From the Department of Pediatrics (RJR), the Department of Medicine (MMG), the Department of Emergency Medicine (BRH, BHR), and the School of Public Health (BHR), University of Alberta, Edmonton, Alberta; Women & Children’s Health Research Institute (RJR), Edmonton, Alberta; the Department of Mathematics and Statistics, Memorial University of Newfoundland (HHM), St. John’s, Newfoundland, Canada; and Alberta Health Services (BRH, BHR), Edmonton, Alberta, Canada. Received December 11, 2015; revision received March 9, 2015; accepted March 21, 2015. Presented at the Canadian Association of Emergency Physicians (CAEP) Conference, Edmonton, Alberta, Canada, May 2015. The study was funded by an Innovation Grant from the Women & Children’s Health Research Institute (www.wchri.org, Edmonton, AB, Canada) and an operating grant from the Canadian Institutes of Health Research (CIHR, Ottawa, AB, Canada). Dr. Rosychuk is salary supported by Alberta Innovates-Health Solutions (AI-HS, Edmonton, AB, Canada; formerly the Alberta Heritage Foundation for Medical Research), as a Health Scholar. Dr. Rowe is supported by a Tier 1 Canada Research Chair in Evidencebased Emergency Medicine from CIHR through the Government of Canada (Ottawa, ON). This study is based in part on data provided by Alberta Health. The interpretation and conclusions are contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the government nor Alberta Health expresses any opinion in relation to this study. The authors have no potential conflicts to disclose. Supervising Editor: Rochelle Fu, PhD. Address for correspondence and reprints: Rhonda J. Rosychuk, PhD; e-mail: [email protected].

© 2015 by the Society for Academic Emergency Medicine doi: 10.1111/acem.12731

ISSN 1069-6563 PII ISSN 1069-6563583

965 965

966

Rosychuk et al. • GEOGRAPHIC CLUSTERING OF ED PRESENTATIONS FOR AFF ACADEMIC EMERGENCY MEDICINE 2015;22:965–975 © 2015 by the Society for Academic Emergency Medicine

trial fibrillation and flutter (AFF) are defined as cardiac tachyarrhythmias characterized by disorganized atrial electrical depolarization leading to an irregular and often a rapid ventricular rate. AFF are the most common arrhythmias seen in outpatient settings. In the emergency department (ED), physicians often manage patients with either recent-onset (first detected or paroxysmal) or permanent (chronic) AFF. AFF are the most common arrhythmias in clinical practice, accounting for approximately one-third of hospitalizations for cardiac rhythm disturbances. There has been a 66% increase in hospital admissions for AFF due to the aging of the population and a rising prevalence of chronic heart disease.1,2 AFF are a costly public health problem due to repeat visits, hospitalizations, drugs, consultations, investigations, lost time at work, and complications.3 For example, AFF are associated with an increased long-term risk of transient ischemic attack and stroke, heart failure, and all-cause mortality, especially in women.4 Very little is known about the epidemiology of AFF in the ED; however, much of what is known arises from data in the United States.5 U.S. and Canadian emergency physicians (EPs) exhibit different practices, as Canadians more often perform electrical cardioversion, less frequently refer to cardiologists, and discharge most patients with AFF from the hospital.6 These marked differences suggest that Canadian data may help provide important comparative results and illustrate the value of population-based administrative data in understanding complex chronic disease presentations to the ED. Further, geographic and temporal variations in the nature and frequency of patients may represent greater severity of illness, lesser availability of health care resources, or variation in health care delivery. The aims of this study were to examine AFF trends over time and geography in ED presentations made by adults (age ≥ 35 years) to EDs in the province of Alberta, Canada, over a 12-year period. We used statistical cluster detection techniques to identify geographic areas with higher numbers of individuals presenting to the ED with AFF, and higher numbers of ED presentations with AFF, than expected by chance alone. We also identified geographic areas with higher numbers of individuals with subsequent physician claims for stroke or heart failure in the 365 days following ED presentations for AFF.

A

METHODS Study Design This is a retrospective cohort study using populationbased administrative health databases in the province of Alberta, Canada, from April 1, 1999, to March 31, 2011. ED presentations for AFF were extracted for all Alberta residents aged 35 years or older. The human research ethics board at the University of Alberta approved this study.

Study Setting and Population The Province of Alberta is located in the western region of Canada. It covers 640,082 km2 and had a population of 3,645,257 in 2011.7 All Alberta residents have access to health care through a uniform, single-payer health system. This government-funded health plan maintains health databases as part of its activities. Emergency care is delivered in over 100 publicly funded hospitals, representing a very wide range of settings varying from those in large, urban cities to those servicing remote, isolated communities. Most urban hospitals are staffed by full-time EPs, and almost all rural hospitals have lower volumes and are staffed mainly by on-call primary care physicians. Three databases were used for the data extraction: the Ambulatory Care Classification System (ACCS)8 for ED presentations, the Alberta Health Care Insurance Plan (AHCIP) cumulative registry file for population counts and demographic data, and the Physician Claims File (PCF) for physician visits in non-ED settings after ED presentations (hereafter called follow-up visits). The ACCS database records ambulatory care visits to government-funded facilities, including 104 EDs, and contributes to the National Ambulatory Care Records System (NACRSs) maintained by the Canadian Institute of Health Information (CIHI).9 All ED encounters are entered into computerized abstracts that constitute the majority of records. Trained and supervised medical records nosologists, using a uniform protocol, code each chart using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM;10 before April 1, 2002), or the Canadian Enhancement of International Classification of Diseases, 10th Revision (ICD-10-CA;11 April 1, 2002, onward) diagnostic codes. Each ACCS record represents a unique service and contains a unique identifier for each Albertan. The AHCIP file includes all persons registered in the plan (99% of the provincial population). The PCF provides data on the date of each follow-up visit and has three diagnosis fields recording ICD-9-CM codes. Study Protocol The ACCS database has a main diagnosis field and either five (ICD-9-CM) or nine (ICD-10-CA) additional fields to capture diagnosis data. To be considered an AFF presentation, the first diagnosis field in the ACCS database had to have diagnostic codes in 427.3x (AFF) or I48.x (AFF) for the ICD-9 and ICD-10 periods, respectively. All ED presentations for AFF made by individuals aged 35 years or older during April 1, 1999, to the March 31, 2011, were extracted. A case was defined as an individual with at least one ED presentation for AFF during the study period. We call these individuals ED cases hereafter. In 2003, Alberta’s nine administrative health areas were divided into 70 subregional health authorities (sRHAs; Figure 1 and Data Supplement S1, available as

ACADEMIC EMERGENCY MEDICINE • August 2015, Vol. 22, No. 8 • www.aemj.org

Figure 1. Alberta’s 70 sRHAs and five zones (from top to bottom: North, Edmonton, Central, Calgary, and South).

967

supporting information in the online version of this paper)12 by the provincial health ministry, Alberta Health. These were created to provide smaller regions for the administration of programs and for surveillance. In particular, the sRHAs provide smaller population sizes that still enable valid rate calculations. The sRHAs have diverse population sizes ranging from 550 to 140,211, with a median of 46,075 in 2011. The geographic data provided by Alberta Health were geocoded to the 70 sRHAs (numbered 1 to 70 for the purposes of this study) based on the 2003 boundaries (using postal codes). The 70 sRHAs were further aggregated into five administrative health zones (North, Edmonton, Central, Calgary, and South) that were formed in 2009. Alberta Health provided latitudes and longitudes for each sRHA’s population-based centroid. Variables for the ED presentations included sex (female, male), age in years at ED presentation, and sRHA of residence at fiscal year-end. Population counts by age by sex, age in years at fiscal year-end, and sRHA of residence at fiscal year-end for the population 35 years or older were also provided. Five-year age groups were formed (35–39, 40–44, . . . 85–89, 90 + ). The follow-up visits were also extracted and classified as stroke or heart failure based on the first diagnosis field (stroke—434.x, 435.x, 436.x, 362.3x; heart failure—402.01, 402.11, 402.91, 428.x, 518.4). An individual with an ED presentation for AFF who had a at least one subsequent follow-up visit for stroke or heart failure within 365 days of an ED presentation was termed a case with stroke or a case with heart failure follow-up, respectively.

Table 1 ED Cases and Presentations by Demographic Variables ED Cases

ED Presentation per Fiscal Year

Demographic

Total (N = 32,101), n (%)

Sex Female 15,419 (48.0) Male 16,682 (52.0) Age group (yr) 35–39 703 (2.2) 40–44 911 (2.8) 45–49 1,316 (4.1) 50–54 1,926 (6.0) 55–59 2,423 (7.5) 2,941 (9.2) 60–64 65–69 3,562 (11.1) 70–74 4,453 (14.0) 75–79 4,927 (15.3) 80–84 4,552 (14.2) 85–89 2,932 (9.1) 90+ 1,455 (4.5) Zone North 4,099 (12.8) Edmonton 9,394 (29.3) Central 5,444 (17.0) Calgary 9,597 (29.9) South 3,567 (11.1)

Median (n = 3,816) 1,841 1,960 69.5 107.0 157.0 224.5 316.0 316.5 414.5 520.0 585.0 547.5 322.0 163.0 460.5 1,133.0 668.0 1,083.0 430.5

per Fiscal Year

Range (n = 3,090–4,292)

Total (N = 63,395), n (%)

1,549–2,009 1,541–2,283

30,547 (48.2) 32,848 (51.8)

46–92 80–129 108–187 157–315 201–381 246–454 389–488 472–596 531–614 425–574 241–422 109–220 383–474 966–1,236 496–720 833–1,261 362–474

Median (n = 5,352) 2,550 2,768

1,214 1,760 2,731 4,185 5,525 6,331 7,573 8,982 9,545 8,361 4,939 2,249

(1.9) (2.8) (4.3) (6.6) (8.7) (10.0) (12.0) (14.0) (15.0) (13.0) (7.8) (3.5)

98.5 148.5 229.0 345.5 471.5 543.5 633.5 734.0 795.0 706.5 413.5 200.0

8,588 18,710 11,384 17,598 7,115

(13.5) (29.5) (18.0) (27.8) (11.2)

668.0 1,579.0 964.0 1,466.0 599.0

Range (n = 4,262–5,953) 2,110–2,830 2,152–3,123

2010/2011 Population ≥35 yr (Total N = 1,953,830), n (%) 983,570 (50.3) 970,260 (49.7)

59–138 103–171 149–276 218–465 318–549 378–656 545–745 642–874 728–868 557–778 297–521 127–257

273,714 269,464 290,847 285,754 238,153 185,082 127,345 95,728 76,809 57,668 34,242 19,024

(14.0) (13.8) (14.9) (14.6) (12.2) (9.5) (6.5) (4.9) (3.9) (3.0) (1.8) (1.0)

557–963 1,321–1,682 706–1,028 1,130–1,756 520–669

212,656 619,455 238,327 735,319 148,073

(10.9) (31.7) (12.2) (37.6) (7.6)

968

Rosychuk et al. • GEOGRAPHIC CLUSTERING OF ED PRESENTATIONS FOR AFF

Figure 2. Atrial fibrillation and flutter ED cases (●), presentations (+), and adult population (▲), by fiscal year.

Data Analysis Numerical summaries (e.g., frequency, percentage, median, range) describe the ED case and presentation counts. The summaries were provided by age group, sex, zone, and fiscal year. Crude rates and sex and age group directly standardized rates (DSRs) are provided for the ED cases and presentations. For ED presentations, the directly standardized visit rates (DSVRs)13 are calculated because cases may have multiple (correlated) presentations. To enable standardized rate calculations to be compared over time (and changing population distributions), the Alberta population 35 years or older as

of March 31, 2000, formed the reference population. DSRs are provided with 95% confidence intervals (CIs). Data were analyzed using S-Plus statistical software (TIBCO Spotfire S+ Version 8.1.1 for Linux). To identify geographic areas of excess numbers of ED cases and presentations, we used two statistical cluster detection methods that adjust for the underlying population counts. Both methods test geographic areas separately with the aim of combining the l nearest neighboring areas that meet a prespecified cluster size (number of ED cases or analogously, number of ED presentations) labeled k and perform subsequent testing.

Figure 3. Sex and age group directly standardized rates (DSRs) per 1,000 population of age 35 + by fiscal year (●) and zones North (*), Edmonton (9), Central (▲), Calgary (+), and South (&).


969

Figure 4. Sex and age group directly standardized visits rates (DSVRs) per 1,000 population of age 35 + by fiscal year (●) and zones North (*), Edmonton (9), Central (▲), Calgary (+), and South (&).

The sRHAs were the geographic unit of analyses. Nearest neighbors were determined by increasing order of pairwise distances between sRHA population-based A

centroids. The Besag and Newell (BN)14 method was used to detect clusters of ED cases, with up to two nearest neighbors in the testing algorithm.15 To examine the B

Figure 5. Atrial fibrillation and flutter clusters identified for (A) cases and (B) presentations for fiscal year 1999/2000: subregional health authorities that are potential clusters alone (dark shading) and when combined with one (medium shading) or two (light shading) nearest neighbors are shown.

970


A

B


clustering of ED presentations, the Rosychuk, Huston, and Prasad (RHP)16 method was used with the analogous testing algorithm (a C/C++ program that performs cluster detection analyses; developed by author RJR). Monte Carlo simulations were used to assess the overall number of clusters identified and provide an overall p-value. More details on the cluster detection tests can be found elsewhere,17 and we illustrate the method with a hypothetical example available in Data Supplement S2 (available as supporting information in the online version of this paper). The analyses were conducted separately, for each year and for ED cases and presentations, using specialized software (developed by author RJR). These analyses were adjusted by sex and age group (based on the year of data analyzed). In addition, the BN method was used to detect clusters of cases of stroke or heart failure with follow-up within 365 days, separately. Data with missing sRHAs were not used in analyses. A p-value less than 0.05 was considered to be statistically significant. RESULTS General Trends During the study period, 63,398 ED presentations for AFF made by 32,104 people (ED cases) were extracted. Of these, only three ED presentations did not have sRHA data recorded. The analyzed data set had 63,395 ED presentations (32,101 ED cases), with a median of

one ED presentation per ED case (range = 1 to 53). The majority of ED cases (n = 20,393, 63.5%) visited the ED only once for AFF, and fewer than 1.3% of ED cases (n = 418) presented more than 10 times. Men outnumbered women for both ED cases and presentations (Table 1). The 70- to 84-year-old age group had the highest numbers of ED cases and presentations. The areas with the largest populations, Calgary and Edmonton zones, had the vast majority of ED cases and presentations (Calgary, 9,597 ED cases; Edmonton, 9,394 ED cases). Geographic and Temporal Trends The yearly number of ED presentations increased over time, from 4,262 in 1999/2000 to 5,953 in 2010/2011 (Data Supplement S3, available as supporting information in the online version of this paper; Figure 2). The Alberta population of adults aged 35 or older grew from 1,457,501 in 1999/2000 to 1,953,830 in 2010/2011. The crude ED presentation rates were relatively stable, varying from 2.9 in 1999/2000 to 3.0 in 2010/2011, with a peak of 3.4 in 2005/2006. When adjusted for sex and age group, the DSRs and DSVRs showed very little variability and were approximately 2.0 and 3.0 per 1,000, respectively (Data Supplement S4, Table a, available as supporting information in the online version of this paper). When zones were examined over time (Figures 3 and 4), more variability was seen over time and geography. The North zone had the highest rates with


A

971

B


DSRs ranging from 2.6 ED cases per 1,000 (95% CI = 2.4 to 2.9) in 1999/2000 to 3.1 per 1,000 (95% CI = 2.9 to 3.4) and DSVRs, ranging from 3.8 ED presentations per 1,000 (95% CI = 3.3 to 4.3, Data Supplement S4, Table b) in 1999/2000 to 4.6 per 1,000 (95% CI = 4.1 to 5.0). The Edmonton and Calgary zones had the lowest and most stable DSRs and DSVRs. Geographical Clustering ED Cases and Presentations for AFF. The cluster detection results adjusted by sex and age group are presented for three fiscal years: 1999/2000 (Data Supplement S5, available as supporting information in the online version of this paper; Figure 5), 2005/2006 (Data Supplement S6, available as supporting information in the online version of this paper; Figure 6), and 2010/ 2011 (Data Supplement S7, available as supporting information in the online version of this paper; Figure 7). The tables show the details on the cluster size k and number of observed and expected ED cases and presentations for the sRHAs and their l nearest neighbors. The number of significant clusters increased over time for both the analyses based on ED cases and the analyses based on ED presentations. In most of the yearly analyses, the statistically significant clusters were identified in the North and South zones. For the entire study period, the cluster tests for each year generally identified the same clusters, although there was some

variability from year to year (Data Supplement S8, available as supporting information in the online version of this paper). For 1999/2000, 17 sRHAs were identified as clusters of ED cases (Data Supplement S5, Figure 5A), with 10 of the sRHAs significant as clusters alone (i.e., without any nearest neighbors combined). For example, sRHA 2 required 0 nearest neighbors to be combined with it (l = 0) to have at least k = 36 ED cases observed. In fact, 36 ED cases were observed where only 26.5 were expected. Based on the test, sRHA 2 had a higher number of ED cases than would be expected by chance (p = 0.044). This sRHA was also identified as a cluster of ED presentations when combined with its two nearest neighbors (l = 2) to have at least k = 248 ED presentations observed. The combined sRHAs actually had 250 ED presentations when 211.6 were expected and when tested, the combined sRHAs had more ED presentation than would be expected by chance (p = 0.049). Twelve of the 17 clusters of ED cases were also identified as clusters of ED presentations. This analysis also identified another four sRHAs that were clusters of ED presentations but not clusters of ED cases. Further, five sRHAs that were clusters of ED cases were not identified as clusters of ED presentations. These results demonstrate that higher numbers of ED presentations may not appear with higher numbers of ED cases and vice versa. Overall, the number of identified clusters of ED

972


Table 2 Clustering Results for Cases With Stroke Follow-up for Each Fiscal Year With Sex and Age Group as Strata 1999/2000

2005/2006

2010/2011

ID

k

l

Obs/Exp

p-value

k

l

Obs/Exp

p-value

k

l

Obs/Exp

p-value

6 7 13 16 17 18 19 20 21 22 26 32 33 36 39 40 61 62 63 64 65 66 67

10 11 10 14 12 18 18 16 13 13 10 14 12 6 12 9 10 7 6 7 8 9 6

0 1 18 19 19 20 20 21 20 21 19 2 1 0 2 4 2 0 5 4 3 3 6

2.1 1.9 0.2 0.4 0.2 0.4 0.4 0.4 0.3 0.3 0.3 1.9 1.9 2.6 1.9 1.3 2.5 2.5 1.7 2.0 1.4 1.8 1.6

0.04* 0.05* 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.03*† 0.03*† 0.03*† 0.03* 0.35 0.03*† 0.03*† 0.43† 0.17† 0.31† 0.10 0.46†

17 18 7 19 17 24 24 24 22 22 19 20 24 12 6 11 14 16 6 7 10 8 6

5 4 0 4 3 5 5 5 4 4 3 5 7 2 0 1 3 3 0 1 1 0 2

0.9 1.3 2.2 0.9 1.2 0.9 0.9 1.0 1.0 1.0 1.0 0.9 0.8 1.9 2.5 2.0 1.5 1.4 2.8 2.4 2.7 2.6 2.3

0.91† 0.28† 0.05* 0.86 0.37 0.93† 0.81† 0.82 0.64 0.71 0.50 0.78† 0.99† 0.04*† 0.03*† 0.02*† 0.31† 0.10† 0.02*† 0.05*† 0.04*† 0.04† 0.05*†

15 19 13 20 8 16 9 13 21 10 16 22 28 13 15 12 15 17 6 10 11 11 7

0 4 1 2 0 1 0 0 1 0 1 5 5 5 9 5 3 8 0 2 1 1 3

1.8 1.2 1.7 1.7 2.1 1.7 2.2 1.9 1.6 2.0 1.9 0.9 0.9 0.8 0.7 0.8 1.4 0.6 2.6 2.5 2.1 2.1 1.8

0.03*† 0.28† 0.05* 0.03* 0.04* 0.05* 0.04* 0.04* 0.03* 0.03* 0.04*† 0.86† 0.87† 0.89 0.99 0.91 0.35† 1.00† 0.03*† 0.03*† 0.04*† 0.04*† 0.11†

Displayed are the significant sRHA (ID) in any of the three fiscal years, the cluster size (k), test statistic (l), the observed-toexpected ratio (Obs/Exp). sRHA = subregional health authority. *Significance at a = 0.05. †Significance from the ED case for AFF analysis.

cases (overall p-value = 0.001) and the number of identified clusters of ED presentations (overall p-value < 0.001) were not likely to have occurred purely by chance. For 2005/2006, 29 clusters of ED cases and 24 clusters of ED presentations were identified (Data Supplement S6, Figure 6). There were six sRHAs that were identified as clusters of ED cases but not as clusters of ED presentations. One sRHA (ID = 41) was not identified as a cluster of ED cases but had enough ED presentations (135 observed, 102.3 expected) to be identified as a cluster of ED presentations. Several sRHAs were identified as clusters without requiring combination with nearest neighbors in both 1999/2000 and 2005/2006. Overall, the number of identified clusters of ED cases (overall p-value < 0.001) and the number of identified clusters of ED presentations (overall p-value < 0.001) were not likely to have occurred purely by chance. For 2010/2011, 30 sRHAs were identified as potential clusters (Data Supplement S7, Figure 7): 28 as clusters of ED cases and 28 as clusters of ED presentations. Discrepant results between ED cases and ED presentations only occur for four sRHAs (25, 26, 31, and 38). For sRHAs 25 and 26, the tests identify them as clusters of ED cases but not as clusters of ED presentations. For sRHAs 31 and 38, they do not have enough ED cases to be considered clusters of ED cases but do have enough ED presentations (43 observed, 26.2 expected for sRHA 38) to be considered clusters of ED presentations. Several sRHAs had observed to expected ratios of ED cases

above 2.0. Most of these sRHAs also had observed to expected ratios of ED presentations above 2.0 except for sRHA 30 (2.13 for ED presentations, 1.40 for ED cases) and sRHA 34 (2.13 for ED presentations, 1.78 for ED cases). The highest observed to expected ratio for ED presentations was seen in sRHA 68, with 21 ED presentations observed (k = 14) and 6.5 expected (ratio = 3.24). The results suggest that most of the northern and southeastern parts of the province have more ED cases and/or ED presentations than expected, and the results also suggest that some sRHAs have higher numbers consistently over time. Overall, the number of identified clusters of ED cases (overall pvalue < 0.001) and the number of identified clusters of ED presentations (overall p-value < 0.001) were not likely to have occurred purely by chance. Cases With Stroke or Heart Failure Follow-up. The cluster results of cases with stroke follow-up within 365 days appear in Table 2 for three fiscal years. In 1999/2000, eight sRHAs were identified as clusters of stroke follow-up and five of these sRHAs were identified as clusters for ED cases of AFF (sRHAs 32, 33, 36, 61, 62). In 2005/2006, seven of the nine clusters of stroke follow-up were also clusters of ED cases of AFF. Of the 14 clusters of ED cases of AFF identified in 2010/2011, six were clusters of stroke follow-up. More clusters of cases with heart failure follow-up were identified than for the stroke outcome (Table 3). There were 32, 23, and 17 clusters identified in 1999/


973

Table 3 Clustering Results for Cases With Heart Failure Follow-up for Each Fiscal Year With Sex and Age Group as Strata 1999/2000

2005/2006

2010/2011

ID

k

l

Obs/Exp

p-value

k

l

Obs/Exp

p-value

k

l

Obs/Exp

p-value

1 2 3 4 5 6 7 17 18 25 27 28 29 30 31 32 33 34 35 36 37 38 39 40 50 53 54 55 56 61 62 63 64 65 66 67 68

18 11 32 40 44 32 37 41 62 10 18 22 13 21 13 32 10 37 26 16 30 9 36 26 18 16 9 25 19 31 20 12 7 8 27 13 8

1 0 0 1 2 0 1 18 20 0 2 1 0 1 0 0 0 3 2 0 2 0 2 4 0 1 0 2 0 2 0 0 0 0 3 2 1

1.8 2.0 1.6 1.6 1.4 1.4 1.5 0.2 0.4 2.2 1.7 1.9 1.7 1.6 1.8 1.6 2.2 1.3 1.5 2.4 1.8 2.7 1.4 1.3 1.6 1.7 2.1 1.7 2.0 2.0 2.6 3.5 3.4 2.3 1.7 3.2 2.9

0.04* 0.05*† 0.05*† 0.04*† 0.04*† 0.04* 0.04* 1.00 1.00 0.05* 0.05* 0.04* 0.05* 0.04* 0.03* 0.04*† 0.03*† 0.98 0.05* 0.05*† 0.05* 0.03*† 0.05* 0.93 0.05* 0.05* 0.04* 0.04* 0.03* 0.04*† 0.03*† 0.03*† 0.05*† 0.03*† 0.38 0.04*† 0.04*†

43 15 45 55 61 44 51 59 30 34 26 39 18 41 70 69 13 19 27 20 39 36 17 32 118 31 42 42 39 44 42 21 9 37 42 18 11

2 0 0 1 2 0 1 3 0 1 2 4 0 4 3 3 0 0 1 0 2 3 0 1 5 3 3 3 3 3 1 1 0 2 3 3 1

1.4 1.8 2.0 1.8 1.6 1.4 1.4 1.0 1.4 1.4 1.6 0.9 1.6 1.0 1.1 1.1 2.2 1.7 1.4 2.6 1.5 1.1 2.0 1.6 0.8 0.9 0.8 0.9 0.6 1.1 1.8 1.6 2.1 1.5 1.2 1.4 1.9

0.04*† 0.03*† 0.04*† 0.05*† 0.05*† 0.04*† 0.04*† 0.82 0.03*† 0.04* 0.05*† 0.93 0.04*† 1.00† 0.40† 0.27† 0.05*† 0.05*† 0.04* 0.04*† 0.05*† 0.46 0.03*† 0.04*† 1.00 1.00 1.00 1.00 1.00 0.87† 0.04*† 0.05*† 0.05*† 0.04*† 0.44 0.33† 0.03*†

41 51 42 14 57 40 47 21 75 59 25 38 30 37 66 66 12 57 29 35 35 33 40 31 113 30 45 45 30 43 36 20 25 10 24 6 11

3 1 0 0 2 0 1 0 5 3 2 3 3 4 2 2 0 3 3 3 3 4 3 3 4 3 3 3 0 3 1 3 2 0 0 1 3

1.s3 1.4 1.4 1.8 1.4 1.4 1.4 1.8 0.8 1.1 1.5 1.2 1.1 1.2 1.3 1.4 2.3 1.2 1.3 1.3 1.1 1.1 1.3 1.1 0.9 0.7 0.8 1.0 1.4 1.2 1.3 1.5 1.7 2.8 1.6 3.1 1.5

1.00† 0.04*† 0.05*† 0.04*† 0.04*† 0.04*† 0.04*† 0.05* 1.00 0.59† 0.03*† 0.58 0.37† 1.00† 0.05* 0.04*† 0.05*† 1.00† 0.68 0.86 0.86 0.92 0.36 0.41 0.96 1.00 1.00 1.00 0.04* 0.78† 0.05*† 0.09† 0.05*† 0.03*† 0.05*† 0.02*† 0.84†

Displayed are the significant sRHA (ID) in any of the three fiscal years, the cluster size (k), test statistic (l), and the observed-toexpected ratio (Obs/Exp). sRHA = subregional health authority. *Significance at a = 0.05. †Significance from the ED case for AFF analysis.

2000, 2005/2006, and 2010/2011, respectively. Of these clusters, 15, 21, and 14 were also previously identified as clusters of ED cases of AFF during the fiscal years. DISCUSSION This study found that while crude rates and directly standardized rates of ED cases and ED presentations for AFF remained relatively stable over time, directly standardized rates of ED cases and ED presentations showed variation over zone and time for some zones. The North zone, a rural region with two relatively large regional EDs, had the highest rates when adjusted for the sex and age group distribution. Such data would be helpful for provincial leaders to explore causes and health planners responsible for health services delivery. In addition, while the administrative data reflect one province’s data, the methods employed have important

implications for other jurisdictions and other disease entities. For clinicians, the importance of diagnostic accuracy and completeness is highlighted in these results. While some studies of AFF have considered geographic variation,17–20 to our knowledge we are the first team to examine geographic clustering of AFF. We used two statistical cluster detection tests to identify geographic clustering of ED cases and ED presentations for AFF. Cluster detection tests are popular for identifying geographic “hot spots” of disease and have been used to examine multiple conditions, including cardiac arrest.21 These cluster detection tests allow for the diverse underlying population distribution of the geographic subareas, the situation present in Alberta. Across years, several of the same sRHAs were identified as clusters of ED cases and/or ED presentations. In particular, the north and southeast portions of the province

974


typically had higher numbers of ED cases and/or ED presentations for AFF than would be expected by chance based on the sex and age group distribution. When we subsequently examined cases with stroke or heart failure follow-up within 365 days of the ED presentation, some of these same potential clusters were identified, suggesting that higher numbers of individuals presenting to EDs for AFF may be indicative of higher numbers of stroke or heart failure patients. The higher rates in some regions may be the result of different causes. The clusters may be because of differential distributions of key variables that are unobserved and/or unavailable for analyses or differential coding of AFF in EDs or may have occurred purely by chance. These clusters could be spurious. Real clusters may be identified because of a greater severity of disease or a lack of availability of non-ED health services. Further investigation could target better education, treatment, and monitoring of patients with AFF to reduce ED presentations and potentially reduce stroke or heart failure associated with AFF. For some areas, there were higher numbers of ED cases and ED presentations, whereas for other areas there were more discrepant results. Such results highlight that the unit of analysis needs to be carefully considered. Only examining ED cases may not truly characterize clustering in the use of ED health services. For health services research, the clusters of cases and the clusters of events related to the case, such as ED presentations, may both be important. Notably, the methods used in this paper account for the potentially repeated number of ED presentations. LIMITATIONS As with all cluster detection methods, the areas of identified as potential clusters may represent real clusters or may be spurious. We believe, however, that these clusters do represent areas of excess ED presentations for AFF, and special intervention and/or education programs could be considered to reduce ED use. Due to the lack of granular detail in the administrative databases (e.g., no information regarding smoking history, alcohol/caffeine intake, body mass index, and other potential AFF confounders), more detailed examination (root cause analysis) would be required to determine the cause of such clusters. While clinicians are able to subgroup different AFF presentations (e.g., paroxysmal, chronic, recurrent), the coding was not sufficiently detailed to examine these categories. Nonetheless, we did use the primary or secondary diagnosis, which limited the inclusion of patients with chronic stable AFF. Other study limitations include that the cluster detection methods have more traditionally been applied to regions with smaller geographic subunits, and when applied to our provincial boundaries, the ability to detect very small geographic clusters is limited. The metric of geographic closeness of the population-based centroids may also not truly represent the geographic closeness of ED cases when the sRHA covers a large geographic area. Changes in sRHA of residence over time or time spent in the sRHA are also not considered in the analysis. Our definition of an AFF ED case was

based on having at least one ED presentation for AFF during the study period and may not be representative of all patients who have AFF or seek health services for AFF. Nonetheless, we do not feel that these limitations have had a substantial effect on our findings. CONCLUSIONS Our population-based study spanned 12 fiscal years and showed variations in the number of people presenting to EDs for atrial fibrillation and flutter and the total number of ED presentations for atrial fibrillation and flutter over geography. In particular, some areas were also potential clusters of stroke or heart failure followup after ED presentation for atrial fibrillation and flutter. The potential clusters identified may represent geographic areas with higher disease severity or lower availability of non-ED health services. The clusters are not all likely to have occurred by chance, and further investigation and intervention could occur to reduce ED presentations for atrial fibrillation and flutter. The authors thank Alberta Health for facilitating access to the data and Xiaoqing Niu, PhD, for assistance with data analysis. References 1. Friberg J, Buch P, Scharling H, Gadsbphioll N, Jensen GB. Rising rates of hospital admissions for atrial fibrillation. Epidemiology 2003;14:666–72. 2. Wattigney WA, Mensah GA, Croft JB. Increasing trends in hospitalization for atrial fibrillation in the United States, 1985 through 1999: implications for primary prevention. Circulation 2003;108:711–6. 3. Fuster V, Ryden LE, Cannom DS, et al. ACC/AHA/ ESC 2006 guidelines for the management of patients with atrial fibrillation. J Am Coll Cardiol 2006;48:149–246. 4. Stewart S, Hart CL, Hole DJ, McMurray JJ. 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: 359–64. 5. McDonald AJ, Pelletier AJ, Ellinor PT, Camargo CA Jr. Increasing US emergency department visit rates and subsequent hospital admissions for atrial fibrillation from 1993 to 2004. Ann Emerg Med 2008;51:58–65. 6. Rogenstein C, Kelly AM, Mason S, et al. An international view of how recent-onset atrial fibrillation is treated in the emergency department. Acad Emerg Med 2012;19:1255–60. 7. Statistics Canada. Population and Dwelling Counts, for Canada, Provinces and Territories, 2011 and 2006 Censuses. Available at: http://www12.statcan.gc.ca/census-recensement/2011/dp-pd/hlt-fst/pdpl/Table-tableau.cfm? LANG=Eng&T = 101&S = 50&O=A. Accessed May 29, 2015. 8. Alberta Health and Wellness. Ambulatory Care in Alberta Using Ambulatory Care Classification System Data. Edmonton, AB: Alberta Health and Wellness, 2004.


9. Canadian Institute for Health Information. National Ambulatory Care Reporting System (NACRS). Available at: http://www.cihi.ca/CIHI-ext-portal/internet/en/ document/types+of+care/hospital+care/emergency+ care/NACRS_METADATA. Accessed May 29, 2015. 10. Practice Management Information Corporation. International Classification of Diseases, 9th revision, 3rd ed, Clinical Modification. Los Angeles, CA: Practice Management Information Corporation, 1989. 11. Canadian Institute of Health Information. The Canadian Enhancement of ICD-10 (International Statistical Classification of Diseases and Related Health Problems, Tenth Revision). Ottawa, ON: Canadian Institute of Health Information, 2001. 12. Ellehoj E, Schopflocher D. Calculating Small Area Analysis: Definition of Sub-regional Geographic Units in Alberta. Edmonton, Alberta: Alberta Health and Wellness, 2003. 13. Carriere KC, Roos LL. Comparing standardized rates of events. Am J Epidemiol 1994;140:472–82. 14. Besag J, Newel J. The detection of clusters in rare diseases. J R Stat Soc Ser A 1991;154:143–55. 15. Le ND, Petkau AJ, Rosychuk RJ. Surveillance of clustering near point sources. Stat Med 1996;15: 727–40. 16. Rosychuk RJ, Huston C, Prasad NG. Spatial event cluster detection using a compound Poisson distribution. Biometrics 2006;62:465–70. 17. Naderi S, Wang Y, Miller AL, et al. The impact of age on the epidemiology of atrial fibrillation hospitalizations. Am J Med 2014;127(158):e1–7. 18. Wetmore JB, Ellerbeck EF, Mahnken JD, et al. Stroke and the “Stroke Belt” in dailysis: contribution of patient characteristics to ischemic stroke rate and its geographic variation. J Am Soc Nephrol 2013;24:2053–61. 19. Tabereaux PB, Brass LM, Concato J, Bravata DM. Hospital admissions for stroke among the very old in the USA. Neuroepidemiology 2008;31:93–9.

975

20. Ayoola AE, Banzal SS, Elamin AK, Gadour MO, Elsammani EW, Al-Hazmi MH. Profile of stroke in Gizan. Kingdom of Saudi Arabia. Neurosci (Riyadh) 2013;8:229–32. 21. Sasson C, Cudnik MT, Nassel A, et al. Identifying high-risk geographic areas for cardiac arrest using three methods for cluster analysis. Acad Emerg Med 2012;10:139–46. Supporting Information The following supporting information is available in the online version of this paper: Data Supplement S1. Sub-Regional Health Authority (sRHA) IDs and names. Data Supplement S2. Brief description of the cluster detection methods. Data Supplement S3. Population, cases with ED presentations for AFF, and crude rates by fiscal year in Alberta during April 1, 1999, to March 31, 2011. Data Supplement S4. (a) Sex and age group directly standardized rates (DSRs) per 1,000 by fiscal year and by zones and (b) sex and age group directly standardized visit rates (DSVRs) per 1,000 by fiscal year and by zones. Data Supplement S5. Clustering results for each sRHA with sex and age group as strata for fiscal year 1999/2000. Data Supplement S6. Clustering results for each sRHA with sex and age group as strata for fiscal year 2005/2006. Data Supplement S7. AFF clusters identified for cases and presentations for fiscal year 2010/2011: sRHAs that are potential clusters alone (dark shading) and when combined with one (medium shading) or two (light shading) nearest neighbors are shown. Data Supplement S8. Atrial fibrillation or flutter clusters of cases (*) and presentations (x) by fiscal year.