The estimated prevalence was calculated based on the age-specific incidences in the same year.
Variation in incidence (A) and prevalence (B) of CRS in 2012-2013. Reproduced with permission from Alberta Health Services.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Xu Y, Quan H, Faris P, et al. Prevalence and Incidence of Diagnosed Chronic Rhinosinusitis in Alberta, Canada. JAMA Otolaryngol Head Neck Surg. 2016;142(11):1063–1069. doi:10.1001/jamaoto.2016.2227
Copyright 2016 American Medical Association. All Rights Reserved.
What is the incidence and prevalence of chronic rhinosinusitis (CRS) from the perspective of the health care system?
In this Canadian population-based analysis of the past 10 years, the estimated prevalence of diagnosed CRS was approximately 2% to 2.5% of the population, and the mean incidence was 2.5 per 1000 population. In addition, there was high geographic variation in the prevalence and incidence of diagnosed CRS.
Although survey-based studies report higher prevalence rates for CRS, results from this study may more accurately reflect the disease burden of CRS to the health care system; the large geographic variation in diagnosed CRS indicates a potential gap in quality of care and justifies further investigation into the reasons for the variation.
Reported prevalence rates of chronic rhinosinusitis (CRS) range from 1% to 12% worldwide. To facilitate appropriate health service delivery and resource allocation, it is important to improve the estimated burden of CRS to the health care system.
To assess the prevalence and incidence of diagnosed CRS in Alberta, Canada, from the perspective of the health care system and to evaluate the 10-year temporal trend and geographic variation of diagnosed CRS.
Design, Setting, and Participants
From provincial-wide physicians’ claim data, a CRS cohort was identified using a validated case definition. The population at the midpoint (2008-2009) of the study period (2 925 930) was used as the reference. The crude as well as age- and sex-standardized incidence and prevalence rates were calculated. The age-specific incidence and prevalence by sex were also assessed in each study year. Small-area variation analysis was conducted using extremal quotient, weighted coefficient of variation, χ2 statistic, systematic component of variation, and empirical Bayes variance estimate.
Of the 2 925 930 individuals in the study at midpoint (2008-2009), 1 451 261 (49.6%) were women, and the mean (SD) age was 45 (17) years. From fiscal year 2004-2005 to fiscal year 2013-2014, the mean age- and sex-standardized incidence of diagnosed CRS was 2.5 (range, 2.3-2.7) per 1000 population. The estimated prevalence based on age-specific incidence varied between 18.8 (95% CI, 18.7-18.9) and 23.3 (95% CI, 23.1-23.5) per 1000 population during 2004-2005 to 2013-2014, and no obvious growing trend was found. There was high geographic variation in the diagnosed incidence and prevalence of CRS (mean systematic component of variation, 19.4 and 12.3, respectively).
Conclusions and Relevance
Although the incidence and prevalence rates of diagnosed CRS were lower compared with earlier published estimates obtained from population-based survey analysis, outcomes from this study may more accurately reflect the disease burden of CRS to the health care system. Given that the prevalence of CRS within a single province is expected to be uniformly distributed, the large geographic variation in diagnosed CRS indicates a potential gap in quality of care and justifies further investigation into the reasons for the variation.
Chronic rhinosinusitis (CRS) is a common, chronic inflammatory disease of the paranasal sinuses.1,2 Given its chronic nature, long-term and consecutive medication control or surgery may be needed for successful outcomes,3 which leads to substantial health resource utilization.4 However, it is challenging for policymakers and health services researchers to design improved systems for care since the currently reported prevalence for CRS varies substantially with rates ranging from 1.0% to 12.1% across the world.5-12 Potential reasons for the wide range of reported prevalence may include the different epidemiologic methods applied (ie, population survey vs administrative database review) and the challenge in diagnosing CRS on a population level given the need to demonstrate objective evidence of inflammation in addition to subjective symptoms.
In general, survey-based approaches relying on patient reporting produce higher prevalence rates and likely overestimate the true prevalence rate owing to reporting bias or recall bias.1 Furthermore, patient self-diagnosis of CRS has been shown to be inaccurate, and survey approaches likely fail to quantify clinically relevant CRS since a proportion of the patients may have mild disease and do not require medical care. However, prevalence rates for CRS using administrative data typically report lower prevalence rates, which may underestimate the true prevalence since the rates depend on the accuracy of physician diagnostic coding and will fail to identify patients with CRS who do not seek medical care.9,10
Within the Canadian population, the best available evidence supports a 3% to 5% prevalence of CRS. This estimate was derived from a single survey study7 delivered to 30 000 Canadian respondents within 10 provinces in 1996. However, the prevalence estimate of CRS was based on a single question assessing whether the respondent had a diagnosis of rhinosinusitis by a health care provider and was not specific to CRS. Therefore, the CRS prevalence estimate of 3% to 5% was likely overestimated since a proportion of cases would include diagnosed acute rhinosinusitis.
Given the limitations of earlier published prevalence estimates of CRS, the primary objective of the present study was to investigate the prevalence and incidence of diagnosed CRS from the perspective of the Canadian health care system. Secondary objectives were to define the temporal trend and geographic variation of diagnosed CRS. Outcomes of this study will help to elucidate the burden of CRS to the health care system and assist health service researchers and policy makers in designing improved methods for health care delivery.
We conducted a retrospective health care administrative database study involving 2 925 930 adults 18 years or older at the midpoint of the study during fiscal year 2004-2005 to fiscal year 2013-2014 in Alberta, Canada. The population-based administrative database from Analytics of Alberta Health Services was used in this study. This provincial database collects data on each health care encounter in Alberta. From this database, we extracted patients’ unique identifier numbers, the International Classification of Diseases, Ninth Revision (ICD-9) diagnosis code assigned to each of the 3 diagnosis fields available per patient encounter, and the date of the claim to identify the CRS cohort. The data were deidentified. This study was approved by the University of Calgary Health Research Ethics Board.
With the unique identifier number, the physician claims database was linked to the provincial health insurance registry data, which included all residents in Alberta with few exceptions, such as the members of federal correctional facilities. We obtained data on the age, sex, date of death, move from the province, and postal code of the residence. The postal code was used to assign participants into the 64 Health Status Areas (HSAs), which were developed for the purpose of status reporting on surveillance and assessment of population health in Alberta. Each HSA includes 25 000 to 55 000 people and was defined on the basis of community similarity and travel patterns.13
A previous study14 evaluated the validity of several potential case definitions for CRS within an administrative database. This previous analysis used the case definition of 2 or more claims coded as ICD-9 codes 471.x or 473.x within a 2-year period. The validity of this case definition was 77.0% sensitivity, 79.0% specificity, 78.6% positive predictive value, and 77.5% negative predictive value.
We applied this case definition to identify the entire CRS cohort in the study period. Patients were defined as new cases within a certain year if the first claim for CRS occurred in that year. One patient could be defined as a new case only once; therefore, the CRS case was regarded as a prevalent case until the person died or moved from the province in each subsequent year. Although the available data were from fiscal year 2001-2002 to fiscal year 2014-2015, we started the incidence calculation from 2004-2005, allowing a 3-year washout period to identify new cases assuming that patients received a new CRS diagnosis if they did not seek any health care service for treatment of CRS in at least the past 3 years.
The incidence of CRS was calculated using the new cases divided by the population at risk in various age and sex groups in the different geographic regions. The population at risk was calculated by excluding individuals who had died or moved as well as prevalent cases at the beginning of the particular year. The prevalent cases at a certain time were the total number of new cases before this point; the prevalence was then calculated using the prevalent cases divided by the population at different age and sex groups and regions.
For chronic diseases, the prevalence climbs as the observation period increases. Thus, we conducted another analysis of prevalence by simulating the life-table approach to estimate the prevalence based on age-specific incidence.15 This method estimated the prevalence by accumulatively summing the population-weighted age-specific incidences. The 95% CI for all prevalence and incidence data was calculated using the binomial method. To compare the provincial prevalence and incidence between different years, we calculated the age- and sex-standardized prevalence and incidence in each year from 2004-2005 to 2013-2014 choosing the middle range (2008-2009) as reference. Age was categorized by 5-year increments from 18 years to more than 85 years.
Geographic variation of prevalence and incidence was assessed by widely used statistics including extremal quotient, weighted coefficient of variation, χ2 statistic,16 systematic component of variation (SCV),17 and empirical Bayes (EB) variance estimate.18 The SCV represented 2 sources of variations: random and nonrandom; thus, it is a robust measure for geographic variation.17 Empirical Bayes variance estimate outperforms SCV because it takes into consideration the few cases in certain regions with small populations.19,20 The formulations of these statistics were described in previous studies.13,21 All statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc).
The overall diagnosed incidence and prevalence of CRS annually from 2004-2005 to 2013-2014 is presented in Table 1. Overall, the mean crude incidence of CRS between 2004-2005 and 2013-2014 was 2.5 (range, 2.3-2.7) per 1000 population. Using the age and sex distribution at the middle study period (2008-2009) as reference, the age- and sex-standardized incidence was stable over the 10-year period, ranging from 2.2 to 2.6 per 1000 population (Table 1 and Figure 1). The population in the midpoint (2008-2009) of the study period was 2 925 930. Of these 1 451 261 individuals (49.6%) were women, and the mean (SD) age was 45 (17) years.
The overall age- and sex-standardized prevalence increased from 8.5 (95% CI, 8.4-8.6) to 30.3 (95% CI, 30.1-30.5) per 1000 population, with a mean annual increase of 2.4 per 1000 population (P < .001). However, the estimated prevalence according to the age-specific incidences ranged from 18.8 (95% CI, 18.7-18.9) to 23.3 (95% CI, 23.1-23.5) per 1000 population (Table 1 and Figure 1).
There were 90 170 patients with a diagnosis of CRS; 8568 of these (9.5%) were newly diagnosed cases in 2012-2013. The age-specific prevalence and incidence of diagnosed CRS in 2012-2013 is presented in Table 2. The crude prevalence among women (25.4 per 1000 population) was higher than that of men (24.3 1000 population), and the crude incidence among women (2.4 per 1000 population) was also higher than that of men (2.3 per 1000 population). Generally, the prevalence continuously increased from the 18- to 20-year age group (13.1 per 1000 population) to the 76- to 80-year age group (35.8 per 1000 population) and then decreased in the oldest age group (>85 years, 20.8 per 1000 population). The incidence rate of CRS grew as the age of the population increased up to age 71 to 75 years (Table 2). The prevalence and incidence of diagnosed CRS within the male and female populations had no significant difference at various age groups except for the 18- to 20-year and 76- to 80-year groups (Table 2).
The HSA-specific prevalences and incidences (standardized to the provincial age and sex distribution) in 2012-2013 are displayed in Figure 2. The regional prevalence ranged from 11.3 to 40.5 per 1000 population, and the regional incidence ranged from 1.0 to 4.2 per 1000 population. Overall, there was large geographic variation in both the prevalence and incidence of diagnosed CRS across the 64 HSAs (Table 3). During the study period, the mean EB statistics were 21.1 and 13.0 for incidence and prevalence, respectively. The mean SCV values were 19.4 and 12.3 for incidence and prevalence, respectively. Empirically, an EB and SCV larger than 10 is regarded as very high variation.21
In this population-based study from the perspective of the Canadian health care system, the age- and sex-standardized incidence of diagnosed CRS fell within a narrow range: 2.3 to 2.7 per 1000 population from 2004-2005 to 2013-2014. Although the crude prevalence increased by year, the estimated prevalence based on age-specific incidences was between 18.8 and 23.3 per 1000 people (ie, 2%-2.5% of the population) during 2004-2005 to 2013-2014, and no obvious growing trend was found. In addition, we found high variation among the 64 geographic HSAs in Alberta in the diagnosed prevalence and incidence of CRS. This large geographic variation implies the need to further investigate the reasons for the variation to ensure the population is receiving high-quality care for this chronic disease.
Currently reported prevalence for CRS varies substantially with rates ranging from 1.0% to 12.1% across the world.5-12 The highest estimated prevalence of CRS was 12.1% reported by the 2012 National Health Interview Survey in a nationwide survey study12 of the US population. However, this US-based survey study likely overestimates the true prevalence of CRS since it failed to differentiate between various forms of rhinosinusitis and included acute rhinosinusitis in the estimate. Another survey study8 conducted by Global Allergy and Asthma European Network reported a 10.9% prevalence of CRS in European countries. However, the European estimate of 10.9% may overestimate the true prevalence rate for CRS since it relied on patient self-reported CRS symptom criteria for longer than 3 months and lacked demonstration of objective evidence of inflammation. When patients reported a physician diagnosis of CRS, the European prevalence of diagnosed CRS was 5.0%. Two administrative database studies9,10 from the United States reported much lower prevalence rates for CRS (2%-2.3%) using ICD-9 coding data. The lowest prevalence estimate for CRS (1%) was reported in a Korean study11 that applied rigorous diagnostic criteria to identify patients with CRS.
Overall, our study identified that the crude prevalence of diagnosed CRS in 2013-2014 was 30.8 per 1000 population (3.1% of population). This value is lower than the reported prevalence of 5% in an earlier published Canadian survey study.7 Several potential reasons could explain the differences in Canadian prevalence estimates. The Canadian survey study conducted by Chen et al.7 included children (aged 12-18 years) and relied on people to answer a single question about being diagnosed with rhinosinusitis without specifying CRS. There is evidence22 to suggest that the prevalence of diagnosed CRS is higher in children than adults in the US population. Furthermore, since acute rhinosinusitis is more prevalent than CRS, asking people to report a nonspecific diagnosis of rhinosinusitis would overestimate the true prevalence of CRS. Our study excluded populations younger than 18 years and included patients with a specific diagnosis of CRS made by a physician and therefore sought medical care for their disease. The prevalence rate reported in our study is that of diagnosed CRS and therefore focuses on patients whose disease is clinically significant enough to warrant physician encounter. Our study would therefore fail to include patients with CRS who have mild disease and do not seek medical care.
Our results showed that the diagnosed prevalence of CRS consistently increased from 2004-2005 to 2013-2014 with an annual increase of 2.4 per 1000 population. However, this increase may not reflect the real change of the disease burden of CRS. Considering data availability, the prevalence was likely underestimated at the beginning years of the study period and the increasing trend of prevalence was probably attributed to accumulation of prevalent cases by adding more data year after year. Thus, we calculated the estimated prevalence based on the age-specific incidences in each year. The estimated prevalence displayed no growing or decreasing pattern during the study period. The observed increasing pattern of prevalence indicated that we underestimated the prevalence at the beginning of the study period and as well for the following years given that we do not have the data before 2001. Therefore, the crude prevalence estimates likely become more accurate in the years closer to the end of the analysis as opposed to toward the beginning of the analysis.
The geographic variation in the prevalence and incidence of diagnosed CRS was consistently high among the 64 geographic HSAs throughout the study period considering that the mean SCV for the past 10 years was 19.4 for incidence and 12.3 for prevalence, respectively. This finding may provide one explanation for the observed variation of medication utilization and endoscopic sinus surgery for CRS in Alberta.13,23 The variation of health care utilization is not always harmful to the systems; sometimes it indicates the patient-centered health care delivery.24 However, when there is no reason to believe a disease should vary within a certain large geographic region (ie, province of Alberta), the high variation in the prevalence and incidence of diagnosed CRS may indicate a gap in quality of care because such a variation may imply disparity in terms of physicians’ understanding of CRS diagnostic criteria and coding, impaired equity and access to health care resources for CRS, or the diverseness on disease awareness among patients. To provide clear action plans to reduce the variation of diagnosed CRS, further study is needed to investigate the reasons for the variation.
Several limitations should be considered when interpreting the results of the present study. First, the findings were derived from 1 province in Canada, which may not be generalizable to other Canadian provinces or other countries given the difference in ethnicity, geographic factors, provider coding, and health care systems. Second, this study did not quantify the prevalence and incidence of CRS within the population using patient-reporting survey measures; the estimates were based on administrative data whereby CRS was diagnosed and coded by a physician after the patient sought medical care. However, the prevalence and incidence of diagnosed CRS provides an important outcome since it quantifies the true burden of CRS on health service utilization. Third, all studies using administrative data are limited by an imperfect disease cohort since identification depends on accurate physician diagnosis coding. For example, a recent study by Novis et al25 demonstrated that most diagnoses of CRS within an emergency medicine or primary care setting were inaccurate. Although we applied a validated case definition for CRS based on provincial data, our cohort will include patients without CRS and exclude some true CRS cases since the positive predictive value and sensitivity of the case definition were 80%. However, 90% of the CRS cohort identified in this study had visited and received an ICD-9 diagnosis of CRS by an otolaryngologist, which helps to reduce the risk of including false cases of CRS. Despite these limitations, this study was strengthened by using a robust population-based administrative database, application of a validated case definition to define our CRS cohort, and rigorous methodology including calculation of estimated prevalence rates using age-specific indices.
This study has reported the prevalence and incidence of diagnosed CRS from the perspective of the Canadian health care system. Over the past 10 years, the estimated prevalence was approximately 2% to 2.5% of the population, and the mean incidence was 2.5 per 1000 population. In addition, there was very high geographic variation in the prevalence and incidence of diagnosed CRS. Given that the prevalence of CRS within a single province is expected to be uniformly distributed, the large geographic variation in diagnosed CRS indicates a potential gap in quality of care and justifies further investigation into the reasons for the variation.
Corresponding Author: Luke Rudmik, MD, MSc, Division of Otolaryngology–Head and Neck Surgery, Department of Surgery, University of Calgary, Richmond Road Diagnostic and Treatment Centre, 1820 Richmond Rd SW, Calgary, AB T2T 5C7, Canada (firstname.lastname@example.org).
Accepted for Publication: July 3, 2016.
Published Online: September 8, 2016. doi:10.1001/jamaoto.2016.2227
Author Contributions: Dr Xu had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Xu, Quan, Faris, Rudmik.
Acquisition, analysis, or interpretation of data: Xu, Garies, Liu, Bird, Kukec, Dean, Rudmik.
Drafting of the manuscript: Xu, Faris, Dean, Rudmik.
Critical revision of the manuscript for important intellectual content: Xu, Quan, Faris, Garies, Liu, Bird, Kukec, Rudmik.
Statistical analysis: Xu, Faris, Liu.
Obtaining funding: Rudmik.
Administrative, technical, or material support: Garies, Bird, Dean.
Study supervision: Quan, Faris, Garies, Rudmik.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Rudmik is a paid consultant for BioInspire. No other conflicts were reported.
Funding/Support:M.S.I. Foundation grant and Petro-Canada Young Innovator in Community Health Sciences of Canada Award (Dr Rudmik).
Role of the Funder/Sponsor: The funding organization had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Create a personal account or sign in to: