Prevalence and prescribing patterns of oral corticosteroids in the United States, Taiwan, and Denmark, 2009–2018

Abstract Oral corticosteroids (OCS) are commonly prescribed for acute, self‐limited conditions, despite studies demonstrating toxicity. Studies evaluating longitudinal OCS prescribing in the general population are scarce and do not compare use across countries. This study investigated and compared OCS prescription patterns from 2009 to 2018 in the general populations of the United States, Taiwan, and Denmark. This international population‐based longitudinal cohort study used nationwide claims databases (United States: Optum Clinformatics Data Mart; de‐identified; Taiwan: National Health Insurance Research Database; and Denmark: National Prescription and Patient Registries/Danish National Patient Registry) to evaluate OCS prescribing. We classified annual OCS duration as short‐term (1–29 days), medium‐term (30–89 days), or long‐term (≥90 days). Longitudinal change in annual prevalence of OCS use and physician prescribing patterns were reported. Among 54,630,437 participants, average annual percentage of overall OCS use was 6.8% in the United States, 17.5% in Taiwan, and 2.2% in Denmark during 2009–2018. Prevalence of OCS prescribing increased at an average annual rate of 0.1%–0.17%, mainly driven by short‐term prescribing to healthy adults. One‐quarter to one‐fifth of OCS prescribing was associated with a diagnosis of respiratory infection. Family practice and internal medicine physicians were among the highest OCS prescribers across countries and durations. Age‐ and sex‐stratified trends mirrored unstratified trends. This study provides real‐world evidence of an ongoing steady increase in OCS use in the general populations of the United States, Taiwan, and Denmark. This increase is largely driven by short‐term OCS prescribing to healthy adults, a practice previously viewed as safe but recently shown to incur substantial population‐level risk.


INTRODUCTION
7][8] In the same studies, one-fifth to one-quarter of individuals in the general populations of both countries were prescribed OCS over a 3-year period.0][11][12] These striking data demonstrate that short-term OCS use is both common among otherwise healthy patients, and associated with rare but severe adverse events.
Several previous studies address the prevalence of OCS use in the general population.However, these focus on long-term OCS use (≥90 days), and most do not include estimates from the past decade.5][16][17] Data from the National Health and Nutrition Examination Surveys from 1999-2008 found an overall prevalence of OCS use of 1.2% among adults in the United States, with approximately two-thirds of users receiving long-term OCS. 18However, a more recent publication by Benard-Laribière et al. reported increasing prevalence of overall OCS use from 14.7 in 2007 to 17.1% in 2014 in a longitudinal cohort of French adults. 19This study, along with the national findings in United States and Taiwan highlighted above, support the hypothesis that global prescribing of OCS may have increased in the general population over the last decade.
To our knowledge, studies evaluating longitudinal trends and prescription patterns of OCS in the general population are scarce, do not compare prevalence of OCS use across countries, and under-represent Asian countries.Utilizing data derived from three nationwide cohorts, this study aimed to investigate and compare longitudinal trends and prescription patterns of OCS from 2009 to 2018 in the general populations of three countries: the United States, Taiwan, and Denmark.

Study design and data sources
This international population-based longitudinal cohort study used nationwide claims databases from the United States (Optum Clinformatics Data Mart; de-identified), Taiwan (National Health Insurance Research Database), 7,18 and Denmark (National Prescription and Patient Registries/Danish National Patient Registry) 19,20

Study participants
For each year of the study period, we classified all enrolled individual with greater than or equal to one pharmacy claim as the total number of subjects enrolled in that year (our denominator).OCS users in that given year included all patients enrolled during that year who had greater than or equal to one pharmacy claim for OCS (our numerator).Non-users in that given year had greater than or equal to one pharmacy claim during this same period, but no claims for systemic (oral or injectable) corticosteroids.We defined an OCS user's index date as the date of their first pharmacy claim for OCS use during the study period, and a non-user's index date as the first pharmacy claim of any kind made except OCS during the study period.Prevalence of OCS users for each year of the study period was calculated as follows: (OCS users in that year)/(total number of subjects enrolled in that year).
We limited our analysis to individuals with greater than or equal to 6 months of medical and pharmacy claims (for OCS users) or greater than or equal to 6 months of medical claims (for non-users) prior to index date to ensure adequate capture of baseline comorbid conditions.We captured these conditions using the Elixhauser Comorbidity Index, a composite comorbidity score of individuals based on International Classification of Disease (ICD) diagnosis codes and healthcare utilization. 20,21We excluded individuals with claims for injectable but not OCS, to avoid issues with misclassification of injectable corticosteroids in claims data.We also excluded individuals who died during the study period to reduce immortal time bias.Cohort selection criteria are outlined in Figure S1.

Oral corticosteroid utilization
To assess OCS utilization, we standardized all oral corticosteroid dosages to prednisone equivalents, and used pharmacy claims data to derive average daily dose, cumulative annual dose, and cumulative annual duration (Table S1). 22Dose conversion to oral prednisone equivalents was performed per standard practice, to account for variation in potency and international utilization across different OCS formulations. 6,7We classified annual duration of OCS into three categories for each year: short-(1-29 days' supply), medium-(30-89 days' supply), and long-term (≥90 days' supply). 6,13

Indication and physician specialty
To evaluate medical diagnoses associated with OCS use, we linked each pharmacy claim for OCS to primary medical claims occurring in the 30 days prior to the OCS dispense date.In the United States and Taiwan, we captured diagnoses made prior to October 1, 2015 (United States) or December 31, 2015 (Taiwan) using ICD, ninth revision (ICD-9) codes, and those made thereafter using tenth revision (ICD-10) codes.ICD-10 codes were used throughout the period 2009-2018 in Denmark.We grouped ICD codes into clinically meaningful categories with Agency for Healthcare Research and Quality Clinical Classification Software (CCS), using CCS version 2015 for ICD-9, and CCS Refined (CCSR) version 2022.21 for ICD-10. 1 We used Medical Expenditure Panel Survey (MEPS) categories to compare data across CCS and CCSR groupings. 23To assess physician specialties in the United States and Taiwan, we used data linked to pharmacy claims to identify the specialty of each physician responsible for an OCS prescription during the study period.

Statistical analysis
Baseline demographic and clinical characteristics were outlined using descriptive statistics.Of note, OCS dose and duration data were not available in Denmark.We calculated the annual prevalence of overall, short-, medium-, and long-term OCS use among individuals with any claim during 2009-2018.We also estimated the annual prevalence of OCS use stratified by age and sex separately.We used linear regression to evaluate time trends in prevalence of OCS use annually during this period.A p value less than 0.05 was declared statistically significant.All analyses were conducted with SAS version 9.4 (SAS Institute).

RESULTS
Table 1 indicates the baseline characteristics of the populations in the United States, Taiwan, and Denmark.A total of 54,630,437 participants (28,386,692, 52.0% women) were included.In all countries, OCS users were older and more often women than nonusers.OCS users in Taiwan had higher comorbidity scores than those in the United States (mean [SD] Elixhauser Comorbidity Index: 1.1 [1.7] in the United States vs. 2.1 [2.4] in Taiwan).A substantial proportion of OCS users had Elixhauser Comorbidity Index equal to 0 in both the United States and Taiwan (53.8% in the United States and 31.9% in Taiwan).OCS users had higher outpatient healthcare utilization than nonusers in all countries; median (interquartile range [IQR]) outpatient visits for OCS users versus non-users were 4 (IQR: 1-7) versus 1 (IQR: 0-3) in the United States, 5 (IQR: 2-11) versus 0 (0-3) in Taiwan, and 1 (IQR: 0-4) versus 0 (IQR: 0-1) in Denmark, respectively.The median per individual user annual number of OCS prescriptions was similar across countries (median [IQR]: 1.0 [1.0-2.0] in the United States; 1.3 [1.0-2.0] in Taiwan; and 1.0 [1.0-2.0] in Denmark).
Figure 1 presents the prevalence and time trends in OCS use over the 10-year study period.The average annual prevalence of overall OCS use was 6.8% in the United States, 17.5% in Taiwan, and 2.2% in Denmark.The overall prevalence of OCS use significantly increased during the study period in all three countries (in the United States: from 6.4% to 7.7%, β = 0.1, p for linear trend = 0.02; in Taiwan: from 16.6% to 18.7%, β = 0.17, p for linear trend = 10 −3 ; and in Denmark: from 1.7% to 2.9%, β = 0.13, p for linear trend < 10 −3 ).When stratified by age and sex, increasing trends were observed across adults of both sexes, but not in children under age 18 (Figure 2).Increasing OCS prevalence over the study period was primarily driven by short-term prescriptions in the United States, whereas short-, medium-, and long-term prescriptions all increased in Taiwan (Figure S2).Trends similar to those shown for overall use were observed for short-, medium-, and long-term use when stratified by age and sex (Figures S3 and S4).
For overall OCS use, the median (IQR) of both annual dose and duration were higher in the United States (200.0 [105.0-310.0]mg and 6.0 [5.0-12.0]days) than in Taiwan (58.2 [37.5-97.5]mg and 5.0 [3.0-9.0]days; Table 1).The exact dose and duration in Denmark are not available.Median dose and duration for short-, medium-, and longterm prescription remained higher in the United States than those in Taiwan, except for the duration for mediumterm prescriptions (Table S2).Most OCS prescriptions were for prednisone or prednisolone (63.9% in the United States; 54.4% in Taiwan; and 78.4% in Denmark).Methylprednisolone accounted for 31.3% of OCS prescriptions in the United States, 12.5% in Taiwan, and 7.7% in Denmark.Dexamethasone accounted for 3.7% of OCS prescriptions in the United States, 23.1% in Taiwan, and less than or equal to 1% in Denmark (Table S3).
Table 2 shows the top 10 indications for overall OCS use in each country during the study period.Acute bronchitis and upper respiratory infection were the most common indications in the United States and Taiwan, accounting for 18.7% and 25.1% of all OCS prescriptions, respectively.Chronic obstructive pulmonary disease (COPD), asthma, and other respiratory conditions were the most common indications in Denmark, accounting for 17.0% of all prescriptions.Five common indications among the top 10 were observed across all countries: COPD, asthma, and other respiratory conditions; allergic reactions; osteoarthritis and other non-traumatic joint disorders; systemic lupus and connective tissue disorders; and skin disorders.In the United States and Taiwan, the top 10 indications for short-and medium-term of OCS use were comparable to those for overall OCS use.Osteoarthritis and non-traumatic joint disorders, COPD, asthma, and other respiratory conditions, and systemic lupus and connective tissue disorders were the top three indications for longterm OCS use in both the United States and Taiwan, cumulatively accounting for 34.7% (the United States) and 45.7% (Taiwan) of long-term prescriptions during the study period.The top 10 indications for OCS use in each year of the study period are summarized in Tables S4-S7.In the United States, overall and short-term OCS use for COPD and asthma and allergic reactions declined over the study period, whereas use for skin and joint disorders increased.Similar trends were seen in Taiwan, with exception of an increase in prescribing for acute rather than chronic respiratory conditions.In contrast, long-term prescribing for both joint disorders and chronic lung conditions declined over the study period in both countries, whereas prescribing for acute lung conditions and allergic reactions remained stable.Long-term OCS prescribing for connective tissue disorders declined in both countries across all durations.
Table 3   In the United States, non-physician providers, such as physician assistants and nurse practitioners, were responsible for 16.0% of OCS prescriptions over the study period (Table S8).The annual frequency of OCS prescribed by these prescribers increased over time, from 6.0% in 2009 to 23.3% in 2018.These statistics are not available for Taiwan.

DISCUSSION
Our study has four key findings.First, in this international population-based longitudinal cohort, the average annual prevalence of overall OCS use in the general population varied substantially between countries during 2009-2018, at 6.8% in the United States, 17.5% in Taiwan, and 2.2% in Denmark.Second, the prevalence of OCS use in each of these three countries increased during the 10-year study period: from 6.4% to 7.7% in the United States, from 16.6% to 18.7% in Taiwan, and from 1.7% to 2.9% in Denmark.This rise was observed only among adults and predominantly driven by increased short-term OCS use.5][26] Fourth, family practice and internal medicine physicians were among the highest OCS prescribers in these countries.
It is important to contextualize the clinical significance of the noted international increase in OCS prescribing over the study period.The percentage increases noted in the United States, Taiwan, and Denmark with an approximately additional 346,800, 471,700, and 65,800 patients in each country, respectively, in 2018 relative to 2009.Our evidence-based, real-world data suggest that the majority of increased OCS prescribing over the past decade is due to short-term use by healthy, working-age adults.This is the first study to demonstrate a longitudinal increase in OCS prescribing among healthy adults internationally, and underscores the need for judicious use of OCS.Recent work has demonstrated significant associations between shortterm OCS use and serious harms, such as gastrointestinal bleeding, heart failure, serious infection, venous thromboembolism, and fracture. 6,9,27As these risks persist in young, otherwise healthy individuals, the populationlevel harms of short-term OCS exposure remain a critical concern even when clinicians avoid prescribing to elderly or comorbid "high-risk" patients.These harms are particularly important to consider as OCS are commonly prescribed for self-limited conditions (e.g., in this study, bronchitis and upper respiratory infections) and when treatment benefit is unclear.Particularly, we found that OCS prescriptions in the United States and Taiwan are commonly administered to patients with respiratory infections, allergic reactions, skin disorders, and back problems, all of which are typically self-limited conditions with effective non-steroidal treatments.These data provide important global context to recent studies evaluating risks attributable to avoidable corticosteroid prescriptions. 6,8,9Similar to successful initiatives to optimize opioid and antibiotic prescriptions, implementing a model of "corticosteroid stewardship" may facilitate reducing excess harms related to avoidable OCS use. 5,28e noted substantial international variation in OCS use.In Denmark, one in 50 individuals on average received an OCS prescription each year, versus one in 14 in

USA a
Taiwan a Denmark a As opposed to Taiwan and Denmark, OCS can be prescribed by mid-level providers in the United States.We noted that such prescribing increased almost fivefold during the study period, with OCS prescriptions written by mid-levels accounting for nearly a tenth of all OCS prescriptions in 2018.0][31] It is unknown whether the rise in OCS prescribing is attributable to midlevels acting as generalists (e.g., in internal medicine or family practice clinics) or as specialists (e.g., in dermatology, gastroenterology, or rheumatology clinics).Additional work to clarify the impact of mid-level prescribing on increasing OCS use in the United States is needed.
5][16][17][18][19] In both the United States and Taiwan, osteoarthritis and other non-traumatic joint disorders were the most common indications for long-term OCS prescriptions.This categorization includes rheumatic diseases, such as inflammatory arthritis.In both the United States and Taiwan, rheumatologists were the specialists most commonly prescribing long-term OCS.However, non-specialist physicians were also responsible for a considerable proportion of long-term OCS; approximately one-third of prescriptions in the United States and onefifth in Taiwan.][17][18][19] Several strengths of this study should be noted.This is the first international study using population-level claims data to compare and contrast OCS prescribing patterns over the past decade.It is one of only a few recent studies evaluating longitudinal trends across different durations of OCS use in the general population, as most previous studies used cross-sectional data to assess prevalence of only long-term use.Limitations of this work include differences in data features which prevent us from comparing certain information across countries, the notable trends observed among mid-level providers in the United States, which are not applicable to Taiwan and Denmark, demographic differences related to the use of population-level medical claims data in Taiwan and Denmark versus commercial claims data in the United States (which may not adequately capture populations eligible for government programs such as Medicare and Medicaid), and the lack of data available for mid-and low income countries, which limits generalizability. 32Although we know of no data to suggest this, it may be theoretically possible that population-level differences in OCS metabolism might influence response to treatment and, ultimately, prescribing patterns.This international population-based longitudinal cohort study provides real-world evidence of an ongoing steady increase in OCS use in the general populations of the United States, Taiwan, and Denmark.The rise seen over the past decade was largely driven by short-term OCS prescribing to healthy adults.This study lends further support to call for a model of "corticosteroid stewardship" when prescribing OCS, particularly as in regard to the judicious use of OCS to treat healthy people for self-limited conditions.It is important to implement effective strategies to prevent avoidable harms caused by OCS use.
to evaluate OCS prescribing.All information used in this study was derived from de-identified medical and pharmacy claims data in United States, Taiwan, and Denmark.Briefly, Optum in the United States is a commercial insurance program containing data for ~87 million individuals between 2001 and 2020, with 15-20 million active members annually.The National Health Insurance Research Database in Taiwan is a single-payer mandatory enrollment insurance program containing data for ~23 million individuals (more than 99% of the Taiwanese population) between 2009 and 2018.The National Prescription and Patient Registries/ Danish National Patient Registry in Denmark is a nationwide hospital registry system containing data for ~5.5 million individuals between 2009 and 2018.This study was approved or deemed exempt by the Institutional Review Boards of the University of Michigan (United States), the National Health Research Institutes (Taiwan), and the Danish Data Protection Agency (Denmark).
Abbreviation: IQR, interquartile range; OCS, oral corticosteroid; SD, standard deviation.a Among patients who were prescribed corticosteroids in a given year.b Reported in prednisone equivalents.

F I G U R E 1
Ten-year trend on prevalence of overall oral corticosteroid use in USA, Taiwan, and Denmark.

F I G U R E 2
Ten-year trend on prevalence of overall oral corticosteroid use in the United States, Taiwan, and Denmark, stratified by age (a-c) and stratified by sex (d-f).

2
Top 10 indications of overall, short-, medium-and long-term oral corticosteroid use in the United States, Taiwan, and Denmark.
Denominator includes all oral corticosteroid prescriptions dispensed during the study period, regardless of prescribing provider type. a

USA Taiwan Top 5 physician specialties % Top 5 physician specialties %
Top five physician specialties of overall, short-, medium-and long-term oral corticosteroid use in the United States and Taiwan.
T A B L E 3