Low agreement between Swedish national registers and parental questionnaires on allergic rhinitis

Allergic rhinitis (AR) has been well documented using questionnaire‐based studies. Here, we examine the agreement between parental‐reported data during childhood with the emphasis on 12‐year‐olds and data from two national Swedish registers to determine whether register data on AR can supplement or replace questionnaire data.


| INTRODUC TI ON
Allergic rhinitis (AR) is one of the most common diseases globally, with an increasing prevalence in most countries and an estimated 500 million affected people. 1,2Epidemiological studies based on wellvalidated questionnaires have expanded our understanding of the prevalence, risk factors, and disease progression of AR. [3][4][5] At the same time, large-scale register-based studies may potentially generate useful disease-related data.However, patients with the disease have to be accurately identified 6 and the importance of misclassification when using registers has to be considered. 7Sweden offers a unique opportunity for register-based research, as a number of national healthcare registers are available.Two examples are the Swedish Prescribed Drug Register (SPDR) 8 and the National Patient Register (NPR) 9 that contain population-based data on all dispensed drugs and information on all inpatient care and the hospital-based outpatient visits, respectively.
3][14][15] Using the Swedish personal identity numbers, questionnaire data from a large, Swedish birth cohort could be compared with SPDR and NPR data over time.This type of comparison is of special interest, as it is currently unclear whether register data on AR are able to supplement or replace questionnaire data.
In the present study, we assessed the agreement between our previous questionnaire data on AR for 12-year-olds 16 and register data on dispensed medication and outpatient visits for AR.Furthermore, we compared SPDR data on dispensed medication and outpatient visits according to the NPR.A secondary endpoint was to study the pattern of dispensed medication and outpatient visits for AR during childhood.

| Study population
7][18] The study population originated from a randomly selected 50% sample of the children born in western Sweden in 2003.In total, 8176 families were approached and 5654 entered the study, participation rate 69.2%.Postal questionnaires were completed by the parents when the children were 6 months and one, four, eight, and 12 years of age.Ninety per cent of the eight-year-old participants were also included in the cohort comprising 12-year-olds (n = 3637/4051), corresponding to a response rate 64.3% (n = 3637/5654) of the families that entered at study start.
The study was approved by the Ethics Committee at the University of Gothenburg (Dnr 846-14), and informed consent was obtained.

| Study design and data sources
Personal identity numbers for 3634 of the 3637 participants were confirmed and used to retrieve information from the SPDR 8 and NPR, 9 as well as to link the questionnaire and register data.An automatic information collection process from pharmacies provides almost 100% of related data on prescribed and dispensed medication to the SPDR, namely the Anatomical Therapeutic Chemical (ATC) codes, date, age, and gender of each patient.However, data on overthe-counter medicines and drugs administered at hospital settings (inpatient or outpatient care) are not included. 8,19The SPDR was established in 2005 when the patients in our cohort were 2 years old.
We were thus able to include SPDR data between 2005 and 2016 accounting for ages 2-13.
The NPR was established in 1964. 9,10Since then, it has gradually expanded and from 1987 the NPR includes all inpatient care.Hospital-based outpatient visits have been recorded since 2001, but primary healthcare has not been included yet.Information on the International Classification of Diseases, 10th revision (ICD-10) codes from all reported physician visits, patient and geographical data as well as date on admission/discharge can be found in the NPR.The underreporting for inpatient data is low, but for outpatient data it is higher, especially for private healthcare and considering the lack of primary healthcare data.The coverage of all included units is generally good, but the exact coverage rate is unknown. 20We analyzed NPR data between 2003 and 2015, accounting for birth until the age of 12.

| Definitions used for the register data
The following definitions of AR medication were based on the SPDR and refer to at least one prescribed medication of the type specified, dispensed between two and 13 years of age.Each medication was covered by specific ATC codes, that is, oral antihistamines (R06AA04, R06AE07, R06AX13, R06AX22, R06AX26, or R06AX27), local antihistamines (S01GX01, S01GX02, S01GX04, S01GX06, S01GX08, S01GX09 [antihistamine eye drops] or R01AC01, R01AC02 [nasal antihistamines]), and nasal corticosteroids (R01AD05, R01AD08, R01AD09, R01AD11, R01AD12, or R01AD58)."Specific allergic rhinitis medication" referred to cases where the SPDR showed at least one dispensation of local antihistamines (as a monotherapy or in combination with oral antihistamines) or nasal corticosteroids (as a monotherapy or in combination with any form of antihistamines),

Key message
This study showed poor agreement between questionnaire and national register data on allergic rhinitis diagnosis and treatment.National registers were found to be incomplete, as they did not provide data on either over-thecounter medication or children treated for allergic rhinitis in primary care.This level of disagreement supports the use of data from well-validated questionnaires to provide an insight into allergic rhinitis.
excluding monotherapy with oral antihistamines.Monotherapy with oral antihistamines was excluded because oral antihistamines can also be dispensed for conditions other than AR, that is, urticaria and food allergy.When referring to "specific allergic rhinitis medication" at 12 years, a wider age span of 11-13 years was chosen to include all 12-year-olds treated for AR, considering that time intervals of up to a year may be seen between prescription dates.

| Definitions used for the questionnaire data
The questionnaire data were used to define "doctor-diagnosed allergic rhinitis" at four (questionnaire-reported medical diagnosis [ever]   plus any symptoms of AR during the last 12 months) and eight or 12 years of age (questionnaire-reported medical diagnosis [ever] plus any symptoms of AR or medication use during the last 12 months).
"Current allergic rhinitis" at 4 years of age was defined as any reported symptoms of AR after the first year of life [ever] plus any symptoms of AR during the last 12 months.Regarding the age of eight and 12, "current allergic rhinitis" was defined as any reported symptoms of AR and medication use during the last 12 months.The questions used to generate these data are shown in Appendix S1.
Parental-reported data on allergy testing (skin prick test and/or specific IgE) were received by the questionnaires at four, eight, and 12 years of age.

| Statistical analysis
The statistical analyses were performed using the IBM SPSS Statistics version 26.0 (IBM Corp, Armonk, NY, USA) with frequencies and cross-tabulations.To estimate the agreement between the register and questionnaire data, we calculated the Cohen's kappa (κ) 21 with 95% confidence intervals (CI).Neither register data nor questionnaires were used as the gold standard in our calculations.
Cohen's kappa and its standard error (SEκ) were estimated using the SPSS kappa value calculator.The 95% CI for Cohen's kappa was calculated using the formula: κ − 1.96 × SEκ to κ + 1.96 × SEκ, where 1.96 is the constant for 95% CI.A kappa over 0.80 was characterized as almost perfect agreement, 0.61-0.80 as substantial, 0.41-0.60 as moderate, 0.21-0.40 as fair, and 0.01-0.20 as slight agreement.The sensitivity and specificity of register compared with questionnaire data were calculated as well.

| RE SULTS
The number of cases and the prevalence of studied variables from the questionnaire, the SPDR and the NPR at four, eight, and 12 years are illustrated in Appendix S2.

| Pattern of dispensed AR medication in the SPDR
To investigate the pattern of dispensed AR medication, SPDR data for 3634 children in western Sweden were analyzed.Approximately 19.5% (n = 709/3634) were dispensed specific AR medication F I G U R E 1 For the 3634 children that responded to the 12 years questionnaire, allergic rhinitis medication was dispensed between two and 13 years of age.Data on oral antihistamines as monotherapy and specific allergic rhinitis medication (excluding oral antihistamines as monotherapy) were obtained from the Swedish Prescribed Drug Register (SPDR).The orange and red dots represent the percentage of current allergic rhinitis and doctor-diagnosed allergic rhinitis, respectively, as reported in the four-, eight-and 12 years questionnaire.

| Comparison between the SPDR and the questionnaire
According to the SPDR, 9   1).The agreement between the NPR data and AR medication according to the SPDR was also slight (data not shown).The sensitivity and specificity of NPR compared with questionnaire data are demonstrated in Table 1.

| DISCUSS ION
In this study, we investigated the agreement between questionnaire and register data on outpatient diagnoses of AR and dispensed AR medication, as well as the pattern of both AR diagnosis and its medical treatment during childhood according to two Swedish national registers.Overall, there was poor agreement between questionnaire and register data on AR and dispensed medication.
According to the 12 years questionnaire, the prevalence of parent-reported current AR was 22%, while that of AR medication during the last 12 months was 23%. 16These data were in line with and Children (ALSPAC) at 11 years (18%). 3,22Other studies reported an even higher prevalence of around 30%. 22 The parent-reported prevalence of doctor-diagnosed AR was lower (9.7%)than current AR, 16 as many children in Sweden use over-the-counter medicines without a doctor's diagnosis. 23e present study showed that the prevalence of AR in the outpatient NPR up to 12 years of age was 3.8%, in line with a previous register-based study (2.11%, age range of 0 months to 13 years). 24e prevalence of AR in the outpatient NPR at 11-12 years of age was 1.8%, diverging significantly from the aforementioned questionnaire data.In addition, the overall agreement between questionnaire and NPR data on outpatient AR was found to be poor.These findings indicate that a considerable percentage of those with an AR diagnosis were missing from the register data, probably explained by the fact that data on diseases diagnosed and treated by a general practitioner are not included in national registers. 9In Sweden, the majority of AR patients, especially milder forms, are investigated and treated by primary healthcare, 25 leading to an underestimation of AR prevalence when national register records are used as reference.Furthermore, patients with milder AR are more likely to buy over-the-counter medication, 25 thus lacking a register-documented diagnosis.
Prescription-based algorithms have been commonly used to identify AR patients in studies from other countries. 12,13,26,27A Danish study found good validity when using register-based information on disease-specific dispensed medication and hospital visits, establishing a criterion of the repeated use of disease-specific medication within 12 months to identify children with allergic rhinoconjunctivitis. 12 The comparison with the register-based algorithms was made by using confirmative answers to questions about physician-diagnosed atopic disease as the gold standard.On the contrary, a Dutch study concluded that the adoption of prescription-derived data for the identification of children with an AR diagnosis was questionable, as sufficiently positive predictive values were only yielded in combination with low sensitivity values. 13In this study by Mulder et al, records of 7439 children, aged 0-10 years, were retrieved by a general practitioners database.
The registered diagnoses were used as gold standards to compute the sensitivity and positive predictive value of 22 medication proxies for identifying the children diagnosed with asthma, atopic dermatitis, or AR.
In our study, the overall agreement between the questionnaire data on AR medication use during the last 12 months at 12 years of age and SPDR data on AR medication at 11-13 years of age was slight.
In Sweden, the majority of drugs used for AR treatment have been available over-the-counter at pharmacies since 1992 and at supermarkets since 2009. 6As a result, AR patients with mild-to-moderate symptoms do not always need to get a prescription filled by their physician to receive treatment, which leads to the SPDR underestimating use.On the contrary, the agreement between the parent-reported use of AR medication and the SPDR might be higher for patients with a moderate to severe form of AR, necessitating physician treatment.Another issue with the SPDR is the lack of information on the underlying diagnosis for each prescribed medication. 8,19art from AR, oral antihistamine monotherapy is also indicated for other conditions such as urticaria and food allergy.In addition, apart from being the first-line treatment for AR, nasal corticosteroids are also prescribed for other diseases such as adenoid hypertrophy and nonallergic rhinitis.As a result, relying on the SPDR to identify AR trends may be misleading.
The data in this study indicate that national registers underestimate the prevalence of AR diagnosis and the use of AR medication.This is in contrast to an earlier study, where data from the same cohort showed good agreement between questionnaire data on asthma medication and the SPDR. 28On the contrary, in that study, the NPR provided incomplete information on asthma diagnoses, most likely because visits to primary care were not included. 28The value of highly validated national registers is undeniable, and they can provide reliable data for many but not all diagnoses. 10Register data on AR can supplement questionnaire data.It might not be a good source for assessing prevalence but can enrich questionnaire data with information on hospital-related main/secondary diagnoses, procedures, and trends of AR medication that is prescribed and dispensed.The specificity of register data on AR was high compared to that of questionnaire data, but at the same time the registers had a low sensitivity.
The main strengths of our study include the large sample size of the participating population, the high response rates up to 12 years of age, and the combination of questionnaire-derived and national register-based data.To further strengthen the validity of our calculations, we designed our questionnaires based on previous major pediatric studies, namely the BAMSE 29 and the International Study of Asthma and Allergies in Childhood (ISAAC) studies. 30As reported earlier, by nonresponders analysis, the questionnaire-derived material appears to be largely representative of the population.Responders were more health conscious and well educated than nonresponders.Atopic heredity was slightly higher among responders, which could lead to overestimation of the prevalence of AR.However, no difference was seen regarding early manifestations of allergic disease in infancy, which is associated with allergic disease such as AR later during childhood. 16This study was limited by the lack of specific information on the exact diagnosis for which the dispensed medication was prescribed and the extent to which it was used.Finally, the SPDR was only established in 2005, 2 years into our study.
In conclusion, we found poor agreement between our questionnaire data and national register data on AR diagnosis and treatment.
National register-derived data were deemed incomplete, as they did not include over-the-counter medication and children treated in primary care.This level of disagreement supports our use of questionnaire data to provide an insight into the prevalence, disease progression, and treatment of AR.

3 . 2 |
rhinitis medication Age (years) Reported current allergic rhinitis from questionnaire Reported doctor-diagnosed allergic rhinitis from questionnaire Allergic rhinitis medication (excluded oral antihistamines as monotherapy) Oral antihistamines as monotherapy (including local antihistamines as a monotherapy or in combination with oral antihistamines or, alternatively, nasal corticosteroids as a monotherapy or in combination with any form of antihistamines) from two to 13 years of age, which increased to 45.3% (n = 1648/3634) if monotherapy with oral antihistamines was included as well.From those that have dispensed nasal corticosteroids at some time from 2 to 13 years of age, 29.9% (n = 165/552) used them as monotherapy (not combined with any form of antihistamines).Up to 8 years of age, oral antihistamines as monotherapy were more common for the children compared with specific AR medication, whereas a switch was noted in favor of specific AR medication from nine to 13 years of age (Figure 1).For children dispensed specific AR medication, local antihistamines were more common among younger children (2-4 years old) as opposed to nasal corticosteroids among older children (5-13 years old, Figure 2).Of the children that were dispensed specific medication for AR at any time from two to 13 years, 77.9% (n = 552/709) collected nasal corticosteroids at least once, 16.5% (n = 117/709) oral antihistamines with local antihistamines, and 5.6% (n = 40/709) local antihistamines as a monotherapy.Pattern of outpatient and inpatient care in the NPR From 0 to 12 years of age, 3.8% (n = 138/3634) received a diagnosis of AR (main or secondary diagnosis) at least once as an outpatient in the NPR.Outpatient visits for AR clearly increased from 8 years of age (Figure 3).Regarding inpatient care in the NPR, there were no cases of allergic rhinitis.
national and international studies of AR prevalence, such as the Children Allergy Milieu Stockholm Epidemiology (BAMSE) study at 12 years of age (22%) and the Avon Longitudinal Study of Parents

F I G U R E 4
Euler diagram showing the observed combinations of children who reported allergic rhinitis (AR) medication use (questionnaire) and/or specific dispensed AR medication (SPDR, excluding oral antihistamines as monotherapy) among the 3634 children that responded to the questionnaire.The prevalence of AR medication according to the questionnaire at the age of 12 is illustrated by the pink circle and the dispensed AR medication in the SPDR at 11-12 years of age is illustrated by the blue circle.The purple area in the middle represents children that both reported AR medication use and had dispensed AR medication in the SPDR.In total, 2649 children neither reported nor had dispensed AR medication according to the questionnaire and the SPDR, respectively.The areas are proportional to the number of children.
Agreement between the questionnaire-and register-based data on allergic rhinitis medication (SPDR) as well as allergic rhinitis diagnosis (NPR) at 12 years of age.Sensitivity and specificity of the register data were calculated in comparison with the questionnaire-based data.Cohen's kappa and its standard error (SEκ) were estimated by using the SPSS kappa value calculator.The 95% CI for Cohen's kappa was calculated using the formula: κ − 1.96 × SEκ to κ + 1.96 × SEκ where 1.96 is the constant for 95% CI.The degree of agreement according to kappa was defined as follows: >0.80 = almost perfect agreement, 0.61-0.80= substantial, 0.41-0.60= moderate, 0.21-0.40= fair and 0.01-0.20 = slight."Specific allergic rhinitis medication" referred to cases where the SPDR showed at least one dispensation of local antihistamines (as monotherapy or in combination with oral antihistamines) or nasal corticosteroids (as monotherapy or in combination with any form of antihistamines), excluding monotherapy with oral antihistamines.
At 11-12 years, 1.8% (n = 65/3634) of the children received an outpatient AR diagnosis (main or secondary) in the NPR.Of these 65 children, 58.5% (n = 38/65; p < .001)had reported doctor-diagnosed AR in the questionnaire and 66.2% (n = 43/65; p < .001)were dispensed specific AR medication according to the SPDR.Among those who reported doctor-diagnosed AR in the questionnaire, 10.9% (n = 38/349; p < .001)had received an outpatient AR diagnosis in the NPR.Of those who were dispensed specific AR medication according to the SPDR, 11.9% (n = 43/360; p < .001) received an outpatient AR diagnosis in the NPR.The overall agreement between the questionnaire data on doctor-diagnosed AR and the NPR data on AR diagnosis was slight (kappa [95% CI]: 0.16 [0.11-0.21];Table