Estimation of the prevalence of thyroid dysfunction in Catalonia through two different registries: Pharmaceutical dispensing and diagnostic registration

Summary Background Population studies on the prevalence of thyroid dysfunctions are costly. The pharmacy dispensing (PDR) and diagnosis (DR) registers allow us to study the epidemiology of these pathologies in a simpler way. Our aims: 1/Estimate the prevalence of thyroid dysfunction in Catalonia based on data from the PDR and the DR, 2/to evaluate the concordance of the results obtained by both strategies. Methods The population studied was the one registered with the public health system in Catalonia(Catsalut). In the PDR analysis, the information obtained through the Pharmaceutical Provision file (during 2012, 2013, 2014) was used regarding the number of patients under treatment (NPT) (levothyroxine and antithyroid medication). The DR analysis (2014) was performed by ICD‐9 codes (hyperthyroidism 242 and hypothyroidism 243, 244). Results According to the NPT in the PDR analysis, the prevalence of treated hypothyroidism increased over 3 years: 2.81%(2012), 2.92%(2013) and 3.07%(2014) (P < .00001). The prevalence of hyperthyroidism in treatment was 0.14%(2012), 0.13%(2013) and 0.14%(2014). According to the DR analysis in 2014, the prevalence of hypothyroidism was 2.54% and 0.35% for hyperthyroidism. The PDR analysis estimated a higher hypothyroidism prevalence compared to that estimated by the DR (P < .0001) and vice versa in the case of hyperthyroidism. Conclusion Both PDR and DR prevalence estimations of thyroid dysfunction show some degree of discordance probably due to undercoding bias in the case of DR and the absence of subclinical pathology in the case of PDR. However, both approaches are valid and complementary for estimating the prevalence of thyroid dysfunction.


| INTRODUC TI ON
Thyroid dysfunction is one of the most frequent endocrine disorders.
However, there are few studies on its prevalence due to the high cost of conducting large-scale cross-sectional epidemiological studies. In addition, there is a relatively high variability in results for epidemiological studies on thyroid diseases due to different factors such as: the population itself; the state of iodination of the population; the laboratory techniques used; or the reference values of thyroid hormones. 1 To address this issue, the EUthyroid project (Towards the elimination of iodine deficiency and related thyroid diseases in Europe), of which our study is part, included, among other objectives, the comparison of the prevalence of different thyroid pathologies among the participating countries (data not yet published). This could give a global vision of how these pathologies are distributed in Europe and in turn relate them to the state of iodine nutrition in each country.
The prevalence of clinical hyperthyroidism (hyperT) varies between 0.2% and 1.3% in iodine-sufficient areas of the world. 1 In Europe and the United States, it is similar (0.7% vs. 0.5%, respectively), while in Australia it is slightly lower (0.3%). The highest rates of hyperT occur in iodine-deficient countries, mainly due to the presence of toxic nodules in older patients. In the case of clinical hypothyroidism (hypoT) it is more prevalent and up to 10 times higher in women than in men.
In Europe it ranges between 0.2%-5.3% and in the United States between 0.3%-3.7%, depending on the population studied. 1 The use of pharmaceutical prescription records is a simple and relatively robust method for studying and monitoring the epidemiology of a specific treated pathology and also for comparing different populations. For this, the defined daily dose (DDD) can be used, which is the dose of a drug established for its main indication in non-pregnant adult subjects. In the case of thyroid hormone treatment, the World Health Organization (WHO) specifies that DDD is 150 µg, 2 a higher dose than the average dose currently consumed in Spain. 3 To study of the prevalence of hypoT, the records of treated patients (NPT) maybe more useful. Similarly, diagnostic coding records from clinical practice allow an estimate of the prevalence of different pathologies to be made. There are no published data in Spain comparing the prevalence of hyperT and hypoT based on these two methods.
Thus, the objective of our study, framed within the EUthyroid Project, was to estimate the prevalence of hypoT and hyperT de-

| MATERIAL AND ME THODS
The population sample of the study was the whole population officially registered in the public health system of Catalonia (Catsalut) For this purpose, the HO3A codes, which define levothyroxine preparations, and the HO3B, which include antithyroid medication preparations were used. The number of patients under treatment (NPT) who had withdrawn the medication was also used. The NPT was used to calculate the prevalence, as it is a more suitable parameter than the DDD, which is used more in population calculations when individual data are not available. 3 The analysis of the results was carried out by age and gender.
The second prevalence analysis was based on the diagnostic cod- in public health care system, which in Spain is practically the whole population. People who are attended in private health care system are also attended in the public health system so they can have financed access to medicines. Registers don't include hospitals inpatients, but after discharge, they are followed-up at Primary Care level, where the corresponding diagnostic is fulfilled by their general practitioner.

| Statistical analysis
Categorical variables were expressed as a percentage. To compare the prevalence obtained by separating groups by gender, by the different years studied and between those obtained from the two different registries, the Chi-square test or Fisher's exact test were used when appropriate. In all cases, P < .05 was considered to be statistically significant.

| Data based on pharmaceutical dispensing
The prevalence of thyroid disorders was estimated from the pharmaceutical prescription database (Register of the Pharmaceutical

| Data according to diagnostic record
The estimated prevalence of hypoT, according to the diagnostic records from considering codes 243 and 244 of the CMBD-AP diagnostic register in 2014, was 2.54%. The distribution was statistically different (P < .0001) by gender, with a much higher prevalence in women (0.81% vs 4.23%) The prevalence of hyperT obtained by considering code 242 on the register was 0.35%. The prevalence was higher in women (0.145% vs 0.552%; P < .0001). The prevalence of hypoT and hyperT significantly increases with age (P < .0001), with a marked increase in people over 50 years old (Figure 1 and Figure 2).

| Comparison of the two methods of estimation of thyroid disorders
When comparing the two methods used to estimate the prevalence of thyroid dysfunction in 2014, a higher prevalence of hypoT was found by using the pharmaceutical dispensing records, when compared to that estimated from the diagnostic registry (3.07% vs 2.54% P < .0001). However, when comparing the prevalence of hyperT obtained from the two methods, a markedly higher rate was found according to DR compared to that found by the pharmaceutical dispensing records (2.4% vs 0.14% P < .0001).

| D ISCUSS I ON
This is the first study carried out in Spain in which the prevalence of thyroid dysfunction was obtained and compared based on two different datasets: one from the PDR and the other from the DR.
The differences observed are due in part to the fact that the estimated prevalence based on PDR includes those cases in which 'clinical' dysfunction was detected and therefore treatment established.
When DR was the criteria, the prevalence estimation included cases of both 'clinical' and 'subclinical' dysfunction; and almost none, or F I G U R E 1 Comparison of the prevalence of hypothyroidism according to DR and PDR in 2014. In both registers, the prevalence is significantly different depending on age (P < .0001) and by gender in each age group (P < .0001). DR: Diagnosis registers, PDR: pharmacy dispensing registers very few of the subclinical cases received specific treatment. On the other hand hypoT is a chronic condition in most cases of life-long duration, while hyperT is usually a transient condition, more or less long-lasting and more or less recurrent in some cases, for which the treatment is obviously only administered during the active phases, but the diagnosis can persist actively on the register, even if the condition has been cured, or as in most cases, has ended up as hypoT.
In general, published studies regarding the epidemiology of thyroid dysfunction also show some heterogeneity in the results due to various factors including the methodology of the study performed, the population studied and the determination or not of thyroid antibodies or iodine status amongst others. 1 The cross-sectional stud-  Regarding drug consumption, our study shows a prevalence of hypoT of 2.81%-3.07% in the time period studied, which is higher than that described in a similar study conducted in a Spanish

F I G U R E 2
Comparison of the prevalence of hyperthyroidism according to DR and PDR in 2014. In both registers, the prevalence is significantly different depending on age (P < .0001) and by gender in each age group (P < .0001). DR: Diagnosis registers, PDR: pharmacy dispensing registers population (Cádiz) by Escribano Serrano et al 3 in 2014. The prevalence of hypoT in this later study using DDD was 1.24% (1.22-1.27).
They also used the DDP, defined as the true average daily dose that each patient takes when using a drug in its main indication adjusted by the DDD, and found a prevalence of hypoT of 2.39% (2.36-2.43) and when the NPT was used the prevalence of hypoT was 2.86% (2.82-2.90). The distribution by age and gender was very similar to that of our study, however they described a higher prevalence in When comparing the prevalence of thyroid dysfunction obtained by the two methods used in our study, we observed that the prevalence of hypoT estimated by means of the PDR registries was higher than that estimated by the DR registry. The most feasible explanation for this is a diagnostic undercoding of hypoT, probably more pronounced in the age groups >50 years as the prevalence of this condition is known to be higher in this age group. In contrast, in the case of hyperT, a higher prevalence was found by DR than by PDR, probably due to the lack of inclusion of subclinical hyperT cases in these registries, due to the fact that in many cases they do not receive pharmacological treatment.
Our study has some limitations and some strengths. Classification of Diseases'. In addition to these difficulties, we can add the lack of time the doctors may suffer from, which can also contribute to undercoding.
In conclusion, our data indicate that the prevalence of thyroid dysfunction found are consistent with those published so far in iodine-sufficient populations, where the prevalence of hypoT is relatively high, reaching almost 10% in women over 50 and the prevalence of hyperT is low, although it increases with age and in women.
Likewise, our study supports the use of the pharmacy dispensing and diagnosis registries for the estimation of the whole population prevalence of thyroid disorders. This is because, despite the inherent biases they present, they could be performed continuously over time and are more feasible than other methods such as cross-sectional studies, which although more precise, require a strong and costly organizational effort.

ACK N OWLED G EM ENTS
The authors appreciate the collaboration of Pere Carbonell

Puigdollers of the Pharmaceutical Benefits Division of the Catalan
Health Service (CatSalut) and Teresa Salas of the CMBD-AP Registry (Division of Demand and Activity Analysis -Area of Health Care) for its contribution to the acquisition of data.

CO N FLI C T O F I NTE R E S T
None of the authors have conflict of interest related to this article.

AUTH O R CO NTR I B UTI O N S
LV and MP initiated and coordinated the study. ST and LV concep- gencat.cat/ca/menu-ajuda/ ajuda/ avis_legal/.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from "Register of the Pharmaceutical Provision and "Register of the