The First Decade with the Swedish Prescribed Drug Register
– A
Systematic Review of the Output in the Scientific Literature
Susanna M. Wallerstedt1, Bj€orn Wettermark2, 3and Mikael Hoffmann4 1
Department of Pharmacology, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden,2Centre for Pharmacoepidemiology, Department of Medicine, Karolinska Institutet, Solna, Sweden,3Department of Healthcare Development, Public Healthcare Services Committee,
Stockholm County Council, Stockholm, Sweden and4NEPI, Network for Pharmacoepidemiology, Link€oping University, Link€oping, Sweden (Received 3 February 2016; Accepted 18 April 2016)
Abstract: The aim of this study was to quantify and characterize the scientific output from the Swedish Prescribed Drug Register (SPDR) the first decade after its establishment. A systematic literature search was performed in Medline, EMBASE and PubMed (2005–2014). Additional publications were identified by personal knowledge, reference lists, contact with active authors and a citation search in Web of Sciences. Publications using SPDR data were included in the analysis and characterized regarding study type, presence of patient-level record linkage, target population and topic. A total of 719 publications were identified in the literature search and an additional 148 by other strategies. Three hundred and thirty-eight studies fulfilled the inclusion crite-ria. The majority were analytic (n= 166; 49.1%) or descriptive (n = 100; 29.5%). The remaining studies focused on validation (n= 20; 5.9%), health economics (n = 16; 4.7%) or miscellaneous (n = 36; 10.7%). The analytic studies investigating effects of drug exposure focused mainly on safety (n= 46) and/or effectiveness (n = 24). The first publications appeared in 2007 (n = 6), and in 2014, 90 articles using SPDR were published. Over the years, linkage with other registers using the personal identity number increased (0–88.9% of the publications). The population was often selected by age (49.7%), condition (45.0%) and/or drug (22.8%) and concerned predominantly psychiatric (29.0%) and cardiovascular (20.4%) diseases. In conclusion, this study illustrates that the establishment of a nationwide individual-based register on dispensed prescription drugs facilitates an encourag-ing development of pharmacoepidemiological research, both regardencourag-ing the number of publications and the scientific level of the analyses.
Increased availability of register data on drug treatment has
contributed to the rapid developments within
pharmacoepi-demiology [1
–3]. In the Nordic prescription databases,
estab-lished 1993 to 2006, detailed and comprehensive nationwide
information on dispensed drugs, patient and prescriber is
avail-able [4]. In Sweden, the Swedish Prescribed Drug Register
(SPDR) was expanded to include the identity of the patient on
1 July 2005 [5]. Thus, since then and for research purposes
after ethics application [6], the unique personal identity
num-ber given to all Swedish citizens has allowed linkage between
drug dispensing data and other registers [7].
The Swedish Prescribed Drug Register now contains one
decade of patient-level data on all dispensed prescription drugs
in Sweden. The first 10 calender years of SPDR include
pur-chases of 891 million prescriptions and 44,829 million defined
daily doses [8]. Each year, drug purchases by more than 6
million inhabitants are registered in SPDR [8], representing
about two-thirds of the Swedish population.
As the establishment of an individual-based register on
dis-pensed prescription drugs can be questioned on the basis of
integrity issues, it is important to balance these risks with
potential benefits, for example the contribution of the register
to scientific knowledge. The scientific output from the Nordic
prescription databases has been summarized [9]. However, that
review was based on publications up to 2010 and included
only 61 studies from SPDR. Thus, considering the
develop-ments within pharmacoepidemiology the latest years, an
update is warranted.
In this study, we aimed to quantify and characterize the
scien-tific output from SPDR from the start in 2005 up to the end of
2014, that is the first decade of patient-level data in this register.
Materials and Methods
Systematic literature searches were performed by research librarians in Medline, EMBASE and PubMed for the period of January 2005 to December 2014. The first search, including references up to 2013, was performed on 25 June 2014. An updating search on the publica-tions of 2014 was performed on 26 May 2015. A broad search strat-egy was used, with combinations of alternative wordings of ‘drug’, ‘register’ and ‘Swedish’. A detailed description of the search strategies is available in Appendix S1.
All authors screened the publications identified in the search, and included, during consensus discussions, studies published within the study period (not ePub ahead of print) where SPDR had been used, regardless of language. Additional studies, published in peer-reviewed journals, were identified by personal knowledge, screening of refer-ence lists and PubMed searches on Swedish authors known by the Author for correspondence: Susanna M. Wallerstedt, Department of
Pharmacology, Sahlgrenska Academy, University of Gothenburg, SE-413 90 Gothenburg, Sweden (fax +46 31 786 31 64, e-mail susanna. wallerstedt@pharm.gu.se).
authors to publish within pharmacoepidemiology. Eventually, first and last authors identified to have authored >3 papers with SPDR data were contacted with an inquiry to provide their SPDR publications, and Web of Sciences was searched to obtain articles which had cited the first paper describing SPDR [5].
Two authors together characterized the publications regarding study type (descriptive, analytic, validation, health economics or miscella-neous; mutually exclusive). The assessments were based on the text in the abstract. When the information therein was not sufficient, we based the assessments on the full publication.
We categorized studies as descriptive if they described drug treat-ment patterns. These studies were further classified as either basic or advanced. The former had a focus on age, sex, regional comparisons and/or time trends, while the latter focused on aspects beyond these.
We categorized studies as analytic if the design and the reporting of results indicated that the authors had expanded the scope of the analy-ses outside pure description. Analytic studies were further classified according to the major role for the SPDR data, that is if the dispensed prescription drug data were used as outcome variables or to determine exposure. The former may be considered as being analytic drug uti-lization studies. For these, we also agreed on the topic/s/ of the study. The following topics were agreed upon beforehand: patient compli-ance/persistence, prescribing quality, equity including socio-econom-ics, abuse, international comparison, evaluation of intervention and other. For analytic studies where SPDR data were used to determine exposure, we agreed as to whether they had effectiveness and/or safety in focus. If not, they were classified as other.
To be categorized as a validation study, the main focus of the study was to explicitly study data or methodology, for example to validate clinical information in registers or to validate methods for analysing register data. We categorized studies with a focus on costs, for exam-ple cost-of-illness or cost-effectiveness, as health economics studies. Studies not fitting the above mentioned types were classified as mis-cellaneous.
The studies were further characterized regarding the prevalence of record linkage using the unique personal identity number, and the reg-ister/s/ or data sources used. For major health registers handled by the National Board of Health and Welfare [10], linkage was recorded sep-arately for each register (Patient register including, e.g. hospital admis-sion and discharge dates, diagnoses and procedures [11], Cause of death register [12], Medical birth register [13], Cancer register [14]). Data retrieved from Statistics Sweden administering civic information including, for example migration [15] and socio-economic characteris-tics [16], healthcare quality registers focusing on certain conditions or procedures [17] or regional databases with detailed administrative information on healthcare consumption [18] were registered as coming from these sources. As medical records include extensive health infor-mation, linkage to this data source was recorded separately [19].
Finally, we determined if the studies included the total population in the country or if the studied population was selected according to age, condition and/or specific drugs/groups of drugs. For all studies, we recorded the therapeutic areas focused upon in the study. We also recorded if the national indicators for prescribing quality in the elderly [20] were used. To explore the dissemination of SPDR in the research society, we summarized individuals first authoring the publications as well as the journals in which the papers were published.
Statistics. Descriptive analyses were performed withSPSS(IBM SPSS Statistics forWINDOWS, Version 22.0, Armonk, NY, USA).
Results
A total of 719 publications were identified in the literature
search and an additional 148 by other strategies (fig. 1). A
total of 338 studies fulfilled the inclusion criteria and were
included in the analysis (Appendix S2). The publications were
written in English (n
= 328; 97.0%) and Swedish (n = 10;
3.0%).
During the whole study period, the majority of the
publica-tions were analytic (n
= 166; 49.1%) or descriptive (n = 100;
29.5%), whereas validation and health economics studies were
less frequent (n
= 20; 5.9% and n = 16; 4.7%, respectively)
(table 1). The majority (63.0%) of the descriptive studies were
advanced, that is focused on aspects beyond age, sex, regional
comparisons and time trends. Although no formal
classifica-tion was made concerning these studies, many concerned
pre-scribing quality (n
= 23; 36.5%). This was also the focus of
11.0% of the analytic studies using drug dispensing data for
outcome variables, but equity and patient
compliance/persis-tence were more common focuses in these studies (33.0% and
16.5%, respectively). In 26 (28.6%) analytic studies with drug
use as outcome, this parameter was used as a proxy for
dis-ease. The focus of the analytic publications studying effects of
drug exposure was primarily safety (61.3%) and/or
effective-ness (32.0%).
The first publications appeared in 2007 (n
= 6), and in
2014, 90 articles using SPDR were published (fig. 2). In
2007, all six studies were descriptive. Three studies were
cate-gorized as basic, that is had a focus on age, sex, regional
com-parisons and/or time trends, whereas the other three studies
were advanced, that is focused on aspects beyond that. In
2014, 25 studies (27.8%) were descriptive (11 basic and 14
advanced), and 44 (48.9%) studies were analytic. Among the
latter, SPDR data were used either for outcome variables
(n
= 17) or to determine exposure (n = 27).
Linkage to other registers by the personal identity number
was increasingly used during the study period (fig. 3). In
2007, none of the studies used this kind of record linkage,
and in 2014, 80 (88.9%) studies contained SPDR data linked
at the patient level with data from other sources. During the
whole decade, 262 (77.5%) studies used record linkage.
Link-age was most often performed with other major health
regis-ters handled within the National Board of Health and
Welfare (n
= 190; 56.2%), that is the agency responsible for
SPDR. Civic information from Statistics Sweden was also
frequently used for patient-level record linkage (n
= 116;
34.2%).
In 306 (90.5%) publications, the population was selected,
and in 32 (9.5%) publications, no population selection was
mentioned (table 1). In 218 (64.5%) publications, the
popula-tion was restricted by one parameter only, that is age,
condi-tion or drugs/groups of drugs. Among populacondi-tions selected
by condition, the most frequent ones were cardiovascular
dis-ease including atrial fibrillation (16.4%) and pregnancy
(11.8%). Among populations selected by dispensed drug/s/,
the most frequent ones concerned prevention for
cardiovascu-lar disease (31.2%) and psychotropic drugs (16.9%). In 16
studies (4.7%), national indicators of prescribing quality were
used.
In all, 209 different persons had authored the publications
as first name, 151 (72.2%) authors first authoring one
publica-tion only. The articles were published in 164 different
journals, most frequently the European Journal of Clinical
Pharmacology (n
= 24) and Pharmacoepidemiology and Drug
Safety (n
= 22). There were also articles published within
high-impact journals with a more general scope, such as the
British Medical Journal (n
= 8), the Journal of American
Medicine Association (n
= 3), the New England Journal of
Medicine (n
= 2) and the Lancet (n = 1).
Discussion
Our systematic review illustrates the contributions to research
upon the establishment of a national register with individual
level data on dispensed prescription drugs. Indeed, the 338
scientific studies published during the first decade after SPDR
was established illustrate the magnitude of this contribution.
Record linkage was increasingly used, and the studies targeted
various populations, covered most therapeutic areas and
included a broad range of study types and topics. Thus, SPDR
has contributed considerably to new knowledge on drug
ther-apy in clinical practice.
Interestingly, the first studies appeared as early as 2007, that
is within 2 years of the establishment of SPDR. As could be
expected, these studies were descriptive. Analytic studies
emerged in 2008, and in 2014, they constituted almost half of
the publications. The later emergence of analytic studies may
be explained by the fact that these studies require more data
and longer follow-up. Further, despite the Nordic tradition
within epidemiology [10], analytic pharmacoepidemiological
studies may require a learning period within the research
soci-ety. Indeed, register data on dispensed prescription drugs may
include challenges even for experienced epidemiologists, for
example on how to determine exposure. The increase in the
proportion of studies linking SPDR data to other data sources
may also indicate a learning period from another perspective.
Indeed, although linkage of data between registers held by
dif-ferent national agencies has been performed for many years, it
can take time to form legally acceptable routines for data
han-dling and linkage with other data sources, such as regional
registers and healthcare quality registers.
As validation of data and methodology is the basis for
research, it is reassuring to note that validation studies started
to appear in 2009 and have continued to appear every year
since then. They include for instance validation of data from
different registers, including linkage with questionnaire data,
and validation of methods to estimate outcomes. Indeed, the
unique personal identity number enabling linkage between
dif-ferent sources of information constitutes a good basis for
vali-dation. Doubtless, there is room for more validation studies
for methodological advancements [1].
Concerning topics of analytic studies where SPDR data
were used for outcome variables, it may not be surprising that
equity was often in focus, concerning for example level of
education, income and country of birth. Indeed, as health care
in Sweden is a public responsibility, financed primarily
through taxes, Swedish regulations stipulate that the healthcare
system shall provide good care on equal terms to all [21]. This
Potentially relevant publications identified in theliterature search (n = 719)
Duplicates removed (n = 201)
References identified by personal knowledge (n = 70)
Abstracts screened for inclusion criteria (n = 290)
Did not use data from SPDR (n = 100)
Publications included in the analysis (n = 338)
Conference abstracts (n = 197)
epub ahead of print (n = 31)
References provided by authors (n = 48)
References identified by Web of Sciences (n = 30)
is a challenge, and both descriptive studies illustrating
differ-ences in drug treatment between patient groups and analytic
studies focusing on underlying mechanisms may be valuable
tools to achieve the goal.
Noteworthy, prescribing quality was often the focus of both
advanced descriptive studies and analytic drug utilization
stud-ies. Further, the national indicators of prescribing quality for
the elderly were used in several publications. In this context,
it may be of interest to point at recent validation studies.
Indeed, although often used in pharmacoepidemiological
research, polypharmacy indicators give limited information on
the quality of drug treatment [22]. However, the number of
drugs in the medication list may be a valuable proxy for
bur-den of disease [23]. Further, drug-specific indicators of
pre-scribing quality, although easy to apply on register data, have
limitations when it comes to the concurrent validity, that is,
how well they correlate with a gold standard for drug
treat-ment quality [24,25]. In fact, the specificity is generally high,
but the sensitivity is low. Thus, general measures of quality of
drug treatment are hard to define, and patient-level linkage
between register data and medical records may contribute to
deeper knowledge within this topic. However, our study
shows that, so far, relatively few studies have linked data
between SPDR and medical records. Although not specifically
investigated and reported in this study, linkage between drug
dispensing data and patient-reported use of drugs may also be
a valuable strategy to learn more about the quality of drug
treatment [26].
The topic of several analytic studies with drug use as
out-come was categorized as
‘Other’. In these cases, the
before-hand set topics were not applicable. An explanation for this is
that our search strategy identified publications where data
from SPDR had been used for other purposes than traditional
pharmacoepidemiology.
Indeed,
our
results
indicate
that
knowledge on the existence of SPDR is quite well spread as
most therapeutic areas were represented in this review, and the
publications were first-authored by many different researchers.
These findings illustrate a successful dissemination within the
academic society of the availability of SPDR data for research
purposes.
Pharmacoepidemiological studies are generally subject to
three sources of bias; information bias, selection bias and
con-founding [27]. There may theoretically be a lower risk of
information bias when SPDR is used as it covers all citizens
from birth to death. Furthermore, the potential of patient-level
linkage, retrieving other relevant data, facilitates the use of
propensity scores and other matching algorithms to minimize
confounding [28,29]. However, access to high-quality data
does not ensure that researchers use appropriate methods to
address methodological challenges. The mix of high- and
low-ranked journals publishing studies using SPDR data in the
pre-sent study may indicate that quality is divergent.
The most important strength of this review is that it gives a
comprehensive picture of the scientific output from an
individ-ual-based register on dispensed prescription drugs during the
first 10 calendar years after its establishment. Although some
publications may have been missed, the fact that we have used
Table 1.
Characteristics of publications based on data from the Swedish Pre-scribed Drug Register in 2005 to 2014.
Publications n (%) Study type1
Validation study 20 (5.9)
Basic descriptive study2 37 (10.9)
Advanced descriptive study3 63 (18.6)
Analytic study, drug use as outcome4
Total 91 (26.9)
Equity5 30 (33.0)
Patient compliance/ persistence5 15 (16.5) Prescribing quality5 10 (11.0)
Evaluation intervention5 7 (7.7) International comparison5 2 (2.2)
Abuse5 1 (1.1)
Other5 41 (45.1)
Analytic study, drug use as exposure
Total 75 (22.2) Safety5 46 (61.3) Effectiveness5 24 (32.0) Other5 8 (10.7) Health economics 16 (4.7) Miscellaneous 36 (10.7) Studied population Age specific Total 168 (49.7) Children 37 (10.9) Elderly 34 (10.1) Condition specific Total 152 (45.0) CVD incl. AF 25 (7.4) Pregnant 18 (5.3) Coeliac disease 15 (4.4) Diabetes 14 (4.1) Cancer 10 (3.0) Drug specific Total 77 (22.8) Cardiovascular prevention 24 (7.1) Psychotropics 13 (3.8) Antidiabetic 9 (2.7) Therapeutic area Psychiatry 98 (29.0) Cardiovascular 69 (20.4) Respiratory 29 (8.6) Infectious diseases 29 (8.6) Pain 25 (7.4) Diabetes 22 (6.5) Neurology 20 (5.9) Gastro-intestinal 19 (5.6) Osteoporosis 10 (3.0) Oncology 9 (2.7) Rheumatology 8 (2.4) Urology 8 (2.4) Dermatology 7 (2.1) Gynaecology 6 (1.8)
Endocrine excl. sex hormones 3 (0.9)
Immunomodulators 1 (0.3)
Nephrology 0
AF, atrial fibrillation, CVD, cardiovascular disease.
1Mutually exclusive at the first categorization level, not the second
one.
2Focus on age, sex, regional comparisons and time trends. 3
Focus beyond age, sex, regional comparisons and time trends.
4
May also be referred to as analytic drug utilization studies.
5
several strategies to identify relevant publications strengthens
the review. Interestingly, the broad librarian search could only
identify 190 of the 339 studies (56%) included in the review.
This may be explained by the fact that no consensus has been
reached within the research society on how to refer to SPDR.
Further, the data source may not be explicitly expressed in the
abstract, in particular if the publication is not a traditional
pharmacoepidemiological study. A limitation of the study is
that the classification of studies was not always crystal clear,
and required, from time to time, intense discussions.
Neverthe-less, we reached consensus on all assessments.
Summarized, this review illustrates the contributions to
pharmacoepidemiological research upon the establishment of a
nationwide individual-based drug register with the potential of
patient-level linkage to other registers.
Acknowledgements
The authors are grateful to the research librarians Ann
Lilje-gren and Yommine Holmberg
–Hjalmarsson who performed
the literature search, and Andrejs Leimanis for valuable
contributions to the development and use of SPDR at the
National Board of Health of Welfare.
Funding
No funding was received for this study.
Ethical considerations
The study contained no sensitive personal data. Thus, the
Ethical Review Act was not applicable and no ethical approval
obtained.
Conflict of i nte re st
The authors declare that they have no conflict of interests
other than being researchers publishing within the field of
pharmacoepidemiology.
References
1 Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epi-demiol 2005;58:323–37. 0 10 20 30 40 50 60 70 80 90 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Num ber o f publ ic at io ns Other Health economics Analytic study, drug use as exposure
Analytic study, drug use as outcome
Advanced descriptive study Basic descriptive study Validation study
Fig. 2. Number of scientific studies within each study type by year.
0 20 40 60 80 100 Proporo n of public aons (% ) 2007 2008 2009 2010 2011 2012 2013 2014
2 Takahashi Y, Nishida Y, Asai S. Utilization of health care data-bases for pharmacoepidemiology. Eur J Clin Pharmacol 2012;68:123–9.
3 Ferrer P, Ballarin E, Sabate M, Laporte JR, Schoonen M, Rot-tenkolber M et al. Sources of European drug consumption data at a country level. Int J Public Health 2014;59:877–87.
4 Furu K, Wettermark B, Andersen M, Martikainen JE, Almarsdottir AB, Sorensen HT. The Nordic countries as a cohort for pharma-coepidemiological research. Basic Clin Pharmacol Toxicol 2010;106:86–94.
5 Wettermark B, Hammar N, Fored CM, Leimanis A, Otterblad Olausson P, Bergman U et al. The new Swedish Prescribed Drug Register– opportunities for pharmacoepidemiological research and experience from the first six months. Pharmacoepidemiol Drug Saf 2007;16:726–35.
6 Ludvigsson JF, Haberg SE, Knudsen GP, Lafolie P, Zoega H, Sarkkola C et al. Ethical aspects of registry-based research in the Nordic countries. Clin Epidemiol 2015;7:491–508.
7 Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, Ekbom A. The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol 2009;24:659– 67.
8 National Board of Health and Welfare. Swedish prescribed drug register. Available at https://www.socialstyrelsen.se/statistik/statis-tikdatabas/lakemedel. (last accessed on 18 August 2015).
9 Wettermark B, Zoega H, Furu K, Korhonen M, Hallas J, Norgaard M et al. The Nordic prescription databases as a resource for phar-macoepidemiological research – a literature review. Pharmacoepi-demiol Drug Saf 2013;22:691–9.
10 Rosen M. National Health Data Registers: a Nordic heritage to public health. Scand J Public Health 2002;30:81–5.
11 Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall C et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:450. 12 Johansson LA, Bjorkenstam C, Westerling R. Unexplained
differ-ences between hospital and mortality data indicated mistakes in death certification: an investigation of 1,094 deaths in Sweden dur-ing 1995. J Clin Epidemiol 2009;62:1202–9.
13 Cnattingius S, Ericson A, Gunnarskog J, Kallen B. A quality study of a medical birth registry. Scand J Soc Med 1990;18:143–8. 14 Barlow L, Westergren K, Holmberg L, Talback M. The
complete-ness of the Swedish Cancer Register: a sample survey for year 1998. Acta Oncol 2009;48:27–33.
15 Ludvigsson JF, Almqvist C, Bonamy AE, Ljung R, Michaelsson K, Neovius M et al. Registers of the Swedish total population and their use in medical research. Eur J Epidemiol 2016;31:125–36. 16 Statistics Sweden. Longitudinal integration database for health
insur-ance and labour market studies (LISA). Available at http:// www.scb.se/en_/Services/Guidance-for-researchers-and-universities/
SCB-Data/Longitudinal-integration-database-for-health-insurance-and-labour-market-studies-LISA-by-Swedish-acronym/#. (last accessed on 3 May 2016).
17 Emilsson L, Lindahl B, Koster M, Lambe M, Ludvigsson JF. Review of 103 Swedish Healthcare Quality Registries. J Intern Med 2015;277:94–136.
18 Carlsson AC, Wandell P, Osby U, Zarrinkoub R, Wettermark B, Ljunggren G. High prevalence of diagnosis of diabetes, depression, anxiety, hypertension, asthma and COPD in the total population of Stockholm, Sweden– a challenge for public health. BMC Public Health 2013;13:670.
19 Hayrinen K, Saranto K, Nykanen P. Definition, structure, content, use and impacts of electronic health records: a review of the research literature. Int J Med Inform 2008;77:291–304.
20 Fastbom J, Johnell K. National indicators for quality of drug ther-apy in older persons: the Swedish experience from the first 10 years. Drugs Aging 2015;32:189–99.
21 Anell A. The public-private pendulum–patient choice and equity in Sweden. N Engl J Med 2015;372:1–4.
22 Belfrage B, Koldestam A, Sj€oberg C, Wallerstedt SM. Number of drugs in the medication list as an indicator of prescribing quality: a validation study of polypharmacy indicators in older hip fracture patients. Eur J Clin Pharmacol 2015;71:363–8.
23 Brilleman SL, Salisbury C. Comparing measures of multimorbidity to predict outcomes in primary care: a cross sectional study. Fam Pract 2013;30:172–8.
24 Wallerstedt SM, Belfrage B, Fastbom J. Association between drug-specific indicators of prescribing quality and quality of drug treat-ment: a validation study. Pharmacoepidemiol Drug Saf 2016;24: 906–14.
25 Wallerstedt SM, Belfrage B, Koldestam A, Sj€oberg C. [Concurrent validity of Swedish drug-specific indicators of prescribing quality] L€akartidningen. 2016;113: DSU4.
26 Dahlen E, Wettermark B, Bergstrom A, Jonsson EW, Almqvist C, Kull I. Medicine use and disease control among adolescents with asthma. Eur J Clin Pharmacol 2016;72:339–47.
27 Wettermark B. The intriguing future of pharmacoepidemiology. Eur J Clin Pharmacol 2013;69(Suppl 1):43–51.
28 Schneeweiss S. Developments in post-marketing comparative effectiveness research. Clin Pharmacol Ther 2007;82:143–56. 29 Suissa S. Statistical methods in pharmacoepidemiology: advances
and challenges. Stat Methods Med Res 2009;18:3–6.
