Article Text

Original research
Healthcare costs of adverse drug reactions and potentially inappropriate prescribing in older adults: a population-based study
  1. Eirin Guldsten Robinson1,
  2. Khedidja Hedna2,3,
  3. Katja M Hakkarainen4,5,
  4. Hanna Gyllensten6
  1. 1Department of Pharmacy, University of Oslo, Oslo, Norway
  2. 2Department of Psychiatry and Neurochemistry, University of Gothenburg, Goteborg, Sweden
  3. 3Statistikkonsulterna AB, Gothenburg, Sweden
  4. 4Global Database Studies (GloDaSt), IQVIA, Mölndal, Sweden
  5. 5Epidemiology & Real-World Science, RWE Scientific Affairs, Parexel International, Gothenburg, Sweden
  6. 6Institute of Health and Care Sciences, Sahlgrenska Academy, University of Gothenburg, Goteborg, Sweden
  1. Correspondence to Professor Hanna Gyllensten; hanna.gyllensten{at}gu.se

Abstract

Objectives To describe the distribution of costs based on potentially inappropriate prescribing (PIP) and adverse drug reaction (ADR) status in terms of total direct costs and costs caused by ADRs, among older adults.

Design A retrospective cohort study was conducted among older adults, identified from a random sample of the general Swedish population. PIP was identified based on the Screening Tool of Older Persons’ Prescriptions (STOPP) criteria and ADRs were identified using the Howard criteria. Causality between PIP and ADRs was evaluated using Hallas’ criteria. Prevalence-based direct healthcare costs were calculated for the 3-month study period, including the total cost for healthcare and drugs, and the cost caused by ADRs.

Setting All care levels, including primary care, other outpatient care and inpatient care.

Participants 813 adults ≥65 years.

Primary outcome measures The prevalence and cost of PIP and ADRs.

Results Total direct cost for persons with PIP was approximately twice the total cost of those without PIP (€1958 (€1428–€2616) vs €881 (€817–€1167), p=0.0020). The costs caused by ADRs was 10 times higher among persons with PIP, compared with those without PIP (€270 (€86–€545) vs €27 (€10–€61), p=0.047). For persons with ADRs caused by PIP, total direct costs were €4646 (€2617–€7931). This group represented 8% of the study population and used 25% of the costs. The main cost driver in all studied patient groups was healthcare contacts.

Conclusions Older persons with PIP and ADRs had high healthcare costs, particularly when ADRs were caused by PIP. Since these costs appear to be substantial, the potential savings by preventing their occurrence may, to a certain degree, cover the added cost of such activities. Further studies should be undertaken to provide further evidence on the costs of PIP, ADRs and ADRs caused by PIP.

  • health economics
  • public health
  • adverse events

Data availability statement

No data are available. The microlevel data in this study are not to be made publicly available due to the sensitive nature. According to the Swedish Ethical Review Act, the Personal Data Act, and the Administrative Procedure Act, data can be made available after legal review for researchers who meet the criteria for access to this type of sensitive and confidential data. For questions about this, please contact HG (hanna.gyllensten@gu.se).

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

STRENGTHS AND LIMITATIONS OF THIS STUDY

  • This study includes direct costs of potentially inappropriate prescribing (PIP), adverse drug reactions (ADRs) and ADRs caused by PIP in the general older population.

  • This study describes full healthcare costs, deviating from the previous focus on costs of drugs.

  • The study combines data from medical records with data from four administrative registers, and the application of three sets of validated clinical criteria (Screening Tool of Older Persons' Prescriptions/STOPP, Howard’s and Hallas’ criteria).

  • A limitation of the study is that the evaluation of PIP, ADRs and costs are based on a dataset generated in 2008.

  • All costs are provided in Euros 2020 value, and supplementary materials include the costs in SEK 2008 value to provide for cost translations.

Introduction

To tackle the challenges of prescribing in older adults and reducing potentially inappropriate prescribing (PIP), explicit prescribing criteria have been developed. The Screening Tool of Older Persons' Prescriptions (STOPP) criteria are screening tools developed to improve the quality and appropriateness of prescribing by reducing potentially inappropriate medications (PIMs).1 Studies have reported that older patients prescribed PIMs had a twofold increase in odds of experiencing adverse drug reactions (ADRs),2 and that the STOPP criteria are useful in linking PIMs to preventable ADRs, in a hospital3–7 or community setting.2 8 9

The occurrence of ADRs generates high healthcare costs,10 and the presence of PIMs is associated with increased healthcare utilisation.9 Yet, several studies reporting costs of PIP in the older population solely include drug costs,3 7 11–15 although some have addressed healthcare costs.16–20 However, the relationship between PIP and costs resulting specifically from related ADRs represents a gap in the knowledge.

The Drug-Related Morbidity in Sweden (DRUMS) project has previously investigated the prevalence and economic impact of adverse drug events (ADEs), including ADRs, among the general population of Swedish adults.10 21 22 This included studying the relationship between PIP and ADRs for the 813 older adults of the cohort.2 However, a combined analysis of PIP, ADRs and costs was not undertaken. By revisiting the data from the DRUMS project, it is possible to provide a new and important knowledge and guide future studies aiming to close the identified knowledge gap. This study aims to describe the distribution of costs based on PIP and ADR status in terms of total direct costs and costs caused by ADRs among older adults. Specifically, the study compares costs between subgroups of the population with PIP vs without PIP, and the subgroup with ADRs vs the subgroup with ADRs caused by PIP.

Methods

Study design and study population

A random sample of 5025 adults in the Swedish county Östergötland was identified by Statistics Sweden from the Total Population Register, of which 813 were older adults (≥65 years)2 (figure 1). Data for the cohort were collected retrospectively from health registries and through a detailed review of medical records, including primary care, other outpatient care and inpatient care. Clinical criteria and causality criteria were applied, compiling information for the cohort about ADEs, drug use, healthcare use and costs for a 3-month period in 2008.10 21 Thereafter, PIP and their potential contribution to ADRs were evaluated for the subcohort of older adults.2

Figure 1

Study flow chart including data sources and clinical criteria employed. ADR, adverse drug reaction; ADR+, ADRs caused by PIP; ADE, adverse drug event; LISA, Longitudinal Integration Database for Health Insurance and Labour Market Studies; PIP, potentially inappropriate prescribing; STOPP, Screening Tool of Older Persons’ Prescriptions.

Data sources

The unique personal identification number was used to link microdata between registers, including sociodemographics collected from the Longitudinal Integration Database for Health Insurance and Labour Market Studies, administered by Statistics Sweden. Healthcare use during the study period was identified through the Regional Care Data Warehouse, in Östergötland County, covering all public and most private healthcare encounters in the county.23 Healthcare costs were identified from the Cost Per Patient Register from Östergötland County Council, providing information about costs divided by resource types.10 Prescribed and dispensed medications and their costs were identified from the Swedish Prescribed Drug Register,24 25 and defined as the reimbursement costs to the counties and the patient out-of-pocket cost. The study population and data collection have been previously described in detail elsewhere.2 10

Identification of PIP

A research pharmacist (KH) identified PIP using the first version of the STOPP criteria, including drug–drug and drug–disease interactions, unnecessary therapeutic duplication and drugs which can increase the risks of cognitive decline and falls in older patients.26 PIP was identified from the Swedish Prescribed Drug Register,24 and the medical records for a period of 6 months, starting 3 months prior the study period for the 813 older adults (figure 1).

Identification of ADRs and ADRs caused by PIP

In this study, ADRs were defined according to the WHO as ‘a response to a drug which is noxious and unintended, and which occurs at doses normally used in man’.27 Medical records were reviewed for suspected ADEs, including ADRs, for a period of 15 months, starting 9 months prior to the study period and ending 3 months after. A standardised data collection sheet was used by research pharmacists to extract information necessary for the assessment and evaluation. ADRs and causal relationship with used medication were independently assessed in a separate process by two expert reviewers; a clinical pharmacologist and a pharmacist, in a stepwise manner using the Howard criteria.28 ADRs assessed to have at least possible causality were considered ADRs. Ongoing ADRs with causal contribution from identified PIP2 are hereafter referred to as ADR+.

Identification of costs

Prevalence-based direct costs for healthcare and drug use during the 3-month study period were calculated, both as the total cost for all healthcare encounters and dispensed drugs, and as the cost caused by ADRs employing the resource use method.29 The ADR costs were derived from a study evaluating costs of ADEs, in mutually exclusive categories including ADRs, and their contribution to healthcare encounters.10 The reviewers were instructed that the ADE contributing most to costs should be listed first. Costs caused by ADRs were costs for identifying, monitoring and treating ADRs, derived by evaluating the association with the prevalent ADRs for each healthcare encounter, using a method developed from the Hallas’ criteria.10 29 The evaluation was as follows: a healthcare encounter was categorised as dominantly caused by ADRs if one or more ADRs were the main reason for the encounter. Further, ADRs could be partly contributing (ie, ADRs had a substantial contribution to the encounter), less important (ie, ADRs had a minor or uncertain contribution to the encounter) or not contributing (ie, other symptoms/circumstances were the main reason for the encounter). Healthcare costs caused by ADRs were the full costs of healthcare encounters dominantly caused by ADRs. For other encounters caused at least partially by ADRs (partly contributing or less important) and for drug costs resulting from ADRs, the specific costs used for diagnosing, treating and monitoring ADRs were identified.

Statistical analyses

The characteristics of the population subgroups were reported with descriptive statistics; those with and without PIP, persons with or without ADRs and persons with ADR+. The subgroups were compared using z-test of proportions.

Overall mean costs were described and compared between subgroups with or without PIP and with or without ADRs. Costs caused by ADRs were described and compared between subgroups where applicable. Costs caused by ADR+ were also described. All cost estimates were described with bias corrected 95% CIs calculated by bootstrap, to account for skewed cost data.30 Results were translated to Euro (€) in 2020 value using the Swedish healthcare inflation index (price index with quality adjusted wages for the county, including medicines),31 and the 2020 exchange rate.32

We also conducted a sensitivity analysis omitting the 12 criteria that had been excluded from the second version of the STOPP criteria published after the data collection was conducted.33

Statistical analyses were performed using STATA V.14.2. This work followed The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.34 Definition of terms used in this study is provided in online supplemental table S1.

Patient and public involvement

This project did not include patient or public involvement in developing the research questions, design, conduct, choice of outcome measures or recruitment.

Results

In total 46% (375 of all 813 older adults) had one or more PIP (table 1) and 20% (159 of 813) experienced one or more ADRs. Among them, 62 persons experienced ADR+, representing 39% of persons experiencing ADRs (62 of 159) and 17% of persons receiving PIP (62 of 375). Significantly fewer individuals in the youngest age group received a PIP, as opposed to the oldest age group, where significantly more individuals received a PIP. Polypharmacy appeared to be equally common regardless of PIP and ADR status. The use of multidose dispensing, an adherence aid widely used in the Nordic countries with machine dispensed disposable sachets individually packaged for the intended time of administration,35 was more common in the subgroup with PIP compared with the subgroup without PIP, and in the subgroup with ADR+ compared with the total subgroup with ADRs. Healthcare use, including primary care visits and hospitalisations, was more common among individuals with PIP compared with those without PIP.

Table 1

Descriptive statistics of the study population (N=813)

Table 2 shows included costs and quantities for healthcare encounters dominantly associated with ADRs and dominantly associated with ADR+. In the total population, nurse visits and other outpatient were the most frequent type of healthcare encounters, followed by home healthcare visits, physician visits and telephone contacts. A similar distribution could be observed among the encounters caused by ADRs and ADR+. Hospitalisations dominantly associated with ADR+ were about twice the cost of hospitalisations caused by ADRs regardless of PIP status, although this represented very few hospitalisations overall.

Table 2

Overview of the included quantities and costs for healthcare encounters dominantly caused by ADRs and ADR+ by types of encounters

In table 3, the direct costs of persons with or without PIP and with or without ADRs are detailed. In the total population, the total direct cost for persons with PIP was approximately double compared with those without PIP (€1958 (€1428–€2616) vs €881 (€817–€1167), p=0.002). The cost caused by ADRs was 10 times higher among the population with PIP, compared with the population without PIP (€270 (€86–€545) vs €27 (€10–€61), p=0.047). In the subgroup with ADRs, there was a tendency towards higher total costs among persons with PIP, although not statistically significant (€4084 (€2714–€6239) vs €2193 (€1527–€3028), p=0.058). For the total population, total direct costs for the subgroup with ADRs was almost four times the total cost of those without ADRs (€3501 (€2564–€5134) vs €929 (€775–€1121), p=0.0001). For persons with ADR+, total direct costs were €4646 (€2617–€7931).

Table 3

Comparing direct costs over a 3-month period among older adults with and without PIP, and with and without ADRs, respectively

Figure 2 shows the prevalence of PIP, ADRs and ADR+ in the total population and the associated distribution of total direct costs for the individuals with PIP, ADRs and ADR+. There was a disproportion between the prevalence and costs for all the subgroups. PIP occurred in 46% of the population who used 63% of the total healthcare costs. Furthermore, the 20% with ADRs used 48% of the total healthcare costs. The subgroup who experienced ADR+ represented 8% of the study population and used 25% of the healthcare costs.

Figure 2

Prevalence of PIP, ADRs and ADR+ and the associated distribution of total direct costs for individuals with PIP, ADRs and ADR+. Costs in €1000. ADR, adverse drug reaction; ADR+, ADRs caused by PIP; PIP, potentially inappropriate prescribing.

Table 4 details the costs caused by ADRs into drug costs and costs of healthcare encounters for the subgroup with ADRs and the subgroup with ADR+. Healthcare costs were clearly the largest contributor to patient costs, whereas drug costs represented a small fraction of the total cost, both for the subgroup with ADRs and the subgroup with ADR+ (table 4). Among those with ADR+, the ADR+ caused approximately half of the costs of healthcare encounters caused by ADRs (mean €448 of €912). Moreover, their ADRs were the main contributor to healthcare encounters (mean €912 of €925). For the group with any ADRs, costs of healthcare encounters caused by ADRs were €705, and ADRs contributed to some extent to other encounters costing €1750.

Table 4

Costs resulting from ADRs among older adults

Sensitivity analysis

In the sensitivity analysis without the 12 criteria that had been excluded in the newer version of the STOPP criteria, 271 persons (33%) had PIP (table 5). Total direct cost for persons with PIP was twice the total cost, compared with those without PIP (€2131 (€1475–€3005) vs €1083 (€890–€1331), p=0.0089). Costs in SEK 2008 value are provided in online supplemental table S5. Online supplemental tables S2–S5 provide all the corresponding costs in SEK 2008 value.

Table 5

Sensitivity analysis of direct costs over a 3-month period among older adults with and without PIP according to the STOPP criteria version 2

Discussion

Principal findings

This population-based study of adults ≥65 years from both primary and specialised healthcare settings adds to the previous knowledge that healthcare costs of individuals who experienced ADR+ were disproportionally high (25% of total costs) compared with the prevalence of such ADRs in the study population (8% of the 813). Individuals with PIP had higher healthcare costs than those without PIP (€1958 (€1428–€2616) vs €881 (€817–€1167), p=0.002), and the costs were mainly driven by the cost of healthcare encounters and not by the cost of drugs.

Strengths and limitations of this study

This study is the first to estimate costs caused by ADRs based on PIP status and PIP causation in the older adults of a random sample of the general population of the county Östergötland, recognised as representative of the general Swedish population.21 The main strength of this study is the combination of medical records with data from four administrative registers, and the application of three sets of validated evaluation clinical criteria (STOPP, Howard’s and Hallas’ criteria). We have, therefore, been able to explore each patient’s health and costs outcomes closely. Nevertheless, the findings should be interpreted with some limitations in mind. The evaluation of ADRs and costs was based on a dataset generated in 2008, and the first version of the STOPP criteria were used to identify PIP. The STOPP criteria have since been updated, including more criteria than the first version, and new drugs have been introduced to the market. The current study finds a PIP prevalence of 46%, which is in line with other studies using the STOPP criteria.36 37 It is unlikely that the ADR prevalence has improved, as a recent Swedish report found the ADR prevalence of 10% for the full Swedish population in the period 2013–2018,38 while the ADR prevalence in our full dataset and in this subset were 7%21 and 20%, respectively, although the difference in time frames calls for a cautious comparison. The sensitivity analysis, employing only 53 of the criteria in STOPP V.2 yielded very similar results as the main analysis in terms of total direct costs. As the costs in this study were register based, reflecting the actual costs for healthcare use translated to 2020 values, and since the main point of this analysis was to compare costs between subgroups, it is unlikely that our findings would have changed significantly in the recent years.

The ADR costs were derived from a study evaluating costs of different types of ADEs and their contribution to healthcare encounters.10 The cost evaluation method employed was the resource use method.29 The reviewers were instructed that the ADE contributing the most to costs should be listed first. If the first ADE listed by a reviewer was an ADR, any additional ADEs contributing to the resource use during that same healthcare encounter could potentially contribute to an overestimation of cost. The method used to evaluate costs caused by ADRs may result in an underestimation in the encounters where ADRs only contributed to a small extent, or were not listed as the first ADE. Moreover, several resources used for diagnosing, treating or monitoring ADEs were not possible to single out as costs in the healthcare register.29 Thus, although the total cost reported is the cost used by the Region in their administrative system and should be well representing the healthcare costs in this patient group, the reported cost resulting from specific ADRs is likely underestimated, and should be viewed as more of an indication of its distribution.29 Costs assessed as dominantly caused by an ADR can be expected to better represent the actual costs to the health system.29 The analyses in this article are descriptive in nature, and the results were not adjusted for other factors, such as comorbidities. Thus, methods for estimating the attributable costs of ADRs were not used, such as adjusted regression analyses or propensity score matching, due to residual confounding limiting the comparability of groups.

Interpretation of results

Comparing healthcare costs: different methods

Comparing healthcare costs between studies is difficult, due to varying use of methods and definitions of outcomes. Some studies only included drug costs,13 14 39 or used other PIP criteria, like the European PIM-list13 14 or Beers criteria.39 Among studies including full healthcare costs, comparison is complicated by the use of different PIP criteria. A German study found a difference in healthcare costs of €1237 the first 3 months after being prescribed a PIM, compared with those not receiving a PIM, defined by the German PRISCUS list.17 Although that study used a different PIM-list and estimated the incidence and not prevalence, its finding corresponds to our result of €1933 mean cost per patient with PIP (table 3). A recent population-based study from Canada found a healthcare cost of €528 ($C773, 2017 value32) attributable to PIP per individual with STOPP/START criteria for a 90-day follow-up.19 This may to a certain degree correspond to our results in persons with ADR+ of €451 per patient caused by ADR+. They also found that 39% of total costs of hospitalisations, emergency department visits and drugs were attributable to PIP, which was similar to our findings, in spite of the difference in methodology. The Canadian study estimated costs attributable to PIP by multiplying the total cost with a population attributable fraction, while our study reported cost data from the cost per patient register for each patient with PIP. Furthermore, they did not associate the cost to ADRs, although the resulting healthcare costs attributable to PIPs are likely to partly reflect ADR costs.

We can draw other knowledge from relating our results to other studies. Patients who experienced ADR+ had relatively high total direct costs compared with other patients with ADRs not caused by PIP or to the total patients with PIP. Several studies have found similar increase in costs due to PIP.20 However, patients with ADR+ can be expected to differ from other patients with either ADRs alone or PIP alone. They are presumably susceptible to developing an ADR due to their PIP, making it likely that there are unmeasured confounding factors influencing the comparisons. This was also deduced from a study where numerous matching criteria were used, and where PIM exposure was associated with polypharmacy and higher healthcare costs,17 which are both associated with ADRs.

Healthcare costs versus drug costs or PIM costs

Previous studies have suggested there is a potential to reduce costs by exchanging PIMs with recommended alternatives,14 or that PIMs have a higher cost than non-PIMs, so that costs can be reduced by exchanging the treatment itself,13 and that in particular drug costs are reduced after multidisciplinary medication reviews to reduce PIMs among patients in nursing homes.40 However, there is a potential to save drug costs by simply removing a drug without replacing it (deprescribing), and removing a PIM may reduce the need to treat ADRs with further medications, hence reducing costs. In our study, drug costs had only a marginal contribution to the total costs of the healthcare encounters mainly caused by ADRs. Only assessing drug costs does not give the full picture on preventable costs of negative health outcomes. Thus, we believe our results presenting the healthcare costs specifically caused by the ADRs associated with PIP use contributes with a new and important perspective.

Distribution and preventability of costs

The distribution of costs compared with the prevalence of PIP, ADRs and ADR+, shows that in particular ADR+represent a considerable burden of cost. Only 8% of the population had ADR+, but the costs for this subgroup is more than triple that level, at 25% of the cost of the total population. 5.7% of the study population ≥65 years had ADRs deemed to be preventable21 and by using STOPP criteria, it may be possible to prevent ADR+.2 However, a cost that has already incurred cannot be reversed. Hence, it is necessary to prevent the occurrence of ADR+ before the ADR occurs in order to save healthcare costs, although primary prevention as well as secondary prevention strategies would also incur a cost. The subgroup with PIP also represents a burden of cost higher than the corresponding prevalence, and so does the subgroup with ADRs. It is possible that other factors like age, polypharmacy and multidose use influences the costs of the subgroup with PIP. As this study is a snapshot of the development in the subgroups, it is also possible that some of the persons with PIP are in the process of developing ADR+, and hence may have increased healthcare costs.

Implications

This study indicates that ADRs caused by PIP generate disproportionally high healthcare costs. However, there is a need to further study which PIPs are more likely to cause ADRs and reduce or prevent their use at an early stage, before the patient develops an ADR, thus preventing negative health outcomes and associated healthcare costs. Targeting PIPs likely to cause ADRs should have implications for future prescribing practice among older adults, and should also be reflected in future prescribing guidelines. These findings warrant further study of the costs associated with PIP, ADR and ADRs causally linked to PIP.

Conclusion

The occurrence of PIPs and ADRs resulted in high healthcare costs among older adults, especially when the ADRs were caused by PIP. Healthcare use and especially hospitalisations were the main cost drivers. Since costs caused by ADRs associated with PIP appear to be substantial, the potential savings by preventing their occurrence and mitigating them may, to a certain degree, cover the added cost of such activities. Further studies on the relationship between PIP, ADRs and healthcare costs should be undertaken to provide updated evidence on the costs of PIP, ADRs and ADRs causally linked to PIP.

Data availability statement

No data are available. The microlevel data in this study are not to be made publicly available due to the sensitive nature. According to the Swedish Ethical Review Act, the Personal Data Act, and the Administrative Procedure Act, data can be made available after legal review for researchers who meet the criteria for access to this type of sensitive and confidential data. For questions about this, please contact HG (hanna.gyllensten@gu.se).

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Regional Ethical Review Board in Gothenburg, Sweden (approval reference number: 644-08).

Acknowledgments

This research is part of the project Drug-Related Morbidity in Sweden (DRUMS). We thank Karolina Andersson Sundell, Anders Carlsten, Staffan Hägg, Ingela Jakobsson, Anna K. Jönsson, Mats Klingberg, Josefina Lindstén, Ellinor Rålid, Johnny Pettersson, Max Petzold, Parshin Saadatirad, Staffan Svensson, Karin Tunér, Annika Yeiter, and Tatiana Zverkova Sandström who were involved in the study design or data collection.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors EGR: methodology; writing—original draft, review and editing; visualisation. KH: conceptualisation; investigation; writing—review and editing. KMH: investigation; writing—review and editing. HG: conceptualisation; methodology; investigation; formal analysis; writing—review and editing; guarantor of the study.

  • Funding This study was supported by the National Corporation of Swedish Pharmacies (Apoteket AB), Region Västra Götaland and Östergötland County Council.

  • Competing interests At the time of data analysis and writing of the manuscript, KMH was employed by IQVIA while at the time of finalising the manuscript, KMH is employed Parexel International. HG is employed part time by IQVIA. Both IQVIA and Parexel International are contract research organisations that performs commissioned pharmacoepidemiological studies, and therefore, are collaborating with several pharmaceutical companies. HG is part of a research group at Karolinska Institutet that is partly funded by research grants from Biogen. The other authors declare that they have no competing interests relevant to this article.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.