Article Text

Download PDFPDF
Informed consent for record linkage: a systematic review
  1. Márcia Elizabeth Marinho da Silva1,
  2. Cláudia Medina Coeli2,
  3. Miriam Ventura2,
  4. Marisa Palacios2,
  5. Mônica Maria Ferreira Magnanini2,
  6. Thais Medina Coeli Rochel Camargo2,
  7. Kenneth Rochel Camargo Jr3
  1. 1General Management of Information Systems, Agency for Health Plans and Insurance (ANS), Rio de Janeiro, Rio de Janeiro, Brazil
  2. 2IESC, UFRJ, Rio de Janeiro, Rio de Janeiro, Brazil
  3. 3IMS, UERJ, Rio de Janeiro, Rio de Janeiro, Brazil
  1. Correspondence to Márcia Elizabeth Marinho da Silva, General Management of Information Systems, Agency for Health Plans and Insurance (ANS), Av. Augusto Severo, 84, 10°. Andar, Glória, Rio de Janeiro, Rio de Janeiro 22021040, Brazil; marcia.marinho.silva{at}gmail.com

Abstract

Background Record linkage is a useful tool for health research. Potential benefits aside, its use raises discussions on privacy issues, such as whether a written informed consent for access to health records and linkage should be obtained. The authors aim to systematically review studies that assess consent proportions to record linkage.

Methods 8 databases were searched up to June 2011 to find articles which presented consent proportions to record linkage. The screening, eligibility and inclusion of articles were conducted by two independent reviewers. The authors carried out meta-regression, subgroup and sensitivity analyses to assess heterogeneity.

Results Of the 141 studies identified, only 11 presented empirical consent proportions and were included in the systematic review. The consent proportion varied widely from 39% to 97%. Seven studies presented consent proportions of 88% or higher, one of 72%, and only three presented consented proportion equal to or lower than 53%. None of the studies' characteristics evaluated explained heterogeneity.

Conclusion The results of this review show that, in general, individuals tend to consent to the use of their data for record linkage, with exceptions in specific populations or minorities. The authors believe that this, as well as the cited literature, lends support to policies that, while keeping relevant ethical controls in place, do not require individual informed consent for each and every study that relies on secondary data.

  • Informed consent
  • record linkage
  • public health research
  • systematic review
  • allocation of health care resources
  • bills
  • laws and cases
  • confidentiality/privacy
  • ethics committees/consultation
  • information technology
  • abortion

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.

Introduction

Record linkage techniques aim to link records believed to refer to the same individual.1 It is a powerful tool for conducting epidemiological studies in a practical and inexpensive way. Its use, however, demands access to personal identifiers, raising privacy concerns such as whether formal consent for access to health records for linkage should be obtained.2

Secondary data research uses health data that are stored without the individual's explicit consent or even awareness. At the moment the data are recorded, it is difficult to predict the type of questions they will be used to answer, especially with linked data. Large numbers of participants would have to be contacted and the contact would only be necessary for obtaining informed consent. Based on these considerations, it has been proposed that alternatives to the traditional opt in consent model, used in primary data based research, should be adopted in the context of secondary data based research.3–6

The discussion about the ethical access to secondary data is part of a wider reconsideration of the limits of the traditional informed consent model.7 Another such context in which several authors are exploring alternative consent models, for example, is biobanking.8

The 4th guideline of the Council for International Organizations of Medical Sciences6 states that the waiver of individual informed consent is to be regarded as exceptional and must in all cases be approved by an ethical review committee, unless otherwise permitted under national legislation that conforms to the ethical principles in those guidelines. Since 1995, governments of developed countries have passed legislation which requires consent for the use of health information in order to protect the privacy of identifiable health information.9 Although the privacy laws allow for a waiver of consent, in some special situations, researchers became concerned that conservative interpretations of the legislation by ethical review committees would make mandatory the request for individual consent.9 Ness,10 in a survey carried out 4 years after the implementation of the Health Insurance Portability and Accountability Act Privacy Rule, showed that American epidemiologists perceived that this legislation impacted health research negatively, increasing project costs and time.

The Brazilian context with a large number of secondary data sources, an important and growing number of studies relying on linkage techniques and a national system for supervising ethics in research established since 1996 demands a broad discussion of this subject. With varying degrees, however, the same could be said even for countries with already established rules; as the technology evolves, new challenges are inevitable. Thus, it is important to know the consent proportions for accessing private health information and record linkage when a traditional opt in consent model is used. We found a systematic review of consent and secondary data in general,11 but this issue has not been addressed specifically in the context of record linkage. Therefore, this study aims to systematically review articles that assess consent proportions to record linkage.

Methods

The systematic review was conducted in accordance with the guidelines.12 Computer-assisted search of eight electronic databases up to June 2011 was performed using the strategy described in online box 1.

The screening, eligibility and inclusion of articles were conducted by two independent reviewers. Disagreements between reviewers were resolved by a third independent reviewer. We excluded editorials, letters and reviews. The eligibility criteria were: full original articles published in Portuguese, Spanish, English and French whose aim was to assess consent proportions to record linkage or which reported consent proportions to record linkage.

The studies were classified according to publication year; country of the study; type and number of databases intended to be linked; study population features; approach used to obtain consent and who consented (subject himself/herself vs parent or care giver).

Analyses were conducted using Stata V.9.0 (http://www.stata.com). We intended to perform a meta-analysis; however, we detected a significant heterogeneity in the data (τ=0.91). We used a mixed-effects meta-regression model to specifically assess the underlying reasons for between study heterogeneity. The following covariates were tested in the meta-regression model: country, age group, population origin, approach to obtaining consent and who consented. Subgroup analysis was performed according to the five variables mentioned above. Finally, in order to identify influential studies, we performed a leave-one-study-out sensitivity analysis.

Results

Online figure 1 depicts the flowchart of this review. After removing duplicated publications, the number of abstracts to be screened was 140. One article identified scanning the reference lists of the selected articles was later included. After reviewing the abstracts, 81 studies were discarded due to language (n= 4), type of publication (n=6) and study aim not compatible (n=71). Of the 60 remaining articles, we discarded 49 because the study aim did not meet the eligibility criteria (n= 42), the full text was not accessible (n=1), the studies reported consent proportions already considered in earlier studies (n=5) and one study reported having replaced the individual consent with a confidentiality protocol approved by an ethics committee on research.

Table 1 shows the general characteristics of the 11 studies included. One article15 was published in 1996 and 10 articles were published after 2000. The studies were carried out in Canada, UK, Australia and Taiwan. The majority of the studies dealt with both adults and seniors and only one study dealt exclusively with women.22 In six articles, the study population came from health services, whereas in the five remaining the participants were enrolled in population surveys. The majority of the studies requested consent from the participant himself/herself and used a face-to-face approach. With the exception of one article,18 the studies had moderate to large study sizes.

Table 1

Summary of included studies evaluating the consent for record linkage (n=11)

The consent proportion varied widely from 39% to 97%. Seven studies presented consent proportions of 88% or higher, one of 72%, and only three presented consented proportion equal to or lower than 53% (table 1). In subgroup analysis as well as in meta-regression, the studies' characteristics did not explain heterogeneity (table 2). Sensitivity analysis did not show significant variation in the results.

Table 2

Study characteristics associated with consent proportion for record linkage

Discussion

We observed an expressive heterogeneity among the studies. The majority of them reported high consent proportions but three presented overall consent proportions equal or lower than 53%.4 ,21 ,22 The target population of the latter studies differed markedly: patients of a stroke registry;4 diabetic patients in an organised association;21 and participants in a longitudinal study of women's health.22

As a limitation of our article, differences in studies design as well as lack of power due to the small number of articles included in the review are possible reasons for our inability to explain the marked difference in proportion of consent for record linkage observed between the two groups of studies. However, a complex interplay between contextual and individual factors might also play a role. Kho et al,11 in a systematic review carried out to evaluate whether informed consent introduces selection bias in studies using medical records, found differences between individuals who gave and did not give consent with regard to age, sex, race, income, education and health status, but the direction of such differences varied among the studies and remained inconclusive.

When we consider the problems of informed consent related to large databases, such as administrative and genetic, what is at stake is a potential conflict of ethical principles: on the one hand, an absolute approach to issues of consent may hinder necessary and useful research that will contribute to the common good; on the other, lack of adequate protection may lead to breaches of subjects' privacy that can affect their lives (and those of their descendants, if we consider issues of genetic screening, for instance) in ways that are unforeseeable. These extremes can be avoided with a sensible ethical approach. There are important variations among authors who discussed this dilemma. A recent report from the Institute of Medicine23 recommended a new framework based on emphasising the requirement that all organisations that collect and use personal health information adopt a variety of procedures aiming to protect privacy including accountability, strong security measures and transparency policies with regard to the purposes for which information is used, instead of relying mainly on informed consent. Other authors defend the importance of obtaining consent, even if in modified forms. Sheehan24 makes a compelling case for the compatibility of broad consent and individual autonomy in choosing. Hansson et al 25 describe the basic requirements for the necessary safeguards to prevent ethical infringements of research subjects' rights: allowing for withdrawal of individual data from databases; providing adequate measures for privacy protection; and requiring IRB oversight of future research.

Conclusion

The results of this review show that in general individuals tend to consent to the use of their data for record linkage, with exceptions in specific populations or minorities. We believe that this, as well the cited literature, lends support to policies that, while keeping relevant ethical controls in place, do not require individual informed consent for each and every study that relies on secondary data.

References

Footnotes

  • Funding The study was supported by Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ- E-26/100.691/2007; E-26/110.465/2007) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq—(473911/2009-4). Cláudia Medina Coeli and Kenneth Rochel de Camargo were partially supported by research fellowship grants from CNPq.

  • Competing interests None.

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

Linked Articles

  • The concise argument
    Imogen Goold

Other content recommended for you