Elsevier

Preventive Medicine

Volume 44, Issue 2, February 2007, Pages 174-177
Preventive Medicine

Testing anonymous link procedures for follow-up of adolescents in a school-based trial: The EU-DAP pilot study

https://doi.org/10.1016/j.ypmed.2006.07.019Get rights and content

Abstract

Objective

To study the feasibility of an anonymous coding procedure linking longitudinal information in a multi-center trial of substance abuse prevention among adolescents.

Methods

A school-based survey with re-test procedure was conducted among 485 students (mean age 13.8 years) from three countries at four study centers in order to study accuracy and repeatability of a self-generated anonymous code.

Results

Errors affected 18% of codes and 3% of all digits required for the code generation, with highest figures for two of the seven generation items. Sixty-one percent of the codes generated at the test were repeated identically at the re-test. Seventy-six percent of the codes could be linked excluding the 2 digits with the highest error rate in code generation, while 92% were linked using the best combination of the remaining seven or six digits. There was substantial variation between the centers in the results.

Conclusions

Self-generation of anonymous codes is a feasible, but not a very efficient procedure to link longitudinal data among adolescents. Easy derivation and iterative matching procedures are crucial for achieving high efficiency of this type of anonymous linkage.

Introduction

Longitudinal studies with repeated measurements require linking together data from the same subjects at different time points. This is usually accomplished by means of personal identifiers such as name, birth date or security number. Because of ethical considerations it is often not possible to keep identified computerized health records of young people without obtaining guardian's consent (Moolchan and Mermelstein, 2002). Although rules may vary between countries, consent procedures are often long, complex and may lead to a substantial dropout from the study base, with consequent loss of efficiency (Severson and Biglan, 1989). However, if record linking would still be possible without the use of personal identifiers, the guardian's consent would in many cases not be mandatory. Computerized procedures making personal data anonymous have been proposed to partially solve issues of data protection in public registers (Quantin et al., 2000), but to date there are no examples applied to field studies involving longitudinal data.

We report on a study aiming to test accuracy and stability over time of an anonymous code self-generated by adolescents, to be used in a large trial of drug prevention in schools.

Section snippets

Methods

The Ethical Boards at the respective study centers approved the protocol of the European Drug Addiction Prevention Trial (EU-DAP study), which included field test procedures.

Accuracy of the anonymous code (AC)

Of the 222 identified forms returned by the students after anonymous code generation 39 (17.6%) carried errors, including missing digits, incorrect derivations and translocations (e.g. a character was assigned to a different digit position than expected). The proportion of erroneous codes was very different between centers. An analysis by digit determined that incorrect inputs came particularly from two particular source items, which accounted for more than 40% of the total errors: own eye

Discussion

Generating a nonsense code following strict rules at different levels (logical, visual–spatial, and typesetting) can be conceived as a task involving well developed executive functions (Anderson, 2002). Indeed a rather difficult endeavor for this geographically heterogeneous sample of adolescents. The cumulative error rate affected one out of six generated codes, which explained the rather low efficiency of the link procedure as originally planned, as just over 60% of the codes could be exactly

Conclusions

Self-generation of anonymous codes is a feasible, but not a very efficient procedure to link longitudinal data among adolescents. The use of these procedures should therefore be confined to studies with major concern for anonymity rather than for study base retention. Optimized link procedures should be employed by distinguishing between “source” and “functional” code, for example the best combination of digits from the source code that still allows for one- to-one matching.

Acknowledgments

We acknowledge all teachers and students who volunteered in this study; Samuela Bighiani, Laura Vitale and Paride Angius for their contribution to the field work in Italy, Tatiana Perez for her outstanding work at EDEX, Renée Stockling (Sweden) for editing assistance and Sara Sanchez (Sweden) for language revision.

References (11)

There are more references available in the full text version of this article.

Cited by (92)

  • Matching anonymous participants in longitudinal research on sensitive topics: Challenges and recommendations

    2020, Evaluation and Program Planning
    Citation Excerpt :

    As DiIorio et al. (2000) point out, it is not realistic to anticipate a 100 % matching rate. In the current study, 88 % of surveys were matched, which is higher than other studies with ranges of 45.8%–78.1% matching rate (DiIorio et al., 2000; Galanti et al., 2007; McGloin et al., 1996). Second, the results from the analysis regressing surveys matched into the longitudinal dataset on several demographic, engagement in campus life, salience of research topic, delinquent behavior, and adherence to survey protocol variables do not provide a lot of evidence of differences between characteristics of students whose surveys matched into the longitudinal dataset compared to those whose surveys were unable to be matched.

View all citing articles on Scopus

European Public Health Programme 2002 grant # SPC 2002376. National funding: Compagnia di San Paolo, grant # 2002–0703 (Novara center); Swedish Council for Working Life and Social Research grant # 2002–0979, and Stockholm County Council, Public Health grant # LS 0401–0117 (Sweden).

1

The EU-DAP study group: Barbara Zunino, Valeria Siliquini (Piemonte Monitoring Center for Drug Abuse, Turin, Italy); Peer Van Der Kreeft, Erwin Coppens (De Sleutel, Merelbeke, Belgium); Gudrun Wiborg (IFT-Nord, Kiel, Germany); Ann-Marie Lindahl (Stockholm Center for Public Health/Tobacco Prevention, Stockholm, Sweden);Vicky Yosidi, Clive Richardson (University Mental Health Research Institute, Athens, Greece); Maro Vassara (Pyxida, Thessaloniki, Greece); Gregor Burkhart (EMCDDA, Lisbon, Portugal); Tatiana Perez (EDEX, Bilbao, Spain).

View full text