To the Editor: There is growing international awareness about the realities and difficulties of obtaining written informed consent from patients to compile and use data recorded in population-based registries. In some cases, such data could be of substantial benefit to the community, while posing relatively low risk of infringing patients’ privacy.1-3

The Canadian Stroke Registry achieved only a 39% participation rate from 4285 eligible patients when applying a policy of written informed consent to their data collection, resulting in significant representation bias.1 The issues and consequences are mirrored in our own endeavour, since 1999, to maintain a prospective register of people diagnosed with autism spectrum disorders in Western Australia.4

As protocol, diagnosing clinicians in Western Australia ask parents for written consent to allow their autistic child’s name, date of birth and postcode to be forwarded to the register. However, for many reasons, parents are not always asked and diagnostic information is not always forwarded. Missed cases are collected annually from the major diagnostic and service provision centres, but only the sex and year of birth are recorded for the child.

Between January 1999 and December 2002, 757 new cases of autism were registered in WA, but only 35% of parents gave written consent to include identifying details. Forty-three per cent of the 757 cases were identified by cross-referencing to diagnostic centres. When clinicians remember to notify new cases to the register, 56% of parents give consent, either at the time of diagnosis, or by sending the forms directly to the register, indicating that clinicians are instrumental in encouraging patient participation. Sometimes clinicians send anonymous diagnostic information without giving parents information about the register, and sometimes parents forget to send permission forms to the register.

Various characteristics of the child or the child’s family have a bearing on whether consent is given. For example, consent is received for only 31% of children living in rural areas, 25% of children with no intellectual disability, 23% of children not Australian-born, and 16% of children from families whose primary language is not English.

Without consent for every case, the register cannot accurately reflect the age or geographic distribution of children with autism. If the register only collected information on consenting cases, there would be severe under-ascertainment and the output would be notably biased. Perhaps more importantly, unidentifiable records cannot be linked to other datasets. Other WA population databases include information on hospitalisations, genetic testing, genealogical links, midwife notifications, birth defects, pharmaceutical history, and people with cerebral palsy. Linkage to these datasets would enormously facilitate population-based autism research to investigate the aetiology, associations and natural progression of autism disorders. The WA autism register is an internationally unique population-based resource, but its application is limited without the inclusion of identifying information. This remains the reality at a time when the research community and affected families desperately seek information about the condition and solutions for their children.