If you haven’t yet got sick of this topic (sorry it has been very hot for me nowadays!), here it is:
Better utilisation of health information is a necessity for delivering on the pressing requirements of modern clinical practice to provide the best available medical care for individuals, yet equitable and sustainable for the society over time. While the ultimate aim is to have the longitudinal and lifelong electronic health record that is accessible whenever and wherever needed, because it doesn’t exist, clinical registries are established to collect information about individuals in areas where improvement in practice is of high importance. These capture observations, diagnoses, procedures, clinical processes and most importantly outcomes and for example may include patients treated with a particular drug, device or surgical procedure (e.g. joint replacement), with a particular illness (diabetes), and utilising a specific healthcare resource (e.g. treated in ICU).
The success of a registry depends on how well and to what extent the collected information is able to answer the original questions. Therefore it is critical to have the right questions from the outset on which the definition of the final dataset will depend. The dataset definition process is usually the one most registry projects struggle and is one of compromise: the right balance between ambition and what data are relevant and feasible to collect. That requires good clinical communication among other things like the effective use of collaborative methods and tools. It is not uncommon that some or all of dataset items be defined previously by another project or by a national standard or meta-data registry – so reinvention should be avoided.
An often overlooked critical factor is to ensure that the registry users are willing to enter data – structured and in good quality. Increasingly health information is collected by different systems, more and more as part of the care delivery process, and healthcare professionals and patients don’t want to enter the same data more than once. Therefore it is imperative to look at integrating registries with such systems before attempting to deploy a solution. Over the years this has been consistently demonstrated, the PREDICT system in New Zealand which is used to capture CVD and diabetes data is a good example. This is essentially a clinical decision support tool for GPs but as a by-product a registry database is established. This is now being deployed in the secondary care for establishing the ACS/PCI cardiac registry.
The technical solution to the registry should be usable and offer as much data validation checks as possible as it is easiest to fix issues with data collection at the time and point of entry. The security of the application, in most cases web based, should comply with current laws and regulations, and in particular should warrant the maximum protection of individuals’ privacy.
Last but not the least, ethics and privacy issues should be thought well in advance and not done in a reactive fashion. The models of individual participation vary, for example from individual patient consent to opt-out type and to mandatory involvement. One thing that is crystal clear, though, opt-in mechanism results in poor data and should be avoided if possible. National or regional level ethics committee approvals seem to be a reasonable way to go with opt-out mechanisms. Like the opt-in mechanism, a model that involves agreement with each provider separately might prove difficult.