Clinical trials are characterized by highly specialized documentation needs, specifically tailored to the requirements of an individual trial. In the context of Good Clinical Practice (GCP) [11] regulations, data monitoring needs to be applied to achieve high data quality. It is necessary to validate electronic documentation systems for clinical trials. General hospital information systems are – at least at present – not capable to address all documentation needs of clinical trials [12].
However, during the evolvement process of electronic patient records more and more clinical information becomes available in HIS. In some cases – for instance laboratory values – electronic data becomes a primary data source. Our approach utilises routine HIS data – captured for administrative and clinical purposes – as an automated screening tool for patient recruitment. Even if not all items needed for inclusion and exclusion are available in the routine HIS, basic information like diagnosis, gender and age appear to be sufficient for an automated screening tool.
There are some reports in the literature regarding HIS-based support for patient recruitment. For instance, Embi [5] describes a system for a diabetes mellitus trial in an outpatient setting which resulted in a significant increase of patient enrollment. Weiner [6] analysed a real-time alerting system in the emergency department and observed improved study investigator notification. Afrin [7] reports mass screening of lab values for a lupus nephritis trial, which also highlights the issue of alert precision: 7 Mio. lab values were screened, 70 potential patients identified, only 3 were enrolled into the trial.
Many false positive alerts can be annoying for physicians and impact user acceptance. In our case, 20 of 41 alerts were related to trial patients and no AML trial patient was missed. A test period of 50 days is certainly not long enough for a valid assessment of improved recruitment, therefore we plan to do this kind of analysis after about one or two years of routine operation. However, it demonstrates a relatively high precision of notifications. We capitalised on a specific, coded diagnosis which is commonly entered into electronic systems after patient encounter. This delayed documentation is acceptable for inpatient alerts, but obviously limits real-time notification in the outpatient setting.
From our perspective, integration into standard HIS is relevant for several reasons: A direct link to the patient record is provided which facilitates eligibility verification. Since all diagnoses and suspected diagnoses are entered into the system at the day of admission (inpatient or outpatient), eligible patients can routinely be identified before start of therapy. Also, redundant data entry is avoided, because routine HIS data is re-used for trial purposes.
Data privacy regulations are covered by standard HIS access controls and protocols, so there is no need for additional authentication systems. Clinical users are familiar with the HIS which limits the need for additional training activities. When informed consent is obtained, additional orders can be placed in the HIS order-entry-system according to each trial protocol.
From a technical perspective, our approach does not need external systems and avoids firewall issues. On the other hand, it requires vendor-specific technical configuration for each trial at each hospital site and is therefore primarily suited for clinical trials with few centers.