How to manage real time data support for registries and tissue banking in clinical practice? Cancer care case study

In the era of personalized medicine, detailed description of the disease has become of the same importance as identification of a patient. It is particularly the case of chronic illnesses, which require multidimensional scoring of their stage and risk status. Personalized medicine strongly increases the value of tissue banking because we can turn back in time searching for useful prognostic factors or performing land mark analyses of therapeutic response or survival. However, with the current dynamic arrival of many new biomarkers, corresponding research activity should address how to link the new dimensions with standard clinical data in order to utilize their incremental value to existing diagnostic methods. Moreover, targeted treatment of many chronic diseases (e.g. breast cancer, chronic leukaemia) brings a new challenge for attempts to measure the therapeutic results, as the patients can experience multiple disease-free periods during the course of the therapy. The added value of the therapeutic progress depends on whether we can link common clinical data with new molecular and genetic markers which are able to cope with multiple disease remissions or progressions in time.

Any banking of clinical material should be coupled with complete diagnostic and therapeutic identification of recorded subjects otherwise the information stored is insufficient for future views. Comprehensive electronic health records (EHR) stand in the position of key factor limiting current progress in therapeutic optimization. There are two principal ways how to manage support for personalized medicine, focus on clinical registries or automated collection of data from hospital information systems (HIS). This presentation proposes a novel approach combining both these data sources in hospital–based data mining system supporting analyses of cancer data and tissue banking.

Common HIS is composed of a number of modules and satellite systems producing large volumes of heterogeneous data. Most of the outcomes are produced to fulfill administrative duties, namely reporting for health care payers. In such administrative outcomes, the clinical information is inevitably insufficient and many clinically important views are not possible. It limits all fields of medicine, particularly the disciplines which require prognostic typology of patients, based not only on diagnosis itself. Cancer care can serve as a typical model of such information demanding field because in most of cancers, the therapeutic standards are related to diagnosis, clinical stage, grade and other prognostic markers. These attributes are however not well standardized in HIS data capture systems. Their retrospective availability can be managed only through external sources like population-based registries.

This study documents functionality of a newly developed system which is able to merge records of the Czech National Cancer Registry (CNCR) and hospital administrative data related to cancer care. A new data mining tool provides solution for this highly required data merge. The proposed system is capable to generate synthesis of different data sources and to extract information relevant for the evaluation of cancer care. Database of CNCR contains of more than 1.6 million of recorded cancer cases collected in a standardized way in past 30 years. In combination with HIS data from any given hospital, the CNCR is able to map epidemiologic load and to identify hospital case mix in this segment of medical care. The CNCR thus supports the weakest point in parametric data collection, i.e. retrospectively available, fully representative diagnostic typology of cases.

Successful implementation of the proposed system will be demonstrated over data from five cancer centers in the Czech Republic. The final merged database consists of more than 500 thousands of cancer care records containing all necessary diagnostic items and records on all therapeutic steps and corresponding outcomes. Available follow-up period exceeds 5 years in each participating hospital. The system can be easily linked to any tissue bank, registry or individually-based set of laboratory data. Comprehensive parametric structure forms a framework enabling all necessary views in the data, e.g. sorting according to demography, diagnoses, tumor morphology, or TNM classification of tumors. Routinely generated hospital records further contribute to the identification of therapeutic episodes, their content, cost and outcomes. Any examination of a biomarker is therefore associated with exactly measured disease status at a given time and its value is attributed to relevant therapeutic steps. The presentation introduces a new version of hospital reporting system which enables on-line analyses of the merged data.

Acknowledgements: Validation of the Czech National Cancer Registry and population–based monitoring of cancer disparities are supported by grant ‚Addressing Cancer Disparities in Central and Eastern Europe‘ Bristol-Myers Squibb Foundation, 2009 – 2011 (Project: National Information System for the Assessment and Communication of Cancer Care Results and Quality in the Czech Republic).