METMBS 2001 The 2001 International Conference on Mathematics and Engineering Techniques in Medicine and Biological Sciences June 25-28, 2001 Monte Carlo Resort & Casino, Las Vegas, Nevada, USA
A Clinical Form Designer W. Gall, G. Duftschmid, E. Eigenbauer, T. Wrba, W. Dorda Department of Medical Computer Sciences University of Vienna, Austria Abstract Computer-assisted documentation of clinical research data has become standard practice nowadays. As an important prerequisite, flexible and readily adaptable clinical forms are needed that streamline workflows and facilitate data retrieval and analysis. In this communication we present the design tool for clinical research data, which forms part of the ArchiMed medical information and retrieval system. This tool is used to generate database-related forms and offers important functions such as pre-filling by derivations from other documents and automatic calculation of variables. Keywords: User interface, form design, data collection, clinical information system 1 Introduction Forms designed to capture clinical data must be user-friendly and efficient [1], and they have to meet all requirements of clinical data acquisition, which can be quite heterogeneous [2]. Support of free text entry offers a high degree of freedom in documentation, whereas structured data formats allow for better retrievability and greater uniformity of the recorded data. Certainly, a structured approach to data entry that is optimized for retrievability is very useful in clinical research. Central item dictionaries and clinical coding systems (e.g. ICD or SNOMED) provide the basis for comparing the recorded data and for data retrieval to address issues raised by clinical researcher and hospital administrators. The design of clinical research forms should be as flexible as possible, to meet the great demand for subsequent modifications [3]. Frequently, however, clinicians who wish to include a new variable into a form need to consult a database administrator, or they have to use a pocket calculator to obtain clinical scores from various recorded values. In many cases they need to formulate extensively queries in lengthy retrieval sessions. Clinical research forms tend to include both, routinely documented variables and specific study variables. Frequently, timeconsuming multiple documentation efforts are necessary and prone to error and thus liable produce false results. 2 The ArchiMed form designer 2.1 Database connection All components of the ArchiMed forms are based on a generic data model [4]. The data model includes, among other things, a directory of items (data dictionary), all forms, variables (i.e. items being represented differently on different forms), documents, and the data contained in those documents. The administration of form-related data (e.g. forms, sheets, tables) in the database, enables the user to use these components during the retrieval process. For example, to group and analyze logically related items (e.g. surgeon and date of surgery) or to distinguish multiple occurring items on the same form.
2.2 Functionality The main functional requirements were to make the form design tool user-friendly, to optimize data entry, to avoid multiple entry of identical data, and to support subsequent data retrieval and analysis. In order to meet these requirements, the following features were implemented in the tool: Various Widget types (e.g. input field, combo box, radio button) depending on the item types (alphanumeric, numeric, temporal) facilitate efficient data entry. Users can adapt the interface themselves to meet their individual requirements. To produce a form variable, user can select an item from the central item directory, or can create a new item immediately. In addition to linking variables to externally administrated catalogues, the selection lists can be edited right during the design procedure. Variables can be pre-filled by simple initialization values or by deriving them from other variables, with or without calculations performed in the process. Variables to be pre-filled can be derived from any forms (e.g. blood pressure from the last follow-up visit). A formula editor offers comprehensive functions for generating additional variables. It provides logical, arithmetic, statistical, and temporal functions to generate values. Plausibility testing is available for individual variables or for complete sheets or forms. Conditional documentation sequencing based on values already entered helps to simplify the data acquisition process. Depending on previous input, branching into individual sheets or fields, as well as skipping or repetitions, are supported. Variables or entire sheets can be dynamically enabled or disabled. They can be set to required, read only or hidden (e.g. to prevent recording pregnancy-related items for male patients). 2.3 Procedure In a typical design procedure, the user first opens a new or existing sheet of a new or existing form (see Figure 1). Using the PALETTE tool, items are selected in various forms of presentation, grouped or summarized in tables as needed, and then put onto the sheet by drag and drop. The LAYOUT tool is used to align the widgets, and the PROPERTY tool is used to set and modify the properties of the included items. The PROPERTY tool offers general properties (e.g. labels) that are available for all item types, and it offers context-specific properties that depend on item type (e.g. predefined units for numeric items) and the selected widget (e.g. values for selection lists). The PROPERTY is also used to define all complex settings such as pre-filling functions. 3 Results 3.1 Use The ArchiMed system has been used in twelve clinical departments in Vienna and Graz since 1997. During that period, more than 100 clinical research forms have been generated that have resulted in a total of around 1 million documents including 20 million individual data values. 3.2 Experience The options of pre-filling and calculating variables have turned out to be particularly powerful feature, notably in combination. The fact that variables can immediately be calculated and displayed using a formula (e.g. difference in days ) facilitates the documentation process and prepares the
basis for subsequent data retrieval and analysis. Variables can be accurately derived from older documents and pre-filled by formulating temporal conditions like "last value in three years". For example, the department of heart surgery at the University of Graz medical school uses this feature to automatically calculate the Euroscore [5] to predict mortality in cardiac surgical patients on the basis of objective risk factors. This function is implemented in the EuroScore form by pre-filling of 20 documented variables with relevant values obtained over the previous 10 years, which are then processed into the variable EuroScore. Figure 1: Defining the variable EuroScore on the Score sheet of the EuroScore form (middle) using the ArchiMed form designer tool. The sheet to be edited is selected from the main window (top left). The PALETTE tool (bottom left) is used to select variables, functions and graphic design elements. The LAYOUT tool (top right) supports the form design process via alignment, grouping and spacing functions. The PROPERTIES tool (bottom right) is used to adjust attributes to the current selected variable. 3.3 Development Platform The tool was developed using the Smalltalk-based object-oriented Visual Works environment. The relational database was implemented with Oracle. 3.4 Planned further steps One of the next improvements to the system will be to generate web-based forms. It is also planned to implement the Archimed system on another clinics.
4 Discussion The quality of form production heavily depends on the feedback already in the development phase [6]. The ArchiMed design tool allows users to quickly create and discuss forms. The forms are linked to the database, so that data documentation can be started immediately. To ensure user acceptance and efficiency, other factors must be considered in addition to the functionality of the form designer tool. For example, we recommend users start by defining a selection list with free text entry and then to retrieve all documented texts (text + number + list ID) that are not included in the first list. In the next step, the user can remove any redundant texts and automatically add the remaining ones to the selection list based on the list ID. Our experience has shown that implementing functions within the form itself (e.g. calculating variables or creating summary forms) is more efficient than subsequent formulation of complex queries. While the option of pre-filling variables with calculated values has proved to be particularly useful, these automatic features need to be handled judiciously. In automatic post-processing, for example, some groups of documents may be processed at different times. It must be guaranteed, however, that only the desired formula-based calculations are executed. There can be no doubt that the issue of free versus structured input of clinical texts will remain a busy area of research for some time to come. One of the main objectives certainly is to make interfaces userfriendly and flexible adaptable for the clinicians, which includes hiding as many technical details as possible. Acknowledgments The authors wish to thank Gerald Boldt for his valuable suggestions. References [1] E. Boralv, B. Goransson, E. Olsson, B. Sandblad. Usability and efficiency. The HELIOS approach to development of user interfaces. Comput Methods Programs Biomed, 45 Suppl: 47-64, 1994. [2] A.M. van Ginneken, H. Stam, E.M. van Mulligen, M. de Wilde, R. van Mastrigt, J.H. van Bemmel. ORCA: the versatile CPR. Methods Inf Med, 38 (4-5): 332-338, 1999. [3] S.W. Singer, C.L. Meinert. Formatindependent data collection forms. Control Clin Trials, 16(6): 363-376, 1995. [4] W. Dorda, T. Wrba, G. Duftschmid, et al. ArchiMed - A Medical Informationand Retrieval System, Methods Inf Med, 38: 16-24, 1999. [5] S.A. Nashef, F. Roques, P. Michel, E. Gauducheau, S. Lemeshow, R. Salamon. European system for cardiac opera-tive risk evaluation (EuroSCORE). Eur J Cardiothorac Surg, 16(1): 9-13, 1999. [6] V.L. Patel, A.W. Kushniruk. Interface design for health care environments: the role of cognitive science. Proc AMIA Symp, 29-37, 1998.