A Software Application for Survey Form Design and Processing for Scientific Use

Abstract

A form processing application (FPA) automates digitization of information contained in forms. Smaller research groups do not use FPAs as they cannot justify operation of an in-house commercial system. This paper describes the design and testing of a new FPA that is targeted toward the needs of this group, and is released as free open-source software. The new FPA covers form design, printing, scanning, and digitization. It has a flexible plug-in architecture and double-keying is used to reduce transcription error. A common content module (CCM) implements the form design based on the format-independent hierarchical content. The scan module has basic handwriting recognition and can process the input fields used by the CCM. The FPA was field-tested using data from a clinical study to compare the error rate with manual processing. A similar comparison was also made between interviewer and self-administered survey forms. The first comparison shows that the FPA with double-keying had no errors while manual transcription had three errors (0.06%) out of 4952 input fields (p=0.083). The second comparison shows that the FPA with double-keying had no errors for the interviewer-administered form (0/3681 fields, 0%), while the self-administered form had three errors (3/6096 fields, 0.05%) (p=0.178). When double-keying was not used, the error rate for tablet-type fields was not significantly different (p=0.120) between the interviewer (2/3400 fields, 0.06%) and self-administered forms (19/6000 fields, 0.32%). There was, however, a highly significant difference (p<;0.001) for handwriting-type fields between the interviewer (11/881 fields, 1.25%) and self-administered forms (75/1896 fields, 3.96%).

Keywords

• Document handling
• handwriting recognition
• image matching
• image processing
• image registration
• object detection