Document-Based Databases for Medical Information Systems and Crisis Management

Introduction

Information systems (IS) are routinely used in the field of healthcare. Running a hospital without a Hospital Information System (HIS) has become impractical and most physicians use Patient Information Systems (PIS). Electric devices such as tablet computers replace paper (Holzinger, Kosec, Schwantzer, Debevc, Hofmann-Wellenhof & Frühauf, 2011). Moreover, IS act as the interface between healthcare and patient-centered medical research. The aforementioned tablets are also used to provide patients with medical services (Abernethy, Herndon, Wheeler, Patwardhan, Shaw, Lyerly, & Weinfurt, 2008).

Also in cases of emergencies and crises, access to medical IS is vital. Almost all IS used in this context rely on a database (DB) that stores data about patients and related information such as diagnoses, treatment plans, and prescriptions. A high number of records is document-based, e.g. x-ray images, electroencephalography wave recordings, and forms with measurement values and patient details. Also data sets may be extended on ad-hoc base during emergency situations e.g. with geographical data (Montoya, 2003)

The predominant kind of databases used in IS are relational (RDBMS) ones (Codd, 1983). Despite the strong focus on documents that can be observed in this context, this also applies to medical IS. The relational concept indisputably has amenities. It is widely understood by IS developers and established in curricula (Robbert & Ricardo, 2003). However, it also has drawbacks. We will not attempt to contribute to a discussion of pros and cons of relational databases¹. We rather deem checking alternative concepts worthwhile. It is the right time to look “beyond relational databases” (Seltzer, 2005). Particularly, the document-based databases caught our interest due to the document-focus we observed in medical IS development projects we were involved in.

Document-based databases can be attributed to the NoSQL (not only SQL) databases (Lith and Mattsoon, 2010). These DBs do not only rely on the Structured Query Language, which is predominately used to access data from relational databases. The idea of using a document-centered database is not new; it has been incorporated into Lotus Notes as early as 1989 (Mohan, 1999).

We have used the document-oriented CouchDB (Anderson, Lehnardt, and Slater, 2010; Holt, 2011a; Lennon, 2009) in a project based on work of colleagues from the field of medical informatics, who have designed and developed DocFrame, a Web-based documentation framework for electronic data capture for use in clinical care (Hartz, Verst, & Ueckert, 2009) as well as patient registries (Brüntrup, Lablans, and Ückert, 2011b; König, Omran, Schmitt, Lablans, and Ückert, 2012). They have also published a generic software library as a reference implementation for compliance with rigid data protection laws (Brüntrup, Lablans, & Ückert, 2011a). The sophisticated patient registry is used in a multi-tier architecture that provides a Web interface. Its layers (e.g. business logic and data storage) are loosely coupled. Brüntrup et al. (2011b) decided to use CouchDB since they found a relational database to be inferior in concept to the document-based approach. Refraining from using an RDBMS was preceded by an elaborate analysis.

The work presented in this article allows learning from using CouchDB and judging the general feasibility of it. At the same time, we want to check the applicability for crisis management. Since CouchDB can be integrated in the database abstraction layers of almost any IS that would use a RDBMS, it might be a choice if new systems are designed. In contrast to a stationary setting, challenges such as a distributed and changing environment, unclear conditions, potential loss of personnel and a host of technical problems have to be taken into account in a setting of crisis management. Each of these problems leads to additional requirements toward the database choice.