A distributed, scalable, community care network architecture for wide-area electronic patient records: modeling and...

Principal systems issues relative to computerizing patient medical records that are yet to be addressed in the scientific literature include (1) the characteristics of networks, i.e. bandwidth and capacity, and their impact on the performance of the system, (2) the architecture and the underlying algorithm of the system, (3) the location and migration of medical records, (4) scalability of the system, and (5) the nature of the performance variation under heavy and light use of the network. Key parameters that affect performance include the number of patients, doctors, frequency of patient visits, and the number of electronic queries and record entries initiated during a patient-doctor interaction episode. This paper presents AMPReD, a Distributed, Scalable, Community Care Network Architecture that aims to provide Real-Time Access to Geographically-Dispersed Patient Medical Records. The AMPReD model includes stationary hospitals and medical clinics, mobile clinics, migrating doctors as well as patients, the communications network, and the patient medical record database. AMPReD's goals include (1) the accurate modeling of the propagation of medical records and (2) providing real-time access to patient medical records from anywhere in the system. To achieve these goals, an asynchronous, distributed algorithm must be developed that achieves concurrent access of multiple, autonomous databases. AMPReD is modeled and simulated for a representative community care network on a network of workstations configured as a loosely-coupled parallel processor, for different parametric combinations of number of doctors, patients, and number of queries or record entries generated corresponding to every patient-doctor interaction episode. AMPReD defines and obtains key performance measures including the idle times of the doctors, patient waiting times, the access times of queries as functions of their sizes, and the growth of the databases. In addition, AMPReD also measures the deviation of the actual time required for a patient-doctor interaction episode from the scheduled interaction interval, as a function of the network load. For the representative system selected, performance measures indicate that the network, utilizing 1/2T1 links, and the database system poses no bottleneck to the system even where the number of doctors and patients within a 30 minute interval are chosen at 192 and 200 respectively. A T1 is a standard, digital, transmission link that is rated at 1.44Mbits/sec.