Matthias Grabert, P. Merkle and F. Schweiggert
University of Ulm
Department of Applied Information Processing (SAI)
Helmholtzstr. 18
D-89069 Ulm
Tel.: +49 731 502 3575
Fax: +49 731 502 3579
R.W. Holl and E. Heinze
University of Ulm
Department of Pediatrics I
Prittwitzstr. 45
D-89070 Ulm
J. Högel and W. Gaus
University of Ulm
Department of Statistic and Documentation
D-89070 Ulm
Keywords: Diabetes, Chronic Disease, Information System, Quality Control
The necessity of using electronic databases in the treatment of chronic diseases is evident and has been demanded for diabetes mellitus since 1989 [1]. Three major goals can be reached with the help of such information systems: monitoring the quality of care, support for daily routine work and scientific investigations. A diabetes information system has been developed during the last five years. Using the system, summaries of patient data for the physician as well as patient reports after an outpatient or inpatient visit can be generated. Patient data are displayed in tabular and graphical form. To monitor the quality of care, statistics of several quality indicators concerning diabetes can be generated on a keystroke serving for internal or external quality control as well as quality circles. The program is currently used in about 35 German, Austrian and Swiss diabetes centers. Due to the open architecture of the information system, new tables and masks can be added to the program at each center. This supports individual investigations as well as joint studies among several diabetes centers.
Introduction
Regarding the current discussion about medicine in Germany, it is conspicious, that beside the demand for better economic efficiency of the public health system, the discussion is focussed on quality assurance. Since 1988, measures for external and internal quality control have been anchored in the German Social Statute Book [2]. This is mostly implemented through the use of indicators, tracers and their interpretation and comparison in quality circles. For efficiency reasons, the data collected should be processed by computer programs. Especially during the course of treatment of chronic diseases like insulin-dependant diabetes mellitus (IDDM), very much patient data is accumulated over the years. To manage this vast amount of data and to obtain information about the outcome of care - and to improve it - , the use of information technology in diabetology was demanded in the declaration of St.Vincente in the year 1989 [1]. Thus the necessity of information systems in diabetes care is evident.
Background and goals
The diabetes information system DPV has been developed at the University of Ulm in a team of statisticians, physicians and information scientists during the years of 1989 until 1995 [3][9]. The first step was made with a self constructed database on a UNIX system, written in C. But this program was a dead-end street. Only the data could be saved from this effort. Now we are using a Xbase-application on an IBM PC. The program was originally designed to support the outpatient and inpatient treatment of pediatric patients with IDDM and was later adapted for adults. A primary goal was to reduce the time for routine work, as writing patient reports and data summaries about one patient. Many general practitioners are never informed about the visits of their patient at a diabetes center due to the lack of time for writing patient reports at the center. Secondly we wanted a flexible but all-in-one database for the many research projects performed at the diabetes center of an university hospital. And, last but not least, we wanted to improve and to assess the quality of care for diabetes patients. These three goals implemented in one currently usable program, seem to be unique in Germany today. Some centers have implemented solutions for writing patient reports (e.g. [4]). The approach at DIABCARE is pursuing external quality control, implementing the plans of St.Vincente [5]. And the MEDWIS project, which will include a documentation and decision system among other things, will not be finished before the late nineteen-nineties [6].
Routine work
The database contains all important parameters of a patient concerning his
diabetes:
anthropometric data (weight and length), treatment modalities (e.g. number of
injections and quantity of insulin), acute and chronic complications (e.g.
hypoglycemic episodes, retinopathy, neuropathy and microalbuminuria) level of
metabolic control (e.g. blood glucose, fructosamine and HbA1c) as well as risk
parameters (e.g. blood pressure and cholesterol). Furthermore, data of diabetes
education and examinations of relatives are collected in the database.
In total, the database consists of 16 tables of patient data and 21 tables of
system control data (including hardware settings, measuring units, normal values
of healthy persons, ICD/9 code, types of insulin, classification of eye
examination etc.).
Data is not information. They have to be structured, summarized and displayed in
relation to healthy persons, if necessary.
We are generating patient summaries and graphical printouts prior to every
outpatient visit. The parameters for the printouts can be configured freely from
a predefined list.
The graphical printouts display length, weight, body-mass-index, HbA1c-level and
insulin units per kilogram. Length, weight and body-mass-index are shown on the
background of the 3rd, 50th and 97th percentile of healthy children. Standard
values can be chosen from different control studies, as for example the Zurich
Longitudinal Growth Study [7].
HbA1c-levels are adjusted for different laboratory methods, which are changing
from time to time. We are using a transformation based on mean and standard
deviation of HbA1c, taken from 100 non-diabetic persons at the laboratory.
After the outpatient or inpatient visit we are automatically generating patient
reports and treatment plans. One or two weeks after their visit, patients are
directly informed in a letter about their degree of metabolic control.
The last two points play an important part in the communication with general
pratitioners and patients.
Furthermore, it is possible to write serial letters to invite groups of patients
for diabetes education or research projects. Patients can be selected by name,
address, age, sex, diabetes duration and so on.
As a result of the DCCT study it is known, that patients with poor metabolic
control have a very high risk for late complications as e.g. blindness and
amputation [10].
Therefore patients with risk factors, dependant on e.g. levels of metabolic
control, blood pressure, body-mass-index, microalbuminuria or cholesterol can be
identified by the program, in order to target special treatment plans to this
high risk groups.
All letters printed by the system have the original letter-head of the hospital
on it, which is reproduced with the help of a scanner.
Quality monitoring
The fear to give away data for external quality control, is wide spread. On the
other hand it is hardly feasible to compare raw data of treatment outcome among
several hospitals, without detailed knowledge about interferring variables
(e.g. different patient population, different laboratory methods).
We decided to integrate quarterly statistical evaluations into our program,
stratifying for chronological age and duration of diabetes.
The following statistics are generated:
overview on percentage of patients with - risk factors body-mass-index > 90th percentile length < 10th percentile blood pressure >90th percentile HbA1c > 5 standard deviations of the mean - acute complicationshypoglycemic episodes, admission for ketoacidosis - chronic retinopathy complications neuropathy micro/macroalbuminuria - completeness of control of blood pressure and eye examination examinations HbA1c, cholesterol, urinary albumine, length and weight measured - therapy educated in diabetes number of insulin injections / day number of blood glucose level determinations / day
Research
The goal is to support retrospective as well as prospective investigations in
the context of diabetes. Therefore the parameters of the datasets are documented
in the manual. The wide-spread dBASE dataformat is used, which allows data
processing by many other commercial software systems marketed for use on PCs.
For prospective investigations it is often necessary to add new parameters to
the database. There are two features to satisfy this demand: several fields of
the main dataset can be customized by the user, and it is possible to add new
tables to the database at each center.
For this job a dBASE IV or dBASE V system for DOS is required. The masks and
tables must be created within the dBASE system, copied into the program
directory and added to the program system table. After this procedure they can
be processed identical to the other predefined tables: adding, editing and
deleting datasets is possible.
These features support also data collection in prospective multi-center studies.
A pilot example was presented at the German Diabetes Congress in 1994 [8].
Hardware and Software
Our major aim was to keep hardware requirements as low as possible. Many small
hospitals can't afford to buy expansive computer equipment for a new program. We
decided to use the IBM-compatible personal computer as the platform for the
program.
The whole system is written in dBASE V for DOS, with one exception: the routine
for the graphical printouts are using a C-program. The operating system is DOS
rather than Windows, in order to keep the RAM requirements minimal. However, a
Windows-version will be developed in 1995.
The current hardware requirements are: IBM-PC 80486/25 SX (better: 80486/33 DX
because of the integrated co-processor), 4 MB of RAM (8 MB are certainly
better), 20 MB space on the harddisk (for about 11.000 datasets) and a
HP-compatible laserprinter.
DOS 3.31 (or higher) or OS/2 as operating system. A local area network like
Novell- Netware is supported by the program. Finally dBASE IV 1.5 or higher is
needed to define additional masks and tables.
Future plans
To use the benefits of existing hospital information systems, we are planning to
integrate an interface for data exchange using the HL7 or the DICOM standard
protocol.
As mentioned above, we will move from DOS to a graphical user environment like
Microsoft Windows, to keep the time of getting used to the program as minimal as
possible.
Last but not least we want to publish an English program version. This work will
be started in spring 1995.
We thank the German Ministry of Health for financial support.
Literature