THREE YEARS EXPERIENCE WITH A PATIENT DATA MANAGEMENT SYSTEM AT A NEONATAL INTENSIVE CARE UNIT M. Urschitz,1 S. Lorenz,2 L. Unterasinger,1 P. Metnitz,3 K. Preyer,2 and C. Popow 1
Urschitz M, Lorenz S, Unterasinger L, Metnitz P, Preyer K, Popow C. Three years experience with a patient data management system at a neonatal intensive care unit. J Clin Monit 1998; 14: 119^125
ABSTRACT. We report about our experience with the patient data management system (PDMS) Hewlett Packard CareVue 9000 at two neonatal ICUs. We describe our PDMS con¢guration (hard- and software), local adjustments and enhancements such as knowledge based systems for calculating the parenteral nutrition of newborn infants (VIE-PNN), for advising medication (VIE-Nmed), and for managing mechanical ventilation (VIE-VENT), and the results of a structured interview with our sta¡ members about the acceptance of the system. Despite some criticism nearly all collaborators liked the system, especially because of its time saving automated documentation of vital data and mechanical ventilation parameters. More than 2/3 preferred the computer assisted documentation to charting by hand, only 1/41 would have liked to return to paper documentation. All sta¡ members possessed excellent (15/39) or good (24/39) knowledge of the system. Main points of critique were the system's therapy planning facilities (medication administration records), the restrictive facilities for documenting patient care and the yet unsolved problems of data evaluation and export. PDM systems have to be constantly adapted to the user's needs and to the changing clinical environment. Living with the system asks for an intensive dialog with the system and its functionalities, for creativity and well de¢ned ideas about the future system development. KEY WORDS. Newborn infants, intensive care, computer applications, patient data management system.
INTRODUCTION
From the 1 Neonatal Unit, Departments of Pediatrics and Adolescent Medicine, University of Vienna, Austria, 2 Hewlett Packard, Austria, and 3 Department of Anesthesiology, University of Vienna, Austria. Received Jan 7, 1997. Accepted for publication Oct 28, 1997. Address correspondence to C. Popow, Univ. Klinik fu«r Kinder- und Jugendheilkunde, A-1090 Wien, Wa«hringer Gu«rtel 18^20, Austria. E-mail:
[email protected] Journal of Clinical Monitoring and Computing 14: 119^125, 1998. ß 1998 Kluwer Academic Publishers. Printed in the Netherlands.
The enormous amount of data accumulating continuously at modern intensive care units (ICU) exerts considerable stress on the caretakers who have to perceive, generate, validate, and select the most important data. Computer assisted patient data management systems (PDMS) facilitate these tasks by automatically collecting, displaying and storing the data for further processing [1] and by improving reliability and accuracy of documentation [2, 3]. Because a neonatal ICU (NICU) could also bene¢t from these advantages [4], we adapted a PDMS which was originally designed to be used in adults (CareVue 9000, Hewlett Packard, Andover, MA) to the speci¢c needs of a NICU and integrated the system into the clinical routine. As there are only very few NICUs equiped with such a PDMS, we report about our experience with the implementation, the clinical use and the enhancements we added to the system at our two NICUs.
120
Journal of Clinical Monitoring and Computing Vol 14 No 2 February 1998
DESCRIPTION AND APPLICATION OF THE PDMS
Description of the NICUs There are two NICUs at the Perinatal Center of the Medical School, University of Vienna, an 18 bed NICU at the Department of Pediatrics, and an eight bed NICU next to the labor ward. We care for about 4000 births and admit more than 500 newborn infants per year to our two NICUs. The technical equipment of both units is equivalent, enabling all common neonatological diagnostic and therapeutic procedures including high frequency ventilation. Each of the 26 beds is equiped with a bedside PDMS workstation. The PDMS is used in identical con¢guration at both NICUs which are interconnected via a local area network (LAN).
Description of the PDMS Hardware CareVue 9000 runs as a client ^ server database application (Allbase2, Hewlett Packard) on a network of Unix workstations (Apollo2 700, 715, or 730, Hewlett Packard, Boblingen, Germany; Figure 1). For safety reasons, the database is mirrored on a second server. The database gets manual input and inputs from the patient monitoring system (Component Monitoring System2, Hewlett Packard, Boblingen, Germany), from a PC network (installed by Medtron, Germany) collecting data from external data sources (e.g. ventilators), and from the lab information system (KNS Inc., Graz, Austria). Selected data are daily exported automatically to a PC (db-Export2, Hewlett Packard, Andover, MA).
Fig. 1. Hardware con¢guration of the PDMS. The PDMS hardware consists of a network of UNIX workstations with two interconnected servers, a network of patient monitors connected to the PDMS via a gateway PC and a PC network collecting data from external devices such as ventilators.
Urschitz et al: ThreeYears Experience with a Patient Data Management System 121
Software The PDMS is delivered in a basic software con¢guration which can be adapted to the individual ICUs needs using a speci¢c software con¢guration tool. A con¢guration team consisting of an intensivist, several nurses, and an HP software engineer discussed and de¢ned the various parameters and parameter groups (e.g. heart rate, oral intake, urine output; vital parameters, in- output balance, general care etc.). The team also described formal aspects of the variables such as format (e.g. numerical) and contents, and de¢ned selection lists for text variables because it is easier to later evaluate structurized than free text. Graphics for vital and mechanical ventilation parameters, and various forms and reports for patients' history, diagnoses etc. were implemented for on-screen display, printouts and data export. Recently, we also implemented the medication administration record (MAR) software for prescribing drugs. Vital parameters (heart rate, transcutaneous oxygen saturation, blood pressure etc.) are automatically sent to the PDMS every 12 seconds. These data are averaged every minute and stored automatically on request every 15^180 minutes, thus helping to reduce the time needed for documentation. For reasons of database performance and storage capability, the patients' data can remain in the PDMS database only a little longer than the patient is admitted to the NICU. In order to save the most important data of Table 1. Documentation tables in the PDMS spreadsheet 1. Overview table
2. Vital parameters 3. Mechanical ventilation 4. Lab values I 5. Lab values II 6. Drugs administered 7. Input/output balance 8. General care 9. Special care 10. Specialists' comments 11. Neurology 12. Daily report
Most important vital parameters, mechanical ventilation parameters, input/output balance, important actions Heart rate, respiratory rate, blood pressure, transcutaneous measurements etc. Ventilator settings, blood gas analyses Electrolytes, blood cell count etc. Results of bacteriological cultures etc.
Skin care, nutrition, feeding behavior, excretions etc. Catheters, drains etc. E.g. ophthalmologist's opinion, radiograms Clinical and sonographical data Enlarged overview table
the patients' ¢les, we daily export selected PDMS reports on paper and to PC text ¢les using the dbExport2 software. There are plans to automatically transfer daily reports to the hospital's digital archive. This will prevent loss of data, will improve the accessibility of past patients' records and will save space in the hospital's paper archive. The daily exported PDMS data are not easily accessible for further processing because there is no simple, ready to use data evaluation tool and because some data are very di¤cult to extract due to the structure of the CareVue database (e.g. temporal course and dosage of bypass medication). We are currently developing a program which will automatically export the patients' data into a new ``scienti¢c'' database. This data set may be further processed, visualized, and statistically analyzed for quality control, patient or ward speci¢c analyses etc. One of the problems of this project is automated data validation and the repair of false or missing data. False data may originate from errors of manual input or automated data input from external sources into the PDMS, and from errors arising during data export or data import to the scienti¢c data base. Such errors are frequently encountered at rates of about 10% [1], a rate which renders manual data validation and repair virtually impossible. Individual PDMS con¢guration The PDMS con¢guration used at our NICUs was worked out by the above mentioned team within a period of three months. Because the contents of a neonatal PDMS di¡ers largely from the contents of a PDMS for adults (e.g. di¡erent parameter ranges, care protocols and items, oral intake, dosage etc.) most of the documented items had to be newly de¢ned and implemented. In order to facilitate the daily routine work, we designed ward speci¢c ``forms'' for infection control, transfusions etc. Neonatal diagnoses may be selected with their corresponding 5 digit ICD-9 code from organ speci¢c selection lists. Our PDMS presently covers 2 graphic displays (vital and ventilation parameters), 11 documentation tables, and 15 ``forms.'' The patient data overview screen can be seen in Figure 2. The documentation groups and the titles of the implemented ``forms'' are listed in Tables 1 and 2. Moreover, we have developed a few knowledge based systems in cooperation with the Institute for Medical Cybernetics and Arti¢cial Intelligence and the Austrian Research Institute for Arti¢cial Intelligence. These programs run as client ^ server applications on an additional workstation which is connected to the local PDMS network and may be called up at any patient
122
Journal of Clinical Monitoring and Computing Vol 14 No 2 February 1998
Table 2. PDMS forms Overview of patient care patient history and clinical state Diagnoses I Diagnoses II Infection control
Monitor trend Ward statistics Transfer report Transfusions Patients' check list
History of previous pregnancies, this pregnancy, birth and resuscitation, clinical state at admission Chronological table of diagnoses Diagnoses grouped according to organ systems, including ICD codes Daily assessment of risk factors, possible source of infection, relevant lab results, results of bacteriological cultures, antibiotic therapy Graphical display of vital parameters Data overview for ward statistics Data overview for transferring a patient Chronological overview of blood transfusions Schedule of routine investigations and therapies
terminal: VIE-PNN [5] interactively calculates the composition of the parenteral nutrition for newborn infants according to standard clinical algorithms. Requested input are administrative (e.g. name, birth weight) and actual data (e.g. actual weight, £uid supply, oral intake, lab results, relevant diseases etc.). Output is a parenteral nutrition sheet which also includes statistical calculations. All data are stored in the system's data base which provides an overview on all previous TPN prescriptions. VIE-NMed [6] provides information about drugs used in newborn infants. The system is based on the pharmacological manual ``NeoFax'' [7] which lists data about commonly used drugs in neonatal care and on local therapy plans. The system provides text searching facilites (e.g. for drugs and side e¡ects), a neonatal dosage calculator, and a facility for ¢nding drugs by organ group (e.g. cardiovascular system) or clinical problems (e.g. supraventricular tachycardia). VIE-VENT [8], is an on-line advisory system for the mechanical ventilation of newborn infants. It uses clin-
Fig. 2. Documentation spreadsheet. The spreadsheet displays the vital parameter graphics and the patient's data overview with selected vital and mechanical ventilation parameters and the input-output balace vs. Time. The column on the left shows the displayable parameter groups (cf.Table 1) which can be activated and deactivated by mouse click. Forward/backward and up/down scrolling is possible by clicking the arrows.
Urschitz et al: ThreeYears Experience with a Patient Data Management System 123
Table 3. Evaluation of the questionnaire Questions rating
1
2
3
Mean (SD)
Median
Are you satis¢ed with the PDMS?
1 Very satis¢ed 5
2 Satis¢ed 35
3 Not satis¢ed 1
1.9 0.4
2
Is computer aided documentation better, equal or worse than hand written documentation?
1 Better
2 Equal
3 Worse 4
1.4 0.7
1
How well do you know how to work with the PDMS?
1 Very good 15
2 Good 24
3 Bad 0
1.6 0.5
2
Would you prefer to use hand-written documentation?
1 Yes 1
2 No 39
3 Don't know 1
29
8
Questions
1 C/P
2 C/P
3 C/P
4 C/P
5 C/P
Mean rating C/P
Median rating C/P
1. Time needed for documentation 2. Comprehensiveness of documentation 3. Accurracy of documentation 4. Availability of data
9/6 13/1 12/5 17/3
15/9 16/11 17/20 16/18
11/5 8/17 8/9 4/15
4/15 3/8 3/6 0/4
1/5 0/4 0/1 5/3
2.3/3.2 2.1/3.1 2.1/2.5 1.7/2.7
2/4 2/3 2/2 3/3
The table lists the questions contained in the questionnaire which was distributed to the doctors and nurses of both NICUs. 1. Subjective rating of the PDMS: Answers are listed as numbers of sta¡ members choosing the proposed rating and by the mean (standard deviation) and median ratings. 2. Ratings of the system's performance: Rating of the performance of the PDMS according to a 5 level progressive scale (1 ^ excellent, 5 ^ bad): Answers are listed as numbers of sta¡ members choosing a speci¢c rating and by the mean (standard deviation); and median ratings. The ratings for computer (C) and paper (P) documentation are separated by a slash.
ical data (e.g. PtcCO2, SaO2 and PtcO2) as input and gives recommendations for ventilator settings. Training the sta¡ The implementation team and the PDMS trainers of the computer ¢rm initially instructed all sta¡ members (nurses and doctors) in how to use the PDMS. After one day basic training, the PDMS was used in parallel with the previous paper documentation for two weeks, since then, only the PDMS is used. New sta¡ members are trained by the team, there are also team training hours for PDMS updates. Keeping the PDMS up-to-date A few persons, a doctor and two nurses of each ward, are responsible for keeping the system up-to-date. Changes, like correcting false value limits, introducing new drugs or therapies, and system enhancements are implemented in close collaboration with the software engineers of the computer company. The technical sta¡
of the hospital and, subsidiarily, the computer company are responsible for solving technical problems, mainly simple hardware problems with monitors, keyboards or trackballs. There were three major program updates and several smaller changes and system enhancements. Although the mean duration of stay at the NICU is about 28 days, some of our patients may stay much longer contributing to extremely large patient data ¢les. Therefore the system's data storage capability has recently been enlarged.
Experience with and evaluation of the PDMS CareVue 9000 runs without system breakdown for three years. System maintenance needed 23 hours, hardware repair, mainly of trackballs and monitors, 406 hours, and software and application maintenance 280 service hours. These expenses were covered by the service contract. In order to evaluate the opinion of the sta¡ on the PDMS, we designed a questionnaire (Table 3.) which was distributed to all available sta¡ members. The ques-
124
Journal of Clinical Monitoring and Computing Vol 14 No 2 February 1998
tionnaire consisted of three parts: the ¢rst asking for a general opinion about the system and its daily use (rating on a three level scale), the second for rating the systems' performance on a ¢ve level scale and to compare it with paper documentation, and the third for listing three positive and three negative features of the PDMS. We counted the number of scores per level and calculated mean and median levels. The second part of the questionnaire was evaluated separately for each NICU because of di¡erences in the personal dynamics of the nurses and di¡erences in hardware (the ¢rst NICU has slower processors). Forty-one out of 68 questionnaires were returned. All sta¡ members possessed excellent (15/39) or good (24/39) knowledge of the system. Only one was not satis¢ed with the system, 4/41 (10%) found the documentation on paper more accurate than the documentation using the PDMS but only one would have liked to return to hand written documentation. There was no di¡erence between the ratings obtained at both NICUs. Simplicity of learning about how to use the system, comprehensiveness of data structure and access, automated documentation of vital parameters, ventilator data, and in-output calculations, avoidance of redundant data input [10], and an increase of the documentation precision were the most frequently cited bene¢ts of using the PDMS at our ward. Kari et al. [2] and Hammond et al. [3] investigating PDMS performance and acceptance also found improvements in the precision of documentation and a high degree of satisfaction with the introduction of a PDMS in adult ICUs. However, the often believed ^ time saving abilities of a PDMS ^ which we did not measure ^ could not be demonstrated in a recent prospective study by Pierpont and Thilgen [9]. The criticism mentioned in our questionnaire was mainly focused on the system's poor therapy planning capabilities, especially the medication administration record (criticized 14 times) and the complicated methods of documenting patients' care (criticized ¢ve times). Other points of critique were the system's slowness in ¢nding older data and the limited overview: backward scrolling at that time was only possible day by day, and the most condensed view was 8h per column or 3 days per screen. Twenty four hours would have been preferred in order to display one week per screen (4 times). These criticized points have already been ameliorated in the latest program revision (H0). Some co-workers disliked the de¢ciencies of useful data export and analysis, the de¢ciencies in graphical visualization (e.g. the missing of a body scheme, of more trend graphics etc.), and the ``snap-shot'' documentation of vital parameters, i.e. a maximum density
of 15 minutes time intervals between automatically sampled vital parameter data (e.g. heart rate) in the ICU version of the program. A more intelligent data sampling ^ e.g. adapting the data density according to the variability of the values ^ would have been preferred. It is also very complicated to track subsequent data corrections in the printouts because these changes are only documented in numerous footnotes. At present, we are not able to comment on the precision of the stored data. A very ¢rst test of our data evaluation program helped us to identify unrepresentative data (e.g. the heart rate at a given time point does not necessarily represent the actual heart rate), wrong data (e.g. manual input errors like using the wrong line for documentation), and missing data. A further thorough evaluation of the data will help to improve the precision of documentation and to ¢nd ways to prevent or correct faulty data. Because the clinical conditions and the users' expectations change constantly, it is essential to adapt the PDMS continuously to the changing clinical conditions without disturbing the system's performance, continuity and safety. On the other hand it is mandatory for the users to adapt their expectations to the reality of a widely used system and to ^ at least temporarily ^ tolerate some of the inconveniencies of a system: a worldwide used PDMS will o¡er the advantages of safety and continuity at the price of a certain degree of in£exibility. This should, however, not keep the program developers from listening to and understanding of the users' wishes, from creativity, from de¢ning targets, and from developing or adjusting necessary and useful enhancements. Computerized systems only live if system developers and users constantly interact about the system's funcionalities. There will never be a perfect PDMS in a constantly changing environment, but there should be an end in view for such a system. CONCLUSION We conclude from our data that most sta¡ members think that it was worth introducing a PDMS at our neonatal intensive care units. The system is well accepted and integrated into the daily routine and has improved our documentation tasks. Future enhancements such as enhanced data retrieving facilities and improved graphical visualization tools will further add to the satisfaction with the system.
Urschitz et al: ThreeYears Experience with a Patient Data Management System 125
REFERENCES 1. Metnitz PG, Laback P, Popow C, Laback O, Lenz K, Hiesmayr M. Computer assisted data analysis in intensive care: The ICDEV project-development of a scienti¢c database system for intensive care (Intensive Care Data Evaluation Project). Int J Clin Monit Comput 1995; 12: 147^159 2. Kari A, Ruokonen E, Takala J. Comparison of acceptance and performance of automated and manual data management systems in intensive care. Int J Clin Monit Comput 1990; 7: 157^162 3. Hammond J, Johnson HM, Varas R, Ward CG. A qualitative comparison of paper £owsheets vs a computer-based clinical information system. Chest 1991; 99: 155^157 4. Frayer WW. Patient data management in neonatal intensive care. Clin Perinatol 1980; 7: 145^154 5. Dobner M, Miksch S, Horn W, Popow C. VIE-PNN: An expert system for calculating parenteral nutrition of intensive care premature and newborn infants. Wien Klin Wochenschr 1995; 107: 128^132 6. Horn W, Popow C, Miksch S. Vie-NMed, a computer assisted drug prescription facility for sick newborn infants. 7. Young TE, Mangum OB. NeoFax #94: A manual of drugs used in neonatal care, ed 7. Columbus, OH: Ross Production Division, Abbott Laboratories, 1994 8. Horn W, Popow C, Miksch S, Seyfang A. Integrating a knowledge-based system for parenteral nutrition of neonates into a clinical intranet. IEEE Expert, Special issue ``Arti¢cial intelligence and the changing face of health care,'' accepted 1997 9. Geiger G, Merrilees K, Walo R, Gordon D, Kunov H. An analysis of the paper-based health record: Information content and its implications for electronic patient records. Medinfo 8 Pt 1295 1995 10. Pierpont GL, Thilgen D. E¡ect of computerized charting on nursing activity in intensive care. Crit Care Med 1995; 23: 1067^1073