Thomas D. East

A Strategy for Development of Computerized Critical Care Decision Support Systems

It is not enough to merely manage medical information. It is difficult to justify the cost of hospital information systems (HIS) or intensive care unit (ICU) patient data management systems (PDMS) on this basis alone. The real benefit of an integrated HIS or PDMS is in decision support. Although there are a variety of HIS and ICU PDMS systems available there are few that provide ICU decision support. The HELP system at the LDS Hospital is an example of a HIS which provides decision support on many different levels. In the ICU there are decision support tools for antibiotic therapy, nutritional management, and management of mechanical ventilation. Computer protocols for the management of mechanical ventilation (respiratory evaluation, ventilation, oxygenation, weaning and extubation) in patients with adult respiratory distress syndrome ((ARDS) have already been developed and clinically validated at the LDS Hospital. These protocols utilize the bedside intensive care unit (ICU) computer terminal to prompt the clinical care team with therapeutic and diagnostic suggestions. The protocols (in paper flow diagram and computerized form) have been used for over 40,000 hours in more than 125 adult respiratory distress syndrome (ARDS) patients. The protocols controlled care for 94% of the time. The remainder of the time patient care was not protocol controlled was a result of the patient being in states not covered by current protocollogic (e.g. hemodynamic instability, or transport for X-Ray studies). 52 of these ARDS patients met extra corporal membrane oxygenation (ECMO) criteria. The survival of the ECMO criteria ARDS patients was 41%, four times that expected (9%) from historical data (p<0.0002). The success of these computer protocols and their acceptance by the clinical staff clearly establishes the feasibility of controlling the therapy of severely ill patients.Over the last four years we have refined the process which we use for generating computerized protocols. The purpose of this paper is to present the six step development strategy which we are successfully using to produce computerized critical care protocols.

Performance of Computerized Protocols for the Management of Arterial Oxygenation in an Intensive Care Unit

Computerized protocols were created to direct the management of arterial oxygenation in critically ill ICU patients and have now been applied routinely, 24 hours a day, in the care of 80 such patients. The protocols used routine clinical information to generate specific instructions for therapy. We evaluated 21,347 instructions by measuring how many were correct and how often they were followed by the clinical staff. Instructions were followed 63.9% of the time in the first 8 patients and 92.3% in the subsequent 72 patients. Instruction accuracy improved after the initial 8 patients, increasing from 71.5% of total instructions to 92.8%. Instruction inaccuracy was primarily caused by software errors and inaccurate and untimely entry of clinical data into the computer. Software errors decreased from 7.2% in the first 8 patients to 0.8% in subsequent patients, while data entry problems decreased from 7.5% to 4.2%. We also assessed compliance with the protocols in a subset of 12 patients (2637 instructions) as a function of 1) the mode of ventilatory support, 2) whether the instruction was to increase or decrease the intensity of therapy or to wait for an interval of time and 3) whether the instruction was ‘correct’ or ‘incorrect’. The mode of ventilatory support did not affect compliance with protocol instructions. Instructions to wait were more likely to be followed than instructions to change therapy. Ninety-seven percent of the correct instructions were followed and 27% of the incorrect instructions were followed. The major problem in creating the protocols was obtaining clinician agreement on protocol logic and their commitment to utilize it clinically. The major problem in implementing the protocols was obtaining accurate and timely data entry. We conclude that computerized protocols can direct the clinical care of critically ill patients in a manner that is acceptable to clinicians.

Real time data acquisition: recommendations for the medical information bus (MIB)

Care of the acutely ill patient requires rapid acquisition, recording and communications of data. In the modern hospital it is not unusual for a patient to be connected to several monitoring and recording devices simultaneously. Each of these devices is typically made by a different manufacturer who may specialize in one sort of measurement, for example, pulse oximetry. Most of the modern monitoring and recording devices are micro-processor based and have communication capabilities. Unfortunately, there is no operable standard communication technology available from all devices. In addition different clinical staff (physicians, nurses, or repiratory therapists) may be responsible for collecting data. As a result there is a need to develop methods, standards, and strategies for timely and automatic collection of data from these monitoring and recording devices. We report on more than 5 years of clinical experience of automated ICU data collection using a prototype of the Medical Information Bus (MIB).

Simultaneous comparison of intraarterial, oscillometric, and finapres monitoring during anesthesia.

In 30 patients (15 with normal peripheral vascular status and 15 with peripheral vascular disease, hypertension, or a heavy smoking history), systolic, mean, and diastolic arterial pressures were recorded simultaneously every 5 min using a radial arterial catheter, an oscillometric arm cuff, and a Finapres finger cuff during 1–6 h of anesthesia and operation. The average accuracy of oscillometric and Finapres pressure measurements was good. Comparisons of arterial, oscillometric, and Finapres pressures showed only a small bias in the oscillometric and Finapres pressure estimations. Finapres pressures underestimated arterial pressures by 1 mm Hg more than oscillometric pressures did. Peripheral vascular status had no effect on comparisons made between pressures measured with these two techniques. Although bias was small, precision was often lacking as shown by the large variability of the difference between individual values from the three monitors. However, the precision of Finapres pressure measurements was about the same order of magnitude as that of oscillometric measurements.

Automated measurement of functional residual capacity by sulfur hexafluoride washout

We have constructed a computerized, totally automated system for measuring functional residual capacity (FRC) during mechanical ventilation, at any positive end-expiratory pressure (PEEP) and fraction of inspired oxygen. This system uses washout of a small amount (0.5 to 1.0%) of an insoluble, nontoxic tracer gas, sulfur hexafluoride, to measure FRC. It requires no modification of the ventilator and only minimal changes in the breathing circuit; it can be programmed to make measurements routinely without manual intervention.The system was evaluated with three tests. (1) The prototype sulfur hexafluoride analyzer characteristic curve was determined, and the analyzer was evaluated to determine carbon dioxide interference. (2) A comparison with nitrogen washout FRC measurements was made in an extensive bench test with a Plexiglas lung model. The bench test was designed to determine the effects of changing gas composition and minute volume. (3) A study was done in six healthy dogs to determine reproducibility of the FRC measurements at four PEEP levels (0, 5, 10, and 15 cm H2O: two repetitions in each animal).The sulfur hexafluoride analyzer was well characterized by an exponential equation with a multiple r2 = 0.996. The analyzer was not affected by the presence of carbon dioxide (pairedt test,t19 = 1.23,P > 0.10). The bench test indicated that FRC (measured) = 0.969 × FRC (true) − 5.3 ml. (Multiple r2 = 0.979.) This was significantly better than the nitrogen washout system, whose regression equation was also a function of minute volume. In the six animals studied, increasing PEEP always increased FRC and did not significantly alter reproducibility of the FRC measurement (P > 0.1).This automated sulfur hexafluoride washout system should make routine FRC measurements both relatively simple and possible without altering normal ventilatory therapy.

Computerized ventilator data selection: artifact rejection and data reduction

Objective: To determine acceptable strategies for automated data acquisition and artifact rejection from computerized ventilators using the Medical Information Bus. Design: Medical practitioners were surveyed to establish ‘clinically important’ ventilator events. A prospective study involving frequent data collection from ventilators was also conducted. Subjects: Data from 10 adult patients were collected every 10 seconds from a Puritan Bennett 7200A ventilator for a total of 617.1 hours. Interventions: Twelve different computerized data selection and artifact algorithms were tested and evaluated. Measurements and Main Results: Data derived from 12 data selection algorithms were compared with each other and with data manually charted by respiratory therapists into a computerized charting system. Ventilator setting data collected by the algorithms, such as FIO2, reduced the amount of data collected to about 25% compared to manually charted data. The amount of data collected for measured parameters, such as tidal volume, from the ventilator had large variability and many artifacts. Automated data capture and selection generally increased the amount of data collected compared to manual charting, for example for the 3 minute median the increase was a modest 1.2 times. Conclusion: Computerized methods for collecting ventilator setting data were relatively straightforward and more efficient than manual methods. However, the method for automated selection and presentation of observed measured parameters is much more difficult. Based on the findings and analysis presented here, the authors recommend recording ventilator setting data after they have existed for three minutes and measured parameters using a three minute median data selection strategy. Such an algorithm rejected most artifacts, required minimal computational time, had minimal time-delay, and provided clinically acceptable data acquisition. The results presented here are but a starting point in developing automated ventilator data selection strategies.

A Microcomputer-Based Differential Lung Ventilation System

In order to provide a versatile means of delivering differential lung ventilation (DLV), a computer-controlled system was constructed to allow a variety of ventilation protocols as well as to record and monitor relevant physiologic parameters. Two Siemens servo ventilators were modified for synchronous operation and computer control of minute volume and respiratory rate. Twenty-five parameters on the two lungs were collected every breath. Feedback control was used to adjust respiratory rate to maintain PaCO2 = 35 torr and to keep total tidal volume equal to 15 mI/kg. Three differential volume delivery protocols were established. The DLV system was evaluated in a study involving eighteen mongrel dogs (six dogs for each volume delivery protocol), each with a unilateral lung injury caused by an infusion of 0.1 N HOC through the endobronchial tube. This system has proven to be a highly effective and versatile means of providing differential ventilation as well as precise feedback control of essential physiologic parameters such as PaCO2 and tidal volume. The system handles automated data collection of all relevant physiological parameters, making graphical as well as statistical analysis extremely easy.

Evaluation of compliance with a computerized protocol : Weaning from mechanical ventilator support using pressure support

STUDY OBJECTIVES To use a computerized consultation system to evaluate the feasibility of a mechanical ventilator weaning protocol which used the rapid shallow breathing index to guide adjustments in pressure support. A program to monitor user compliance and reasons for noncompliance was built into the computerized consultation system. METHODS A total of nine critically ill patients (ten weaning episodes) were enrolled in the protocol. The respiratory therapists performed routine computer charting in the electronic database. They accepted or declined the explicit instructions generated by the computerized protocol and displayed on the bedside terminal. The consultation program monitored whether accepted instructions were implemented by the user. RESULTS Patients therapy was controlled by protocol for a total of 1075 h (mean 108 h, range 4 to 339 h) and 94.8% (1321/1394) of instructions were followed by the clinical staff. Of the 42 instructions clinical staff refused to follow, 23 (55%) were extubation instructions. There were 52 (3.7%) incorrect instructions generated with 24 software errors, 21 errors in underlying logic, and seven user misunderstanding errors. CONCLUSIONS A high level of user compliance with this protocol was achieved. The methods described herein to monitor compliance and reasons for noncompliance within a protocol are reusable in the domain of mechanical ventilation and possibly in other domains.

Computer-controlled optimization of positive end-expiratory pressure.

Positive end-expiratory pressure (PEEP) is a standard treatment for patients with refractory hypoxemia due to an acute restrictive pathology. The therapeutic range of PEEP can be quite narrow. PEEP therapy has been optimized using invasive variables such as oxygen transport and pulmonary shunt, and noninvasive variables such as compliance; however, the measurements are complex. We constructed a computerized PEEP-optimization system consisting of a Siemens 900C ventilator, Siemens prototype sulfur hexafluoride analyzer, Siemens 940 lung mechanics analyzer, and a DEC 11/ 23 microcomputer. The user may choose from three different noninvasive PEEP titration algorithms: maximizing static total respiratory system compliance (CTR), maximizing functional residual capacity(FRC)-based compliance (CFRC), and normalizing FRC. The device was tested in six dogs with pulmonary injury induced by oleic acid. The system was constrained to 3-cm H2O PEEP steps at 20-min intervals. The algorithm normalizing FRC reached optimal PEEP levels in 40 min, with a mean difference from the desired FRC of 15 ± 48 (SEM) ml. This corresponds to a mean percent error of 1.0% ± 2.63%. The CFRC and CTR algorithms reached optimal PEEP levels in 60 and 40 min, respectively, and maintained a maximal compliance for 85% of the time. This system provides fully automated noninvasive PEEP titration and is flexible enough to incorporate easily any other PEEP titration algorithms. It should improve patient care by guaranteeing that PEEP therapy is truly optimized throughout the patients recovery.

The evolution of eProtocols that enable reproducible clinical research and care methods

Unnecessary variation in clinical care and clinical research reduces our ability to determine what healthcare interventions are effective. Reducing this unnecessary variation could lead to further healthcare quality improvement and more effective clinical research. We have developed and used electronic decision support tools (eProtocols) to reduce unnecessary variation. Our eProtocols have progressed from a locally developed mainframe computer application in one clinical site (LDS Hospital) to web-based applications available in multiple languages and used internationally. We use eProtocol-insulin as an example to illustrate this evolution. We initially developed eProtocol-insulin as a local quality improvement effort to manage stress hyperglycemia in the adult intensive care unit (ICU). We extended eProtocol-insulin use to translate our quality improvement results into usual clinical care at Intermountain Healthcare ICUs. We exported eProtocol-insulin to support research in other US and international institutions, and extended our work to the pediatric ICU. We iteratively refined eProtocol-insulin throughout these transitions, and incorporated new knowledge about managing stress hyperglycemia in the ICU. Based on our experience in the development and clinical use of eProtocols, we outline remaining challenges to eProtocol development, widespread distribution and use, and suggest a process for eProtocol development. Technical and regulatory issues, as well as standardization of protocol development, validation and maintenance, need to be addressed. Resolution of these issues should facilitate general use of eProtocols to improve patient care.