**4. Decision Support Systems for Intensive Care Unit**

Intensive care medicine represents an equally challenging field for physicians. Patients are usually affected by multiple disease and need to be constantly supervised; often in a state of unconsciousness, patients cannot directly communicate with the practitioner and have to be monitored using a multitude of parameters, producing a significant amount of data. There are significant costs involved.

We will focus on general DSSs, DSSs for artificial ventilation and DSSs for infections.

#### **4.1. General DSSs for ICU**

One DSS designed to deal with general issues of intensive care medicine is ACUDES (Archi‐ tecture for Intensive Care Unit Decision Support), which takes into account the evolution of the patient following her/his diseases over time and gives information about illness and con‐ comitant signs [84]. Another general DSS for intensive care unit is RHEA [85]; like ACUDES, it collects data from patients and gets information about adverse events and nosocomial in‐ fection risk for each patient. Data are entered manually; the system launches useful messag‐ es or alerts about the therapy by prediction models.

The acceptability of DSSs by medical staff in the ICU is quite good; in his review [88], East found that DSSs were well accepted by ICU physicians and provided good perform‐

Decision Support Systems in Medicine - Anesthesia, Critical Care and Intensive Care Medicine

http://dx.doi.org/10.5772/51756

37

A DSS for artificial ventilation and weaning is the knowledge-based closed-loop system SmartCare™ [89]. Based on respiratory rate, tidal volume and end-tidal CO2, it takes a pic‐ ture of the patient's present state, and using this picture, it continuously adapts the level of pressure support to maintain the patient in a respiratory comfort zone. The system contains a weaning protocol that gradually decreases the level of pressure support when patient's respiratory status improves (Figure 21). The use of the DSS resulted in more efficient wean‐ ing, with a decrease of the total duration of artificial ventilation from 12 to 7.5 days [90].

**Figure 21.** Working principles of NeoGanesh/SmartCare. PS, pressure support; SBT, spontaneous breathing trial [91].

Another closed-loop ventilation system is IntelliVent-ASV. The system provides automat‐ ic setting of ventilator parameters and closed-loop regulation based on the inputs of ET‐ CO2, SpO2 and FiO2 after individualization of correct target ranges. Initial results appear

A hybrid knowledge-based and physiological model-based DSS is the Sheffield Intelligent Ventilator Advisor (SIVA). It gives advices and adaptive patient-specific decision support using FiO2, PEEP, inspiratory pressure and ventilatory rate. When compared with expert de‐ cision making, appropriated decision making was obtained with the system and the control

Another DSS for artificial ventilation is the INVENT project, based on physiological model‐ ling. This system uses two software modules to process patient data to suggest specific ven‐

ance improving outcome.

interesting [92].

of blood gases was similar in both groups [93].

A general DSS designed for neuro-intensive care unit is iSyNCC (intelligent System for Neu‐ ro-Critical-Care) [86]. This system collects data from patients in a continuous way and uses them to provide decision support in terms of alerts or therapeutic messages, predicting also the patients' recovery. This is possible by the integration of four modules: data acquisition module, data storage module, data transmission module and user interface (Figure 20).

**Figure 20.** Graphical user interface [86].

#### **4.2. Decision support systems for artificial ventilation**

Systems for artificial ventilation have to be effective, safe and easy to use at the patient bed‐ side. To be useful, its software must remove all noise and artifacts, processing only validate data. Ventilation requires constant monitoring, in order to ensure a timely weaning. Tehrani [87] suggests four possible barriers:


The acceptability of DSSs by medical staff in the ICU is quite good; in his review [88], East found that DSSs were well accepted by ICU physicians and provided good perform‐ ance improving outcome.

**4.1. General DSSs for ICU**

36 Decision Support Systems

**Figure 20.** Graphical user interface [86].

[87] suggests four possible barriers:

**2.** no immunity to noise and erroneous data,

**3.** inadequate training for use of the systems, and

**4.** lack of implementation in commercial ventilators.

**1.** lack of accessibility,

**4.2. Decision support systems for artificial ventilation**

es or alerts about the therapy by prediction models.

One DSS designed to deal with general issues of intensive care medicine is ACUDES (Archi‐ tecture for Intensive Care Unit Decision Support), which takes into account the evolution of the patient following her/his diseases over time and gives information about illness and con‐ comitant signs [84]. Another general DSS for intensive care unit is RHEA [85]; like ACUDES, it collects data from patients and gets information about adverse events and nosocomial in‐ fection risk for each patient. Data are entered manually; the system launches useful messag‐

A general DSS designed for neuro-intensive care unit is iSyNCC (intelligent System for Neu‐ ro-Critical-Care) [86]. This system collects data from patients in a continuous way and uses them to provide decision support in terms of alerts or therapeutic messages, predicting also the patients' recovery. This is possible by the integration of four modules: data acquisition module, data storage module, data transmission module and user interface (Figure 20).

Systems for artificial ventilation have to be effective, safe and easy to use at the patient bed‐ side. To be useful, its software must remove all noise and artifacts, processing only validate data. Ventilation requires constant monitoring, in order to ensure a timely weaning. Tehrani A DSS for artificial ventilation and weaning is the knowledge-based closed-loop system SmartCare™ [89]. Based on respiratory rate, tidal volume and end-tidal CO2, it takes a pic‐ ture of the patient's present state, and using this picture, it continuously adapts the level of pressure support to maintain the patient in a respiratory comfort zone. The system contains a weaning protocol that gradually decreases the level of pressure support when patient's respiratory status improves (Figure 21). The use of the DSS resulted in more efficient wean‐ ing, with a decrease of the total duration of artificial ventilation from 12 to 7.5 days [90].

**Figure 21.** Working principles of NeoGanesh/SmartCare. PS, pressure support; SBT, spontaneous breathing trial [91].

Another closed-loop ventilation system is IntelliVent-ASV. The system provides automat‐ ic setting of ventilator parameters and closed-loop regulation based on the inputs of ET‐ CO2, SpO2 and FiO2 after individualization of correct target ranges. Initial results appear interesting [92].

A hybrid knowledge-based and physiological model-based DSS is the Sheffield Intelligent Ventilator Advisor (SIVA). It gives advices and adaptive patient-specific decision support using FiO2, PEEP, inspiratory pressure and ventilatory rate. When compared with expert de‐ cision making, appropriated decision making was obtained with the system and the control of blood gases was similar in both groups [93].

Another DSS for artificial ventilation is the INVENT project, based on physiological model‐ ling. This system uses two software modules to process patient data to suggest specific ven‐ tilator settings. These modules are the Automatic Lung Parameters Estimator (ALPE system), that is based on a model of pulmonary gas exchange with particular attention to the oxygen transport, and the system for the arterializations of venous blood (ARTY system), in‐ tegrating arterial values from venous blood. All data produced are used to send suggestions of ventilator settings based on the prediction of outcome (Figure 22).

After observation of infection management in a 21-bed mixed medical/surgical adult ICU, an Australian group [96] designed a software, ADVISE, that allowed the digitalization of patient data to create appropriate antibiotic recommendations in real-time. Messages are not only strictly about therapy, but also about the general theory of the specific pathogens or other pa‐ tient-related information (Figure 23). Comparison between 6 months before and 6 months af‐ ter the use of this DSS revealed a better rationalization of the antibiotic use, with a reduction of 10.5% in the overall prescription of antiobiotics, mostly cephalosporin and vancomycin [97].

Decision Support Systems in Medicine - Anesthesia, Critical Care and Intensive Care Medicine

http://dx.doi.org/10.5772/51756

39

**Figure 23.** Screenshot of ADVISE showing with microbiology results, allergy profile, and antimicrobial/isolate matrix. It is displayed the medical review panel, alerting the user to potential allergy risks or overlapping antibiotic coverage,

A sepsis computer protocol was implemented in a 27-bed surgical ICU to manage sepsis [98]. International validated scores were used to make a daily assessment of outcome togeth‐ er with all patient data. All this material was placed in a computer protocol, with nine logi‐ cal diagrams displayed individually to be used at bedside. Computer instructions had to be confirmed by the clinician before the execution of interventions. The system's performance was evaluated via the comparison with paper guidelines: the DSS improved the administra‐

and the rule-based recommendation generated by clicking on the isolate in the tabular view [96].

tion of antibiotics, supported therapeutic decision making and reduced mortality.


**Figure 22.** Graphical user interface with its 3 sections. The left side shows the ventilator settings and penalties, dis‐ played as current, simulated and optimal. The right side displays variables of lungs, arterial and venous blood descri‐ bed as current, simulated and optimal value. On the bottom, patient specific parameters and related organ systems. In this illustration, data of a single post-operative cardiac patient are showed [94].

#### **4.3. Decision support systems for infections in ICU**

Infections in ICU represent 26% of nosocomial infections [95] and usually are accompanied by a high rate of serious problems like sepsis and mortality. Bacteria in the ICU are more resistant to antibiotic therapy. Patients in ICU are monitored by a variety of invasive devi‐ ces; these devices can transmit germs directly. The most frequent nosocomial infections are catheter-associated urinary infections, and occur in about 35% of the cases, characterized by low mortality and costs, while bloodstream infections and respiratory ventilator-associated infections occur at about 15% of the cases, but are associated with high mortality and costs [34, 95]. For all these reasons, antimicrobial infection surveillance in ICU has to be very strict, usually more than in other hospital wards [95].

After observation of infection management in a 21-bed mixed medical/surgical adult ICU, an Australian group [96] designed a software, ADVISE, that allowed the digitalization of patient data to create appropriate antibiotic recommendations in real-time. Messages are not only strictly about therapy, but also about the general theory of the specific pathogens or other pa‐ tient-related information (Figure 23). Comparison between 6 months before and 6 months af‐ ter the use of this DSS revealed a better rationalization of the antibiotic use, with a reduction of 10.5% in the overall prescription of antiobiotics, mostly cephalosporin and vancomycin [97].

tilator settings. These modules are the Automatic Lung Parameters Estimator (ALPE system), that is based on a model of pulmonary gas exchange with particular attention to the oxygen transport, and the system for the arterializations of venous blood (ARTY system), in‐ tegrating arterial values from venous blood. All data produced are used to send suggestions

**Figure 22.** Graphical user interface with its 3 sections. The left side shows the ventilator settings and penalties, dis‐ played as current, simulated and optimal. The right side displays variables of lungs, arterial and venous blood descri‐ bed as current, simulated and optimal value. On the bottom, patient specific parameters and related organ systems. In

Infections in ICU represent 26% of nosocomial infections [95] and usually are accompanied by a high rate of serious problems like sepsis and mortality. Bacteria in the ICU are more resistant to antibiotic therapy. Patients in ICU are monitored by a variety of invasive devi‐ ces; these devices can transmit germs directly. The most frequent nosocomial infections are catheter-associated urinary infections, and occur in about 35% of the cases, characterized by low mortality and costs, while bloodstream infections and respiratory ventilator-associated infections occur at about 15% of the cases, but are associated with high mortality and costs [34, 95]. For all these reasons, antimicrobial infection surveillance in ICU has to be very

this illustration, data of a single post-operative cardiac patient are showed [94].

**4.3. Decision support systems for infections in ICU**

strict, usually more than in other hospital wards [95].

of ventilator settings based on the prediction of outcome (Figure 22).

38 Decision Support Systems


**Figure 23.** Screenshot of ADVISE showing with microbiology results, allergy profile, and antimicrobial/isolate matrix. It is displayed the medical review panel, alerting the user to potential allergy risks or overlapping antibiotic coverage, and the rule-based recommendation generated by clicking on the isolate in the tabular view [96].

A sepsis computer protocol was implemented in a 27-bed surgical ICU to manage sepsis [98]. International validated scores were used to make a daily assessment of outcome togeth‐ er with all patient data. All this material was placed in a computer protocol, with nine logi‐ cal diagrams displayed individually to be used at bedside. Computer instructions had to be confirmed by the clinician before the execution of interventions. The system's performance was evaluated via the comparison with paper guidelines: the DSS improved the administra‐ tion of antibiotics, supported therapeutic decision making and reduced mortality.

The COSARA research project (Computer-based Surveillance and Alerting of nosocomial infections, Antimicrobial Resistance and Antibiotic consumption in ICU) is a complex sys‐ tem that automatically acquires patient data and records all information about the pa‐ tient's clinical history, therapy and antibiotic resistance, providing decision support through visual presentation of graphs, icons, visual bars, pop-ups and audible alerts [99]. The DSS consists of different modules, having each one the ability to manage some func‐ tions, as e.g. X-ray photos. For each antibiotic prescription, there is a pop-up menu prompting the physician to register the motivation for starting therapy (Figure 24). Clini‐ cal decision support is delivered through mail or messages based on guidelines. After 3 months of implementation, clinical outcome was improved.

cessity to react quickly and adjust the reactions throughout a short time interval. More and more, DSSs are integrated in AIMS in the perioperative period. These have been successfully used to configure checklists to reduce the incidence of intraoperative problems, thus even re‐ ducing costs significantly. Decision support systems in that context can reduce the amount of laboratory investigations performed. More anesthetic specific DSSs are also being developed, allowing risk stratification for difficult intubations, better PONV prophylaxis, appropriate timing of antibiotic prophylaxis and thus reducing the incidence of wound infections. In the arena of emergency medicine, both in- and out-of-hospital DSSs have been created and suc‐ cessfully tested. They range from systems to help identify microemboli or pulmonary embo‐ lism, to systems to help coordinating the pre-hospital logistics. Some of these systems are met with skepticism by physicians because they fear of handing over decision making to these sys‐ tems. In ICU units, DSSs can help with the setup of artificial ventilation and significantly re‐ duce infections: this is specifically important in the complex decision making and

Decision Support Systems in Medicine - Anesthesia, Critical Care and Intensive Care Medicine

http://dx.doi.org/10.5772/51756

41

establishment of appropriate treatment options in patients suffering from sepsis.

Decision support systems are useful tools in modern medicine. They can improve clinical prac‐ tice, adherence to best evidence based medicine and, in some cases, clinical and health educa‐

In the operating room, DSSs can be effective to deal with a variety of problems. In emergen‐ cy medicine, they have shown good performance, but their acceptability by clinicians has not been sufficient. Decision support systems in intensive care medicine allow the possibili‐ ty to manage a large number of data whilst allowing the clinician to follow them in a more efficient way. However, a widespread use of these systems is obstructed by economical and cultural barriers; a greater involvement of the medical users in the design of the systems, in‐ cluding better user interfaces, could improve their clinical acceptance. DSSs are software de‐

**2.** There is an increased development of DSSs in any medical field, specifically in the area

**3.** Despite the increasing request, cultural and economical barriers obstruct the diffusion

signed to assist the physician but are not a substitutes of this figure.

**1.** Decision support systems improve quality of practice and patient safety.

**6. Conclusion**

Key messages

of critical care medicine.

of these systems.

tion and patients outcomes.

**Figure 24.** Screenshot with infections and antibiotics history; there also graphs that represent common used values [99].

#### **5. Discussion**

Decision support systems have been successfully developed in the areas of intensive care, emergency medicine and anesthesia. These are areas with an overload of information and ne‐ cessity to react quickly and adjust the reactions throughout a short time interval. More and more, DSSs are integrated in AIMS in the perioperative period. These have been successfully used to configure checklists to reduce the incidence of intraoperative problems, thus even re‐ ducing costs significantly. Decision support systems in that context can reduce the amount of laboratory investigations performed. More anesthetic specific DSSs are also being developed, allowing risk stratification for difficult intubations, better PONV prophylaxis, appropriate timing of antibiotic prophylaxis and thus reducing the incidence of wound infections. In the arena of emergency medicine, both in- and out-of-hospital DSSs have been created and suc‐ cessfully tested. They range from systems to help identify microemboli or pulmonary embo‐ lism, to systems to help coordinating the pre-hospital logistics. Some of these systems are met with skepticism by physicians because they fear of handing over decision making to these sys‐ tems. In ICU units, DSSs can help with the setup of artificial ventilation and significantly re‐ duce infections: this is specifically important in the complex decision making and establishment of appropriate treatment options in patients suffering from sepsis.

### **6. Conclusion**

The COSARA research project (Computer-based Surveillance and Alerting of nosocomial infections, Antimicrobial Resistance and Antibiotic consumption in ICU) is a complex sys‐ tem that automatically acquires patient data and records all information about the pa‐ tient's clinical history, therapy and antibiotic resistance, providing decision support through visual presentation of graphs, icons, visual bars, pop-ups and audible alerts [99]. The DSS consists of different modules, having each one the ability to manage some func‐ tions, as e.g. X-ray photos. For each antibiotic prescription, there is a pop-up menu prompting the physician to register the motivation for starting therapy (Figure 24). Clini‐ cal decision support is delivered through mail or messages based on guidelines. After 3

**Figure 24.** Screenshot with infections and antibiotics history; there also graphs that represent common used values [99].

Decision support systems have been successfully developed in the areas of intensive care, emergency medicine and anesthesia. These are areas with an overload of information and ne‐

**5. Discussion**

40 Decision Support Systems

months of implementation, clinical outcome was improved.

Decision support systems are useful tools in modern medicine. They can improve clinical prac‐ tice, adherence to best evidence based medicine and, in some cases, clinical and health educa‐ tion and patients outcomes.

In the operating room, DSSs can be effective to deal with a variety of problems. In emergen‐ cy medicine, they have shown good performance, but their acceptability by clinicians has not been sufficient. Decision support systems in intensive care medicine allow the possibili‐ ty to manage a large number of data whilst allowing the clinician to follow them in a more efficient way. However, a widespread use of these systems is obstructed by economical and cultural barriers; a greater involvement of the medical users in the design of the systems, in‐ cluding better user interfaces, could improve their clinical acceptance. DSSs are software de‐ signed to assist the physician but are not a substitutes of this figure.

#### Key messages

