1 Introduction

Managing and processing medical data can be very time consuming and this is very obvious in intensive care units (ICU) settings where physicians are confronted with a vast amount of information generated by their patients. Because of the large data volume and the management complexities while caring for critically ill patients, a number of promising technologies have been developed and introduced in the hope that they can help the physicians to increase their productivity as well as efficiency of care they provide.

During the past 20 years, ICU equipped with clinical information systems, also known as intensive care information systems (ICIS) have rapidly increased in number [1, 2]. These systems were developed to meet the specific needs to optimise data processing and workflow in critical care units.

The end effect of ICIS on the quality of care remains largely unknown, although selectively it has been shown that the use of such systems can decrease the heavy administrative workload of clinicians or nurses [3–6], improve or increase compliance with existing treatment guidelines or protocols [7] and also they serve as early warning systems for specific problems often encountered in ICU patients [8]. In addition, the use of ICIS has also resulted in the development of a variety of databases.

Because of data concerning the impact of the use of ICIS on the outcome of critically ill patients are scarce [9], we conducted the present study in order to evaluate the impact of ICIS implementation upon standard ICU patient outcomes.

2 Materials and methods

2.1 Intensive care unit characteristics

This observational study was carried out in the Liver Intensive Care Unit at Paul Brousse University Hospital. This is a 15-bed, closed-format ICU with approximately 750 medical and surgical admissions each year. This is an ICU specialised in medical and surgical Liver diseases [10, 11]. The team of intensivists is directly responsible for medical or surgical patient’s admission in the ICU. The team includes four permanent intensivists and two intensive care fellows. A physician (hepatologist) is always present on the unit. The nursing staff consists of 52 nurses working on 8-h shifts, with a patient-to-nurse ratio 2:1, but this ratio may vary depending of the severity of the patient’s illness.

All the patients included in the study were admitted to the ICU between January 2004 and August 2006. The study was approved by an institution review board. The need for informed consent was waived as this study was observational and required no additional intervention.

2.2 Intensive care information systems

As from April 2005, our ICU was equipped with a bedside computerised information system (ICIS, Métavision®, iMDSoft, Israel, and Medi-Lan®, France). All ICU rooms (single), the central monitoring desk, as well as physicians and intensivists offices are equipped with ICIS workstations. The medical and nursing staff used the ICIS. They were requested to complete all documentations for the admitted patients as well as to record the daily observations, patient’s management and specific treatments/interventions. No handwritten paper documentation as primary storage mechanism was allowed. Thus, the entire medical, nursing and physiotherapy documentation was carried out via the ICIS.

The ICIS system was used to collect and store data from patient monitoring (medical management and data collection) (Fig. 1). This system connected to bedside monitor (Spacelab Healthcare®, Creteil, France), ventilator (Engtrom ventilator®, Datex-Ohmeda, GE Healthcare, France or Taema® Horus, Bonneuil-sur-Marne, France) and the dialysis machine (Prisma® and Prismaflex®, Hospal, Meyzieu, France) captured data every minute. Similarly vital signs (hemodynamic and respiratory parameters), fluid intake (perfusion, drugs) are automatically saved, thus limiting the risk of human error which is known to happen in handwritten flow sheets. Every single hour the nurses validated the vital signs (from patient’s monitor) as well as ventilator and/or dialysis data recorded by the ICIS. Several other parameters as fluid output (urinary, blood loss, digestive losses…) were entered manually by nurses every hour. In addition several automated script were systematically printed on a daily basis at 7 AM, then validated by medical staff. For example, one of these prints concerned patient’s computed daily fluid balance.

Fig. 1
figure 1

The ICIS system is used to collect and store data from patient monitoring (medical management and data collection)

Full size image

The ICIS included a drug prescription component. Indeed, all orders were prescribed through the same interface in the system. Assistance to prescription software was parameterized allowing individual optimization of drug prescription for criteria such as dose, route of administration, rate of administration and automatically proposing real time warnings to be given (Fig. 2).

Fig. 2
figure 2

The ICIS includes a drug prescription component. Assistance to prescription software is parameterized allowing individual optimization of drug prescription for criteria such as dose, route of administration, rate of administration and automatically proposing real time warnings

Full size image

In addition, all documentation and protocol updates were completed using the ICIS (by nurses and/or intensivists).

The ICIS was adjustable, adaptable and evolutive system. We customized the system in order to trace the model we followed before implementation of ICIS. For example, after defining various clinical, biological and radiological parameters to diagnose and follow-up any organ dysfunction/failure, several monitor screens have been created, one for each organ. Moreover, we have developed specific screens concerning internally validated guidelines, procedures and policies such as blood glucose control, enteral and parenteral feeding, insulin infusion rate, antibiotics administration.

The system was installed March 2005, customized and then used for all daily patient observations, treatment decisions and investigation results. No hand-written documents were compiled as primary documentation. In addition, a team of specialised nurses and ICU doctors was responsible for maintaining, applying and developing the system.

2.3 Study design

This retrospective analysis included two distinct periods.

The first period concerned patients admitted to the ICU before ICIS implementation (BEFORE) beginning January 2004 and ending March 2005. During BEFORE all observations, drug prescriptions, medical and nursing notes were compiled using specific paper charts.

The second period concerned patients admitted to the ICU 3 month after ICIS implementation (AFTER) elapsing between June 2005 and August 2006. During AFTER all documentation was compiled using the ICIS.

Patients admitted in between April 2005 and May 2005 were excluded from the analysis in order to exclude the difficulties inherent to the implementation of a new computerised system and to limit variability following the changes to prescriptions and monitoring that resulted from computerisation of the ICU.

Were compared the standard clinical outcomes: ICU mortality rate, ICU and hospital stays, ICU- readmission rate depending upon BEFORE and AFTER. The severity of acute disease was determined using the Simplified Acute Physiology Score II (SAPS II) [12].

2.4 Statistical analysis

Data are expressed as mean ± SD. Statistical analyses were performed using Statview 5.0 for Macintosh (SAS Institute inc., Cary, NC, USA). To compare the two periods Mann–Whitney test for continuous variables, and Chi square tests or Fischer test for categorical variables, were performed. SAPS II was used to calculate the standardised mortality ratio (SMR). The SMR is calculated as the ratio of the observed mortality of the study’s population above the mortality predicted by SAPS II. In addition, binary logistic regression was used to explore the differences in mortality before and after ICIS implementation, controlled for the risk of mortality (SAPS II). A p value of less than 0.05 was considered to be statistically significant.

3 Results

3.1 Overall patient population characteristics

Details of 1,397 adult patients (BEFORE, n = 662 and AFTER n = 735) referred to the Liver Intensive Care Unit between January 2004–March 2005 and June 2005–September 2006 were analysed. The mean age at ICU admission was 54.5 ± 14.7 years (range 12–96), 66 % (922/1397) were men and the mean SAPS II score was 31.3 ± 18.0 (range 0–113). The mean length of ICU and hospital stay were 7.5 ± 14.1 and 28.2 ± 33.9 days, respectively.

The medical indications for ICU admission included: acute hepatitis, cirrhotic patients with severe liver decompensation (variceal bleeding, encephalopathy, sepsis, and hepatorenal syndrome). The surgical indications included elective surgery (monitoring after major interventions such as hepatectomy), emergency surgery and liver transplantation. Other surgical indications included complications such as postoperative sepsis (wound or intra-abdominal infection or bronchopneumonia) and haemorrhage.

During the stay in ICU, a total of 574 (41.1 %) patients required mechanical ventilation. Two hundred fifty-eight patients (18.5 %) needed vasopressor therapy (epinephrine, norepinephrine dobutamine or dopamine >5 µg/kg/min) mainly for septic or hemorrhagic shock. Nineteen (7.1 %) patients required renal replacement therapy (RRT) or Molecular Adsorbent Recirculating System (MARS®) during their ICU stay. During the study period, the cumulative incidence of mortality was 10.4 % (145 of 1,397 patients) in our liver ICU. Mean overall readmission rate was 4.3 % (60/1,397 admissions).

3.2 BEFORE and AFTER period comparisons

The patient characteristics for both groups are described in Table 1. There were no baseline differences between the two groups, except for their SAPS II values. Indeed, patients of AFTER had a higher SAPS II value score than those of BEFORE (32.1 ± 17.5 vs. 30.5 ± 18.5, p < 0.014). Although patients of AFTER had a higher predicted death rate than that of BEFORE, there were no significant differences in organ supports (mechanical ventilation, catecholamine support, RRT or MARS) during the ICU stay between the periods (Table 1).

Table 1 Patient characteristics
Full size table

Implementation of ICIS resulted in statistically significant decrease in the length of ICU stay 8.5 ± 15.2 day (BEFORE) versus 6.8 ± 12.9 days AFTER; p = 0.048) (Table 2). The ICU re-admission rate was similar between the 2 periods (4.4 % BEFORE vs. 4.2 % AFTER; p = 0.86). In addition, a decrease of the length of stay has also been observed in each sub-group of patients (surgical, medical or liver transplant).

Table 2 Patient’s outcome
Full size table

The mortality rate did not differ significantly between the two periods (11.2 % BEFORE and 9.6 % AFTER; p = 0.35). Subgroup analysis of patients requiring organ support (one or more organ failure) confirmed significant decrease in length of ICU stay without impact on ICU mortality rate (22.8 % BEFORE vs. 20.0 % AFTER, p = 0.32) (Table 3)

Table 3 Outcome of the subgroups of patients requiring support of one or more organs
Full size table

In addition, the standardised mortality ratio (SMR) between the observed number of deaths in our study sample and the number of expected deaths based on SAPS II values was calculated. During BEFORE the observed mortality was not significantly different from the predicted mortality (SMR 1.06). In contrast, during AFTER the observed mortality was significantly lower (p < 0.001) than predicted by SAPSII (SMR 0.75) (Table 4). In addition, the SMR for AFTER was 29 % lower than the SMR for BEFORE.

Table 4 Observed mortality, standardized mortality rate per group
Full size table

4 Discussion

The present analysis describes the experience of a single centre following implementation of an Intensive Care Information System in a medical and surgical intensive care unit. The analysis shows that, although both the ICU staff and policies were not changed during the study period, the implementation of an ICIS was associated with a significant shortening of the ICU length of stay. To our knowledge, this is the first study which has evaluated the impact of ICIS on the outcome of patients admitted to an ICU. In the study by McCambridge et al. [13], the authors used multifaceted intervention (including ICIS) on patient outcomes and observed that the use of health information technology was associated with lower mortality and less use of ventilatory support.

We believe that the clinically relevant differences concerning the length of stay in our ICU resulted from an improved quality of care following the implementation of ICIS. Reductions in medical errors [6, 14, 15], improved protocol compliance [16, 17], and changes to medical organisation [18, 19] have constituted the crucial relationship between quality of care and computerisation. We must acknowledge that the statistically significant association between the implementation of ICIS and the shortened ICU stay cannot be per se attributed solely to a direct care effect.

Medical errors in critically ill patients are common and can influence ICU mortality and morbidity rates [20]. The critically ill patients are at greater risk for medication errors and subsequent adverse drug events. It is reported that 19 % of medication doses are incorrectly prescribed (dose or indication errors) in hospitalized patients [21]. For example, in a cardiothoracic intensive care unit, errors in prescribing were common including omission of route of administration, date of prescription, time to be given and no dose or incorrect dose prescribed [22]. To limit the number of the known potential errors, in our institution, all orders are prescribed though a similar interface and all orders are pre-entered in the ICIS where parameters such as dose, route of administration, frequency and time to be given are already available to be selected automatically. The user only has to validate the prescription. Although we have not yet evaluated this aspect, its implementation was motivated by the many publications showing that the computerized entry of instructions by the physician and the use of decision support can improve the safety of drug prescriptions and administration and reduce medication errors [22–24].

In intensive care, the most documented example is the impact of blood glucose control on mortality and morbidity [25]. However, as showed in the study by Garrouste-Orgeas et al. [20] the most common medical error was associated with insulin administration. To limit potential complications related to insulin dosing errors, Rood et al. [16] implemented computerised guidelines to regulate blood glucose levels using an integrated decision support module. They observed that with the computerised version (compared to a paper version), blood glucose measurements and regulation were much more accurate, reducing complications and thus improving overall patient care. Other authors have also demonstrated the superiority of computerised glucose monitoring in terms of the efficiency of glycaemia control, with patients experiencing no further episodes of hypoglycaemia [26]. In our unit, we have developed patient follow-up systems and implemented them in ICIS, one of them includes guidelines on blood glucose strict control.

Another important aspect to consider is the impact of implementation of an ICIS on the medical organisation [27]. It is generally accepted that the organisational quality of a unit can have a major impact on patient outcomes. Indeed, because of the enormous volume of information generated in ICUs, it is clear that computing technology can help to manage the large quantity of data produced by each patient. Our ICIS uses a high level of interfacing with both bedside devices (invasive and non-invasive ventilator supports, monitors and haemodialysis machines) and the hospital’s main information system. The link to the latter system also enables to retrieve of a large number of clinical, biological and radiological data, to integrate all this information, allowing diagnosis and alarming tools to be activated, resulting in necessary medical actions. These facilities given to the physicians have certainly participated to improved efficiency of care and reduced ICU stay duration we have observed. Moreover automatic data acquisition reduces human copying error and is more rapid than manual data entry. The reduction in documentation time achieved with an ICIS (less time spent on hourly checks, entering laboratory results and prescribing medications) can increase the time spent on patient care [3–6, 17]. For example, in our unit, we have showed in a previous study that the use of ICIS decreased significantly the time dedicated to the coding process [28]. The installation/use of an ICIS changes medical and nursing activities as well as influences cross-disciplinary communication during ICU ward rounds [19]. Such a system is operator-dependent, relying on the accuracy of the data entered, and time must be spent on training the medical/nursing personnel. A rapid personnel (medical and paramedical) turnover rate can thus affect the efficiency of an ICIS system. We found that the system was quickly accepted in our ICU [29] once the users (medical and nursing personnel) realised the advantages of its use.

However, the introduction of ICIS in an ICU is associated with implementation and maintenance costs. In a recent European study evaluating the use of ICIS, the authors reported that the main drawback to the purchase of an ICIS was the substantial financial cost associated with its initial implementation [2]. This conclusion has also been reached by other authors [30]. Using a simple configuration (program and screen), in a recent study, we have showed that following implementation of the ICIS, a rapid return on investment could be achieved. The extra cost per bed of installing an ICIS is between €20,000 and €25,000, or €300,000 to €375,000 for our entire unit (15 beds). Based on the current billing rates in our institution, these improvements alone have increased the ICU revenue uptake by nearly €200,000 in its first year of use [28].

We are aware that our study had some limitations, the first being its retrospective nature. This was a single and specialised centre evaluation (Liver Intensive Care Unit for medical and surgical patients) so the generalizability of the result could be limited. In addition, our evaluation is based on an observational study over a period of 3 years during which some therapeutic strategies and treatment protocol changes could have influence the outcome of the patients. However, we checked for this confounding factor. No major changes of our practices and procedures occurred during 2004–2006 time periods. In addition, when comparing the length of ICU stay after implementation over the following years, we did not observe a significant difference (2007: 6.49 ± 11.95 days, 2008: 6.39 ± 9.78 days, 2009: 6.18 ± 10.51 days). Hence, to prove that the ICIS is a key contributing factor to improved outcome in ICU patients, further evaluation should be conducted to confirm our results.

In conclusion, the results of the present analysis show that the implementation of ICIS in the ICU impacts upon the outcome of ICU patients suggest that it may improve the quality of care as well as.

The authors would like to thank the Department of Medical Information (DMI) at Paul Brousse Hospital for assistance with data management and the Liver Intensive Care Unit Metavision Team at Paul Brousse (Rashmi Tripathy, Beatrice Nieda, Emmanuelle Brisseaux, Maël Ryan and Benoit Vincent) and G Dhonneur (M.D., PhD, Hôpital Henri Mondor, Creteil France) for reviewing this article and making insightful comments and suggestions.