Mechanical ventilation (MV) is probably the most widely used therapeutic procedure in Intensive Care Units (ICUs), and is a technique with a long history. Following a period of ventilation with negative pressure, induced by the invention of the iron lung in 1929, Ibsen introduced positive pressure ventilation outside the operating room in 1952.1 This marked the birth of the Intensive Care Unit. Observational studies2–10 have shown the percentage of patients admitted to the ICU who require MV to be between 33–53%. Studies in the general population11 in turn have found that about 2% of all adults receive MV at one time or other (39% during more than 96 h), and the percentage is moreover rising.12–16

Although the use of MV is associated to a decrease in mortality among patients with acute respiratory failure (ARF), the technique is not without complications17 — the most important of which is ventilator-induced lung injury, which can sustain or worsen lung dysfunction. As a result, in the last few decades many clinical trials have been carried out with the purpose of preventing or minimizing ventilator-induced lung injury, based on the use of noninvasive positive pressure ventilation,18 lung protective ventilation strategies,19,20 adjustment of positive end-expiratory pressure (PEEP),21 ventilation in prone decubitus22 and the early use of neuromuscular blockers,23 and of reducing the duration of MV through the adjustment of sedation24,25 and early identification of the moment for starting patient weaning from MV.26 Some of these interventions, which initially focused on the management of patients with acute respiratory distress syndrome (ARDS), now appear to be applicable to all patients subjected to MV.27

The present study was carried out to evaluate the changes that have taken place in the way in which MV is applied in patients admitted to Spanish ICUs participating in four international studies on MV. In addition, an analysis was made to determine whether such changes are accompanied by changes in patient outcomes, and whether variability is observed conditioned to the geographical setting involved, according to the different regional societies of the Spanish Society of Intensive and Critical Care Medicine and Coronary Units (Sociedad Española de Medicina Intensiva, Crítica y Unidades Coronarias [SEMICYUC]). Some of the results of this analysis have been previously published in a Doctoral Thesis.28

A post hoc analysis was made of four prospective, multicenter observational studies carried out during a one-month period in the years 1998,4,29,30 2004,31 201032 and 2016.33 The studies included patients requiring invasive MV during more than 12 h or noninvasive ventilation during more than one hour. For the purpose of the present analysis, we only included those patients admitted to ICUs in Spain.

The methodology used was similar in all four studies, with some differences in the recorded variables (the list of variables compiled in each study are described in Table S1 of the Supplementary material). All the studies recorded the following variables: basal demographic data (age, gender, estimated weight and height, severity estimated by the SAPS II score), and reason for starting MV. Likewise, the following parameters were recorded on a daily basis and for as long as the patient was subjected to MV (up to a maximum of 28 days): arterial gases, parameters programmed and measured by the ventilator (mode, tidal volume, respiratory frequency, fraction of inspired oxygen, PEEP, peak pressure, plateau pressure), administration of sedatives and neuromuscular blockers, and the appearance of complications such as ARDS, ventilator-associated pneumonia (VAP), sepsis and organ dysfunction (cardiovascular, renal, hepatic, hematological), and the date of and situation at discharge from the ICU and from hospital. Table S2 of the Supplementary material describes the operating definitions used. The Ethics Committees of each hospital approved the study protocol, and the need for informed consent was adjusted to the decision of each Committee.

In the studies of 1998 and 2004, each investigator received a manual describing the data to be recorded and the definitions used, together with data recording forms in paper format. The investigators completed a form for each patient included in the study and forwarded it to the coordinating center (Hospital Universitario de Getafe, Madrid) for inclusion in an electronic database. In the studies of 2010 and 2016, data registry took place through a secure website. In addition, before the analysis, all the data were checked for atypical and potentially erroneous values and information. The cases with missing information referred to principal study variables were not included in the analysis. The STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement on observational cohort studies was followed.34

The main finding of our study is that the mortality rate among patients requiring MV in Spanish ICUs has decreased over the last 18 years, though the figure has remained stable at 27% since the year 2010 - albeit with a slightly higher severity index in 2016 (47 points, predicted mortality rate of 39%) than in 2010 (46 points, predicted mortality rate of 37%). These changes in mortality have been associated to changes in clinical practice, with the generalized adoption (though with differences depending on the geographical setting) of protective ventilation strategies, reducing tidal volume to allow airway pressure to be kept below the level considered to be damaging to the lungs.36

The mortality observed in this study corresponds to the lower mortality range reported in both studies including patients from different countries9,37 and in national studies.38–45 In the course of the 18-year period between the first and the last study, we recorded a progressive decrease in mortality, though the figures remained similar in the last two studies. Furthermore, this decrease has occurred despite a change in the case-mix of the ventilated patients, characterized by an increase in the latest studies of patients with neurological disease, whose mortality rate is higher than in patients with other disease conditions such as heart failure or COPD – the presence of which has gradually decreased in the Units that have participated in our studies. A relevant finding is the fact that no geographical variability of mortality in the ICU was observed.

The investigation of MV in the last two decades evidences changes in routine clinical practice,31–33 though there is still significant heterogeneity in its implementation.46 One of the changes would be an increased utilization of noninvasive ventilation as the first ventilatory support measure. Few studies have evaluated this evolution over time. A French study reported an increase from 16% in 1997 to 24% in 2011.18 In Spain we have also observed an increase in the use of noninvasive mechanical ventilation as first ventilatory support mode, with a percentage use similar to that reported in other countries47 – though in our case significant variability is seen among the different regional societies of the SEMICYUC.

Technological advances in the field of MV have caused ventilators to evolve from simple machines offering few options to microprocessor-based systems allowing a broad range of ventilation modes. At present there are almost 200 ventilatory mode denominations, though many of them are similar in terms of their operating characteristics.48 While to date no studies have demonstrated the superiority of one mode over the rest,49 the preferred mode – as in other studies39,42,44 – remains the volume assist-control strategy, though its use has gradually decreased in favor of partial support modes such as pressure support, or dual modes such as volume control ventilation with pressure regulation. Mention must be made of the gradual decrease in the use of synchronized intermittent mandatory ventilation (SIMV), with or without pressure support, to the point where it has now become a marginal mode.50 On the other hand, as in other countries,51 little use is made of more complex modes (proportional assist ventilation [PAV], neurally adjusted ventilatory assist [NAVA]) – possibly because these proportional modes are protected by manufacturing patent rights and are not available in all commercially available mechanical ventilators.

Probably the greatest change produced as a result of clinical research is the so-called lung protective ventilation strategy, characterized by ventilation with low tidal volume settings (<6 ml/kg ideal body weight) and high PEEP, with the aim of maintaining a plateau pressure <30 cmH2O.36 This strategy, initially indicated for patients with ARDS,19,20 has been extended to the general population of patients subjected to MV.27 It gradually has been incorporated to routine clinical practice. Observational studies have recorded a progressive decrease from 8 and 10 ml/kg reported in the studies published in the closing years of the last century3,5,29 to tidal volumes <8 ml/kg but >6 ml/kg, maintaining a plateau pressure at the safety limits in the most recent studies.38–40,42–45 We observed a similar trend in our case. There has been a gradual decrease in tidal volume, though settings of <6 ml/kg are not reached probably because the current volumes secure the target plateau pressure < 30 cmH2O and/or driving pressure <15 cmH2O.52 Fewer changes are seen in PEEP level. In the implementation of this strategy we have also recorded differences between the different geographical regions.

The changes observed in the ventilation strategies (increased use of noninvasive ventilation and progressive implementation of protective ventilation) may have influenced the observed changes in patient mortality. A meta-analysis including individual data from patients participating in 9 randomized clinical trials has reported that a decrease in daily water balance, plateau pressure and tidal volume, and an increase in PEEP, reduced the mortality rate among patients with ARDS over a period of 17 years.53 In a mediation analysis of the global patients included in the four international studies to which our own cohort belongs, we observed a moderate effect (< 25%) upon mortality on the part of the protective ventilation strategy in the subgroup of patients with PaO2/FiO2between 100–200.33

Our study has a number of limitations that may influence interpretation of the results. Although a significant number of ICUs participated in the study, they do not represent all such Units in Spain. Thus, although we do consider the participating Units to be highly representative, there may have been selection bias causing limitations in the drawing of conclusions. However, the changes observed in the 14 Units that participated in all four studies were similar.

In conclusion, over the last two decades there has been a significant decrease in mortality among ventilated patients in Spanish ICUs. These changes in turn may have been influenced by the increased use of ventilatory strategies designed to minimize ventilator-induced lung damage.

Oscar Peñuelas, Fernando Frutos-Vivar and Andrés Esteban, as coordinators of the four studies, had access to all the data of the studies and accept responsibility for their reliability.

Study design: Andrés Esteban, Fernando Frutos-Vivar, Oscar Peñuelas, Antonio Anzueto.

Data compilation (investigators in 3 or 4 studies): Antonio García-Jiménez, Raúl de Pablo, Manuel Valledor, Miquel Ferrer, Miguel León, José María Quiroga, Susana Temprano, Inmaculada Vallverdú, Rafael Fernández, Federico Gordo.

Statistical analysis: Alfonso Muriel, Oscar Peñuelas, Fernando Frutos-Vivar.

Data interpretation: Alfonso Muriel, Oscar Peñuelas, Fernando Frutos-Vivar.

Drafting of the manuscript: Fernando Frutos-Vivar.

Critical review of the manuscript: Oscar Peñuelas, Fernando Frutos-Vivar, Alfonso Muriel, Jordi Mancebo, Antonio García-Jiménez, Raúl de Pablo, Manuel Valledor, Miquel Ferrer, Miguel León, José María Quiroga, Susana Temprano, Inmaculada Vallverdú, Rafael Fernández, Federico Gordo, Antonio Anzueto, Andrés Esteban.

Final approval of the manuscript: Oscar Peñuelas, Fernando Frutos-Vivar, Alfonso Muriel, Jordi Mancebo, Antonio García-Jiménez, Raúl de Pablo, Manuel Valledor, Miquel Ferrer, Miguel León, José María Quiroga, Susana Temprano, Inmaculada Vallverdú, Rafael Fernández, Federico Gordo, Antonio Anzueto, Andrés Esteban.

The authors declare that they have no conflicts of interest.