Review of Respiratory Medicine - Volumen 24, Número 2 - June 2024

Original Articles

Mortality in Patients Requiring Prolonged Mechanical Ventilation: A Retrospective Cohort Study with Propensity Score Analysis

Mortalidad de pacientes en ventilación mecánica prolongada: Estudio de cohorte retrospectiva con análisis de propensión

ABSTRACT

Background: An increase has been observed in the number of patients requiring spe­cialized care in mechanical ventilation weaning and rehabilitation centers (MVWRCs).

Methods: An observational study with propensity score analysis was conducted on a 13- year cohort of patients in a MVWRC in Argentina. Predictors of mortality were analyzed.

Results: Mortality assessed using the inverse probability of treatment weighting was as­sociated with age [OR=1.037 (95% CI: 1.023-1.052), p<0.001], weaning from mechanical ventilation (MV) [OR=0.398 (95% CI: 0.282-0.560), p<0.001], decannulation [OR=0.059 (95% CI: 0.038-0.091), p<0.001], history of cardiovascular disease [OR=1.684 (95% CI: 1.146-2.474), p<0.001], pneumonia in non-chronic obstructive pulmonary disease (non-COPD) [OR=2.649 (95% CI: 1.631-4.302), p<0.001], and COPD [OR=0.477 (95% CI: 0.298-0.762), p=0.002].

Multiple logistic regression analysis in the propensity score-matched sample indicated that weaning from MV [OR=0.313 (95% CI: 0.137-0.715), p=0.006] and decannulation [OR=0.057 (95% CI: 0.021-0.155), p=<0.001] remained associated with lower mortal­ity, whereas age [OR=1.056 (95% CI: 1.026-1.087), p=<0.001] remained a predictor associated with higher mortality.

Conclusion: Mortality in patients requiring MV in a MVWRC was independently associ­ated with older age, failed weaning from MV, and non-decannulation. It is very important to identify such predictors in order to plan attainable treatment goals.

Key words: Critical Care, Assisted ventilation, Tracheostomy

RESUMEN

Introducción: Se ha observado un aumento en el número de pacientes que requieren ser derivados a centros de desvinculación de ventilación mecánica y rehabilitación.

Material y métodos: Se realizó un estudio observacional con análisis por puntaje de propensión en el que se analizaron los predictores de mortalidad en una cohorte de trece años internados en un centro de desvinculación de ventilación mecánica y rehabilitación.

Resultados: La mortalidad analizada mediante ponderación por el inverso de la proba­bilidad de tratamiento se asoció a edad [OR = 1,037 (IC95 % 1,023-1,052), p < 0,001], desvinculación de la ventilación mecánica (VM) [OR = 0,398 (IC95 % 0,282-0,560), p < 0,001], decanulación [OR = 0,059 (IC95 % 0,038-0,091), p < 0,001], antecedentes cardiovasculares [OR = 1,684 (IC95 % 1,146-2,474), p < 0,001], neumonía en no en fermedad pulmonar obstructiva crónica [OR = 2,649 (IC95 % 1,631-4,302), p < 0,001] y la presencia de enfermedad pulmonar obstructiva crónica [OR = 0,477 (IC95 %0,298- 0,762), p = 0,002] El análisis de regresión logística múltiple de la muestra emparejada mantuvo la aso ciación entre la desvinculación de la ventilación mecánica [OR = 0,313 (IC95 % 0,137- 0,715), p = 0,006] y la decanulación [OR = 0,057 (IC95 % 0,021-0,155), p ≤ 0,001] como variables asociadas a una menor mortalidad y a la edad [OR = 1,056 (IC95 % 1,026-1,087), p ≤ 0,001] como predictora asociada a mayor mortalidad.

Conclusión: La mortalidad en pacientes con ventilación mecánica en un centro de des­vinculación de ventilación mecánica y rehabilitación se asoció de manera independiente a una mayor edad, imposibilidad para la desvinculación de la ventilación mecánica y la no decanulación. Es importante contar con dichos predictores para poder planificar objetivos de tratamiento reales.

Palabras claves: Cuidados críticos, Ventilación asistida, Traqueostomía

Received: 04/11/2024

Accepted: 06/16/2024

INTRODUCTION

In recent decades, greater interest has been observed in patients with physical, psychological, and cognitive sequelae who were discharged from the Intensive Care Unit (ICU) requiring prolonged mechanical ventilation (PMV), that is to say, using mechani­cal ventilation for more than 21 days, more than 6 hours a day.¹ These patients are often referred to as “chronic critical patients” due to their clinical condi­tion overlapping with a state of persistent inflamma­tion and high catabolism.² Functional dependence, the need for skilled care, and high mortality justify the implementation of measures to try to improve clinical practice in order to achieve the best possible outcomes in the shortest period of time.

In our country, this population of patients dis­charged from the ICU can be treated in mechanical ventilation weaning and rehabilitation centers (MVWRCs). It is difficult to compare outcome descriptions, as there is a great diversity of admis­sion and discharge criteria in different countries. However, better percentages of mechanical ven­tilation weaning and higher survival rates have been described in patients admitted to MVWRCs compared to those who remained in the ICU.^3-6

Identifying and recognizing predictors of mor­tality in chronic critical patients would facilitate decision-making for the patient, the physician, and the family. Advanced age, number of comorbidities, the etiology of the acute respiratory failure, and failed weaning from mechanical ventilation (MV) have been recognized in various studies as predic­tors of mortality.^{1, 3-6} In our country, there is limited information on the population of patients on PMV (prolonged mechanical ventilation).^{7, 8}

The objective of this study was to analyze the re­lationship between overall mortality and weaning in patients on PMV in a MVWRC and to identify factors associated with mortality.

MATERIALS AND METHODS

A retrospective observational cohort study was conduc­ted using propensity score analysis from January 2007 to December 2019 at Clínica Basilea, a privately funded mechanical ventilation weaning and rehabilitation center with 60 beds located in Buenos Aires, Argentina.

The study included all patients recorded in a database who were admitted during the study period and had signed the consent for the use of anonymized data for epidemio­logical research.

Exclusion criteria: patients admitted to the MVWRC without a tracheostomy cannula, not requiring MV, with missing data in their medical records, or those who were discharged from the MVWRC for reasons other than death or medical discharge (i.e., patients discharged because they were transferred to the ICU, by decision of the health coverage, or the family).

The primary variable of analysis was mortality during hospitalization at the MVWRC. Weaning from MV was considered an independent variable.

The institution’s weaning protocol stipulates that a pa­tient on MV can begin a nocturnal spontaneous ventilation (SV) test after having ventilated spontaneously for 12 hours during the day, for the previous three days. Additionally, we used the definition established at the 2005 consensus conference,¹ where a patient is considered successfully weaned from MV after being without it for seven consecu­tive days.⁹ Patients were considered “unweaned from MV” if they required continuous or partial MV at the time of discharge from the MVWRC.

The following predictive variables were collected: pre- ICU history (age, gender, medical history); ICU admission data (reason for admission: medical, surgical, or polytrau­ma; diagnosis at ICU admission; days in ICU; and days on MV in the ICU); MVWRC hospitalization data (days on MV and days hospitalized in the MVWRC; decannulation; recannulation); and reason for discharge (discharge to home or death).

Statistical analysis

A description of the complete sample was performed. Continuous variables that assumed a normal distribution were reported as mean and standard deviation (SD); otherwise, the median (Med) and interquartile range (IQR) were used. Categorical variables are reported as frequency and percentage. The Kolmogorov-Smirnov test was used to determine the distribution of the sample.

Baseline comparison of the cohort

Clinical and demographic variables were compared between patients who achieved weaning and those who failed. The Mann-Whitney test was used for numerical variables, while Pearson’s Chi-square test or Fisher’s exact test, as appropriate, were used for categorical variables in the distribution of the double entry table.

Calculation of propensity score (PS), inverse probability of treatment weighting, and PS matching

Given the observational, uncontrolled nature of the study and the suspected differences between the groups to be compared, a propensity score (PS) for weaning was calcu­lated as an independent variable. This aimed to minimize differences between weaned and non-weaned patients and create comparable groups.

The PS was calculated using a binary logistic regres­sion model that included variables related to weaning and mortality, namely: age, gender, successful decannulation, recannulation, total days on MV in the ICU, total days on MV in the MVWRC, total days of ICU hospitalization, total days of hospitalization in the MVWRC, medical his­tory (cardiological, respiratory, neurological, metabolic, or oncological), and reason for admission (medical, surgical, or polytrauma). The sample size considered was 10 cases of the dependent variable per variable included in the bi­nary logistic regression model for the construction of the propensity score.

The PS should be considered as the individual proba­bility of each subject being weaned according to the pre­dictive variables. The PS was used in two ways: as inverse probability of treatment weighting (IPTW) applied to the entire cohort, and outcomes between the weaned and not-weaned groups were compared. The weight was defined as the inverse of the propensity score for the exposed group (weaned) and the inverse of one minus the propensity score for the unexposed group (not-weaned).^{10, 11}

On the other hand, in order to confirm the obtained result, a tolerance of 0.05 standard deviation of the logit of the PS was used without replacement to generate PS matching through the radius method.

Multivariate analysis for mortality

A sensitivity analysis was conducted on the sample both before and after matching to identify factors related to the development of adverse events. To do that, a logis­tic regression model with stepwise selection was used, using the maximum likelihood method. The goodness of fit of both final predictive models was verified using the Hosmer-Lemeshow test. Additionally, the discriminatory power was evaluated through the ROC (Receiver Operating Characteristic) curve analysis. The area under the curve (AUC) was used to determine the level of precision. The AUC ranges from 0 to 1; an AUC < 0.5 indicates that the model’s performance is worse than chance, while an AUC= 1 indicates perfect performance. AUC values > 0.7 and > 0.9 are considered to represent acceptable and excellent performance, respectively.

Power analysis

A sample size of 421 subjects was calculated, which were sufficient to detect a 17% difference in mortality between the two exposure groups (weaned and unweaned), based on previous publications, considering the lower limits of the confidence intervals for mortality in this population according to the exposure factor (46% in unweaned and 29% in unstable weaned patients).¹² The parameters used for the sample size calculation were set at a 95% power to detect differences between the two groups and a probability of type I error or alpha of 5%.

Matching was performed using the PS and statistical analysis with SPSS software, version 25 (SPSS, Chicago, USA), and R Version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria; available through https://cran.r-project.org/mirrors.html). For the power calculation, G*power software was used, version 3.1.9.2 (University of Düsseldorf, Düsseldorf, Germany).¹³ A p-value < 0.05 was considered statistically significant.

RESULTS

Characteristics of the entire sample

During the study period, 1,103 patients were ad­mitted. 76 patients were excluded due to missing data, 65 for not having a tracheostomy, 280 for not requiring MV upon admission to the MVWRC, and 5 for still being hospitalized at study closure (Figure 1). 223 patients who were discharged from the MVWRC for reasons other than death or medical discharge were also excluded. 454 patients were analyzed. The median age was 72 years (IQR 61-78); 55.1% were men. The median number of hospitalization days in the ICU was 33 (IQR 25 - 45). 87% (395) of the patients were admitted to the ICU for medical reasons, 11% (51) for surgical reasons, and the remaining 2% (8) for polytrauma.

Baseline comparison of the entire cohort

The comparison of the entire cohort between patients weaned and unweaned from MV showed differences in the median age [56 years (45 - 70) versus 64 years (51 - 73); p=0.003]. With regard to the reasons for ICU admission, weaning was less frequent among patients with exacerbated COPD and more frequent among patients with stroke and Guillain-Barré syndrome. Unweaned patients had a median of 31.5 days on MV in the MVWRC (IQR 13.5-38), and 61 total days includ­ing both ICU and MVWRC (IQR 61-94), compared to weaned patients who had a median of 14 days (IQR 13.5-63.2), and 49 days (IQR 37-76.5), re­spectively. Also, a higher proportion of weaned patients achieved decannulation compared to not weaned patients. Mortality in the MVWRC was significantly higher in the not weaned group compared to the weaned group [70.9 vs. 27.5%; p<0.001] (Table 1).

Mortality analysis of the entire cohort

Overall mortality was independently associated with older age [OR=1.042 (95% CI: 1.023-1.062), p<0.001] and a history of cardiovascular disease [OR=1.748 (95% CI: 1.035-2.951), p=0.037].

Conversely, the presence of a history of respiratory disease [OR=0.588 (95% CI: 0.356-0.971), p=0.038], mechanical ventilation weaning [OR=0.439 (95% CI: 0.247-0.777), p=0.005], and decannulation [OR=0.096 (95% CI: 0.051-0.178), p<0.001] de­creased the chances of death in our sample (Table 2).

Mortality analysis of the entire cohort by inverse probability of treatment weighting (IPTW)

Mortality analyzed using IPTW yielded results consistent with those observed in the initial unweighted cohort analysis. In addition to the previously mentioned variables, the presence of pneumonia in non-COPD patients was added as an independent factor for higher mortality. Besides the same factors as in the unweighted model, lower mortality was associated with MV wean­ing [OR=0.398 (95% CI: 0.282-0.560), p<0.001], tracheostomy decannulation [OR=0.059 (95% CI: 0.038-0.091), p<0.001], and the presence of COPD instead of any respiratory history [OR=0.477 (95% CI: 0.298-0.762), p=0.002] (Table 3).

Mortality analysis of the matched sample using the propensity score (PS)

The probability of weaning before and after match­ing by propensity score is shown in Figures 2A and 2B, respectively.

The performance of the PS in matching weaned and unweaned patients and creating comparable groups is illustrated in Figure 3. All variables included in the PS exhibit a mean difference of less than 0.2.

The characteristics of the sample after matching are detailed in Table 4.

The multiple logistic regression analysis of the matched sample maintained the association between MV weaning [OR=0.313 (95% CI: 0.137- 0.715), p=0.006] and tracheostomy decannulation [OR=0.057 (95% CI: 0.021-0.155), p<0.001] as variables associated with lower mortality and age [OR=1.056 (95% CI: 1.026-1.087), p<0.001] as a predictor associated with higher mortality. (Table No. 5)

 

DISCUSSION

In recent decades, there has been a substantial in­crease in both the number of patients on prolonged mechanical (PMV) and in mechanical ventilation weaning and rehabilitation centers (MVWRCs) in our country. A variable number of patients achieve complete recovery and are subsequently discharged, but there is also a significant group of patients who do not succeed and die during hospitalization in the MVWRCs. ^3-9 In Argentina, patients who cannot be weaned from mechanical ventilation do not have access to institutions for their hospitalization, nor is there home hospital­ization available for MV-dependent patients. As a result, they usually remain hospitalized until their death.

Being able to predict which patients have a higher or lower chance of being discharged could benefit both the planning of patient recovery pro­cesses and family organization.

We have been able to analyze 13 years of follow-up data from our MVWRC database and apply the propensity analysis in order to balance the groups of weaned and unweaned patients, with a significant sample of patients. Three factors were related to mortality.

In all the models we used, age remained an in­dependent predictor associated with mortality. The older the age, the higher the mortality. This find­ing aligns with numerous published studies.^{1, 17,22}

On the other hand, achieving MV weaning and decannulation in the MVWRC are protective fac­tors against mortality. Considering the tracheos­tomy decannulation as a protective factor against mortality is consistent with the findings of Scrigna et al (p = 0.0001, OR 7.51; CI: 2.77 – 20.38). In this study, 80% of the 44% of patients who were decannulated were medically discharged, while only 15% of patients in the non-decannulated group were discharged. Additionally, follow-up after discharge showed that the median survival among decannulated patients was 45 months (CI: 15.1 to 75.1) compared to 11 months (CI: 2.1 to 19.6) for non-decannulated patients (p = 0.004).²⁰

The study by Jubran et al found that one-year survival was 66.9% for patients disconnected from MV and 16.4% for those still on MV, which is consistent with our findings that MV weaning is a protective factor against mortality.²¹ This study followed a cohort of patients discharged from a weaning center for one year and highlighted in its discussion that “instead of limiting (or abandon­ing) weaning efforts based on the perception of a poor prognosis, physicians should adopt a more aggressive approach and assess the patient’s per­formance in the total absence of ventilator support, which facilitates early ventilator discontinuation. This approach minimizes the risk of failing to identify patients who can be separated from the ventilator.” All of this knowing that patients who are successfully weaned off have a higher chance of being discharged. This method of carrying out the weaning process is similar to the protocol at our institution, where patients are tested during spontaneous ventilation (SV) without MV support, based on prior work by the same group.²³

It is not possible to assert that the higher mor­tality observed in the group dependent on MV and non-decannulation of the tracheostomy is solely dependent on these two conditions. It may, perhaps, reflect an overall poorer health status of the patients associated with worse prognosis and survival. Patients who do not achieve MV weaning and tracheostomy decannulation do not get medi­cally discharged, thereby prolonging their stay in the MVWRC.

Some countries have intermediate care centers or home hospitalization systems for patient refer­ral in cases where the patient doesn’t improve within a certain period, facilitating the discharge of patients with artificial airways and need for MV. This could be the reason why such centers report higher survival rates and certainly do not attribute their mortality rates to weaning and decannulation failure.

Another reason preventing the discharge of tracheostomized patients is the inability to pro­vide home support, depending on the geographic area where they live and their family’s capacity to accommodate home hospitalization. If this is the reason for prolonged hospitalization, it is possible that mortality is more associated with the length of stay than with the patient’s severity or instability.

Knowing the patient’s probabilities based on their characteristics at admission to the MVWRC and their progress there (success in MV weaning and decannulation) allows for better decision-making and improving communication with the patient and/or their family. This helps avoid cre­ating false expectations and prioritizes the best quality of life.

The main limitation of this work is that it is a retrospective analysis, where only data obtained from medical records were analyzed. However, including a propensity analysis allowed for a balanced sample concerning the variable of MV weaning, leading to a better analysis of mortality within the MVWRC. As the study was conducted in a single institution, it also limits the extrapola­tion of the results.

CONCLUSION

This study showed that mortality in patients requiring MV in a MVWRC was independently associated with older age, failure to wean from MV, and non-decannulation. It is vitally important to have these predictors in order to plan achievable treatment goals in collaboration with the patient and their family.

Conflict of interest

None to declare

REFERENCES

1. MacIntyre NR, Epstein SK, Carson S, Scheinhorn D, Chris­topher K, Muldoon S. Management of patients requiring prolonged mechanical ventilation - report of a NAMDRC consensus conference. Chest 2005;128:3937-54. https://doi.org/10.1378/chest.128.6.3937

2. Carson SS, Bach PB. The epidemiology and costs of chronic critical illness. Crit Care Clin. 2002;18:461-76. https://doi.org/10.1016/S0749-0704(02)00015-5

3. Mauri T, Pivi S, Bigatello L. Prolonged mechanical ventila­tion after critical illness. Minerva Anestesiol. 2008;74:297- 301.

4. Mamary A, Kondapaneni S, Vance G, Gaughan JP, Martin UJ, Criner GJ. Survival in Patients Receiving Prolonged Ventilation: Factors that Influence Outcome. Clin Med Insights Circ Respir Pulm Med. 2011;5:17-26. https://doi.org/10.4137/CCRPM.S6649

5. Scheinhorn DJ, Artinian B, Catlin J. Weaning from prolonged mechanical ventilation: the experience at a re­gional weaning center. Chest 1994;105:534-9. https://doi.org/10.1378/chest.105.2.534

6. Lone N, Walsh T. Prolonged mechanical ventilation in critically ill patients: epidemiology, outcomes and modelling the potencial cost consequences of establishing a regional weaning unit. Critical Care 2011;15: R102. https://doi.org/10.1186/cc10117

7. Ambrosino N, Goldstein R. Ventilatory Support for Chronic Respiratory Failure. Informa Healthcare. 2008. https://doi.org/10.3109/9781420020229

8. Villalba D, Plotnikow G, Feld V, RiveroVairo N, Scapellato J, Díaz Nielsen E. Weaning from prolonged mechanical ventilation at 72 hours of spontaneous breathing. Medicina (B Aires). 2015;75:11-7.

9. Villalba D, Gil Rossetti G, Scrigna M, et al. Prevalence of and Risk Factors for Mechanical Ventilation Reinstitution in Patients Weaned from Prolonged Mechanical Ventila­tion. Respir Care. 2020;65:210-6. https://doi.org/10.4187/respcare.06807

10. Schafer JL, Kang J. Average causal effects from nonran­domized studies: a practical guide and simulated example. Psychol Methods. 2008;13:279-313. https://doi.org/10.1037/a0014268

11. Newgard CD, Hedges JR, Arthur M, Mullins RJ. Ad­vanced statistics: the propensity score--a method for estimating treatment effect in observational research. Acad Emerg Med. 2004;11:953-61. https://doi.org/10.1197/j.aem.2004.02.530

12. Schönhofer B, Euteneuer S, Nava S, Suchi S, Köhler D. Survival of mechanically ventilated patients admit­ted to a specialised weaning centre. Intensive Care Med 2002;28:908-16. https://doi.org/10.1007/s00134-002-1287-5

13. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods. 2007;39:175-91. https://doi.org/10.3758/BF03193146

14. Rose L, Fraser IM. Patient characteristics and outcomes of a provincial prolonged-ventilation weaning centre: a retrospective cohort study. Can Respir J 2012;19:216-20. https://doi.org/10.1155/2012/358265

15. Scheinhorn DJ, Hassenpflug MS, Votto JJ, et al. Ventilator dependent survivors of catastrophic illness transferred to 23 long-term care hospitals for weaning from prolonged mechanical ventilation. Chest 2007;131:76-84. https://doi.org/10.1378/chest.06-1079

16. Rojek-Jarmuła A, Hombach R, Gierek D, Krzych ŁJ. A single centre seven-year experience with mechanical venti­lation weaning. Anaesthesiol Intensive Ther 2015;47:204-9. https://doi.org/10.5603/AIT.2015.0033

17. Mifsud Bonnici D, Sanctuary T, Warren A, et al. Prospective observational cohort study of patients with weaning failure admitted to a specialist weaning, rehabilitation and home mechanical ventilation centre. BMJ Open 2016;6:e010025. https://doi.org/10.1136/bmjopen-2015-010025

18. Hong-Joon Shin, Jin-Sun Chang, SeongAhn, et al Clini­cal factors associated with weaning failure in patients requiring prolonged mechanical ventilation. J Thorac Dis. 2017;9:143-50. https://doi.org/10.21037/jtd.2017.01.14

19. Bigatello LM, Stelfox HT, Berra L, Schmidt U, Gettings EM. Outcome of patients undergoing prolonged mechanical ven­tilation after critical illness. Crit Care Med 2007;35:2491-7. https://doi.org/10.1097/01.CCM.0000287589.16724.B2

20. Scrigna M, Plotnikow G, Feld V, et. al. Decannulation after the ICU Stay. Analysis of 181 Tracheostomized Patients. Rev Am Med Resp 2013;2:58-63.

21. Jubran A, Grant B, Duffner LA, et. al. Long-Term Outcome after Prolonged Mechanical Ventilation. A Long-Term Acute-Care Hospital Study.Am J RespirCrit Care Med. 2019;199:1508-16. https://doi.org/10.1164/rccm.201806-1131OC

22. Engoren M, Arslanian-Engoren C, Fenn-Buderer N. Hospi­tal and long-term outcome after tracheostomy for respira­tory failure. Chest. 2004;125:220-7. https://doi.org/10.1378/chest.125.1.220

23. Jubran A, Grant BJ, Duffner LA, et al. Effect of pressure support vs unassisted breathing through a tracheos­tomy collar on weaning duration in patients requir­ing prolonged mechanical ventilation: a randomized trial. JAMA. 2013;309:671-7. https://doi.org/10.1001/jama.2013.159