COPD: Mortality and Readmissions in Relation to Number of Admissions with Noninvasive Ventilation

ABSTRACT

In severe COPD, patients having survived acute hypercapnic respiratory failure (AHRF) treated with noninvasive ventilation (NIV) have a high mortality and risk of readmissions. The aim was to analyze the prognosis for patients with COPD having survived AHRF and to assess whether previous admissions with NIV predict new ones.We conducted a retrospective follow-up analysis of 201 patients two years after NIV treatment of AHRF. Comparison of time-to-event in patients previously treated with NIV versus patients with no previous NIV treatment. We found a one-year mortality of 33.8% and high risks of: readmission (53.2%), any event (67.7%), and life-threatening events (49.8%). Patients with previous NIV treatments had an increased hazard ratio for life-threatening events: 1.60, p = 0.023 despite having lower in-hospital mortality than patients with no previous NIV treatment (18.9% vs. 33.1%, p = 0.043). We found that having survived one episode of AHRF considerably worsened the prognosis for the affected patients.The prognosis for patients having survived AHRF with NIV treatment is poor: the prognosis worsens with additional episodes of AHRF. Future research and treatment should focus on patients with repeated episodes of AHRF.

KEYWORDS:

• Exacerbations
• follow-up
• prognosis
• ventilatory failure

Introduction

According to the World Health Organization (WHO), chronic obstructive pulmonary disease (COPD) is the third leading cause of death by 2012, causing 3.1 million or 5.6% of the annual worldwide deaths (1).

In Denmark, COPD is the most frequent cause of admission to a medical ward with almost 23,000 admissions per year; this disease accounts for approximately 7% of the annual deaths (2–4).

As the disease progresses, the affected individuals have an increasing risk of acute exacerbations; with further progression, individuals having an exacerbation are at risk of developing acute hypercapnic respiratory failure (AHRF) with acidosis (pH < 7.35) and hypercapnia (P_a,CO2 > 6 kPa).

Noninvasive positive-pressure ventilation at the general ward is shown to reduce in-hospital mortality and need for intubation in patients with COPD, admitted with AHRF, first in a 2000 randomized controlled trial (RCT) by Plant et al. (5) and then in a systematic review (6). Acute noninvasive ventilation (NIV) at the general ward is now recommended as a first line treatment of AHRF due to COPD (7).

Attempts have been made to predict exacerbations of COPD. The large Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) study attempted to predict the exacerbations of COPD in 2138 individuals. It identified that AHRF due to COPD increased overall mortality and need for intubation in the patients. It also estimated that this led to approximately 7% of the deaths annually with increased hazard ratio (HR, −3.29; P < .001) (9), and that hospitalization for exacerbation significantly increased the risk of mortality (P < .001) (9).

Likewise, attempts have been made to predict AHRF and the need for NIV due to COPD:

In 2004, Chu et al. published their findings on the prognosis for patients having survived a NIV-treated AHRF due to COPD (10). The analysis was based on 118 patients who were admitted with a pH < 7.35, a P_a,CO2 > 6 kPa, and a P_a,O2 < 8 kPa. Patients who were admitted to an intensive care unit or who had a sleep-disordered breathing or AHRF due to other reasons than COPD were excluded. The 110 patients who survived after the admission had a one-year mortality of 49.1% and one-year risks of another life-threatening event (repeat AHRF with need of ventilation or death) of 63.3% and readmission of 79.9%.

Many studies on patients who have survived a NIV-treated AHRF due to COPD have compared their results to the mortality and readmission rates found in this study; and RCTs have based their power calculations on the study results. Among these were Struik et al. who conducted an RCT on Long-Term NIV (LTNIV) for patients having survived an AHRF due to COPD (11). Struik et al. argued that the rates presented by Chu et al. were too high compared to European standards and that the relatively high age and especially the low body mass index (BMI) in Chu et al.'s were not representative of European patients admitted due to an acute exacerbation of COPD (AECOPD) and AHRF (11). Thus, the rates presented by Chu et al. would not be transferable to a European setting.

We retrospectively reviewed the records of patients admitted in 2012 and 2013 to two of the largest pulmonary wards in Greater Copenhagen with need of NIV treatment for AHRF due to COPD.

Aims

‐	To characterize the population of patients with COPD who have been admitted for NIV treatment of an AHRF.
‐	To investigate the patients treated for an AHRF in 2012 and 2013.
‐	Inspired by the findings presented by Hurst et al. and Men with need of NIV s and that the relatively high age and especially evious NIV treatments of AHRF have an increased risk of readmissions, repeat AHRF and of death (8,9).

Methods

Three of the authors, SHS, DBR and KLA, reviewed the records of all patients admitted for acute NIV for any cause at the pulmonary wards at Bispebjerg and Gentofte Hospitals in Greater Copenhagen, 2012–2013.

Data extraction

The admission of patients in the study period was denoted the LA, reviewession.”

In accordance with a predefined protocol, the following data were extracted from the reviewed records: Baseline characteristics at the index admission, data concerning the index admission, and the patients' prognosis, i.e., number of and time to readmission to a pulmonary ward, to recurrent AHRF, and to death. The records were reviewed two years after discharge from the index admission.

Inclusion criteria

This analysis included any patient, treated with NIV, who survived AHRF due to COPD. COPD was diagnosed in accordance with Global initiative for chronic Obstructive Lung Disease (GOLD) guidelines (7); likewise, AHRF was defined in accordance with GOLD with pH < 7.35 and P_a,CO2 > 6 kPa despite initial treatment with oxygen, nebulized bronchodilators and intravenous corticosteroids (7). Patients who were treated successfully and who were discharged were offered a follow-up program at the outpatient department; if appropriate, the patients were encouraged to participate in rehabilitation and smoking cessation.

We also included patients who had been admitted for NIV treatment at the intensive care unit (ICU) and who were moved to the pulmonary ward for NIV weaning, provided they had been treated in accordance with GOLD and if they had not been intubated.

Comorbidity was not an exclusion criterion except if a chief physician had deemed the comorbidity to be the primary cause of AHRF. Likewise, patients who had a “do-not-intubate” (DNI) or “do-not-resuscitate” (DNR) order were not excluded.

Data analysis

For all dichotomous variables, the percentage of the total patients was denoted. The continuous variables were described using parametric statistics (mean, 95% confidence interval).

Subgroup analysis

We specifically compared the patients who had previously been treated with NIV due to AHRF prior to the index admission to the patients for whom the index admission was their first admission with NIV treatment.

For this comparison, all dichotomous variables were compared using simple X² statistics. Continuous variables were compared using t-tests. As in the original paper by Chu et al., we used Kaplan–Meier, log rank and Cox proportional hazards regression to compute HRs for the following events: Death, al hazards regression to compute simple or to,” and the following events: Death, al hazards regression to compute simp.”

Comparison to the results by Chu et al.

For all comparable items, we compared our results to the results by Chu et al. In order to make the data comparable, we applied the inclusion and exclusion criteria from Chu et al.'s “Introduction.”

Data processing

We used SAS (the SAS institute, version 9.4) for statistical analyses.

Ethics

All data and patient information were extracted and handled in accordance with Danish law, with guidelines from the Scientific Ethics Committee and the Data Protection Board.^{Table 1}

Results

Three hundred and four patients were admitted for acute NIV at the pulmonary wards at Gentofte and Bispebjerg Hospitals in 2012 and 2013.

Of these, 15 patients were excluded from the analysis as they had AHRF due to other causes than COPD (3 congestive heart failure, 1 asthma, 6 pneumonia, 2 pneumonia and pulmonary edema, 1 pneumonia and overdose of opiates, 1 malignant pleural effusion, and 1 idiopathic pulmonary fibrosis and pneumonia).

Thus, 289 patients were admitted for NIV treatment for AHRF due to COPD. Eighty eight of these patients (30.4%) died during the admission. Only three of the patients who died did not have a DNI/DNR order at the time of death. Among the patients who died, 73 patients (83.0%) had failure of NIV; 5 of these died during NIV.

Of the 289 admitted patients, 53 had previously been treated with NIV due to AHRF prior to the index admission, whereas 236 patients had not. Ten (18.9%) of the 53 patients with previous NIV treatment and 78 (33.1%) of the patients with no previous NIV treatment died during the index admission (p = 0.043). Two hundred and one patients were discharged and eligible for the general analysis. Forty three of the patients had previously been treated with NIV due to AHRF prior to the index admission, whereas 158 patients had not.

Baseline characteristics

The included patients were eligible for the general analysis. Notably, the patients who had been previously treated with NIV prior to the index admission differed from the patients who had no previous NIV treatment in multiple variables: They were younger (68.3 vs. 72.3 years, p = 0.039); a larger proportion of them received long-term oxygen therapy (LTOT) (41.9% vs. 15.3%, p = 0.0002); to a larger extent, they were ex-smokers than smokers (p = 0.04); and they had been admitted more frequently due to COPD in the last month (34.9% vs. 12.7%, p = 0.0007) and year (67.4% vs. 35.4%, p = 0.0002) and had more admissions (2.1 vs. 0.6, p < 0.0001) prior to the index admission.

The index admission

Data concerning the index admission are listed in Table 2.

Table 1. Baseline characteristics of the included patients.

	No. of patients analyzed	Total Mean (95%CI) or (%) n = 201	Previous NIV Mean (95%CI) or (%) n = 43	No previous NIV Mean (95%CI) or (%) n = 158	p
Age, years	201	71.5 (69.9–73.0)	68.3 (65.0–71.7)	72.3 (70.6–74.1)	0.039
Male (%)	201	87 (43.3%)	20 (46.5%)	67 (42.4%)	0.63
Moved from ICU for weaning	201	8 (2.5%)	1 (2.3%)	7 (4.4%)	0.66
BMI	117	26.2 (24.6–27.7)	27.6 (24.6–30.7)	25.7 (23.9–27.5)	0.28
FEV^1% of predicted	165	35.0 (32.7–37.3)	32.8 (37.4)	35.7 (33.1–38.3)	0.28
Median MRC score	141	4	4	4	0.57
LTOT	157	42 (21.0%)	18 (41.9%)	24 (15.3%)	0.0002
Smoker Never	199	10 (5.0%)	2 (4.7%)	8 (5.1%)	0.04
Ex		94 (47.2%)	27 (62.8%)	67 (42.4%)
Current		95 (47.7%)	13 (31.0%)	82 (51.9%)
Pack-years	189	43.6 (40.6–46.6)	47.6 (41.2–53.9)	42.5 (39.2–45.8)	0.16
COPD admission within 1 month prior to enrollment	201	35 (17.4%)	15 (34.9%)	20 (12.7%)	0.0007
COPD admission within 1 year prior to enrollment	201	85 (42.3%)	29 (67.4%)	56 (35.4%)	0.0002
No. of admissions due to COPD within 1 year	201	1.0 (0.7–1.2)	2.1 (1.6–2.6)	0.6 (0.4–0.9)	<0.0001

Table 2. Data concerning the index admission.

	No. of patients analyzed	Total Mean (95%CI) or (%) n = 201	Previous NIV Mean (95%CI) or (%) n = 43	No previous NIV Mean (95%CI) or (%) n = 158	p
Duration of admission, days	201	7.8 (7.0–8.6)	6.0 (4.2–7.7)	8.3 (7.4–9.2)	0.02
Pneumonia on chest X-ray	201	96 (47.8%)	16 (37.3%)	80 (50.8%)	0.12
Max CRP during NIV, mg/L	201	99.3 (83.8–114.7)	104 (70.1–137.1)	98.1 (80.6–115.5)	0.77
Fever (>38°C)	201	16 (8.0%)	3 (7.0%)	13 (8.2%)	0.79
RF upon NIV initiation	179	27.1 (26.1–28.1)	27.9 (25.8–30.0)	26.9 (25.8–28.0)	0.40
Antibiotics none	201	31 (15.4%)	12 (27.9%)	19 (12.0%)	0.04
Oral		22 (10.9&)	4 (9.3%)	18 (11.4%)
intravenous		148 (73.6%)	27 (62.8)	121 (76.6%)
Consciousness Awake	199	142 (71.3%)	34 (79.1%)	108 (69.2%)	0.55
at NIV Drowsy		28 (14.1%)	6 (14.0%)	22 (14.1%)
Awakened by speech		22 (11.1%)	2 (4.7%)	20 (12.8%)
Awakened by stimulation		3 (1.5%)	1 (2.3%)	2 (1.3%)
Unconscious		0	0	0
Acceptable SpO2% at 90–92% ordered	201	55 (27.4%)	12 (27.9%)	43 (27.2%)	0.92
DNI/DNR ordered No	201	143 (71.1%)	28 (65.1%)	115 (72.8%)	0.19
Upon/before admission		42 (20.9%)	13 (30.2%)	29 (18.4%)
> 2 hours after NIV initiation		16 (8.0%)	2 (4.6%)	14 (8.9%)
Arterial gases at pH	198	7.27 (7.26–7.28)	7.27 (7.25–7.29)	7.27 (7.26–7.28)	0.95
Admission CO2, kPa	196	9.6 (9.3–10.0)	9.6 (8.8–10.4)	9.6 (9.2–10.0)	0.97
O2, kPa	191	10.4 (9.7–11.0)	10.8 (9.3–12.3)	10.2 (9.4–11.0)	0.48
HCO3^−, mmol/L	167	27.0 (26.2–27.8)	28.4 (26.7–30.1)	26.6 (25.8–27.5)	0.07
Base Excess	174	5.1 (4.1–6.0)	6.8 (4.7–8.8)	4.6 (3.6–5.7)	0.07
Lactate, mmol/L	167	1.4 (1.3–1.6)	1.3 (1.0–1.6)	1.4 (1.3–1.6)	0.49
Arterial gases at NIV pH	153	7.27 (7.26–7.28)	7.28 (7.26–7.30)	7.27 (7.26–7.28)	0.28
CO2, kPa	152	9.5 (9.1–9.9)	9.3 (8.4–10.2)	9.6 (9.1–10.0)	0.65
O2, kPa	132	10.1 (9.3–10.9)	11.7 (9.9–13.5)	9.8 (8.9–10.6)	0.06
HCO3^−, mmol/L	129	26.8 (25.9–27.6)	27.4 (25.4–29.4)	26.6 (25.6–27.6)	0.47
Base Excess	130	4.7 (3.6–5.7)	5.6 (3.2–8.0)	4.5 (3.3–5.6)	0.39
Lactate, mmol/L	121	1.4 (1.3–1.6)	1.5 (1.0–1.9)	1.5 (1.3–1.6)	0.91
Delay from NIV indicated to treatment, hours	199	3.0 (2.5–3.5)	3.5 (2.4–4.7)	2.9 (2.3–3.5)	0.35
Other acute diagnoses than AECOPD	201				0.49
Congestive heart failure		20 (10.0%)	3 (7.1%)	17 (11.4%)
Asthma		7 (3.5%)	3 (7.1%)	4 (2.5%)
Acute coronary syndrome		1 (0.5%)	0	1 (0.6%)
Overdose		2 (1.0%)	0	2 (1.3%)
Max IPAP, cm H2O	201	12.9 (12.6–13.3)	13.4 (12.7–14.2)	12.8 (12.4–13.2)	0.13
Max EPAP, cm H2O	201	4.7 (4.6–4.9)	4.8 (4.4–5.1)	4.7 (4.5–4.9)	0.76
NIV time, hours	195	35.2	34.0 (20.6–47.5)	35.4 (28.8–42.1)	0.85

The two subgroups were alike. However, the patients who had a history of a previous NIV treatment were admitted for fewer days (6.0 vs. 8.3 days, p = 0.02) and they were treated with antibiotics to a lesser extent (p = 0.04). Furthermore, they tended to have higher concentrations and values of arterial bicarbonate and Base Excess (BE) at admission (28.4 vs. 26.6 mmol/L and 6.8 vs. 4.6 mmol/L, p = 0.07), but not at NIV initiation.^{Table 2Table 3}

Readmissions, events, life-threatening events and deaths

Data concerning the prognosis of the included patients are listed in Table 3.

Table 3. Data concerning the prognosis of the included patients.

	No. of patients analyzed	Total Mean (95%CI) or (%) n = 201	Previous NIV Mean (95%CI) or (%) n = 43	No previous NIV Mean (95%CI) or (%) n = 158	p
≥ 1 readmission	201	29 (14.4%)	6 (14.0%)	23 (14.6%)	0.92
1 year		107 (53.2%)	22 (51.2%)	85 (53.8%)	0.76
2 years		117 (58.2%)	24 (55.8%)	93 (58.9%)	0.72
≥1 event	201	49 (24.4%)	15 (34.9%)	34 (21.5%)	0.07
1 year		136 (67.7%)	32 (74.4%)	104 (65.8%)	0.29
2 years		150 (74.6%)	36 (83.7%)	114 (72.1%)	0.12
Hazard Ratio	1.33 (0.92–1.94), log likelihood −715.4				0.132
≥1 life-threatening event	201	34 (16.9%)	11 (25.6%)	23 (14.6%)	0.09
1 year		100 (49.8%)	28 (66.7%)	72 (45.6%)	0.02
2 years		119 (59.2%)	31 (72.1%)	88 (55.7%)	0.052
Hazard Ratio	1.60 (1.06–2.41), log likelihood −538.9				0.023
Mortality	201	25 (12.4%)	9 (20.9%)	16 (10.1%)	0.06
1-year		68 (33.8%)	18 (41.8%)	50 (31.6%)	0.14
2-years		88 (43.8%)	23 (58.2%)	65 (41.1%)	0.09
2-years hazard ratio	1.53 (0.95–2.46), log likelihood −442.7				0.078

The patients who had a history of a previous NIV treatment had a significantly higher risk of a new life-threatening event (death or repeat AHRF, treated with NIV) within the year following the index admission (66.7% vs. 45.6%, p = 0.02); and in a two-year Cox proportional hazards regression, the HR was 1.60 (1.06–2.41, p = 0.023), see Figure 1.

Figure 1. Kaplan–Meier curves of the time to a life-threatening event, i.e., repeat acute hypercapnic respiratory failure (AHRF) or death.

Overall, this group of patients tended to have a poorer prognosis in terms of mortality and events (COPD-related admissions including AHRF or death): 41.8% of the patients with a previous NIV treatment died whereas 31.6% of the patients with no previous NIV treatment died one year following the index admission (p = 0.14) (2-year HR = 1.53, p = 0.078).

Likewise, 74.4% of the patients with a previous NIV treatment had an event one year following the index admission whereas 65.8% of the patients with no previous NIV treatment had an event (p = 0.29) (2-year HR = 1.33, p = 0.132). Among these events, 23.3% of the patients with a previous NIV treatment died before having an admission to a pulmonary ward; among the patients with no previous NIV treatment, 12.0% died before having an admission (p = 0.06); there was no difference between the groups in regard to admissions due to COPD (51.2% vs. 53.8%, p = 0.76). Consequently, it makes no sense to look at readmission rates without taking mortality into consideration; arguably, the total number of events (i.e., a COPD-related admission (including AHRF admissions) or death) is a more valid and useful prognostic parameter.^{Figure 1}

Thus, we compared the number of COPD-related admissions in the year prior to the index admission with the number of events in the year following the index admission: Overall, 42.3% of the total 201 patients had had a COPD-related admission in the year prior to index admission whereas 67.7% of the 201 had at least one event in the following year (p < 0.0001). This increase was most remarkable among the patients with no previous NIV treatment: 35.4% had an admission in the year prior to the index admission and 65.8% had an event in the following year (p < 0.0001); the increase from 67.4% to 74.4% among the patients with a previous NIV treatment was not significant (p = 0.46).

Among the 158 patients with no previous NIV treatment, 56 patients (35.4%) had had a COPD-related admission in the year prior to the index admission whereas 102 patients (64.6%) had not. The 56 former had higher one-year risks of events (80.4% vs. 57.8%, p = 0.004) and of life-threatening events (57.1% vs. 39.2%, p = 0.03) whereas there was no difference between the groups in regard to one-year mortality (35.7% vs. 29.4%, p = 0.42).

A similar comparison among the 43 patients who were previously treated with NIV showed no differences.

The 42.3% of the 201 patients who had had a COPD-related admission within one year prior to the index admission had a 78.8% risk of an event one year after the index admission, whereas the 57.7% who had not been admitted had a 59.5% risk of an event the following year (p = 0.004).

Comparison to the results by Chu et al.

When applying the inclusion and exclusion criteria from the study by Chu et al. (see “Introduction”) on our patients, 133 patients were eligible. Eight of these patients died during the index admission. In the study by Chu et al., eight of the 118 eligible patients died (p = 0.80) (10).

Thus, our comparison was based on 125 surviving patients versus 110 patients in the study by Chu et al.

The 125 patients who were included for this comparison had similar values when compared to all the 201 patients who were described and analyzed above.

The 125 patients from our analysis and the 110 patients from the study by Chu et al. presented similar values in regard to age (71.5 (10.7) vs. 73.2 (7.6), p = 0.16), FEV¹_%predicted (33.8 (14.8) vs. 33.3 (16.6), p = 0.81), and respiratory rate at NIV initiation (27.1 (6.6) vs. 26 (6), p = 0.18). However, they differed in the following variables: The patients in our analysis consisted of less male patients (42.4% vs. 79.1%, p < 0.0001); had a much higher BMI (26.2 (8.6) vs. 20.2 (4.5) kg/m², p < 0.0001); were treated with LTOT to a lesser extent (20.0% vs. 52%, p < 0.0001); were current smokers to a larger extent (44.8% vs. 7.3%, p < 0.0001); had been less treated with NIV previously (20.0% vs. 33%, p = 0.037); had a higher incidence of pneumonias on the chest X-ray (48.0% vs. 30.9%, p < 0.0001); had a less severe degree of acidosis (pH = 7.26 (0.058) vs. 7.24 (0.005), p < 0.0001) at NIV initiation; were treated with lower inspiratory positive airway pressures (IPAPs) (13.1 vs. 16.2 cm H₂O, p = 0.009); and were treated NIV for a much shorter time (1 day (median, interquartile range [IQR] = 0–3) vs. 5 days (median, IQR = 3–7)).

Furthermore, the patients from our analysis presented a much lower occurrence of readmissions (56.0% vs. 79.9%, p < 0.0001), of life-threatening events (44.0% vs. 63.3%, p < 0.0001), and of deaths (29.6% vs. 49.1%, p < 0.0001).

Discussion

In this large, retrospective study we found that patients with COPD who survived AHRF after treated with NIV have a poor prognosis with high one-year risks of readmission, repeat AHRF and death. Compared to the original study by Chu et al., our study presented considerably lower incidences of readmissions, life-threatening events and mortality (10). The patients of the two studies were alike in regard to age and lung function but differed in regard to BMI, smoking status and use of LTOT. Arguably, the patients in Chu's study had a poorer general health than the patients in our study, explaining the poorer prognoses in Chu's study; namely, a BMI < 21 kg/m² is shown to predict mortality in COPD (12,13).

The patients in our study compare well to the patients in the control group of a RCT by Struik et al. (11): Here, 201 patients with COPD who survived an AHRF, treated with NIV, were randomized to either LTNIV or usual care (control) and followed for one year. As a criterion for inclusion, the patients had to be persistently hypercapnic, i.e., a P_a,CO2 > 6 kPa without acidosis.

Compared to the patients in our analysis, the included patients in Struik et al.'et al.d to the pager, 63.5 versus 71.5 years, and had poorer values of FEV¹_%predicted, 25.7% versus 35.0%; besides this and the persistent hypercapnia immediately after the AECOPD, which we did not measure in our patients, the patients in the two studies were alike.

In the year of follow-up, the study's 100 control patients presented rates of readmissions (57%), events (admissions, AHRFs and death, 64%) and mortality (29%) similar to the rates presented in our present analysis.

The rates of mortality and readmissions found in our study are similar to the rates for patients with COPD and chronic hypercapnia:

In a study on 252 patients receiving LTOT, Foucher et al. found that patients with a P_a,CO2 CO43 mmHg (n 252 pkPa) had a one-year mortality of 23.5% (14). Likewise, in an RCT on LTNIV for patients with COPD and stable, chronic hypercapnic respiratory failure, Köhnlein et al. found that the control patients—who were treated according to usual care and who had a mean Pa,CO₂ of 7.7 kPa—had a one-year mortality of 33% and a mean of 3.1 emergency hospital admissions per year (15). The authors did not clarify whether these admissions were strictly due to COPD or all-cause.

In a subgroup analysis, we compared the patients who had at least one previous admission with NIV treatment of AHRF prior to the index admission to the patients who had not. The patients with previous NIV treatment were younger and were to a larger extent treated with LTOT; they had been admitted more frequently prior to the index admission.

The patients with previous NIV treatment had a lower in-hospital mortality than the patients with no previous NIV treatment. Interestingly, the patients with previous NIV treatment had a higher post-discharge risk of a new life-threatening event (repeated AHRF or death); though not significant, the patients with previous NIV treatment tended to have higher HRs in regard to mortality and events (readmissions, repeat AHRF, or death).

To our knowledge, this correlation between previous NIV treatments of AHRF and new AHRFs or death has not been shown before.

The paradoxical finding that the patients with previous NIV treatment had a higher in-hospital survival and a poorer post-discharge prognosis than the patients with no previous NIV treatment is highly suggestive that the two groups are, in fact, different.

We believe the finding that multiple previous NIV treatments predict a new AHRF or death is in accordance with the findings by Hurst et al. (8) and Müllerova et al. (9), which is described in “Introduction.” Based on the finding of a higher in-hospital survival and a poor post-discharge prognosis, we argue that the patients with multiple previous NIV treatments of AHRF represent a distinct subgroup with a higher in-hospital survival and a poor post-discharge prognosis.

Likewise, we found that a previous COPD-related admission within the year prior to the index admission predicted events in the following year. This is also in accordance with Hurst's and Müllerovami findings (8,9).

In a large study on 6628 individuals with COPD, Lange et al. found that individuals with the most severe degree of COPD—categorized in the GOLD classification group D due to a high “risk” (airway flow limitation or a history of at least two exacerbations) and a high symptom score (modified Medical Research Council dyspnoea score [mMRC] ≥ 2 or COPD Assessment Test [CAT] ≥ 10)—had a one-year mortality of 3.4% and a one-year risk of a COPD-related admission of 13.6% (16). Furthermore, when analyzing the subgroups within the GOLD group D, the individuals with a FEV¹_predicted < 50% and a history of at least two exacerbations and a high symptom score had a one-year mortality of 7.1% and a one-year risk of a COPD-related admission of 36.4%.

The rates of mortality and admissions—found in our present analysis—for patients having survived an admission with NIV treatment of AHRF, exceed the rates of the patients within the most severe subgroup of the GOLD classification considerably. We found that the included patients had an increased rate of events in the year following the index admission compared to the rate of COPD-related admissions in the previous year. This increased rate was highly attributable to the patients who had not been treated with NIV prior to the index admission. 35.4% of these patients had had a COPD-related admission in the year prior to the index admission these patients had had a COPD-related admission in the year prior to the index admiion compared to the (16). In the year following the index admission 64.6% had an event. This is suggestive that patients with COPD who develop AHRF pass a threshold; after having survived AHRF, the patient's risk of new events increase considerably.

Based on these findings, we argue that if one were to redefine the GOLD subgroups it would be reasonable to consider previous admissions with need of NIV due to AHRF.

It would be interesting to confirm our findings in a larger study on patients with multiple previous NIV treatments and to assess whether previous NIV treatments actually predict death, which we were only able to show as a trend. However, a large study on patients with multiple NIV treatments would be difficult to conduct, as approximately one-third of the patients would die after each admission with NIV treatment, as shown by Struik et al. (11) and in the present study.

We argue that the group of patients who have survived an AHRF due to COPD, treated with NIV, should be paid special attention, as they are highly susceptible to readmissions, repeat AHRFs and death. We believe that, in a longer perspective, long-term noninvasive ventilation (LTNIV) should be paid special attention, as they have a highly susceptible prognosis for these vulnerable patients.

Limitations of the study

The study was based on a retrospective review of the records of patients who had been admitted for NIV at the two largest departments of pulmonary medicine in Greater Copenhagen. A retrospective setup is susceptible to selection bias, detection bias, reporting bias and confounding; arguably, the results of our study are not comparable to the results of Chu et al.'s study, the latter being prospective. However, by including any patient with COPD admitted for NIV treatment of AHRF and by reviewing the patients' records and obtaining data in accordance with a predefined study protocol, the risk of bias should be minimal. On the other hand, it could be argued that a prospective study could exclude some patients who do not wish to participate in a medical study for various reasons; this could cause a selection bias, not making the study population representative of the background population.

Although only patients admitted to Bispebjerg and Gentofte Hospitals were included in the study, the criteria for admission and for initiation of NIV treatment of AHRF are the same throughout Greater Copenhagen. Thus, the included patients should be representative of all similar patients in the region.

Conclusion

In this large follow-up study, we found that patients having survived acute hypercapnic respiratory failure treated with NIV due to COPD have a poor prognosis with high one-year risks of death, recurrent AHRF and COPD-related readmissions, which is considerably higher than the overall rates in GOLD group D and similar to the mortality in patients with chronic hypercapnia due to COPD. We found that after surviving one AHRF treated with NIV, the patient's risk for a new event increases considerably; this is suggestive that the development of AHRF indicates that the disease has exceeded a threshold and thereafter the affected individual becomes considerably more vulnerable and susceptible to serious COPD-related events.

In a subgroup analysis, we found that the patients with a history of multiple NIV treatments of AHRF had a lower in-hospital mortality but an increased post-discharge risk of repeated AHRF or death, compared to the patients who have had only one episode of AHRF and NIV. This is highly suggestive that the patients with multiple previous NIV treatments of AHRF represent a specific phenotype of severe COPD.