Behavioral Interventions and Symptom Cluster Change in Adults with Chronic Obstructive Pulmonary Disease and Insomnia

Abstract

Background:

People with chronic obstructive pulmonary disease (COPD) and insomnia experience multiple co-occurring symptoms, but no studies have examined symptom cluster change over time in this population.

Objectives:

This study explored longitudinal patterns of symptom cluster profiles for adults with COPD and insomnia and evaluated whether behavioral interventions were associated with changes in symptom cluster profiles.

Methods:

This study included 91 adults with COPD and insomnia who participated in a randomized trial of cognitive behavioral therapy for insomnia (CBT-I) and COPD education. The pre-specified symptom cluster included insomnia, dyspnea, fatigue, anxiety, and depression. Latent profile analysis identified participant groups with distinct symptom cluster profiles at baseline, immediately post-intervention, and at 3-month follow-up; latent transition analysis then estimated the probability of group membership change over time. Multinomial logistic regression was used to determine whether the interventions were associated with changes in symptom cluster profiles.

Results:

Three groups were identified at each of three time-points: Class 1 (low symptom burden), Class 2 (intermediate), and Class 3 (high). Classes 1 and 2 showed less movement to other classes (16% and 38%, respectively), whereas Class 3 showed greater transition (64%). The CBT-I intervention was significantly associated with movement to a lower symptom burden group (Class 3 to 2 or 2 to 1).

Conclusions:

CBT-I, with or without COPD education, shows promise as a tailored intervention to reduce symptom burden in the study population. Study findings will facilitate development of interventions to reduce the severity of multiple co-occurring symptoms in people with COPD and insomnia.

Clinical Trial Registration:

Registry: ClinicalTrials.gov; Name: A Behavioral Therapy for Insomnia Co-existing with COPD; Identifier: NCT01973647.

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality.¹ Over 40% of people with COPD are estimated to have insomnia, and the presence of insomnia in people with COPD has been associated with the prevalence of other symptoms such as dyspnea, fatigue, anxiety, and depression.^2,3 Although progressive respiratory symptoms of COPD predicted individuals’ daily function, quality of life, and mortality, lung function measurements may not suffice to predict negative health outcomes.^4–8 Therefore, more attention should be paid to self-reported symptom experiences among people with COPD and insomnia in order to identify their symptom burden.

People with comorbid COPD and insomnia experience multiple co-existing symptoms that can be grouped together as a symptom cluster. However, due to the heterogeneity of this population, a systematic review of symptom cluster research among people with COPD revealed inconsistent results across the few studies performed.⁹ Therefore, symptom cluster research focusing on particular groups of people with COPD can provide evidence for developing more targeted symptom management strategies. Additionally, one of the underlying assumptions for a symptom cluster is its temporal dimension, meaning that a symptom cluster can remain relatively stable or vary over time.¹⁰ However, only a limited number of studies have examined time trends of symptom clusters in people with COPD. Therefore, examination of longitudinal patterns of symptom clusters can provide better understanding of the characteristics of symptom cluster change over time.^8,9,11

Cognitive behavioral therapy for insomnia (CBT-I) is the first-line treatment for insomnia in people with COPD due to the serious or unknown side-effects of pharmacotherapy on the respiratory system.^12,13 Acute insomnia occurs as a result of medical or psychiatric illness, hyperarousal, worry, rumination, or an irregular sleep schedule, but chronic insomnia results from accumulated maladaptive coping behaviors and a mismatch between sleep opportunity and sleep ability.^14,15 CBT-I is a multicomponent therapy consisting of stimulus control, sleep restriction, sleep hygiene, relaxation training, and cognitive therapy.¹⁴ The efficacy of CBT-I has been evaluated for insomnia comorbid with a range of medical and psychiatric conditions,¹⁶ including cancer,¹⁷ chronic pain,¹⁸ and obstructive sleep apnea.¹⁹ The findings from these studies showed that CBT-I had moderate to large effects on insomnia severity as well as modest effects on comorbid outcomes.^2,16,20 Also, CBT-I was found to reliably reduce approximately 50% of symptoms during acute treatment, and this effect lasted over follow-up periods as long as 12 months.¹⁴ However, no study has evaluated the effect of CBT-I on symptom clusters in people with COPD and insomnia.

Most symptom management intervention studies have focused on a single symptom rather than on symptom clusters.¹⁰ Intervention for one symptom, however, may “cross-over” and reduce the severity of other symptoms included in a cluster.²¹ Although CBT-I focuses on insomnia, this treatment has contributed to secondary improvements in depression, fatigue, dyspnea, sleep quality, and quality of life.^6,13,22 Consequently, CBT-I may offer an optimal approach to managing or improving multiple concurrent symptoms clustered in people with COPD and insomnia. In addition, COPD-related education treatment was found to improve depression and anxiety in people with COPD, presumably by increasing their self-efficacy to exert control over their symptoms.^2,23 Therefore, examination of the effects of these behavioral interventions on a symptom cluster can provide new knowledge for development of interventions for symptom clusters in people with COPD and insomnia.

To date, no research has investigated how symptom clusters change over time during or after intervention in people with COPD and insomnia. Addressing these gaps is important because evaluation of the impact of behavioral interventions on longitudinal patterns of symptom clusters can determine the most efficacious and efficient intervention for specific groups of people with COPD and insomnia who promise to show the maximum treatment response. Additionally, such symptom cluster research can provide a useful model for advancing personalized and clinically effective interventions. However, a research gap exists as to which specific groups will maximally benefit from behavioral intervention. Therefore, the aims of this study were to explore longitudinal patterns of symptom cluster profiles for adults with concurrent COPD and insomnia and evaluate whether behavioral interventions were associated with changes in the symptom cluster profiles.

METHODS

Participants

This longitudinal study employed secondary analysis of data for adults with COPD and insomnia who participated in a randomized trial of behavioral therapy for insomnia and fatigue at three locations in Chicago, Illinois.²² Five pre-specified symptom variables were included as indicators of a symptom cluster based on empirical evidence of the relationship among the variables and the current literature. All five variables were measured at baseline, immediately post-intervention, and at 3-month follow-up.

The inclusion and exclusion criteria were reported earlier.²² In brief, people with mild to very severe COPD were included if they had self-reported insomnia (i.e., an Insomnia Severity Index [ISI] score of at least 8), were aged ≥45 years and had no other major health problems, and were clinically stable without major COPD exacerbation. This study excluded individuals who did not complete one or more of the measures employed at baseline, immediately post-intervention, and at 3-month follow-up. Data collection was performed from June 2014 to July 2019.

Behavioral interventions

Under the parent study’s randomized 2×2 factorial design, two intervention components, CBT-I (yes or no) and COPD education (COPD-ED) (yes or no), generated four intervention groups: (1) CBT-I + COPD-ED, (2) CBT-I, (3) COPD-ED, and (4) attention control (AC). Participants were randomly assigned to one of the four groups after baseline assessment.

The participants in each group were provided with six weekly individual sessions, each lasting 75 minutes. The CBT-I sessions included a program of stimulus control, sleep restriction, relaxation, cognitive therapy, and sleep hygiene.¹⁴ The COPD-ED sessions addressed COPD-related topics such as lung function and COPD medications or treatment equipment. The AC sessions provided general health education that was not related to sleep or COPD. The CBT-I + COPD-ED group received a combination of CBT-I and COPD-ED sessions, whereas the other three groups (i.e., CBT-I, COPD-ED, and AC) received AC sessions. The intervention protocol and the numbers of participants per group at each of the three time-points were detailed elsewhere.²²

Measures

Demographic and clinical characteristics

Demographic and clinical characteristics were assessed at baseline. Demographic characteristics included age, gender, race, ethnicity, marital status, number of years of school, and number of comorbidities measured using the Functional Comorbidity Index (FCI). Spirometry was used to measure the ratio of forced expiratory volume in 1 second to forced vital capacity (FEV₁/FVC) and FEV₁ percent predicted (FEV₁pp) to confirm the presence of COPD and to classify COPD severity, respectively. According to the COPD GOLD guidelines,²⁴ COPD was defined as FEV₁/FVC <0.7, and COPD severity was determined using a value of FEV₁pp: (1) mild (FEV₁pp ≥80%), (2) moderate (50%≤ FEV₁pp <80%), (3) severe (30%≤ FEV₁pp <50%), and (4) very severe (FEV₁pp <30% or FEV₁pp <50% plus chronic respiratory failure).

Symptom cluster

Insomnia.

The Insomnia Severity Index (ISI) was used to assess the nature, severity, and impact of insomnia during the previous 2 weeks.²⁵ Each item is scored using a 5-point Likert scale ranging from 0 (none) to 4 (very severe). Total scores range from 0 to 28, and higher scores indicate greater insomnia severity. The ISI has been demonstrated to have good reliability, with Cronbach’s alpha values ranging from 0.79 to 0.85, and validity for measuring subjective sleep experience in the COPD population.^2,26

Dyspnea and fatigue.

Dyspnea and fatigue were measured using two domains of the Chronic Respiratory Questionnaire (CRQ).²⁷ The dyspnea (CRQ-D) and fatigue (CRQ-F) domains assess the degree of dyspnea and fatigue experienced during certain daily activities using a 7-point scale. The CRQ-D uses a scale of 1 (extremely short of breath) to 7 (not at all short of breath), and the CRQ-F uses a scale of 1 (all of the time) to 7 (none of the time). Total scores for dyspnea and fatigue are calculated as the summed scores divided by the number of items in each domain, and lower scores indicate greater severity of dyspnea and fatigue. Among people with COPD, the CRQ-D (Cronbach’s alpha=0.86) and CRQ-F (Cronbach’s alpha=0.85) showed good reliability.²² The validity of the CRQ-D and CRQ-F has been demonstrated for the COPD population.²⁸

Anxiety and depression.

Anxiety and depression were measured using the Patient-Reported Outcomes Measurement Information System (PROMIS) computerized adaptive testing (CAT) versions for anxiety and depression.²⁹ The scores for PROMIS-A and PROMIS-D are converted into standardized T scores (general U.S. population mean=50, standard deviation [SD]=10), with higher scores indicating higher levels of anxiety and depression, respectively.³⁰ The validity of PROMIS-A and PROMIS-D has been supported in intervention studies conducted with the COPD population.³¹

Ethical approval

Study approval was obtained from the Institutional Review Board of the University of Illinois Chicago (2013–0626).

Data analysis

Data were analyzed with Stata/IC 16.1 and Mplus Version 8.3. Descriptive statistics were used to check for normality, outliers, and missing data. Latent profile analysis (LPA) was used to identify latent profiles (i.e., classes or groups) of the symptom cluster. Subsequently, latent transition analysis (LTA) was used to examine the stability and transition probability of latent profiles over time and to evaluate the effect of behavioral interventions on transitions. All tests were two-tailed, and a type I error rate of α<0.05 was accepted as being statistically significant.

LPA is a person-centered analysis approach to identify distinct latent classes of individuals based on scores for continuous indicator variables.³² In this study, a symptom cluster was modeled at each time point as a categorical latent variable indicated by the five continuous symptom variables: insomnia, dyspnea, fatigue, anxiety, and depression. LTA is a longitudinal extension of LPA performed to model transitions in latent class membership, which refers to the probability of belonging to a given class.³² The specific steps of the latent profile and transition analyses are summarized below.

Step 1. Identify latent classes for each time-point.

LPA was used to predict the number and proportions of latent classes of people with COPD and insomnia. Robust maximum likelihood estimation was used for LPA, and the best-fitting latent profile model was determined using the following fit and diagnostic statistics: (1) Akaike information criterion (AIC), (2) Bayesian information criterion (BIC), (3) sample-size-adjusted BIC (SABIC), (4) entropy, (5) bootstrap likelihood ratio test (BLRT), and (6) Lo-Mendell-Rubin adjusted likelihood ratio test (LMRT). Lower values of AIC, BIC, and SABIC indicate better model fit, and entropy over 0.8 is regarded as acceptable.^32,33 In addition, a significant p-value for BLRT and LMRT indicates that the k classes provide a better model fit than the k-1 classes.³⁴ When conflicts arise within and/or between model fits and diagnostic statistics, BIC is considered the most reliable criterion.³⁴ Once the number of latent classes was determined by means of LPA, the t-test or one-way analysis of variance (ANOVA) was used to determine whether scores for the five indicators differed significantly among the classes identified at each time point.

Step 2. Explore longitudinal patterns of latent profiles across three time points.

The prevalence of latent profiles and incidences of transitions between latent profiles were examined across the baseline (T1), post-intervention (T2), and 3-month follow-up (T3) time points. All possible pathways of the classes across the three time points were considered to determine transition probabilities for the classes. A row conditional matrix showed probabilities that summed to 1 across each row, given the class participants belonged to at one time point and the probability that participants belonged to a particular class at another time point. Therefore, transition probabilities were calculated based on each of three pairs of time points: T1→T2, T2→T3, and T1→T3.

Step 3. Determine the effects of interventions on transitions of latent profiles.

Based on the row conditional matrix between T1 and T3, unique patterns of latent profiles were identified and then categorized into several groups. Multinomial logistics regression was then used to determine whether behavioral interventions were significantly associated with the transition probabilities. In the regression, four intervention groups were employed as an independent variable, and groups of unique patterns of latent profiles were entered as a dependent variable. Covariates controlled for in the regression analysis included age, gender, race, ethnicity, FCI score, and FEV₁pp.

RESULTS

Participant characteristics

A total of 91 participants were included in this study, and their characteristics are summarized in Table 1. Their mean age was 64.6 years (SD=8.0), and 60.4% were male. Almost three-quarters of the participants (74.7%, n=68) were African-American, and the mean number of comorbidities was 4.4 (SD=2.3). Slightly over half the participants (51.7%, n=47) had moderate COPD.

Table 1.

Descriptive statistics for all study variables (N=91)

Variable	N (%)	Mean (SD)	Missing value (%)
Age (years)		64.6 (8.0)
Gender
Male	55 (60.4)
Female	36 (39.6)
Race
White	19 (20.9)
African-American	68 (74.7)
Other	4 (4.4)
Ethnicity
Hispanic or Latino	5 (5.5)
Non-Hispanic or non-Latino	86 (94.5)
Marital status
Never married	23 (25.3)
Married	16 (17.6)
Widowed	17 (18.7)
Separated	8 (8.8)
Divorced	27 (29.7)
Number of years of school		13.8 (2.6)
FCI [1–18]		4.4 (2.3)
FEV1pp		67.6 (22.0)
COPD severity
Mild	26 (28.6)
Moderate	47 (51.7)
Severe	12 (13.2)
Very severe	6 (6.6)
Intervention groups
CBT-I	23 (25.3)
CBT-I + COPD-ED	24 (26.4)
COPD-ED	20 (22.0)
AC	24 (26.4)
Indicators (baseline)
ISI		15.9 (4.3)
CRQ-D		4.4 (1.4)
CRQ-F		3.7 (1.1)
PROMIS-A		59.3 (7.9)	1 (0.01)
PROMIS-D		54.4 (8.2)
Indicators (post-intervention)
ISI		11.9 (5.3)
CRQ-D		4.8 (1.5)
CRQ-F		4.2 (1.0)
PROMIS-A		55.2 (8.2)
PROMIS-D		51.2 (7.6)	1 (0.01)
Indicators (3-month follow-up)
ISI		11.0 (5.8)
CRQ-D		4.8 (1.5)
CRQ-F		4.3 (1.1)
PROMIS-A		55.2 (8.4)
PROMIS-D		51.1 (8.2)

Note. AC=attention control, CBT-I=cognitive behavioral therapy for insomnia, COPD=chronic obstructive pulmonary disease, COPD-ED=chronic obstructive pulmonary disease education, CRQ-D=Chronic Respiratory Questionnaire-Dyspnea, CRQ-F=Chronic Respiratory Questionnaire-Fatigue, FCI=Functional Comorbidity Index, FEV₁pp=forced expiratory volume in 1 second percent predicted, ISI=Insomnia Severity Index, PROMIS-A=Patient-Reported Outcomes Measurement Information System-Anxiety, PROMIS-D= Patient-Reported Outcomes Measurement Information System-Depression, SD=standard deviation.

Symptom cluster profiles at each time point

The correlation coefficients among the pre-specified symptom variables at baseline are shown in Table A.1. The three-class model had the lowest BIC at all three time points, indicating the best model fit. Fit indices for LPA at each of the three time points are presented in Table A.2. The scores for the five indicators among the three classes at the three time points are shown in Table 2 and Figure A.1. All indicator scores differed significantly among the classes across the three time points (p<.001).

Table 2.

Differences in indicator scores among three classes at three time points

Indicator	Mean (SD)			F
	Class 1 (low)	Class 2 (intermediate)	Class 3 (high)
T1
ISI	13.0 (4.0)	15.0 (3.0)	20.0 (4.0)	21.62***
CRQ-D	5.6 (1.1)	4.3 (1.3)	3.7 (1.2)	12.56***
CRQ-F	4.7 (0.7)	3.9 (0.6)	2.6 (0.8)	52.72***
PROMIS-A	47.6 (5.1)	59.7 (3.6)	67.0 (4.9)	107.47***
PROMIS-D	43.5 (6.0)	54.8 (5.7)	61.8 (3.6)	61.59***
T2
ISI	7.0 (5.0)	12.0 (4.0)	16.0 (5.0)	19.56***
CRQ-D	5.8 (1.0)	4.8 (1.3)	3.8 (1.5)	10.64***
CRQ-F	5.1 (0.6)	4.3 (0.7)	2.9 (0.9)	42.56***
PROMIS-A	42.3 (3.9)	56.1 (4.1)	64.3 (3.0)	163.88***
PROMIS-D	40.7 (5.1)	52.0 (5.1)	58.3 (4.0)	65.00***
T3
ISI	8.0 (5.0)	11.0 (5.0)	18.0 (6.0)	18.56***
CRQ-D	5.6 (1.2)	4.7 (1.3)	3.5 (1.8)	11.20***
CRQ-F	5.3 (0.6)	4.0 (0.8)	2.8 (0.7)	62.61***
PROMIS-A	46.2 (5.0)	57.8 (4.4)	67.0 (3.5)	115.07***
PROMIS-D	42.5 (5.4)	54.0 (4.2)	61.2 (5.4)	88.29***

Note. CRQ-D=Chronic Respiratory Questionnaire-Dyspnea, CRQ-F= Chronic Respiratory Questionnaire-Fatigue, ISI=Insomnia Severity Index, PROMIS-A=Patient-Reported Outcomes Measurement Information System-Anxiety, PROMIS-D=Patient-Reported Outcomes Measurement Information System-Depression, SD=standard deviation, T1=baseline, T2=post-intervention, T3=3-month follow-up.

^*p <.05,

^**p <.01,

^***p <.001

Each class was labeled according to the distinct characteristics of the indicators in that class at each of the three time points. Across all time points, participants in Class 1 showed the lowest severity of all five symptoms among the three classes. Therefore, Class 1 was labeled the “low symptom burden group.” Across all time points, participants in Class 3 had the highest severity of all five symptoms among the three classes and was therefore labeled the “high symptom burden group.” Given the symptom severity scores in Class 2 across all time points, it was labeled the “intermediate symptom burden group.”

The number of participants in and the probability of belonging to each class (i.e., class membership) at each of the three time points are shown in Figure 1. At all three time points, the greatest proportion of participants was likely to belong to Class 2. Furthermore, the probability of belonging to Class 1 slightly decreased from T1 (20.9%) to T2 (19.8%) but significantly increased from T2 (19.8%) to T3 (34.1%). Finally, the probability of belonging to Class 3 showed a steadily decreasing trend, representing 27.5% of participants at T1, 23.1% at T2, and 14.3% at T3.

Figure 1. Probability of class membership at each time point.

Note. T1=baseline, T2=post-intervention, T3=3-month follow-up.

Longitudinal patterns of symptom cluster profiles

Stability and transition

The probabilities of stability and transitions among the classes across the three time points are represented in Table 3 and Figure A.2. Stability was defined as the probability of belonging to the same class over time, whereas transition was defined as the probability of belonging to different classes over time. The likely patterns of movement between profiles across the three time points are shown in Figure 2. From T1 to T3, the highest stability was shown by Class 1. Of participants in Class 2 at T1, 61.7% remained in Class 2, 29.8% moved to Class 1, and 8.5% moved to Class 3 at T3. Lastly, participants in Class 3 at T1 showed less stability, with 36% remaining in Class 3 at T3 but 60% moving to Class 2 at T3; also, 4% moved to Class 1 at T3.

Table 3.

Row conditional matrix: Longitudinal patterns of symptom cluster profiles

Transition probability from T1 to T2
		T2
		Class 1	Class 2	Class 3
	Class 1	0.684	0.263	0.053
T1	Class 2	0.106	0.787	0.106
	Class 3	0	0.400	0.600
Transition probability from T2 to T3
		T3
		Class 1	Class 2	Class 3
	Class 1	1	0	0
T2	Class 2	0.231	0.731	0.039
	Class 3	0.048	0.429	0.524
Transition probability from T1 to T3
		T3
		Class 1	Class 2	Class 3
	Class 1	0.842	0.158	0
T1	Class 2	0.298	0.617	0.085
	Class 3	0.040	0.600	0.360

Note. T1=baseline, T2=post-intervention, T3=3-month follow-up.

A transition probability of 0 means that the pattern is rarely likely to occur and actually reflects a probability close to 0. A transition probability of 1 reflects a probability close to 1.

Figure 2.

Stability and transition among three classes across three time points.

Stable and transition groups

Based on the transition patterns of symptom cluster profiles from T1 to T3, participants were recategorized into three groups—stable, better transition, and worse transition. The “stable group” was defined as participants having the probability of belonging to the same class at T1 and T3. The “better transition group” referred to participants having a probability of moving toward improvement in symptom cluster severity (i.e., Class 2→Class 1, Class 3→Class 1, and Class 3→Class 2). On the other hand, the “worse transition group” referred to participants having a probability of moving toward worse symptom cluster severity (i.e., Class 1→Class 2, Class 1→Class 3, and Class 2→Class 3). The number and percentage of participants in each group are summarized in Table 4. In terms of transition groups, almost 60% of the participants (n=54) remained in the same class; 33% showed better transition (n=30), while 7% (n=7) had worse transition. In terms of behavioral intervention groups, the CBT-I group had the highest percentage of participants (52%) belonging to the better transition group, followed by the CBT-I + COPD-ED (37.5%), COPD-ED (30%), and AC (12.5%) groups.

Table 4.

Symptom cluster patterns in four intervention groups

Stable and transition groups (n)
		Stable	Better transition	Worse transition	N (%)
Patterns of classes from T1 to T3		Class 1→Class 1 (16)	Class 2→Class 1 (14)	Class 1→Class 2 (3)
		Class 2→Class 2 (29)	Class 3→Class 1 (1)	Class 1→Class 3 (0)
		Class 3→Class 3 (9)	Class 3→Class 2 (15)	Class 2→Class 3 (4)
	n (%)	54 (59.3)	30 (33.0)	7 (7.7)	91 (100)
Intervention group		Stable	Better transition	Worse transition	N (%)
CBT-I		Class 1→Class 1 (1)	Class 2→Class 1 (7)	Class 1→Class 2 (0)
		Class 2→Class 2 (9)	Class 3→Class 1 (0)	Class 1→Class 3 (0)
		Class 3→Class 3 (1)	Class 3→Class 2 (5)	Class 2→Class 3 (0)
	n (%)	11 (48.0)	12 (52.0)	0 (0)	23 (100)
CBT-I + COPD-ED		Class 1→Class 1 (3)	Class 2→Class 1 (5)	Class 1→Class 2 (1)
		Class 2→Class 2 (6)	Class 3→Class 1 (1)	Class 1→Class 3 (0)
		Class 3→Class 3 (3)	Class 3→Class 2 (3)	Class 2→Class 3 (2)
	n (%)	12 (50.0)	9 (37.5)	3 (12.5)	24 (100)
COPD-ED		Class 1→Class 1 (5)	Class 2→Class 1 (1)	Class 1→Class 2 (0)
		Class 2→Class 2 (5)	Class 3→Class 1 (0)	Class 1→Class 3 (0)
		Class 3→Class 3 (3)	Class 3→Class 2 (5)	Class 2→Class 3 (1)
	n (%)	13 (65.0)	6 (30.0)	1 (5.0)	20 (100)
AC		Class 1→Class 1 (7)	Class 2→Class 1 (1)	Class 1→Class 2 (2)
		Class 2→Class 2 (9)	Class 3→Class 1 (0)	Class 1→Class 3 (0)
		Class 3→Class 3 (2)	Class 3→Class 2 (2)	Class 2→Class 3 (1)
	n (%)	18 (75.0)	3 (12.5)	3 (12.5)	24 (100)

Note. AC=attention control, CBT-I=cognitive behavioral therapy for insomnia, COPD-ED=chronic obstructive pulmonary disease education, T1=baseline, T3=3-month follow-up.

Effects of interventions on transition probabilities

Our study evaluated whether behavioral interventions were associated with changes in the symptom cluster profiles. The findings of the multinomial logistic regression are shown in Table 5. Compared to the AC group, two intervention groups (CBT-I and CBT-I + COPD-ED) were significantly associated with the probability of belonging to the better transition group versus the stable group. Additionally, compared to the AC group, participants who received COPD-ED were less likely to belong to the worse transition group versus the better transition group.

Table 5.

Multinomial logistic regression: Factors associated with transitions

Group	Variable	RRR	SE	95% CI	p-value
Stable (ref.)
Better transition	Intervention groups
	AC (ref.)
	CBT-I	11.917	10.557	2.099, 67.641	0.005 **
	CBT-I + COPD-ED	10.484	9.474	1.783, 61.625	0.009 **
	COPD-ED	2.204	2.052	0.355, 13.669	0.396
	Gender
	Male (ref.)
	Female	0.217	0.134	0.065, 0.730	0.014 *
	FCI	1.541	0.234	1.145, 2.074	0.004 **
	FEV1pp	1.005	0.013	0.979, 1.031	0.727
Worse transition	Race
	White (ref.)
	African-American	0.018	0.028	0.001, 0.387	0.010 *
	Other	10.699	22.067	0.188, 609.443	0.250
Better transition (ref.)
Worse transition	Intervention groups
	AC (ref.)
	CBT-I	<.001	<.001	N/A	0.992
	CBT-I + COPD-ED	0.369	0.435	0.025, 4.375	0.400
	COPD-ED	0.023	0.043	0.001, 0.953	0.047 *
	Race
	White (ref.)
	African-American	0.039	0.063	0.002, 0.919	0.044 *
	Other	0.788	1.358	0.027, 23.084	0.890

Note. AC=attention control, CBT-I=cognitive behavioral therapy for insomnia, CI=confidence interval, COPD-ED=chronic obstructive pulmonary disease education, FCI=Functional Comorbidity Index, FEV₁pp=forced expiratory volume in 1 second percent predicted, ref.=reference, RRR=relative risk ratio, SE=standard error.

^*p <.05,

^**p <.01,

^***p <.001

DISCUSSION

This study explored longitudinal patterns of symptom cluster profiles and evaluated whether behavioral interventions were associated with changes in the profiles of people with COPD and insomnia up to 3 months post-intervention. Three symptom cluster profiles were identified that represented different severities of the symptom cluster—low (Class 1), intermediate (Class 2), and high (Class 3)—at three time points. Class 2 had the largest proportions of participants among the three classes, consistently accounting for more than half the sample at each of the three time points (51.6%, 57.1%, and 51.6%). Although the five indicators’ mean scores in Class 2 differed across the three time points, this finding suggests that more than half the participants experienced subthreshold or moderate insomnia and greater anxiety and depression than the U.S. population norm and stroke survivors.³⁵

The severity of the five symptoms decreased or stayed the same in Classes 1 and 2 through the 3-month follow-up, but Class 3 showed increased dyspnea severity while anxiety severity stayed the same. This finding provides evidence that people with COPD and insomnia who have high symptom burden are more likely to have difficulties managing their dyspnea and anxiety in the long term. A systematic review of symptom clusters in the COPD population revealed that difficulty breathing and anxiety-related symptoms were the most common symptoms in people with COPD.⁹ Difficulty breathing is the hallmark of COPD and is the most distressing symptom in people with this disease.^9,36 Additionally, increasing evidence indicates that dyspnea leads to other COPD comorbidities such as anxiety and depression, worse disease prognosis, and negative health outcomes.³⁶ In addition, anxiety is one of the factors most strongly associated with poor quality of life in people with COPD.^7,37 Therefore, long-term symptom management strategies should give more attention to reducing dyspnea and anxiety in people with COPD and insomnia, as such reductions may play an important role in improving their health outcomes.

This study revealed that individuals in Classes 1 and 2 were more likely to remain stable over time while those in Class 3 at T1 were more likely to move to the other two classes. Therefore, identification of factors associated with transitions from Class 3 is especially important to reduce the severity of multiple symptoms in people with COPD and insomnia who experience high symptom burden. On the whole, understanding transition patterns among people with distinct symptom cluster profiles can support development of targeted interventions that maximize treatment response to reduce their high symptom burden.^8,9

Our study showed that CBT-I, with or without COPD-ED, was significantly associated with better transition (i.e., toward reduced symptom cluster severity) in people with COPD and insomnia. This finding may indicate that COPD-ED was not sufficient to increase CBT-I’s effectiveness in reducing symptom cluster severity. However, due to the small sample size of the study, additional research employing larger samples is needed to examine the interaction effects of CBT-I and COPD-ED on transition among symptom cluster profiles. Previous systematic reviews have reported that CBT-I was effective for reducing the severity of individual symptoms such as insomnia, anxiety, depression, and fatigue in various populations up to 1 year after therapy, although its effects decreased over time.^20,38 In people with COPD and insomnia, CBT-I was effective in improving fatigue, dyspnea, and insomnia symptoms up to 3 months post-intervention.²² Therefore, our findings offer new evidence that CBT-I can be effective for reducing the severity of multiple symptoms within a cluster, especially for people with COPD and insomnia who have a high symptom burden. Additionally, this study found that gender, race, and FCI score were significantly associated with transitions between classes. Given the possibility of health inequities, future studies should examine the potential for sociodemographic health disparities in the outcomes of symptom cluster research.

This study’s results have implications for prevention and intervention programs designed to mitigate symptom cluster severity in people with COPD and insomnia. Behavioral interventions have been commonly implemented in populations without considering the characteristics of individuals that could predict their response to treatment.³⁸ More recently, however, interest has grown in administering the most suitable program to individuals in order to achieve the maximum treatment effect.³⁹ The increasing research interest in symptom management strategies underscores the need to administer the right program to the right individual.⁴⁰ Given the paucity of intervention studies for symptom clusters, optimization of a multicomponent intervention should result in a more effective, economical, efficient, and scalable intervention.¹⁰ With its person-centered approach, this study provides evidence that CBT-I reduces symptom cluster severity in people with COPD and insomnia who face a high symptom burden.

The principal strengths of this study lie in its employment of longitudinal data for behavioral interventions and its person-centered analysis method. The study findings will provide clinicians with evidence-based knowledge that they can apply in assessing or treating multiple co-occurring symptoms in people with COPD and insomnia. Ultimately, the findings will contribute to development of targeted interventions for people with COPD and insomnia who are at high risk of experiencing multiple co-occurring symptoms in order to decrease their symptom burden. As an example, our findings can contribute to design of intervention studies to evaluate whether CBT-I can be used as a tailored intervention to simultaneously treat multiple symptoms in people having high symptom burden.

The limitations of this study should be acknowledged. First, the study findings may have limited generalizability, as the study participants were largely African-American (74.7%). Second, although no sample size requirement for LTA has been established in previous symptom cluster research, the 91 participants included in this study constitute a smaller sample than was used in previous LTA studies conducted in the epidemiology domain or targeting the general population. Due to the characteristics of our study participants and the design of the parent study, there were many limitations on our ability to recruit a larger sample. Therefore, the study findings should be interpreted with caution. Third, although this study examined the effects of behavioral interventions on symptom cluster transition up to 3 months post-intervention, it did not investigate the longer-term effects.

CONCLUSIONS

This study underscores the value of a person-centered approach to understanding stability and transition among symptom cluster profiles in people with COPD and insomnia before and after behavioral interventions. CBT-I intervention, either alone or combined with COPD-ED, was associated with the probability of transition toward reduced symptom burden in people with COPD and insomnia. Therefore, CBT-I may be a promising intervention for simultaneously reducing the severity of multiple symptoms in this population.

Highlights.

• Three symptom cluster classes were identified at each of three time points

• Classes 1 to 3 were designated as low, intermediate, and high symptom burden groups

• People in Classes 1 and 2 at baseline tended to remain in the same class over time

• People in Class 3 at baseline showed greater transition to Class 1 or 2

• CBT-I was significantly associated with movement to a lower symptom burden group

Abbreviations

AC

attention control

AIC

Akaike information criterion

BIC

Bayesian information criterion

BLRT

bootstrap likelihood ratio test

CBT-I

cognitive behavioral therapy for insomnia

COPD

chronic obstructive pulmonary disease

COPD-ED

chronic obstructive pulmonary disease education

CRQ-D

Chronic Respiratory Questionnaire-Dyspnea

CRQ-F

Chronic Respiratory Questionnaire-Fatigue

FCI

Functional Comorbidity Index

FEV₁

forced expiratory volume in 1 second

FEV₁pp

forced expiratory volume in 1 second percent predicted

FVC

forced vital capacity

ISI

Insomnia Severity Index

LMRT

Lo-Mendell-Rubin adjusted likelihood ratio test

LPA

latent profile analysis

LTA

latent transition analysis

PROMIS-A

Patient-Reported Outcomes Measurement Information System-Anxiety

PROMIS-D

Patient-Reported Outcomes Measurement Information System-Depression

RRR

relative risk ratio

SABIC

sample-size-adjusted Bayesian information criterion

SD

standard deviation

T1

baseline

T2

post-intervention

T3

3-month follow-up