Determinants of generalized anxiety and construction of a predictive model in patients with chronic obstructive pulmonary disease

Abstract

BACKGROUND

Patients with chronic obstructive pulmonary disease (COPD) frequently experience exacerbations requiring multiple hospitalizations over prolonged disease courses, which predispose them to generalized anxiety disorder (GAD). This comorbidity exacerbates breathing difficulties, activity limitations, and social isolation. While previous studies predominantly employed the GAD 7-item scale for screening, this approach is somewhat subjective. The current literature on predictive models for GAD risk in patients with COPD is limited.

AIM

To construct and validate a GAD risk prediction model to aid healthcare professionals in preventing the onset of GAD.

METHODS

This retrospective analysis encompassed patients with COPD treated at our institution from July 2021 to February 2024. The patients were categorized into a modeling (MO) group and a validation (VA) group in a 7:3 ratio on the basis of the occurrence of GAD. Univariate and multivariate logistic regression analyses were utilized to construct the risk prediction model, which was visualized using forest plots. The model’s performance was evaluated using Hosmer-Lemeshow (H-L) goodness-of-fit test and receiver operating characteristic (ROC) curve analysis.

RESULTS

A total of 271 subjects were included, with 190 in the MO group and 81 in the VA group. GAD was identified in 67 patients with COPD, resulting in a prevalence rate of 24.72% (67/271), with 49 cases (18.08%) in the MO group and 18 cases (22.22%) in the VA group. Significant differences were observed between patients with and without GAD in terms of educational level, average household income, smoking history, smoking index, number of exacerbations in the past year, cardiovascular comorbidities, disease knowledge, and personality traits (P < 0.05). Multivariate logistic regression analysis revealed that lower education levels, household income < 3000 China yuan, smoking history, smoking index ≥ 400 cigarettes/year, ≥ two exacerbations in the past year, cardiovascular comorbidities, complete lack of disease information, and introverted personality were significant risk factors for GAD in the MO group (P < 0.05). ROC analysis indicated that the area under the curve for predicting GAD in the MO and VA groups was 0.978 and 0.960. The H-L test yielded χ² values of 6.511 and 5.179, with P = 0.275 and 0.274. Calibration curves demonstrated good agreement between predicted and actual GAD occurrence risks.

CONCLUSION

The developed predictive model includes eight independent risk factors: Educational level, household income, smoking history, smoking index, number of exacerbations in the past year, presence of cardiovascular comorbidities, level of disease knowledge, and personality traits. This model effectively predicts the onset of GAD in patients with COPD, enabling early identification of high-risk individuals and providing a basis for early preventive interventions by nursing staff.

Key Words: Chronic obstructive pulmonary disease; Generalized anxiety disorder; Predictive model; Determinants analysis; Forest plot

Core Tip: This study constructs and validates a predictive model for generalized anxiety disorder in patients with chronic obstructive pulmonary disease. Utilizing a retrospective design, we identified eight independent risk factors, including educational level, household income, smoking history, smoking index, number of exacerbations, cardiovascular comorbidities, disease knowledge, and personality traits. The model demonstrates high accuracy with area under the curve values of 0.978 and 0.960 in the modeling and validation groups, respectively. This predictive tool can facilitate early identification of high-risk patients, enabling timely interventions to improve their quality of life.

INTRODUCTION

Chronic obstructive pulmonary disease (COPD) is a progressively developing chronic lung condition characterized by respiratory symptoms and irreversible airflow limitation[1]. Epidemiological studies indicate[2] that the prevalence of COPD in China stands at 8.6%, increasing to 13.7% among individuals over the age of 40 years and affecting nearly 100 million people’s health. The incidence and mortality rates of COPD have been increasing annually, and it is projected to become one of the leading causes of death globally by 2030, thus representing a remarkable public health challenge in China. Research has demonstrated[3,4] that patients with COPD frequently present with various extrapulmonary diseases, such as cardiovascular alterations, metabolic syndrome, lung cancer, and anxiety, all of which exacerbate the severity of the disease, diminish quality of life, and serve as independent risk factors for hospitalization and mortality.

Generalized anxiety disorder (GAD) is a prevalent psychiatric condition marked by non-specific or unfixed anxiety objects, accompanied by restlessness. Its core features include uncontrollable chronic worry and a cognitive bias towards perceiving threats and risks, effecting interpersonal relations, work capacity, and overall physical and mental health[5]. Due to recurrent exacerbations requiring multiple hospitalizations and a prolonged disease course, patients with COPD experience progressively worsening clinical symptoms, which affect their daily functioning and social activities, leading to economic and familial burdens and increased psychological distress, thus making them prone to developing anxiety. Anxiety, in turn, exacerbates health issues, such as breathing difficulties, mobility impairments, and social isolation, thereby creating a vicious cycle[6]. Previous studies have noted[7] that patients with COPD with concurrent GAD exhibit more acute symptoms and reduced exercise tolerance. Hence, early screening and intervention for GAD can significantly improve their quality of life.

Past research predominantly utilized the GAD 7-item scale (GAD-7) for screening although GAD-7 primarily relies on self-reported data from patients, which may be subjective and influenced by personal perceptions[8]. Logistic regression analysis is a statistical method that integrates the study of independent disease risk factors to establish predictive models, quantifying the contributions of covariates and realizing the synergistic effects of multiple variables[9]. The current literature on predictive models for GAD risk among patients with COPD is limited. The present study primarily analyzed COPD patient data retrospectively, identified risk factors for GAD, constructed a GAD risk prediction model, and validated its utility to assist healthcare professionals in preventing the onset of GAD and enhancing the emotional well-being and quality of life of patients with COPD.

MATERIALS AND METHODS

Study subjects

This retrospective study included patients with COPD treated at our institution from July 2021 to February 2024. The inclusion criteria were as follows: Confirmed diagnosis of COPD, age > 18 years, complete GAD-7 clinical data available, and coherent and able to communicate verbally or in writing. The exclusion criteria were as follows: Concurrent malignancies, tuberculosis, or other diseases; liver or kidney dysfunction; and alcohol or psychotropic drug dependency. The sample size was calculated using the following formula: N = Z² [π (1 - π)]/δ², where Z = 1.96 at an alpha level of 0.05, δ is the allowable error, and π = 0.17, based on the documented incidence of GAD in patients with COPD being 17.0%[10], with an allowable error of 5%. This calculation indicated a need for 190 samples in the modeling (MO) group and 81 samples in the validation (VA) group based on a 7:3 ratio, totaling at least 271 subjects. This study was approved by the Ethics Committee of Henan Provincial People’s Hospital (No. 2022-42), and the Ethics Committee agreed to waive informed consent.

Study methods

GAD assessment methods: According to the “Chinese Classification and Diagnostic Criteria of Mental Disorders, 3^rd Edition[11]” guidelines, the assessment of GAD is based on the following criteria: (1) Persistent excessive anxiety and worry about everyday life for at least 6 months, with a GAD-7 score ≥ 5; (2) Difficulty managing this worry; (3) Presence of at least one of the following symptoms: Restlessness or feeling tense, easy fatigability, difficulty concentrating, irritability, muscle tension, and sleep disturbances; (4) Clinically significant distress or impairment in social, occupational, or other critical areas of functioning; and (5) Persistent fear of non-specific objects and content. GAD was assessed by a physician unaware of the study specifics.

Data collection: Patients with COPD diagnosed at our hospital were identified via the electronic medical record system. Patient follow-ups were conducted through phone calls and outpatient reviews to ascertain occurrences of GAD. Clinical data, including age, gender, body mass index, educational level, marital status, average household monthly income, living arrangements, smoking history, smoking index, alcohol consumption history, duration of COPD, Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage, number of exacerbations in the past year, last pulmonary function test results before enrollment (forced expiratory volume in the first second), cardiovascular comorbidities, home oxygen therapy status, understanding of the disease, and personality traits, were collected from electronic health records.

Statistical analysis

Data collected were analyzed using SPSS (version 27.0). Normally distributed quantitative data were expressed as mean ± SD, and comparisons between groups were conducted using independent samples t-tests. Non-normally distributed quantitative data were presented as median (P25, P75) and analyzed using the non-parametric Mann-Whitney U test. Categorical data were reported as number or percentage, and comparisons were made using χ² tests. Multifactorial logistic regression analysis was employed to identify risk factors for GAD occurrence in patients with COPD, with model visualization using forest plots. The predictive model’s fit and efficacy were assessed using the Hosmer-Lemeshow (H-L) goodness-of-fit test and the receiver operating characteristic (ROC) curve. The significance level was set at α = 0.05.

RESULTS

Incidence of GAD in patients with COPD

This study ultimately included 271 subjects, comprising 190 in the MO group and 81 in the VA group. A total of 67 patients with COPD developed GAD, representing an incidence rate of 24.72% (67/271). Specifically, 49 cases (18.08%) occurred in the MO group and 18 cases (22.22%) in the VA group, as depicted in Figure 1.

Figure 1 Incidence rate of generalized anxiety disorder in patients with chronic obstructive pulmonary disease. GAD: Generalized anxiety disorder.

Comparison of clinical data between MO and VA groups

No statistically significant differences were observed in age, gender, and other clinical data between the MO and VA groups (P > 0.05), as shown in Table 1.

Table 1 Comparison of clinical data between modeling and validation groups.

Category	MO groups (n = 190)	VA groups (n = 81)	χ2/t	P value
Age (years)
18-60	28	13	0.247	0.619
> 60	122	68
Gender
Male	118	52	0.106	0.744
Female	72	29
BMI (kg/m²)	22.35 ± 1.81	22.09 ± 1.94	1.059	0.290
Educational level
Junior high school and below	57	24	0.912	0.634
Technical/vocational or high school	112	51
College and above	21	6
Marital status
Married	151	64	0.007	0.932
Unmarried, divorced, or widowed	39	17
Average household income, CNY
< 3000	98	42	0.627	0.731
3000-5000	62	29
> 5000	30	10
Living arrangement
Living alone	19	9	0.799	0.671
Living with spouse	126	56
Living with children or others	45	26
Smoking history
Yes	55	22	0.089	0.765
No	135	59
Smoking index, cigarettes/year
≥ 400	92	38	0.052	0.820
< 400	98	43
Alcohol history
Yes	36	14	0.104	0.747
No	154	67
COPD duration (years)
< 5	44	19	0.724	0.696
5-10	59	29
> 10	87	33
GOLD Stage
Stage 1	13	6	1.233	0.745
Stage 2	84	39
Stage 3	54	24
Stage 4	39	12
Acute exacerbations in the past year
0 or 1 time	104	41	0.387	0.534
≥ 2 times	86	40
FEV1 (L)	1.29 ± 0.45	1.17 ± 0.58	1.837	0.067
Cardiovascular comorbidities
Yes	76	31	0.020	0.889
No	118	50
Home oxygen therapy
Yes	29	13	0.470	0.493
No	118	68
Understanding of the disease
Completely unaware	27	12	1.770	0.413
Moderately aware	127	59
Very aware	36	10
Personality
Introverted	116	47	0.217	0.641
Extroverted	74	34
GAD
Yes	49	18	0.041	0.840
No	141	63

BMI: Body mass index; COPD: Chronic obstructive pulmonary disease; GOLD: Global Initiative for Chronic Obstructive Lung Disease; FEV₁: Forced expiratory volume in the first second; GAD: Generalized anxiety disorder; MO: Modeling group; VA: Validation group; CNY: China yuan.

Comparison of clinical data between GAD and non-GAD groups in the MO group

In the MO group, the non-GAD group had higher educational level (χ² = 12.452, P = 0.002), higher average household income (χ² = 7.039, P = 0.030), lower smoking history (χ² = 6.211, P = 0.013), lower smoking index (χ² = 7.538, P = 0.006), lower number of acute exacerbations in the past year (χ² = 10.711, P = 0.001), fewer presence of cardiovascular comorbidities (χ² = 9.966, P = 0.002), better understanding of the disease (χ² = 49.452, P < 0.001), and more extroverted personalities (χ² = 10.561, P = 0.001) compared to GAD group, as shown in Table 2.

Table 2 Comparison of clinical data between generalized anxiety disorder and non-generalized anxiety disorder groups.

Category	GAD group (n = 49)	Non-GAD group (n = 141)	χ2/t	P value
Age (yeas)
18-60	8	20	0.133	0.716
> 60	41	121
Gender
Male	31	87	0.038	0.846
Female	18	54
BMI (kg/m²)	21.97 ± 1.95	22.34 ± 1.87	1.180	0.239
Educational level
Junior high school and below	24	33	12.452	0.002
Technical/vocational or high school	23	89
College and above	2	19
Marital status
Married	37	114	0.636	0.425
Unmarried, divorced, or widowed	12	27
Average household income, CNY
< 3000	24	74	7.039	0.030
3000-5000	22	40
> 5000	3	27
Living arrangement
Living alone	5	14	0.106	0.949
Living with spouse	32	94
Living with children or others	12	31
Smoking history
Yes	21	34	6.211	0.013
No	28	107
Smoking index, cigarettes/year
≥ 400	32	60	7.538	0.006
< 400	17	81
Alcohol history
Yes	9	27	0.014	0.904
No	40	114
COPD duration (years)
< 5	11	33	0.080	0.961
5-10	16	43
> 10	22	65
GOLD Stage
Stage 1	3	10	0.259	0.968
Stage 2	22	62
Stage 3	13	41
Stage 4	11	28
Acute exacerbations in the past year
0 or 1 time	17	87	10.711	0.001
≥ 2 times	32	54
FEV1 (L)	1.19 ± 0.36	1.28 ± 0.41	1.364	0.174
Cardiovascular comorbidities
Yes	30	46	9.966	0.002
No	19	85
Home oxygen therapy
Yes	8	21	0.058	0.810
No	41	120
Understanding of the disease
Completely unaware	21	6	49.452	< 0.001
Moderately aware	27	100
Very aware	1	35
Personality
Introverted	39	75	10.561	0.001
Extroverted	10	66

BMI: Body mass index; COPD: Chronic obstructive pulmonary disease; GOLD: Global Initiative for Chronic Obstructive Lung Disease; FEV₁: Forced expiratory volume in the first second; GAD: Generalized anxiety disorder; CNY: China yuan.

Multifactorial logistic regression analysis in the MO group

Variables that showed significant differences in the univariate analysis were selected as independent variables in a multifactorial logistic regression analysis, with the occurrence of GAD as the dependent variable (assigned values: Non-GAD = 0, GAD = 1). The results indicated that educational level of middle school or below, household income < 3000 China yuan (CNY), smoking history, smoking index ≥ 400 cigarettes/year, ≥ two acute exacerbations in the past year, presence of cardiovascular comorbidities, complete lack of disease information, and introverted personality type were significant risk factors for the development of GAD in the MO group (P < 0.05), as shown in Table 3.

Table 3 Multifactorial logistic regression analysis for the development of generalized anxiety disorder in the modeling group.

Factors	β	SE	Ward χ2	P value	OR	95%CI
Educational level (junior high school and below)	1.198	0.451	7.061	0.009	3.315	1.370-8.024
Average household monthly income (< 3000 CNY)	0.832	0.339	6.024	0.014	2.298	1.182-4.466
Smoking history	0.650	0.274	5.632	0.019	1.916	1.120-3.278
Smoking index (≥ 400 cigarettes/year)	0.633	0.282	5.045	0.024	1.884	1.084-3.274
Acute exacerbations in the past year (≥ 2 times)	1.295	0.368	12.389	< 0.001	3.652	1.775-7.512
Presence of cardiovascular comorbidities	1.047	0.415	6.365	0.010	2.849	1.263-6.426
Understanding of the disease (completely unaware)	0.986	0.433	5.187	0.021	2.681	1.147-6.264
Personality (introverted type)	0.848	0.309	7.539	0.005	2.336	1.275-4.281

OR: Odds ratio; CNY: China yuan.

Establishment and evaluation of the GAD risk prediction model

The results of multifactorial logistic regression analysis indicated that variables including educational level, household income, smoking history, smoking index, number of acute exacerbations in the past year, cardiovascular comorbidities, understanding of the disease, and personality were incorporated into the prediction model. The forest plot model was established, as depicted in Figure 2.

Figure 2 Forest plot model of generalized anxiety disorder risk prediction.

Evaluation of the GAD risk prediction model

The ROC analysis showed that the area under the curve (AUC) values for predicting GAD occurrence in the MO and VA groups were 0.978 and 0.960, respectively, indicating good internal and external validity and predictive ability of the model, as shown in Figure 3. The predictive model underwent H-L testing, yielding χ² values of 6.511 and 5.179, with P values of 0.275 and 0.274 (P > 0.05), indicating good fit of the model in both groups, with no significant deviation from actual conditions. The calibration curves further demonstrated the model’s accuracy in predicting the risk of GAD occurrence compared with actual events, as illustrated in Figure 4.

Figure 3 Receiver operating characteristic curves for the predictive model in two groups. A: The area under the curve (AUC) of modeling group; B: AUC of validation group. AUC: The area under the curve.

Figure 4 Calibration curves for the predictive model in two groups. A: Calibration curves of modeling group; B: Calibration curves of validation group.

DISCUSSION

Patients with COPD frequently face challenges related to anxiety disorders, which are linked to disease exacerbation. Epidemiological data suggest a high incidence of anxiety comorbidity in patients with COPD, with rates of associated depression ranging from 10% to 42% during stable phases and from 10% to 86% during exacerbations. Anxiety rates range from 13% to 46% during stable periods and from 10% to 55% during exacerbations[12]. In the present study, out of 271 patients, 67 developed GAD, an incidence of 24.72% (67/271). This rate is comparable to findings reported in the literature[13]. Studies indicate[14] that patients with COPD with anxiety not only lack confidence and have reduced disease coping and self-management capabilities but also exhibit decreased compliance, physical capacity, quality of life, and work capacity. Moreover, these patients face increased risks of acute exacerbations and mortality. Hence, the selection of reliable assessment tools is critical. While tools, such as the GAD-7 scale, are commonly used for screening, their reliance on self-reported patient data introduces a degree of subjectivity[15]. Logistic regression analysis, a statistical method that evaluates independent disease risk factors and quantifies the contributions of covariates through a predictive model, has been widely employed to predict complications in patients with chronic diseases[16,17].

Following univariate and multivariate logistic regression analyses, this study identified factors influencing the onset of GAD in the MO group. Significant risk factors included an educational level of middle school or below, household income < 3000 CNY, smoking history, a smoking index of ≥ 400 cigarettes/year, ≥ two acute exacerbations in the past year, cardiovascular comorbidities, a complete lack of disease knowledge, and an introverted personality. These findings are consistent with those of Husain et al[18]. Educational level reflects the degree of understanding of the disease and the mastery of disease-specific knowledge. Due to COPD’s unique clinical manifestations, the chronic and worsening nature of the condition, and the experience of breathlessness, patients often perceive their condition as severe, leading to complex and unclear emotional responses. Additionally, patients with lower educational levels tend to have poor emotional self-regulation capabilities, which can lead to anxiety[19]. In the present study, patients with lower educational levels or a complete lack of disease knowledge had a significantly higher incidence of GAD, likely because these patients have limited understanding of the disease’s pathogenesis and progression, and they are unable to effectively manage their condition, thus experiencing increased anxiety and fear. Research also shows a close association between smoking and anxiety[20]. In a study by Uchmanowicz et al[21], among 102 patients with COPD, 50% had been smoking for over 15 years, with 39.3% smoking 10-20 cigarettes per day and 31.4% smoking more than 20 cigarettes per day. Spearman analysis identified smoking as an independent predictor of anxiety comorbidity in COPD. Lou et al’s study[22] indicated that the interaction between smoking and anxiety increased mortality risk in patients with COPD, with risk escalating with the number of years or packs smoked. Cardiovascular comorbidities, such as hypertension and arrhythmias, may disrupt physiological functions and affect neural transmission and chemical release in the brain, thus increasing the risk of anxiety disorders[23]. A survey on the social characteristics of patients with COPD noted[24] that introverted patients tend to internalize emotions, are less adept at expressing and managing negative emotions, and are more likely to accumulate anxiety and stress than extroverted individuals, thus increasing their risk of developing GAD.

This study utilized multifactorial logistic regression analysis results as predictors to develop a forest plot predictive model, evaluated its performance by using ROC curves, H-L goodness-of-fit tests, and calibration curves. The ROC analysis revealed that the AUC values for the prediction of GAD occurrence in the MO and VA groups were 0.978 and 0.960, respectively, indicating good efficacy in internal and external VAs and robust predictive capability. Clinically, the H-L test is employed to assess the performance of a model. P > 0.05 suggests that the model predictions closely match actual occurrences, indicating high calibration and an ideal model fit. Conversely, P ≤ 0.05 indicates a discrepancy between model predictions and actual occurrences, signifying low calibration. In this study, the H-L test produced χ² values of 6.511 and 5.179, with P = 0.275 and 0.274, respectively, suggesting that the model fits well in the MO and VA groups, with no significant differences from actual conditions. Furthermore, the calibration curves demonstrated good consistency between the predicted risks of GAD occurrence and actual events.

Despite successfully identifying important predictors of GAD and constructing a prediction model with high accuracy, this study has several limitations. First, the retrospective design of the study may be subject to recall bias. Second, due to limitations in data collection, we were unable to perform detailed subgroup analyses based on patient characteristics such as age, sex, and disease severity, which may have overlooked risk factors in specific subpopulations. Future studies should consider using a prospective design and conducting more detailed subgroup analyses to further explore how these variables influence the risk of GAD in COPD patients.

CONCLUSION

Establishing a predictive model for GAD risk in patients with COPD is essential, and it can provide a scientific basis for clinical practice to reduce the risk of anxiety disorders. The risk prediction model developed in this study includes eight independent risk factors: Educational level, household income, smoking history, smoking index, number of acute exacerbations in the past year, presence of cardiovascular comorbidities, understanding of the disease, and personality traits. The model performs well, effectively predicting the occurrence of GAD in patients with COPD, and it can aid in the early screening of high-risk individuals, thus providing a basis for early preventive interventions by nursing staff. Future research could integrate new indicators and technologies to further refine and optimize the predictive model, enhancing its accuracy and practicality.