Chronic obstructive pulmonary disease (COPD) is a chronic respiratory disorder characterized by persistent airflow limitation, imposing a substantial global disease burden (1, 2). As the third leading cause of death worldwide, COPD disproportionately affects low- and middle-income countries, accounting for 90% of premature COPD-related deaths (3). The 2019 global prevalence reached 10.3% (approximately 391.9 million cases) in adults aged 30–79 years (4), while Chinese data report higher rates (13.7% in adults ≥40 years) with significant regional disparities (5). As the disease progresses, symptoms such as dyspnea and restricted activity frequently lead to high disability-adjusted life years and diminished quality of life among COPD patients (6–8).

Research indicates that early COPD screening enables smoking cessation interventions to delay lung function decline and reduces hospitalization costs associated with acute exacerbations (9–11), demonstrating significant health economic value. Although spirometry remains the diagnostic gold standard for COPD (1), its utility as a population-wide screening tool is debated (12–14). In resource-constrained primary care settings, spirometry faces substantial implementation barriers, including limited accessibility, high technical requirements, and potential suboptimal patient cooperation, resulting in underutilization of high-quality testing and diminishing its efficiency as a case-finding instrument (15–17). In contrast, brief, user-friendly self-administered questionnaires offer a pragmatic alternative for identifying high-risk individuals requiring spirometry referral, leveraging advantages such as low cost, ease of operation, and capacity for dynamic symptom monitoring (e.g., dyspnea grading).

Although multiple studies have reported the diagnostic performance of COPD screening questionnaires, significant variability exists across different investigations (18–22). This heterogeneity complicates end-users’ ability to determine which questionnaire optimally achieves the highest screening efficacy within specific target populations and clinical contexts. A key contributor to this dilemma is the pronounced dependence of questionnaire performance–particularly PPV and NPV–on the underlying disease prevalence of the target population. Previous evaluations predominantly assessed performance under single, fixed prevalence conditions, neglecting to examine diagnostic robustness across varying prevalence rates. Consequently, their findings demonstrate limited generalizability to real-world settings with divergent prevalence profiles, restricting the broader applicability of screening strategies across diverse implementation scenarios. Therefore, a systematic evaluation of COPD questionnaire performance within a unified cohort–accounting for different usage scenarios and prevalence conditions–is imperative to generate evidence-based support for precision-based screening strategies.

This study conducted a cross-sectional comparison of the static diagnostic performance of six commonly used COPD screening questionnaires (LFQ, Modified-IPAG, COPD-PS, IPAG, COPS-SQ, SCSQ) within a single high-risk cohort (COPD detection rate: 16.63%). Innovatively, by dynamically simulating a wide range of prevalence rates (5%–95%), it systematically quantified the sensitivity of each questionnaire’s screening accuracy to changes in prevalence (i.e., “robustness”). The results revealed distinct response patterns in the performance of different COPD screening questionnaires to dynamic changes in prevalence.

The most striking finding was the exceptional robustness demonstrated by the COPD-PS. In the static assessment, PS showed a balanced diagnostic performance with sensitivity (79.85%) and specificity (79.32%) (AUC 0.796). More crucially, under dynamic simulation across varying prevalence rates, its screening accuracy exhibited minimal fluctuation (ΔACC = 0.06; β = −0.018, P = 0.114). This indicates that COPS-PS’s performance remains relatively stable across different prevalence scenarios, largely unaffected by the level of prevalence in the target screening population. This universal robustness likely stems from its balanced sensitivity and specificity (both ≈79%). Furthermore, its items focus on core risk factors (age, smoking history) and core symptoms (breathlessness, productive cough, activity limitation) with globally recognized associations to COPD (1, 34), and it has undergone rigorous cognitive testing to ensure consistent comprehension across diverse populations (25).

In stark contrast to COPS-PS, other questionnaires exhibited significant “scenario dependency.” Although the COPD-SQ questionnaire achieved the highest specificity (87.05%) and the lowest sensitivity (50.75%) in the static assessment (a pattern maintained in the dynamic evaluation), its accuracy showed a sharp decline as prevalence increased (β = −0.263, P < 0.001), with a substantial fluctuation magnitude of 26% (ΔACC = 0.26). This makes COPD-SQ the optimal tool for low-prevalence scenarios. When the estimated prevalence is below 30%, its high specificity effectively controls false positives and unnecessary referrals while maintaining the highest accuracy (ACC > 80%). However, its accuracy significantly decreases in high-prevalence scenarios (>50%). Conversely, the Modified-IPAG and LFQ questionnaires, leveraging their ultra-high sensitivity (94.78% and 91.79%, respectively) and NPV (98.11% and 97.45%, respectively), performed optimally for rule-out screening, minimizing missed diagnoses (false negatives). It is important to note, however, that their accuracy improves with increasing prevalence (Modified-IPAG β = 0.293, LFQ β = 0.200, both P < 0.001), and their fluctuation range is also larger (ΔACC = 0.31 and 0.24, respectively). This implies that in very low prevalence scenarios, their relatively lower specificity may lead to an increased false positive rate. The IPAG and SCSQ performed relatively poorly in both this study and the dynamic simulations (lower AUC, poor robustness), highlighting inherent limitations of the tools or insufficient validation.

Based on a comprehensive analysis of the static performance characteristics and dynamic robustness of the six screening questionnaires, this study proposes a “Scenario-First” adaptive screening strategy for COPD. Specifically, when the core objective is “ruling out COPD” (prioritizing minimizing missed diagnosis risk), such as during large-scale initial screening in primary care or in settings with extremely limited referral resources needing rapid identification of individuals highly unlikely to have COPD to conserve advanced diagnostics, the Modified-IPAG or LFQ questionnaires should be prioritized; leveraging their ultra-high sensitivity (both >90%) and NPV (both >97%), these efficiently identify non-COPD individuals, minimizing missed diagnoses, though this may result in a higher false positive rate manageable through subsequent spirometry. Conversely, when the core objective is “ruling in COPD or identifying high-risk populations” (prioritizing referral accuracy and controlling false positives), such as during pre-referral screening in specialist clinics or screening high-exposure groups (e.g., smokers, occupational dust), the choice must consider estimated prevalence: if prevalence is low (<30%), the COPD-SQ questionnaire is optimal due to its outstanding high specificity (>87%), effectively reducing false positives and unnecessary referrals; if prevalence is high (≥30%), difficult to determine, likely to fluctuate significantly (e.g., respiratory clinic patients, high-exposure occupations), or a more stable tool is needed, the COPD-PS questionnaire should be prioritized, as it maintains relatively good specificity and PPV at high prevalence (outperforming COPD-SQ’s sharp decline) and its exceptional robustness makes it the “safe choice” and versatile first-line tool for uncertain or variable contexts. Furthermore, for institutions needing a single, universal tool to adapt to diverse scenarios or simplify processes, COPD-PS is undoubtedly the most reliable and adaptable choice due to its exceptional robustness. This strategy transcends the limitations of traditional comparisons based solely on single, fixed prevalence scenarios. Its core principle lies in precisely matching the optimal screening tool to the specific screening objective and the epidemiological background of the target population (primarily estimated prevalence), providing actionable, evidence-based decision-making support for clinical and public health practice (Figure 2).

However, this study has certain limitations. Firstly, the prevalence simulation assumed a linear effect, whereas real-world COPD prevalence may exhibit non-linear patterns. Secondly, the study is based on single-center data; future external validation in multi-center cohorts is warranted. Finally, the cultural adaptation of questionnaire items and participants’ comprehension ability may also impact performance, a common challenge for all self-administered screening tools. Future research could explore integrating risk prediction models with questionnaire results to develop more intelligent dynamic decision support tools and evaluate the cost-effectiveness and implementation feasibility of this strategy in real-world application settings.

By dynamically simulating prevalence rates, this study quantitatively revealed a key reason for the heterogeneity observed in the diagnostic performance of COPD screening questionnaires across previous studies: the significant influence of target population prevalence on tool performance. Based on this insight, we innovatively propose a “Scenario-First” screening strategy. This strategy dynamically matches the optimal tool based on estimated prevalence and the screening objective, thereby transcending the limitations of the traditional “tool-first” static framework and providing an evidence-based decision-making foundation for addressing heterogeneity in screening scenarios.