Bronchiectasis Phenotype Identification Model

Assessment of the validation performance of the Bronchiectasis Phenotype Identification Model (BPIM) for prediction of accepted latent class trajectory analysis (LCTA)-derived bronchiectasis phenotype classes in adults with non-cystic fibrosis bronchiectasis (NCFB). The Bronchiectasis Phenotype Identification Model (BPIM) is a supervised baseline phenotype classifier trained in the development cohort and applied unchanged to the validation cohort. The accepted latent class trajectory analysis (LCTA)-derived phenotype class will serve as the phenotype ground truth. Higher classification accuracy indicates better agreement between Bronchiectasis Phenotype Identification Model (BPIM)-predicted phenotype class and accepted latent class trajectory analysis (LCTA)-derived phenotype class. Unit of Measure: Percentage of participants correctly classified.

Identification of the accepted latent class trajectory analysis (LCTA)-derived bronchiectasis phenotype class using prospectively collected longitudinal disease-signature data. Latent class trajectory analysis (LCTA) will use prespecified functional, oxygenation, and inflammatory longitudinal domains according to the locked variable-combination hierarchy. Depending on the accepted class-number solution, phenotype classes may include Stable phenotype, Progressive/Frequent Exacerbator phenotype, Frequent Exacerbator/Inflammatory phenotype, Advanced Multidomain phenotype, and End-stage/Terminal-risk phenotype according to the locked labeling convention. Unit of Measure: Percentage of participants in each phenotype class.

Agreement between Bronchiectasis Phenotype Identification Model (BPIM)-predicted phenotype class and accepted latent class trajectory analysis (LCTA)-derived phenotype class. Agreement may be assessed using agreement statistics appropriate to the accepted class structure. Higher agreement indicates better reproducibility of latent class trajectory analysis (LCTA)-derived phenotype classes using the supervised Bronchiectasis Phenotype Identification Model (BPIM) baseline classifier. Unit of Measure: Kappa coefficient

Assessment of class-specific sensitivity of the Bronchiectasis Phenotype Identification Model (BPIM) for each accepted latent class trajectory analysis (LCTA)-derived phenotype class. Sensitivity represents the proportion of participants in a given accepted phenotype class who are correctly classified by Bronchiectasis Phenotype Identification Model (BPIM) into that same class. Higher values indicate better class-specific recognition. Unit of Measure: Percentage.

Assessment of class-specific specificity of the Bronchiectasis Phenotype Identification Model (BPIM) for each accepted latent class trajectory analysis (LCTA)-derived phenotype class. Specificity represents the proportion of participants not belonging to a given phenotype class who are correctly classified by Bronchiectasis Phenotype Identification Model (BPIM) as not belonging to that class. Higher values indicate better class-specific exclusion. Unit of Measure: Percentage.

Assessment of positive predictive value for each Bronchiectasis Phenotype Identification Model (BPIM)-predicted phenotype class. Positive predictive value represents the proportion of participants predicted by Bronchiectasis Phenotype Identification Model (BPIM) to belong to a given phenotype class who truly belong to the accepted latent class trajectory analysis (LCTA)-derived class. Higher values indicate better precision of phenotype assignment. Unit of Measure: Percentage.

Assessment of negative predictive value for each Bronchiectasis Phenotype Identification Model (BPIM)-predicted phenotype class. Negative predictive value represents the proportion of participants not predicted by Bronchiectasis Phenotype Identification Model (BPIM) to belong to a given phenotype class who truly do not belong to that accepted latent class trajectory analysis (LCTA)-derived class. Higher values indicate better exclusion of phenotype membership. Unit of Measure: Percentage.

Assessment of macro-average F1 score for Bronchiectasis Phenotype Identification Model (BPIM) phenotype classification. Macro-average F1 score summarizes balanced classifier performance across accepted phenotype classes by combining class-specific precision and recall while giving each class equal weight. Higher values indicate better balanced multiclass classification performance. Unit of Measure: F1 score, 0 to 1 scale.

Assessment of calibration of Bronchiectasis Phenotype Identification Model (BPIM)-predicted class probabilities where predicted phenotype probabilities are generated. Calibration will evaluate agreement between predicted class probabilities and observed accepted latent class trajectory analysis (LCTA)-derived phenotype membership. Better calibration indicates closer agreement between predicted probability and observed phenotype membership. Unit of Measure: Calibration slope

Confusion matrix comparing Bronchiectasis Phenotype Identification Model (BPIM)-predicted phenotype class with accepted latent class trajectory analysis (LCTA)-derived phenotype class. The confusion matrix will summarize correct and incorrect class assignment across the accepted phenotype classes. Unit of Measure: Percentage of participants

Assessment of the relationship between accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes and Bronchiectasis Assessment of Severity and Exacerbations Severity model (BASE-S) categories. This analysis evaluates whether trajectory-derived phenotype classes show clinically coherent patterns across baseline severity categories. Unit of Measure: Percentage of participants across categories.

Assessment of the relationship between accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes and Bronchiectasis Assessment of Severity and Exacerbations Prognostic model (BASE-P) risk categories. This analysis evaluates whether trajectory-derived phenotype classes show clinically coherent patterns across 12-month exacerbation-risk categories. Unit of Measure: Percentage of participants across categories.

Occurrence of at least one bronchiectasis exacerbation during follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Higher exacerbation occurrence in higher-burden phenotype classes indicates clinically coherent phenotype-outcome association. Unit of Measure: Percentage of participants.

Time from baseline assessment to first recorded bronchiectasis exacerbation according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Participants without exacerbation will be censored at last available follow-up or at 12 months. Shorter time to first exacerbation indicates earlier clinical deterioration. Unit of Measure: Days.

Total number of bronchiectasis exacerbations during follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Higher exacerbation count indicates greater disease instability. Unit of Measure: Number of events.

Occurrence of at least one severe bronchiectasis exacerbation requiring hospital admission during follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Hospitalization-requiring exacerbation indicates more severe clinical deterioration. Unit of Measure: Percentage of participants.

Occurrence of bronchiectasis-related hospital admission during follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Bronchiectasis-related hospital admission includes admission related to bronchiectasis deterioration, respiratory infection, bronchiectasis exacerbation, worsening respiratory symptoms, or related respiratory failure. Unit of Measure: Percentage of participants.

Assessment of Bayesian Information Criterion (BIC) for prespecified latent class trajectory analysis (LCTA) class-number solutions within each variable-combination option. Bayesian Information Criterion (BIC) will guide class-number selection among acceptable models but will not override convergence, entropy, minimum class size, or clinical interpretability. Unit of Measure: Bayesian Information Criterion (BIC), unitless.

Assessment of entropy for latent class trajectory analysis (LCTA) class solutions. Entropy reflects classification quality of latent class assignment. Higher entropy indicates better class separation, with entropy of at least 0.70 required for an acceptable primary latent class trajectory analysis (LCTA) solution. Unit of Measure: Entropy, 0 to 1 scale.

Assessment of the minimum class size across accepted or attempted latent class trajectory analysis (LCTA) solutions. No accepted latent class trajectory analysis (LCTA) solution may contain a class smaller than 5% of the total analyzable cohort. Unit of Measure: Percentage of total analyzable cohort.

Assessment of posterior class-assignment probability for accepted latent class trajectory analysis (LCTA)-derived phenotype classes where supported by the modeling framework. Higher posterior assignment probability indicates greater certainty of latent phenotype class assignment. Unit of Measure: Probability, 0 to 1 scale.

Assessment of discrimination performance of the Bronchiectasis Phenotype Identification Model (BPIM) classifier for prediction of accepted latent class trajectory analysis (LCTA)-derived phenotype classes. For two-class solutions, discrimination will be assessed using area under the receiver operating characteristic curve. For three-class or four-class solutions, discrimination will be assessed using one-versus-rest or multiclass area under the receiver operating characteristic curve where applicable. Higher values indicate better discrimination performance. Unit of Measure: Area under the receiver operating characteristic curve, 0 to 1 scale.

Assessment of change in Bronchiectasis Phenotype Identification Model (BPIM) classification accuracy after sequential removal of individual baseline disease-signature domains from the classifier. The full locked baseline disease-signature architecture will remain the primary classifier regardless of ablation findings. Larger decreases in accuracy after removing a domain indicate greater contribution of that domain to phenotype classification. Unit of Measure: Difference in classification accuracy, percentage points

Assessment of repeated forced expiratory volume in one second percent predicted (FEV1% predicted) measurements according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class using longitudinal mixed-effects modeling. This analysis will evaluate whether FEV1% predicted trajectory differs across accepted BPIM phenotype classes during follow-up. Lower or declining FEV1% predicted indicates worsening airflow limitation and worse functional trajectory. Unit of Measure: FEV1% predicted percentage points

Assessment of repeated resting room-air oxygen saturation measured by pulse oximetry (SpO2) according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class using longitudinal mixed-effects modeling. This analysis will evaluate whether resting room-air SpO2 trajectory differs across accepted BPIM phenotype classes during follow-up. Lower or declining SpO2 indicates worse oxygenation reserve. Unit of Measure: SpO2 percentage points

Assessment of repeated C-reactive protein (CRP) measurements according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class using longitudinal mixed-effects modeling. This analysis will evaluate whether CRP trajectory differs across accepted BPIM phenotype classes during follow-up. Higher or increasing CRP indicates greater systemic inflammatory activity. Unit of Measure: Milligrams per liter (mg/L)

Assessment of repeated neutrophil-to-lymphocyte ratio (NLR) measurements according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class using longitudinal mixed-effects modeling. Neutrophil-to-lymphocyte ratio (NLR) is calculated as absolute neutrophil count divided by absolute lymphocyte count. Higher or increasing NLR indicates greater neutrophilic inflammatory activity. Unit of Measure: Ratio, unitless

Assessment of time from baseline to first bronchiectasis exacerbation across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes using survival modeling. Shorter time to first exacerbation indicates earlier clinical deterioration. Unit of Measure: Hazard ratio.

Assessment of time from baseline to first severe bronchiectasis exacerbation requiring hospitalization across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes using survival modeling. Shorter time to severe exacerbation indicates earlier severe clinical deterioration. Unit of Measure: Hazard ratio.

Assessment of time from baseline to first bronchiectasis-related hospital admission across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes using survival modeling. Shorter time to hospitalization indicates earlier clinically severe disease instability. Unit of Measure: Hazard ratio.

Assessment of the association between longitudinal forced expiratory volume in one second percent predicted (FEV1% predicted) trajectory and time to first bronchiectasis exacerbation using joint longitudinal-survival modeling. This analysis is supportive and will not define the accepted latent class trajectory analysis (LCTA) solution or Bronchiectasis Phenotype Identification Model (BPIM) classifier. Unit of Measure: Hazard ratio

Assessment of the association between longitudinal resting room-air oxygen saturation measured by pulse oximetry (SpO2) trajectory and time to first bronchiectasis exacerbation using joint longitudinal-survival modeling. This analysis is supportive and will not define the accepted latent class trajectory analysis (LCTA) solution or Bronchiectasis Phenotype Identification Model (BPIM) classifier. Unit of Measure: Hazard ratio

Assessment of the association between longitudinal C-reactive protein (CRP) trajectory and time to first bronchiectasis exacerbation using joint longitudinal-survival modeling. This analysis is supportive and will not define the accepted latent class trajectory analysis (LCTA) solution or Bronchiectasis Phenotype Identification Model (BPIM) classifier. Unit of Measure: Hazard ratio

Assessment of the association between longitudinal neutrophil-to-lymphocyte ratio (NLR) trajectory and time to first bronchiectasis exacerbation using joint longitudinal-survival modeling. Neutrophil-to-lymphocyte ratio (NLR) is calculated as absolute neutrophil count divided by absolute lymphocyte count. This analysis is supportive and will not define the accepted latent class trajectory analysis (LCTA) solution or Bronchiectasis Phenotype Identification Model (BPIM) classifier. Unit of Measure: Hazard ratio

Assessment of change in forced expiratory volume in one second percent predicted (FEV1% predicted) from baseline to follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Change will be calculated as follow-up value minus baseline value. Lower or declining FEV1% predicted indicates worse functional trajectory. Unit of Measure: Percentage points.

Assessment of change in resting room-air oxygen saturation measured by pulse oximetry (SpO2) from baseline to follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Change will be calculated as follow-up value minus baseline value. Lower or declining SpO2 indicates worse oxygenation reserve. Unit of Measure: Percentage points.

Assessment of change in C-reactive protein (CRP) from baseline to follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Change will be calculated as follow-up value minus baseline value. Higher or increasing CRP indicates greater systemic inflammatory activity. Unit of Measure: Milligrams per liter (mg/L).

Assessment of change in neutrophil-to-lymphocyte ratio (NLR) from baseline to follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Change will be calculated as follow-up value minus baseline value. Higher or increasing NLR indicates greater neutrophilic inflammatory activity. Unit of Measure: Unitless ratio.

Assessment of change in Modified Reiff score from baseline to follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class, where repeat imaging is clinically available. This contextual analysis will not define the primary latent class trajectory analysis (LCTA) solution or the primary Bronchiectasis Phenotype Identification Model (BPIM) classifier. The Modified Reiff score assesses radiological bronchiectasis extent and bronchial dilatation severity. Higher or increasing Modified Reiff score indicates greater structural radiological bronchiectasis burden or progression. Unit of Measure: Score change on Modified Reiff score scale.

Assessment of the distribution of participants across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes. This analysis describes phenotype occupancy within the analyzable cohort. Unit of Measure: Percentage of participants.

Assessment of predicted class-membership probabilities for accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes. Higher probability indicates greater certainty of predicted phenotype membership. Unit of Measure: Probability, 0 to 1 scale.

Assessment of dyspnea burden using the Modified Medical Research Council (mMRC) dyspnea scale across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes where complete data are available. The Modified Medical Research Council (mMRC) dyspnea scale ranges from 0 to 4, with higher scores indicating greater dyspnea burden and worse functional symptom status. This analysis is contextual and will not define the Bronchiectasis Phenotype Identification Model (BPIM) input engine or accepted latent class trajectory analysis (LCTA) solution. Unit of Measure: Modified Medical Research Council (mMRC) dyspnea score, 0 to 4 scale.

Assessment of Bronchiectasis Severity Index (BSI) values or categories across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes where complete data are available. This analysis is contextual and will not define the Bronchiectasis Phenotype Identification Model (BPIM) input engine or accepted latent class trajectory analysis (LCTA) solution. Higher scores indicate greater bronchiectasis severity or risk according to the respective scoring system. Unit of Measure: Score on the BSI scale, points

Assessment of FACED score values or categories across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes where complete data are available. This analysis is contextual and will not define the Bronchiectasis Phenotype Identification Model (BPIM) input engine or accepted latent class trajectory analysis (LCTA) solution. Higher scores indicate greater bronchiectasis severity or risk according to the respective scoring system. Unit of Measure: Score on the FACED scale, points.

Assessment of E-FACED score values or categories across accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype classes where complete data are available. This analysis is contextual and will not define the Bronchiectasis Phenotype Identification Model (BPIM) input engine or accepted latent class trajectory analysis (LCTA) solution. Higher scores indicate greater bronchiectasis severity or risk according to the respective scoring system. Unit of Measure: Score on the E-FACED scale, points.

Assessment of the number of bronchiectasis-related outpatient or emergency healthcare visits during follow-up according to accepted Bronchiectasis Phenotype Identification Model (BPIM) phenotype class. Higher visit count indicates greater healthcare-use burden and greater clinical instability. This analysis is contextual and will not define the Bronchiectasis Phenotype Identification Model (BPIM) input engine or accepted latent class trajectory analysis (LCTA) solution. Unit of Measure: Number of visits.