(1) COPD exacerba-tions requiring hospital admission are relatively fre-quent events occurring in about 30% of patients duringthe 3-year follow-up; (2) past history of hospitalizedexacerbations is most predictive of future events, andother risk factors include the severity of airflow limita-tion, poor health status, radiologic evidence of emphy-sema, older age, and presence of systemic inflammation;and (3) a history of hospitalized exacerbations heraldspoor survival

(1) baseline differences between patients with and withouthospitalized exacerbation during follow-up were tested using analysisof variance or Wilcoxon rank-sum test for continuous variables, andx2 test for categorical variables; (2) the incidence (first hospitalizedexacerbation during the prospective follow-up) and recurrence (secondhospitalized exacerbation during the prospective follow-up) of hos-pitalized exacerbations was summarized as a rate per person per year(PPPY), using a sum of individual patient’s person-time in the studyand standardized per year, accompanied by 95% CIs; (3) factors asso-ciated with first hospitalized (and recurrent) exacerbations during the3-year follow-up, were explored using Cox proportional hazards models,adjusted for a wide range of demographics, and clinical and biologicmarkers.

Subjects were followed-up at 3 months, 6 months, and every 6 monthsthereafter for a maximum of 3 years. All patients had their vital statusconfirmed 3 years after recruitment.Information on COPD exacerbations was collected at scheduled visitsby investigators using the case report forms and based on either sub-jects’ recall of exacerbation events or available medical records for exac-erbation events, supplemented by monthly phone calls. For the purposeof the current analysis, we focused on those exacerbation episodes thatrequired hospital admission (hospitalized exacerbation).

The methodology used in the ECLIPSE study has been described indetail elsewhere.9

Exacerbations of COPD accel-erate disease progression5-7

logisticmultivariate regression tests

xacerbation wasdefined on the basis of symptom-based diagnosis such asincreased cough and sputum production, a change of sputumcolor, and worsening of dyspnea from a stable state andbeyond-normal day-to-day variations, i.e., showing acute onsetand necessitating a change in regular medication, in accor-dance with a previous report [21]. Moderate exacerbationsrequired a prescription for antibiotics and/or systemic corticos-teroids, and severe exacerbations required hospitalization [22].

the CODEX indexis the most useful in predicting survival, hospital read-missions, and a combination of the two in the short andmedium term in patients hospitalized for AECOPD.

Multicomponent scales have been developed toimprove prognosis prediction in COPD, and they haveproved to be better predictors of survival than anyisolated variable.

This newly proposed CODEX index is essentially anevolution of the BODE and BODEX indexes, retain-ing their cutoffs for dyspnea, obstruction, and previousexacerbations, but replacing BMI with comorbiditymeasured using the original Charlson index modifiedby age.

Second, we reportedthe usefulness of this CODEX index in evaluating therisk of readmission, as well as the composite end point(readmission and/or mortality).

Time-dependentvariables from hospital discharge were analyzed with Cox logisticregression and Kaplan-Meier statistics.

Short- and Medium-term Prognosisin Patients Hospitalized for COPDExacerbation

Full methodology is avail-able elsewhere and is summarized in e-Appendix 1. 16,17

The mostcommon statistical method was logistic regression (11 out of 25 different statistical methods analysed)followed by Cox regression (10), and correlation analysis between an index (or a multivariable regressionequation) with the outcome (three). Finally, Poisson regression model, negative binomial regression modeland random forest model were each used once.

In order to come up with high-quality prediction models for exacerbations in COPD patients, a standardmethodology for developing the models should be adopted [55]

Prediction models for exacerbations inpatients with COPD

CODEX was designed to predict mortality and hospitalreadmission in 3–12 months after discharge of patients hospi-talized for AECOPD.6

BODE2 (body mass index [BMI], airflow obstruction, dyspnea, andexercise capacity), BODEX3 (BMI, airflow obstruction, dyspnea, and previous severeexacerbations), ADO4 (age, dyspnea, and airflow obstruction), and DOSE5 (dyspnea,airflow obstruction, smoking status, and exacerbation frequency

Prediction of short term re-exacerbation inpatients with acute exacerbation of chronicobstructive pulmonary disease

When the recordings met criteria number 1 a red lineappeared on the corresponding day of the patient’s time-score plot (Fig. 2) and an electronic mail message wasautomatically sent to the research team.

To establish a baseline all patients entered an exacer-bation-free run-in period of 14 days where they recordedsymptom score and lung function as described above.Baseline for each symptom score and FEV 1 was the medianand the mean value respectively of the 14-day run-in periodrecordings

. AECOPD was, therefore, regarded to be present ifat least one of the following criteria was encountered:1. An increase of at least 1 degree of 2 symptom scoresand/or a decline in FEV 1  10% from baseline for 2successive days.2. A patient presenting with symptoms they felt to bethose of AECOPD and sought help that resulted in thembeing given a course of antibiotics and/or prednisolone.

Remote daily real-time monitoring in patients withCOPD e A feasibility study using a novel device

Conventional threshold-basedalgorithms, adopted in the majority of reviewed stud-ies (n ¼ 12), show poor performance in early detect-ing respiratory exacerbations or identifying severityand duration in COPD, 13,53 with the best reportedaccuracy being 73% of exacerbations detected. 52 ; bestsensitivity/specificity 66%/93%55 24 h before hospi-talization.

Whilst the 7-day recall scoresand the daily diary scores have been found to beequivalent in detecting changes over time of theimpact of COPD symptoms, only daily data seemto be suitable if the outcome of interest is detectingthe onset of exacerbations. 85

In COPD, physiological measurements have notproved to be able to predict deteriorations, eitherbecause they change late in the time course ofexacerbation, they cannot be measured reliably orbecause therapeutic interventions during the experi-ment alter the outcomes hindering the accuracy ofalgorithms. 17

One study used a com-promise solution, 56 whereby the patient with a lowrisk of exacerbation monitored on a weekly basis andwhen risk increased according to the predictivemodel, reporting tasks could be scheduled daily toensure timely detection.

A common limitation in the selected studies was theexamination of a relatively small group of patientswho presented a high rate of exacerbations. Winterperiods were extensively selected for trials becauseexacerbations are more likely than in other sea-sons. 51 Although this may have enabled recruit-ment of ‘at-risk’ populations it may have affectedgeneralizability of results.

EXACT-PRO was specifically designed and validated fordetecting COPD exacerbation

(EXACT)

An ‘event-based’ definition of an exacerbation, 70including self-administration of medication orunscheduled visits to emergency units and/or admis-sions, was used in nine studies (n ¼ 6 in COPD andn ¼ 3 in asthma studies). Symptom-based definitionsof exacerbation (e.g. Anthonisen criteria 71 ) were usedin seven COPD studies.

Exacerbation criteria

Common classification algorithms for supervisedlearning in the healthcare field include artificialneural networks, 36 decision trees, 37 random forests, 38Bayesian networks, 39 k-nearest neighbors,40 supportvector machines, 41 linear discriminant analysis, 42 k-means clustering 43 and logistic regression. 44

Use of predictive algorithms in-homemonitoring of chronic obstructivepulmonary disease and asthma:A systematic review

n most cases, if censoring isnegligible and the follow-up period clearly defined, logistic regres-sion is used; if censoring is significant or time to event is important,then a survival time approach using a Cox proportional Hazardsmodel is preferred. Other more complex model approaches such asmachine learning or competing risk models exist but are beyond thescope of this chapter [18, 19].

To ensure stability of the model coefficients in logistic and Coxregression, an event frequency of at least 10/events per degree offreedom in the model is advised [13]. For example, in a cohort of1000 patients where 100 outcomes have been observed, the pre-diction model should include at most 10 variables. Ratios of lessthan 10 events per variable can result in overfitting of the data,leading to poor generalizability in other patient cohorts. All thesegeneral aspects of study and model specification should bedescribed in the methods to allow assessment of internal validity.

Internal validity refers to the concept that a prediction model mustbe derived from the study sample in such a way that the modelcoefficients accurately reflect the true relationships between thepredictor variables and the outcome of interest. Internal validitytherefore requires that the prediction model be derived from anappropriately structured and assembled cohort.

ClinicalEpidemiology

Cox proportional hazards modeling would alsohave been a valid approach for risk estimation, but we choselogistic regression analysis because we considered each exac-erbation within our predefined time frame of 2 years to beof equal importance, regardless of whether this exacerbationoccurred early or late in the follow-up period.

We used logistic regression modeling to estimate the riskof occurrence of COPD exacerbations within the proceeding24 months.

We were not able to include all potential predictorsavailable from the literature, such as exercise capacity, theSt George’s Respiratory Questionnaire, oxygen therapy,and gastroesophageal reflux

Unfortunately, wecould not evaluate the widely accepted BODE index,5 becausewe did not perform a 6-minute walking test in any of ourcohorts.

For both cohorts we used the same “operational” definitionfor exacerbation of COPD, that is, symptomatic deteriorationrequiring pulsed oral steroid use or hospitalization.15–1

For both cohorts we used the same “operational” definitionfor exacerbation of COPD, that is, symptomatic deteriorationrequiring pulsed oral steroid use or hospitalization.15–1

Development and validation of a model to predictthe risk of exacerbations in chronic obstructivepulmonary disease

An increase in respiratory rates and thepercentage of respiratory cycles triggered by the patients nearly systematically precededexacerbation in patients with COPD treated by home NIV [ 25].

mean decrease of 700 steps per day was associated with an increasein the EXACT score indicating the start of an exacerbation [21 ].

The COPD “frequent exacerbator” phenotype is consistently defined by atleast two treated exacerbations per year and is associated with poor long-term outcomesand an accelerated decline in lung function

Remote Monitoring for Prediction and Management of AcuteExacerbations in Chronic Obstructive PulmonaryDisease (AECOPD)

COPD exacerbations are more common in females, patients with car-diometabolic or psychiatric comorbidities—in particular depression—and at the mostsevere spectrum of the disease. A history of prior exacerbations has been demonstrated tohave by far the strongest association with the risk of future exacerbations [7,8]

The primary composite outcome was death or admissionsfrom the baseline data collection until 60 months. Thesecondary outcome was exacerbations from the baselinedata collection until 60 months.

Prediction of severe exacerbations and mortalityin COPD: the role of exacerbation history andinspiratory capacity/total lung capacity ratio

Seven out of nine studies foundin the literature confirmed the finding that female patientshave a higher number of total exacerbations.15–17,19,21,22,31

Five different prediction models for the annual exacerba-tion rate were estimated using negative binomial regression2

Participants were asked to use the following definition foran exacerbation: a moderate exacerbation was defined asan increase in symptoms requiring a visit to a health careprovider and a course of antibiotics and/or oral steroids.A severe exacerbation was defined as an exacerbation requir-ing hospitalization.

Prediction models for exacerbations in differentCOPD patient populations: comparing results offive large data sources

History of Exacerbations

The test set AUROC of 0.86 was relatively high,18–24 butour models may be of limited value for prediction ofsevere exacerbations due to high false positive rate

These methods included logistic regression with multipleregularization methods (lasso, ridge and elastic net), ran-dom forest and gradient boosted trees models (XGBoost).

This set of models(along with support vector machines and neural networks,which were not taken into consideration being more chal-lenging to interpret) are considered gold standard formachine learning classification studies done on tabulardata. Resampling was applied during cross-validation,making sure that only training folds of each cross-valida-tion iteration are affected, and the effect of resampling istested on the non-resampled test fold in each cross-valida-tion iteration.

The prediction period was divided into non-overlap-ping 10-day prediction windows for each patient (Figure3). Before the start of each 10-day prediction window,lookback periods of 10, 30, 60, 90,180, 365 days or theentire patient history were set up depending on the vari-able. For details refer to Supplemental Table 1.

Predicting Hospitalization Due to COPDExacerbations in Swedish Primary Care PatientsUsing Machine Learning – Based on the ARCTICStudy

Development and Validation of a MultivariablePrediction Model to Identify Acute Exacerbationof COPD and Its Severity for COPD Managementin China (DETECT Study): A Multicenter,Observational, Cross-Sectional Study

he outcome was the occurrence of AECOPD (ie occurrence versus no occurrence)

Seasonality isknown to affect COPD exacerbations, most frequently occur-ring in the winter months,37–39 which is likely to reflect anincreased prevalence of respiratory infections, reduced immu-nity, altered environmental conditions, and physiologicalresponses during these months.40,41

Timing of Re-Exacerbation RiskTo understand the timing and dynamics of re-exacerbationrisk, and to evaluate the choice of cut points for early andlate re-exacerbations, Kaplan–Meier cumulative incidencecurves of time to first re-exacerbation amongst all patientsin cohort A were plotted.

The cut-off pointfor early and late re-exacerbations was chosen based on clinicaljudgment, and was informed by an analysis of dailyre-hospitalization risk in the US.18 The 90-day cut-off pointwas also chosen based on clinical development support for anAECOPD therapy. Patients were classed as having nore-exacerbations if they did not re-exacerbate within the180-day period after the index date.

Re-exacerbations were defined a priori as either early (occur-ring within 1–90 days after the index date) or late (occurringbetween 91–180 days after the index date

This included start date, end (reso-lution) date, treatment (oral corticosteroid/antibiotics),severity (including hospitalization), and outcome (includ-ing death). Data was collected at scheduled visits (baselineand 3, 6, 12, 18, 24, 30, and 36 months), using an electro-nic case-report form and based on either patients’ recall ofexacerbation events or available medical records forexacerbation events, and was supplemented by monthlyphone calls to ECLIPSE participants

First, we aimed to develop and validatea predictive model capable of identifying factors potentiallypredictive of experiencing early, late, or no re-exacerbationwithin 180 days.

Predicting Re-Exacerbation Timing andUnderstanding Prolonged Exacerbations: AnAnalysis of Patients with COPD in the ECLIPSECohort

A challenge in COPD is the variation between patients and how to set alarm limits for anindividual patient. Of the 16 articles included in this review, only eight studies (three at high riskof bias, one at low quality and two at moderate quality) [ 5, 13, 14 ,18 – 21, 25] mentioned that they hadcustomised the alarm limits for each individual. Methods used were reported in six out of the eightstudies

Monitoring of Physiological Parameters to PredictExacerbations of Chronic Obstructive PulmonaryDisease (COPD): A Systematic Review

ACCEPT 2¢0: Recalibrating and externally validatingthe Acute COPD exacerbation prediction tool (ACCEPT)

Prediction of readmission in patientswith acute exacerbation of chronic obstructivepulmonary disease within one yearafter treatment and discharge

Exacerbations were defined as subject-reported worsen-ing in respiratory health, requiring therapy with systemic cortico-steroids and/or antibiotics. Frequent exacerbations were definedas 2 or more exacerbations in 1 year, and severe exacerbations weredefined as those leading to hospitalization or emergency room vis-its [9]

Prediction of Acute COPD Exacerbation in theSwiss Multicenter COPD Cohort Study (TOPDOCS)by Clinical Parameters, Medication Use, andImmunological Biomarkers

Prediction of 30-day risk of acuteexacerbation of readmission in elderly patientswith COPD based on support vector machinemodel

Acute exacerbationreadmission within 30 days due to the short acute exacer-bation cycle, not only severely damages lung function andincreases the risk of death, but also occupies a large num-ber of medical resources [4]

Our study assumes that the SVM model can achieve acertain prediction effect in predicting the risk of readmis-sion in COPD patients, and the results have certain refer-ence value. Therefore, it is proposed to use SVM to builda 30-day acute exacerbation readmission risk predictionmodel for elderly COPD patients, and evaluate its predic-tion effect, so as to provide a basis for early identificationof patients with high risk of readmission in the future.

Neither the developmental nor the validation datasetsincluded patients with mild (GOLD 1) severity and, assuch, we could not establish the accuracy of predictionsfor this subgroup.

ACCEPT is externally validated in anindependent cohort extending its generalisability beyondtherapeutic clinical trials.

Azithromycin reduces annual exacer-bation rate by 27%.8

The Acute COPD Exacerbation Prediction Tool (ACCEPT):a modelling study

ACCEPT can combine predicted risk with effectestimates from randomised trials to enable personalisedtreatment.

We used a joint accelerated failure time and logistic modelto characterise rate and severity of exacerbations. We havepreviously published details of this approach elsewhere.14

A 2017 systematic review of clinical prediction modelsfor COPD exacerbations found that only two models18,19 ofthe 27 reviewed reported on any external validation. Whenavailability of predictors and practical applicability werealso considered, none of the models were deemed readyfor clinical implementation.6

ACCEPT predicts rate and severity of exacerbations.

For example, ACCEPT can predict thenumber of exacerbations at a given time period, time tonext exacerbation, and probability of having a specificnumber of non-severe or severe exacerbations within agiven follow-up time (up to 1 year). By contrast, logisticregression models, used in most previous clinicalprediction models, predict the probability of having at leastone exacerbation in a single timeframe.6

Outcomes of interest were rates of exacerbations andsevere exacerbations over 1 year.

The Association Between Rate and Severity of Exacerbations in ChronicObstructive Pulmonary Disease: An Application of a Joint Frailty-Logistic Model

Prevention, early detection, and appropriate treatment ofexacerbations are a focal point of attention in COPD careand research.

Use of the ubiquitous proportional hazardsmodel, with time to first exacerbation as the outcome, is acommon mode of inference in contemporary clinical trialsof COPD. While it is robust in estimating treatment effectin randomized controlled trials, this analytical method fallsshort of providing other features, such as background rateof exacerbations or the shape of the incidence function, toenable predictions about the rate and (absolute or relative)duration of time to future events for a given patient. Asmentioned by Cox et al. (11), making such informative pre-dictions has been hindered by the widespread use of semi-parametric proportional hazards models.

In the present work, we used a joint para-metric recurrent-event and logistic regression model toenable full quantification of exacerbation incidence andseverity and their correlation

such as bloodeosinophil count, chronic bronchitis, gastroesophagealreflux, socioeconomic status, and insurance coverageimprove risk prediction remains to be investigated.

Second, ACCEPT does not advance ourunderstanding of how various risk factors operatein combination. Many variables appear to be simplymarkers of severity rather than biological predictors ofreal risk. The authors address one of these: contrary toestablished literature, current smoking confers reducedrisk of exacerbation, probably because patients withsevere disease and high frequency of exacerbations aremore likely to have quit smoking.

Finally, preventionof severe exacerbations is needed before they occur,and performance of this risk tool in individuals whohave never had an exacerbation is not clear and needsto be tested.

COPD exacerbations: finally, a more than ACCEPTable risk score

First, although ACCEPTshowed good-to-excellent discrimination overalland appears superior to exacerbation history alone,improvement in risk prediction was smaller in thosewith previous history of exacerbations than in thosewithout.

Thus, the major improvement in risk prediction withACCEPT appears to be for severe exacerbations.

Usinga joint survival–logistic model, this risk tool providesan individualised risk estimate for exacerbations in thesubsequent year and their severity, as well as the rate offuture events.

Severe exacerbations are especially important asthey are associated with rapid decline of lung functionand 50% of patients die within 2 years after admissionto hospital.

COPD exacerbations: Prognosis, discharge planning,and prevention

Vitamin D supplementation — Adhering to current guidelines regarding vitamin Dsupplementation in patients with a 25-hydroxyvitamin D level <20 or 30 ng/mL (50 or 75nmol/L) reduces COPD exacerbations in addition to benefits in reducing falls and fractures

Noninvasive ventilation — For patients who require noninvasive ventilation (NIV) during ahospitalization for a COPD exacerbations and who remain hypercapnic, nocturnal NIV athome significantly reduces the risk of rehospitalization

Prophylactic azithromycin

defined as "an acuteevent characterized by a worsening of the patient's respiratory symptoms that is beyondnormal day-to-day variations and leads to a change in medication

PREVENTION

Pulmonary rehabilitation

Smoking cessation (see "Overview of smoking cessation management in adults")•Proper use of medications (including inhaler technique) ( table 5 and table 6and table 7 and table 8) (see "The use of inhaler devices in adults")•Vaccination against seasonal influenza (see "Seasonal influenza vaccination inadults")•Vaccination against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)(see "COVID-19: Vaccines")•Pneumococcal vaccination ( figure 1) (see "Pneumococcal vaccination in adults")

COPD exacerbations: Management

HOME OR OFFICE MANAGEMENT OF COPD EXACERBATIONS

More than 80percent of exacerbations of COPD can be managed on an outpatient basis, sometimes afterinitial treatment in the office or emergency department.

ADVICE RELATED TO COVID-19

The point is that the probability (1.3) does not depend on the parameter θ bydefinition of “pivotal quantity.

Patients were contacted by telephone by atrained clinical research assistant at months 1, 2, and 3 afterrecruitment and were asked about potential hospitalizationsbecause of ECOPD. Whenever this was identified, hospital,data of admission, and reason for hospitalization were recordedusing standardized questionnaires. The attending pulmonolo-gist of each patient, blinded to the respiratory results, reviewedall available information and validated the episode of ECOPDhospitalization

First, weanalyzed individual changes in mean respiratory frequency usingtime series analysis of breathing rate in each patient whorequired ECOPD hospitalization during follow-up. This timeseries included 5 baseline days and the 5 days that precededhospitalization. Baseline was defined as the first 5 consecutivefollow-up days with valid recorded data and a minimum com-pliance with oxygen therapy of 4 h/d, before any exacerbationhad occurred. Individual changes were analyzed using the YoungC statistic, which is appropriate for the study of changing trendsin short series with a small number of measures. 16 Second, weanalyzed the change in mean respiratory frequency between the5 baseline days and the 5 days that preceded hospitalization ofall individual time series considered together using analysis ofvariance for repeated measures. Third, we analyzed the discrimi-nating power of the change in the mean respiratory frequency topredict ECOPD hospitalizations using receiver operating char-acteristic (ROC) curves in two different potentially relevantclinical scenarios (increase of breathing rate from baseline to24 or 48 h before hospitalization).

Patients were contacted by telephone by atrained clinical research assistant at months 1, 2, and 3 afterrecruitment and were asked about potential hospitalizationsbecause of ECOPD. Whenever this was identified, hospital,data of admission, and reason for hospitalization were recordedusing standardized questionnaires. The attending pulmonolo-gist of each patient, blinded to the respiratory results, reviewedall available information and validated the episode of ECOPDhospitalization.