Mortality and Symptom Burden Post Hospitalisation With COPD

The PEARL (Previous admissions, Extended Medical Research Council Dyspnoea Score, Age, Right-sided heart failure and Left-sided heart failure) score predicts readmission or death within 90 days of discharge following an exacerbation of Chronic Obstructive Pulmonary Disease (ECOPD), and was developed by the lead supervisor's research team. Within the PEARL derivation cohort (2 hospitals, 824 patients), they have developed a novel tool to predict one year survival. The performance of this novel tool and existing Chronic Obstructive Pulmonary Disease (COPD) prognostic tools will be compared in the PEARL validation cohort (6 hospitals, 1,593 patients). The existing tools are the Body Mass Index, Airflow Obstruction, Dyspnoea and Exacerbations (BODEX) index (score range 0-9); PEARL score (score range 0-9); COPD Gold Standards Framework Prognostic Indicator Guidance (COPD PIG); Comorbidity, Obstruction, Dyspnoea and Exacerbations (CODEX) index (score range 0-10); Age, Dyspnoea and Obstruction (ADO) index (score range 0-14); Dyspnoea, Obstruction, Smoking, Exacerbation (DOSE) index (score range 0-8), the Supportive and Palliative Care Indicators Tool (SPICT) and another tool recently developed by another centre (PubMed reference not yet available). COPD-PIG and SPICT are not scored in the same manner and therefore do not have a range. Performance of all tools will also be further prospectively assessed in a minimum of 310 patients. The COPD-PIG is intended to only be scored in patients who the clinician "would not be surprised" if they died within one year; the performance of this tool cannot be fully assessed retrospectively. Prospective validation will also allow an assessment of the ease of data collection; this is not available for the existing PEARL cohort but is a key consideration in selection of the final tool to ensure it is appropriate for widespread use on hospital wards. Consecutive, unique patients admitted to Northumbria Healthcare NHS Foundation Trust and Newcastle Upon Tyne Hospitals NHS Foundation Trust with ECOPD will be identified. Demographic and clinical indices, including the components of the nine prognostic tools, will be collected. Caldicott and ethics committee approval are in place (REC:18/NE/0226). Hospital readmissions, utilisation of other healthcare services such as the local palliative care teams and Hospice and survival will be assessed over one year. In common with similar studies, this validation study will be non-consenting. Patients identified as high-risk by the selected tool who survive beyond one year may still warrant palliative care input if their symptom burden is high. The patients enrolled in the validation study will be invited to participate in a consenting longitudinal outcomes study, aiming for at least 50% participation. Symptom burden and functional status will be assessed using the St George's Respiratory Questionnaire (SGRQ), Hospital Anxiety and Depression Score (HADS), Australian modified Karnofsky Score (AKPS), Extended Medical Research Council Dyspnoea (eMRCD) score, and modified Borg scale at baseline, and 1, 3, 6 and 12 months. The baseline and 3 month assessments will be conducted via face to face meetings. The additional 1, 6 and 12 month assessments will be conducted over the phone. This data will be used both to calculate the relationship between symptom burden and death, and to identify the characteristics of patients who are especially symptomatic. The characteristics of the cohort will be summarised using standard descriptive statistics appropriate to the level and distribution of the data. Groups will be compared (including by mortality outcome) using standard tests of statistical inference (e.g. t-test, Mann-Whitney U test, Fishers exact test). The performance of candidate prognostic tools will be compared by 1) assessing the positive and negative predictive value within the high-risk and low risk groups respectively; and 2) Area Under the curve of the Receiver Operating Characteristic (AUROC) analysis, with performance compared using the method of DeLong et al. Where data imputation is required, this will be done using multiple imputation methods. Statistical significance will be set at 5% throughout. The statistical analysis plan will be finalised prior to end of recruitment to avoid potential bias.