Identifying Early Chronic Obstructive Pulmonary Disease (COPD) Using Health Administrative Data

Study Objectives 1. Identifying a cohort of individuals with a new onset of COPD in the between April 1, 2016 and March 31, 2019. 2. To determine factors associated with a new diagnosis of COPD through using traditional mixed-model regression. 3. To evaluate whether information collected within administrative data can be used to create a prediction model for a COPD diagnosis. 4. To determine whether machine learning methodology improves the prediction of a new COPD diagnosis from administrative data. Measures: 1. The cohort of individuals with COPD in Alberta has already been defined, and this data exists within the Alberta Health Services, Respiratory Health Strategic Clinical Network (RHSCN) dataset. It will be used to further identify individuals with a new diagnosis of COPD within the three year study time period. 2. In order to conduct this study, a variety of data sets will be used including: * Inpatient Discharge Abstract Database; * Practitioner Claims Database; * Provincial Registry Database; * Population Health Database and * Pharmaceutical information Network Project Hypothesis We anticipate individuals with a diagnosis of COPD in the last three years will have identifiable markers associated with lung disease in the five years prior to their diagnosis. These markers may include: diagnosis of acute respiratory disease (such as pneumonia, bronchitis, upper respiratory infections), increased health care utilization, and the use of medications such as antibiotics. The project plan will address the specific project goals as follows: 1. Identifying a cohort of individuals with a new onset of COPD from April 1, 2016 to March 30, 2019. Through the RHSCN, a cohort of individuals with COPD has been identified of over 200,000 individuals with COPD in Alberta. This cohort will be refined to identify only those individuals that have been diagnosed within the specified time period. This time period was chosen due to data availability. Given our most recent data, we know that approximate 19,000 individuals have been diagnosed with COPD per year, over the last five years. Thus we can approximate that our dataset will include approximately 55,000 individuals with COPD diagnosed in a three year time period. 2. Retrospectively review the pattern of health care utilization for individuals with a new diagnosis of COPD in the five years prior to their diagnosis. The health care utilization (ED visits, hospitalization visits, physician visits) for each case in the cohort for the previous five years will be identified. 3. Explore the medication use for individuals with COPD for five years prior to their diagnosis. Lastly, the medication use for each case identified in the cohort during the time period five years prior to their diagnosis will be explored. The Pharmaceutical Information Network (PIN) database will be used to identify all medications (both respiratory and non-respiratory) for individuals in the cohort to assess medication use prior to diagnosis. 4. Working with the machine learning provider (AltaML) we will additionally conduct a machine learning based analysis to further explore this data set regarding the same variables.