Start Date Jan 2020
Code W1-I, PhD
The purpose of this study will be to examine the diagnostic pathways through which people are diagnosed with lung cancer in the United States, in order to increase our understanding of opportunities for improving early diagnosis. This proposed research hopes to examine how the cue to action in the Health Belief Model can be seen in the Modified Andersen Model of Total Patient Delay throughout the continuum of a patient’s diagnostic journey. Boundaries of this inquiry will not limit to an individual step in the continuum of delay in order to provide a holistic context of lung cancer diagnostics. As per the Modified Andersen Model of Total Patient Delay, the process may not unfold in a linear progression and examination of variables through this lens may identify a specific period of delay and/or the contributory factors (Walter et al., 2012).
There is limited data on the diagnostic pathway of individuals with lung cancer in the USA, including basic questions regarding the location of presentation. Examination of existing patient data through this lens is needed because the pathway in which someone is diagnosed impacts their stage of diagnosis, and likely prognosis, and can add to the current knowledge about how to best implement care practices. Zhou et al. (2017) determined a significant difference in survival rates among those diagnosed with cancer in an emergency presentation compared to non-emergency care. In their review of cancer diagnoses in England and the United States, authors concluded receipt of a cancer diagnosis associated with an emergency presentation correlated with a worse prognosis among similar cancer types and stages (Zhou et al., 2017). This indicates the awareness of symptoms and action taken prior to emergency escalation is vital for increasing survival, and must occur throughout the patient, system, and provider continuum. The issue remains understudied and more research is needed to contextualize results in different geographical areas (Zhou et al., 2017). Research is currently underway by Caroline Thompson et al (University of SanDiego) who is using claims data to explore pathways to presentation, but claims data is limited in terms of its granularity for the pre-diagnosis phase of illnesses.
Examining the pathways to diagnosis of patients with lung cancer will help to identify the proportion of individuals who are identified by recommended LDCT screening, compared to those diagnosed through symptomatic presentation (i.e., presenting to family medicine with symptoms) or through opportunistic diagnosis (i.e., detection of lung cancer incidentally in patient being worked up for an unrelated condition). It is likely that the proportions diagnosed through these different routes will differ from research from the UK, given the far greater use of imaging in the US, easier access to specialists (at least among insured populations), more haphazard referral patterns, and greater use of ED services.
Methodology includes a nonexperimental-analytical-retrospective study. The scope of the methodology includes data analysis of de-identified data. Data will be obtained from the University of Washington Medicine (UWM) data warehouse, which includes outpatient and inpatient data from the UWM network of hospitals and outpatient clinics. There are approximately 13 primary care clinics (UW Neighborhood Clinics or UWNCs) in the Puget Sound region. The data also includes hospital-based data from the UW Medical Center, and its partners including Fred Hutchinson Cancer Research Center. Data will primarily be obtained on patients diagnosed with lung cancers who are patients with an established relationship with the UWNC primary care clinics, and will therefore include patients diagnosed in the UWNCs (e.g. through LDCT screening, or CXR or CT scanning for investigation of chest complaints), as well as UWNC patients who are seen in the emergency departments.
The available data should include patient demographics, prescriptions, radiology, lab investigations, as well as outcomes, as well as hospitalizations and complications. The researcher will attempt to exclude patients who are referred in from outside facilities to receive specialized care at the UWMC, in order to try to ensure the data is as representative as possible of primary care practice. Additional efforts, beyond the scope of the proposed CanTest project, will be exploring additional data that may be obtained from unstructured/uncoded data/ free text comments (i.e., the doctor’s notes in the presenting complaint/history of presenting complaint, review of systems, examination findings, assessment and plan), and CPT codes for procedures such as imaging tests. We intend to use the same IRB (ethics) application to obtain a data set that will be used for several studies, including the proposed CanTest study.
The participant cohort of included data will be limiting to those who meet the following criteria:
- Adults 18 years-old and older who were diagnosed with any type of lung cancer between 2013–2019. Lung cancer ICD 10 code is C34.0-C34.92.
- Patients with an existing relationship with the UWNC, ie one or more primary care visit in the 18 months prior to diagnosis
- Extracted data will include: Diagnostic setting (e.g., emergency room, specialist referral, primary care encounter), demographics, insurance type, medications, comorbidities, number of encounters, screening or non-screening, symptoms reported, and lab values.
- Extracted data will exclude: Outcome of lung cancer based on ICD codes as noted above.
Statistical analysis will occur with use of SPSS Statistics software (version 25). Descriptive statistics will be used to summarize the data. Pearson’s chi-squared test will be used to compare the proportions in the different categories across groups. A simple analysis of variance will be used to examine differences between groups. A factor analysis will be used to examine relationships between variables and a correlation will be used to compute the relationship between variables. All statistical analyses will use the .05 level of significance. This will include: numbers and types of lung cancers, patient demographics and comorbidities, location of diagnosis, duration of follow up/mortality.