Identifying missed opportunities in the diagnosis of cancers in the United States Veterans population

Start Date Oct 2018

Code B1-Aff

Status Ongoing

Project Lead
Senior Lead
Prof Yoryos Lyratzopoulos, Dr Monica Koo (both UCL), Prof Greg Rubin (Newcastle University)


Delays in cancer diagnosis can result from lack of timely follow-up of positive test or imaging results. Issues from these delays, such as later cancer stage at diagnosis and lack of curative treatment options can contribute to poor mortality. To study factors that may influence timely cancer diagnosis and health outcomes, we would like to examine three hypotheses:

Hypothesis 1: Patients with emergency presentation (new cancer diagnosis made through ED and hospitalization) are more likely to have later, advanced cancer stages at diagnosis and poor survival than those with non-emergency presentation

Hypothesis 2: Patients with “multiple consultations” (3 or more visits to primary care per Lyratzopoulos et al. 2014) are more likely to have later, advanced cancer stages at diagnosis and poor survival than those without “multiple consultations”

Hypothesis 3: Patients with electronically flagged missed clinical findings or suspected delays in diagnostic evaluation* are more likely to have later, advanced cancer stages at diagnosis and poor survival than those without missed findings. We will study two types of electronically flagged missed clinical findings or delays in diagnostic evaluation.

Aims & objectives

Project Aim: To characterize the relationship between indicators of quality of cancer diagnosis and outcomes in the US Veteran population

a) Validated triggers for delays in diagnostic evaluation in five types of cancers (based on prior studies): Validated trigger algorithms, previously developed by the US team, will be applied to the CDW to identify patients who are “trigger positive” i.e. patients who potentially experienced missed opportunities for cancer diagnosis.

b) Chronological patterns of certain ‘clinical event-cancer’ dyads that are suggestive of quality problems (new diagnosis of endometrial cancer that was preceded by postmenopausal bleeding episode > 6 months prior). We will target 8 cancers: Bladder Cancer, Breast Cancer, Cervical Cancer, Colorectal Cancer, Hepatocellular Carcinoma, Lung Cancer, Ovarian Cancer, and Pancreatic Cancer


Key indicators of quality include missed “red flag” symptoms, number of consultations prior to the diagnosis of cancer(s), and emergency presentations. Outcomes include cancer stage at time of diagnosis and survival.

We will conduct a large retrospective database study using longitudinal, administrative patient data obtained from the Department of Veterans Affairs (VA) Corporate Data Warehouse (CDW). Our sample will include patients who were diagnosed with either 1) Bladder Cancer 2) Breast Cancer 3) Cervical Cancer 4) Colorectal Cancer 5) Hepatocellular Carcinoma 6) Lung Cancer 7) Ovarian Cancer 8) Pancreatic Cancer from 2006 to the most recent available data (2017 as of July 2019).

Analysis:  Multivariable logistic regression and Cox proportional-hazards model will be used to assess factors that may be associated with VA patients who were diagnosed with late stage cancer and survival, respectively. Patient factors such as age, gender, race/ethnicity, SES etc. will be collected to adjust for confounding, as well as to determine whether certain cancer subpopulations are more susceptible to poor outcomes.

Outputs & impact

Anticipated outputs or potential impact/implementation include huge societal benefits for improving the detection and measurement of patient safety events such as missed opportunities and delays in cancer diagnosis, as most health care organizations and systems in the US and UK have converted to electronic record systems with databases ripe for mining.

It is essential to use these large sums of aggregated data to push the measurement and patient safety agendas forward, as once we properly use the tools to find and describe patient safety events, we can be proactive about preventing these errors or delays from occurring. Overall, using measures of quality of cancer diagnosis proposed here (emergency presentation, multiple consultations, and missed clinical findings or suspected delays in diagnostic evaluation) across various countries with different healthcare systems could help stimulate efforts to reduce missed opportunities in cancer diagnosis.

Next steps

Next steps in this project include refining variable definitions for emergency presentation and multiple consultations, constructing data tables for analyses, analyses and report of findings disseminated at international conferences, as well as peer-reviewed publications.

Back to Research