Approximately one-third (n = 161) of 513 cases of colorectal cancer (CRC) might have had one or more missed opportunities to initiate an earlier diagnostic endoscopic test. All missed opportunities by definition were associated with the presence of a predefined clinical clue that would ideally warrant a diagnostic evaluation for CRC. There was a mean of 4.2 missed opportunities and 5.3 clues per patient with at least one missed opportunity. Patients who were older than 75 years, African American and those with congestive heart failure or coronary artery disease were more likely to experience missed opportunities. Suspected or confirmed iron deficiency anemia was the most common clue associated with missed opportunities, and had the longest lag time from first appearance to first endoscopy referral. Hemoglobin value in cases associated with missed opportunities was higher compared to those without, suggesting that mild iron deficiency anemia may be an under-recognized clue for CRC diagnosis. Breakdown in several steps of the diagnostic process was identified, most commonly in the provider–patient clinic encounter, but also in ordering, performance and interpretation of diagnostic tests and in follow-up of patients and abnormal diagnostic test results.
One advantage of this study is that it outlines a new approach to understand and reduce potentially preventable delays in CRC diagnosis. Previous studies, most of which are from outside the US, described the magnitude of delays and their association with both patient and physician factors (15
), but did not present a comprehensive assessment of the types of care breakdowns associated with delays. We focused on missed opportunities in diagnostic care rather than delays in diagnosis because these opportunities precede most delays in care and can be targeted for improvement. For instance, if a clinician missed the presence of iron deficiency anemia on a patient visit and a month later the patient was diagnosed with CRC, we would still denote the first visit as a missed opportunity. We also minimized subjectivity by defining missed opportunities beforehand and the types of clinical clues that would warrant a CRC workup. The presence of an integrated electronic medical record provides valuable clinical information about signs and symptoms, diagnostic tests, consultations and follow-up, and is superior and more detailed than administrative data and paper record review.
Previous studies that addressed delays in CRC diagnosis have found 27–62% of these delays to be provider related (20
). Multiple factors have been implicated in such delays, including misdiagnosis (20
), lack of appropriate physical examination (21
), observation of symptoms suggestive of CRC without subsequent investigation (20
), lack of awareness of symptoms suggestive of CRC (7
) and false-negative barium enema (21
). Our study provides a more comprehensive view of these factors and demonstrates that patients who experience diagnostic delays do so in the presence of multiple definitive clues and likely encounter multiple missed opportunities. In few patients, colonoscopy evaluation was delayed for many years despite the presence of an abnormal test (such as flexible sigmoidoscopy and DCBE) for which colonoscopy was indicated. However, most of these tests were performed before 1996, before the electronic medical record was universally used in the institution and well before notification procedures were put in place to inform providers of abnormal test results. This may explain why providers did not notice the tests on subsequent visits. Other reasons for this unusual delay include change in primary care providers and patients' reduced number of follow-up visits to the health-care system.
Disparities in CRC diagnosis have been described, with African Americans more likely to be diagnosed with more advanced CRC regardless of whether they have health insurance (27
). Our study further supports this phenomenon in terms of disparities in missed diagnostic opportunities. Although the elderly patients differ in their presentation from younger age groups and perhaps have screening related disparities (29
), to our knowledge there is no previous published literature on this finding (22
); in fact one study suggested the opposite (25
The most common clue as well as the one most significantly associated with missed opportunities was suspected or confirmed iron deficiency anemia. We found failure to document, follow-up, or take an action on suspected or confirmed anemia in about half (52%) of the cases with missed opportunities. Given that most studies show approximately 10% prevalence of CRC among patients with iron deficiency anemia (34
) and that the majority of patients with new CRC diagnosis are iron deficient at presentation of CRC, this is a potential area of improvement. Previous studies have described delays in CRC diagnosis in patients with anemia (26
), but determinants and duration of such delays have not been comprehensively studied. In our study, suspected or confirmed iron deficiency anemia had the longest lag time to endoscopy referral (median of 393 days) and was the single most common clue across all types of diagnostic-process breakdowns, that is, those related to the clinical encounter, diagnostic test results and follow-up. We also found that milder cases were more likely to be missed, consistent with findings of a previous study from the United Kingdom (35
Several missed opportunities are likely to have occurred in the clinical encounter. Both systems and cognitive factors have been recently proposed to explain diagnostic breakdowns due to inadequate documentation of history or physical exam or from not requesting certain tests such as for iron deficiency (1
). Consistent with previous work in other types of diagnostic breakdowns in primary care (41
), we found deficiencies in history or physical exam to be common in the medical encounter. However, retrospective reviews such as ours can only obtain limited information about the responsible systems and cognitive factors. Future studies are needed to understand and prevent process breakdowns related to the clinical encounter in CRC diagnosis (42
Our study findings should be interpreted with caution. Although the study was conducted in a single institution, it involved a large number of patients and providers. We did not collect identifiable information on providers and hence were unable to evaluate or adjust for providers' effects. However, patients are allocated to providers non-selectively at the institution (i.e., there is no specific provider assignment pattern for new patients), so it is reasonable to assume that cancer cases are distributed equally among most primary care providers. Subjects in the study do not represent the entire cohort of patients eligible for screening or the cohort of patients who present with signs or symptoms suggestive of CRC but no eventual CRC diagnosis. Hence, we could not account for the impact of false-positive tests, which may also compete for time and resources for diagnostic evaluation. Several clues (signs, symptoms, lab test) that we used to define missed opportunities have unclear and probably low predictive value (and hence high false-positive results). Therefore, chasing these clues in every patient may not be cost-effective in most practice settings. However, the two most common clues (positive-screening FOBT and iron deficiency anemia) are hardly disagreed upon criteria for evaluating the presence of colon lesions (polyps or cancer). For policy makers, most benefit could be garnered by focusing on missed opportunities related to these two clues. We also may have missed clues or follow-up actions that were not documented in the chart. Because of recall bias, we did not interview providers to further analyze why opportunities were missed and what cognitive processes were involved; several were missed many years ago and some providers had since left the institution.
To overcome a methodological limitation of low reliability in studies of diagnostic breakdown (44
), we used two independent reviewers and further analyzed cases only where both agreed with high reliability that missed opportunities were present. We could have added more reviews or provided a venue for increasing agreement but we opted to use only cases where complete agreement was achieved in the first round. Despite a low reproducibility agreement for certain clues, we were assured by the high agreement for several others and a substantial (κ
= 0.75) agreement on the presence or absence of missed opportunities. This level of agreement is much higher than that achieved in previous studies using the similar methodology (41
). Lastly, to address hindsight bias (14
), we did not ask the reviewers to make judgements on outcomes such as stage at diagnosis and patient harm.