Unrelated Death After Colorectal Cancer Screening: Implications for Improving Colonoscopy Referrals

Potential Predictors of Early Death

To better define potential predictors of early death, we focused on the 258 persons (5.3%) who died within 2 years after the index procedure and paired them with matched controls. One patient underwent a colonoscopy for surveillance of previously treated cancer and was excluded due to very advanced age, as no matched control could be identified. The mean (SD) age of this male-predominant cohort was 68.2 (9.6) vs 67.9 (9.4) years for cases and controls, respectively. At the time of referral for the test, 29 persons (11.3%) were aged > 80 years, which is significantly more than seen for the overall cohort with 306 (6.3%; P < .001). While primary care providers accounted for most referrals in cases (85.2%) and controls (93.0%), the fraction of veterans referred by gastroenterologists or other specialty care providers was significantly higher in the case group compared with that in the controls (14.8% vs 7.0%; P < .05).

In our age-matched analysis, we examined other potential factors that could influence survival. The burden of comorbid conditions summarized in the Charlson Comorbidity Index significantly correlated with survival status (Table 4). As this composite index does not include psychiatric conditions, we separately examined the impact of anxiety, depression, bipolar disease, psychotic disorders, and substance abuse. The diagnoses of depression and substance use disorders (SUDs) were associated with higher rates of early death. Considering concerns about SUDs, we also assessed the association between prescription for opioids or benzodiazepines and survival status, which showed a marginal correlation. Anticoagulant use, a likely surrogate for cardiovascular disorders, were more commonly listed in the cases than they were in the controls.

Looking at specific comorbid conditions, significant problems affecting key organ systems from heart to lung, liver, kidneys, or brain (dementia) were all predictors of poor outcome. Similarly, DM with secondary complication correlated with early death after the index procedure. In contrast, a history of prior myocardial infarction, prior cancer treatment without evidence of persistent or recurrent disease, or prior peptic ulcer disease did not differ between cases and controls. Focusing on routine blood tests, we noted marginal, but statistically different results for Hgb, serum creatinine, and albumin in cases compared with controls.

Next we performed a logistic regression to identify independent predictors of survival status. The referring provider specialty, Charlson Comorbidity Index, the diagnosis of a SUD, current benzodiazepine use, and significant anemia or hypoalbuminemia independently predicted death within 2 years of the index examination (Table 5). Considering the composite nature of the Charlson Comorbidity Index, we separately examined the relative importance of different comorbid conditions using a logistic regression analysis. Consistent with the univariate analyses, a known malignancy; severe liver, lung, or kidney disease; and DM with secondary complications were associated with poor outcome. Only arrhythmias other than AF were independent marginal predictors of early death, whereas other variables related to cardiac performance did not reach the level of significance (Table 6). As was true for our analysis examining the composite comorbidity index, the diagnosis of a SUD remained significant as a predictor of death within 2 years of the index colonoscopy.