The study inclusion criteria identified 5,086 patients for the analysis of receipt of flow cytometry and of survival (full sample). Because not all patients had a claim for a sign or symptom prior to diagnosis, 4,081 (80%) were included in the analysis of delays (delay sample). shows the characteristics of study patients by sample. The median age for both samples was 75 years; approximately 30% of patients in both had at least one comorbidity, as measured by the modified Charlson Comorbidity Index. Chi-square and t-tests revealed no significant differences in patient characteristics between the full and the delay samples(results not shown).
Characteristics of study patients, by analytic sample (n=5,086)
shows the presenting signs, symptoms, and diagnostic tests reported in the full sample for the year preceding the diagnosis date. The most common Medicare claim was for infection (32.2%), followed by lymphocytosis (28.7%), anemia (23.9%), and fatigue (16.6%). Just over half had a complete blood count performed during the year prior to diagnosis.
Presenting signs, symptoms, and diagnostic tests prior to SEER diagnosis date: full sample (n=5,086)
In the delay sample (n=4,081), the mean time between first sign or symptom and CLL diagnosis was 121.5 days, with a median of 63 days (Interquartile range = 0 to 251 days). The results from a multivariable logistic regression model estimating the likelihood of delay between sign or symptom and diagnosis are shown in , and were adjusted for year of diagnosis. The presence of one or more comorbidities was strongly predictive of diagnostic delay (Odds Ratio (OR) 2.83 [95% Confidence Interval (95% CI) 2.45–3.28]). Patients aged 75 years or older were 45% more likely to experience delay (95% CI 1.27–1.65). Females (OR 1.22 [95% CI 1.07–1.39]) and urban residents (OR 1.46 [95% CI 1.19–1.79]) were also more likely to experience delay. Patients who had received care in a teaching hospital prior to CLL diagnosis were 20% more likely to experience diagnostic delay (95% CI 1.05-1.38).
Factors predicting likelihood of diagnostic delay (n=4,081)
Approximately half (2,282) of the study patients had a claim for flow cytometry at any time during the study period; 1,965 (38.6%) patients had their initial flow cytometry performed within 30 days before or after the SEER diagnosis date. The use of flow cytometry increased significantly over the study period (see ; Cochran-Armitage test for trend significant at p < .01. It should be noted that Medicare initiated a national coverage decision for flow cytometry in 2000). The results from the logistic regression predicting flow cytometry at any time during the study period are shown in , and were also adjusted for year of diagnosis. Significant predictors of flow cytometry were age below 75 (OR 1.46 [95% CI 1.30-1.66]), urban residence (OR 1.27 [95% CI 1.05–1.53]), and residence in the Northeast (OR 2.01 [95% CI 1.69–2.39]) or South (OR 1.51 [95% CI 1.22–1.89]), when compared with the West(the Western classification includes the state of Hawaii). Of note, because of low numbers of cases from some registries, we report these results by geographic region of the US; however, our parameter estimates did not change when SEER registries were included as covariates. Finally, an increased number of signs or symptoms prior to diagnosis was associated with an increased likelihood of having flow cytometry performed (OR 1.15 [95% CI 1.08–1.22]).Patients with comorbidities were less likely to receive flow cytometry (OR 0.88 [95% CI 0.76–1.02]), but the difference was not statistically significant.
Diagnostic flow cytometry and diagnostic delay for Medicare recipients with CLL by diagnosis year, 1991–2002
Factors predicting likelihood of flow cytometry (n=5,086)
The median survival time for the full sample was 10.4 years. Patients who survived beyond the end of the observation period were censored at 5,056 days(13.8 years, the end of the study period). The results of the multivariable Cox regression model are shown in ; hazard ratios (HR) and 95% confidence intervals are reported. This model estimated the effect of patient characteristics, as well as processes of care, on overall survival. Patients diagnosed at the age of 75 or higher, males, and patients with comorbidities had significantly shorter survival times. While the receipt of systemic chemotherapy within six months of diagnosis (HR 1.58 [95% CI 1.42–1.77]) and history of care in teaching hospitals (HR 1.23 [95% CI 1.14–1.34]) were also associated with shorter survival times, patients who had flow cytometry performed had significantly increased survival times (HR 0.84 [95% CI 0.76–0.91]). We did not observe a significant effect of delay between sign or symptom and diagnosis on overall survival (HR 1.11[ 95% CI 0. 99–1.25]).
Effect of selected covariates on overall survival
In the sensitivity analyses describe d above, the relationships reported were similar when delay was treated as a continuous measure, as tertiles, or as quartiles, as well as if time from sign or symptom to diagnosis exceeded the mean (as opposed to the median)for the sample. When we compared parameter estimates, hazard ratios and statistical significance for the dependent variables in a re-estimated Cox model with empirically-derived independent variables, these also did not change appreciably. While diagnostic delay remained not significant, an interaction term between delay and comorbidity was marginally significant (OR 1.21, 95% CI 1.01–1.44, p = .034).