This retrospective study examined diagnostic patterns and diagnostic, clinical and economic outcomes for patients with suspected hematologic cancers/conditions. Overall, the diagnostic outcomes examined in the study generally favored Genoptix relative to the OL cohort, with fewer changes in diagnosis, fewer repeat bone biopsies or changes in chemotherapy treatments. Differences between the Genoptix and the LL groups were smaller and were not compared directly.
Population characteristics of the Genoptix and LL cohorts differed from the OL, including initial suspected diagnosis listed on the bone marrow biopsy. Multiple regression analyses adjusted for final or suspected diagnoses as appropriate, but the differences in diagnoses may reflect not only underlying differences in the actual diseases of the patient populations, but also differences in coding practices across institutions. However, other differences in severity or complexity of the patients’ conditions may not be adequately reflected in claims data.
The distribution of test types differed for Genoptix compared with the other cohorts. A higher percentage of Genoptix patients were more likely to have undergone complex diagnostic tests and Genoptix was consistently recorded as the provider on the biopsy-related tests during the 30-day testing period. For the OL, the complex diagnostic tests were more likely to have been performed at a lab type other than the one providing bone marrow morphology assessment, requiring clinicians and pathologists using the other labs to integrate test results from multiple sources.
The multiple regression analyses suggested that there may be an advantage for Genoptix over other lab types in reaching a final diagnosis earlier. There was also an advantage for LL compared with the OL in this regard, but the effect for Genoptix was slightly more than double that of the LL. However, the difference in mean time to diagnosis was only a few days in the unadjusted numbers. The clinical impact of a difference that a few days make on the diagnosis is unknown; yet, the earlier the diagnosis is made the more rapidly the patient can be assigned to an appropriate treatment plan. The difference between mean and median time to diagnosis indicates there was a large amount of variability in the time to diagnosis, suggesting that even if on average the difference was a few days, there was a subset of patients who were subject to much lengthier delays in diagnosis. Since the definitions for identifying final diagnosis were created independently of the cohorts we do not expect that any effects due to specific definitions used would vary across cohorts other than the variation due to the initial and final diagnoses.
The Genoptix cohort had the lowest rates of unstable diagnosis and unexpected changes in final diagnosis, followed by the LL and then the OL cohorts. This finding parallels the hypothesis that improvements in diagnostic certainty and completeness occurs with initial hematopathology specialty laboratory assessment. Claims data may not adequately capture this type of diagnostic instability and the algorithm developed in the study may not reflect actual changes in diagnoses. However, the results of the study are consistent with the ranges reported in the literature, particularly with an NCCN study that identified a 6% discordance rate in diagnoses for B-cell NHL .
The changes in chemotherapy treatments favored the Genoptix cohort in descriptive statistics; however, the relationship was not statistically significant in multiple regression analyses after adjusting for potential confounding variables in the model. This may also be due to small size in number of patients that experienced a change in chemotherapy. Since the relationship between the laboratory testing and specific clinical outcomes is unclear, these results should be interpreted with caution.
Results for cost and utilization suggest some differences between the cohorts. The source of costs differed across the cohorts, with the ‘other’ (i.e. laboratory) services being the biggest driver for Genoptix costs compared with hospitalization costs for the OL cohort. The adjusted analyses suggested that overall, paid costs for the GX cohort were lower than for the OL cohort. The LL cohort was not the analytic reference group and was not compared directly to the Genoptix cohort. In general, cost differences between the Genoptix and LL groups were either small, or costs were slightly lower for the LL group.
Overall, our results support that compared to the OL for diagnosis of hematologic malignancies and MDS, a hematopathology specialty laboratory may result in more rapid final diagnosis, fewer changes in diagnoses, reduction in need for follow-on testing including repeat biopsy procedures, and may result in lower overall health care costs. Additional research will be needed to confirm whether the use of a hematopathology specialty laboratory minimizes potential harm from misdiagnosis compared to other types of labs and whether there is an efficiency benefit in using a specialty laboratory compared to large laboratories.
Certain limitations are associated with using claims data for research. Reasons for laboratory tests and other clinical parameters are not readily apparent in claims data. Presence of a diagnosis code on a medical claim is not proof positive of the disease; the disease may have been coded incorrectly. For some hematologic malignancies, the ICD-9-CM coding schematic does not distinguish between disease subsets and the heterogeneity and diversity of the conditions. The study did not evaluate quality of life as this type of data is not available in claims data. Due to the overall survival differences between the various hematologic diseases and too few patients for a specific diagnosis, follow-up times were not considered adequate to assess survival in this study.
While this study used multiple regression analyses to adjust for differences in patient populations between the laboratory cohorts, there may have been differences that could not be identified. Clinician characteristics, such as varying degrees of expertise in recognizing these conditions, may have impacted outcomes as well. Selection bias may account for where biopsies were sent for evaluation: more complicated or confusing cases may have been sent to a specialty laboratory initially. In other instances, contractual considerations and insurance coverage may have determined laboratory selection. In addition, it is not clear if a repeat bone marrow biopsy was necessitated by an initial inadequate sample (i.e., poor technical quality).
Differences in treatment patterns cannot be evaluated as appropriate care since reasons for those treatment patterns were not available. Similarly, while the study developed an algorithm to identify progression and change in diagnoses in claims data, the algorithm cannot be verified based upon claims data alone. This study also excluded academic centers with hematopathology fellowships, as the patients seen in academic centers could differ significantly in their underlying disease in ways unlikely to be measurable in claims data. Furthermore, claims data do not contain quality of life and disease severity information. Thus, it is unclear if the differences observed in this study would also be observed in clinical outcomes.