Archives

  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • br Another important reason to uncover the

    2020-08-30


    Another important reason to uncover the true estimate of mortality is to appropriately counsel patients. Using only a 30-day mortality estimate would severely underestimate their true risk of death following an operative resection. Others studying colorectal, lung, and 331731-18-1 resections have suggested the implementation of 90-day postoperative
    mortality outcomes for outcome comparison, quality improvement, and patient mortality risk discussions.7,13,14
    A final aim of this project was to measure the degree of per-center variability in ranking when alternative postoper-ative mortality rates were employed. The STS star ranking system ranks institutions using outcomes and clinical data to distinguish the top-performing 3-star hospitals (mortal-ity/morbidity rates 2 standard deviations higher than the STS mean), from 2-star hospitals (middle 95%), and lower-performing 1-star hospitals (mortality/morbidity
    576 The Journal of Thoracic and Cardiovascular Surgery c August 2019
    Moore et al Thoracic: Lung Cancer
    Top
    Tier
    Ranked Middle
    Bottom
    Mortality Timeline
    FIGURE 2. Line graph indicating how facilities change ranking groups over 12 interval mortality time points. Density corresponds to the number of fa-cilities changing groups. Values below the graph indicate the percent of facilities that changed ranking groups between time points. For example, 2.8% of facilities changed ranking groups between 30 days and 60 days. 
    THOR
    rates 2 standard deviations lower than the STS mean).3 Loosely based on that format, in this model, high-performing hospitals were distinguished from middle- and lower-performing hospitals at multiple time points using mortality. As shown in Table 3, the top echelon of facilities (ie, top 2.5%) change by almost 50% between 30 and 90 days. Figure 2 shows the percent changes between time intervals. The maximum fluctuation between ranking groups is between 30 days and 60 days and 90 days and 120 days. One explanation for this is that 30-day and 90-day mortality are commonly assessed time points. There-fore, immediately after those dates, patients may be less likely to follow-up or may even transition their care to another facility for further treatment. The changes seem to stabilize after 120 days. The high degree of change in the first 60 days in concerning and could potentially lead to inaccurate comparisons if later mortality figures are ignored. In addition, lower mortality (<3%) at 30 days makes Filter hybridization difficult to stratify programs for quality improve-ment. If only a single mortality time point is to be used, likely 90-day mortality would be more accurate to compare rankings between hospitals.
    Study Strengths and Limitations
    The NCDB is among the largest cancer registries in the world and is very useful for studying outcomes in general populations. This investigation using NCDB data is the most comprehensive review of NSCLC surgical mortality at multiple time points to date. Patients with missing vital status or last known contact information had to be excluded from this study. Although our mortality estimates align with 
    published mortality rates, missing data can represent a po-tential source of bias. Additionally, many patient- and treatment-related factors are not captured in the NCDB. For example, a summative Charlson/Deyo score is pro-vided, but additional details regarding the severity of co-morbidities are not available. There is also limited information available in the NCDB regarding in-hospital postoperative course. For example, there are no data detail-ing major postoperative morbidities, reoperation for onco-logic resection, reoperation to address postoperative complications, readmission to hospitals other than the index hospital, and/or transfers between facilities. More granular data for comorbidities, postoperative course, and detailed cancer therapy would improve our ability to accurately adjust for these factors and improve the quality of models generated from the NCDB.15
    CONCLUSIONS
    This study echoes concern that 30-day mortality may not the most appropriate metric to compare outcomes, to facil-itate risk–benefit discussions with patients, or to use as the sole variable in a ranking system to distinguish institutions of excellence. This study provides further evidence that 30-day mortality underestimates the incidence of death following lung resection for NSCLC, which doubles by 90 days. Evaluation of hospital performance shows as quan-tifiable fluctuation in the top tier facilities between 30 and 90 days. Quality improvement projects and ranking algo-rithms should consider incorporation of 90-day mortality data to obtain a more accurate understanding of postopera-tive mortality.