• 2018-07
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br Our logistic regression model estimates of any ED


    Our logistic regression model estimates of any ED occurrence within 30 and 365 days showed that older age, stage IV disease at diagnosis, and greater number of comorbid chronic conditions (ie, higher CCI score) were all associated with significantly higher odds of an ED visit (Table 3). In addition, urban residence at diagnosis (compared to big metro) and receiving systemic tar-geted therapy after surgery were associated with a significant increase in the odds of an ED visit within 365 days. For 365-day mortality, we found significantly higher odds of mortality for stage II, III, or IV disease at diagnosis, older age, male subjects, nonmarried, renal cell carcinoma and other histology, CCI score of 3 to 4 or 5 to 10, and receipt of systemic targeted therapy after surgery. For 6-month and 12-month costs, significantly higher costs were associated with stage III or IV disease, older age, black race, presence of additional comorbidities, and receipt of systemic targeted therapy before or after surgery.
    Generally, the estimated odds ratios from our ED visit, mortality, and cost prediction models match the literature.27-31 Across the
    Clinical Genitourinary Cancer June 2019 - e653
    Risk-Adjusted ED Visits
    Table 2 ED Visits, Mortality Rates, and Costs by Disease Stage at Diagnosis
    Stage I Stage II Stage III Stage IV Total
    Total Costs Within 6
    Total Costs Less ED
    Visits Within 6 Months
    Total Costs Within 12
    Total Costs Less ED
    Visits Within
    All costs are presented as US$ and are rounded to nearest $100.
    Abbreviation: ED ¼ emergency department. aStages I and II combined due to reporting requirement related to small sample size.
    models, greater stage of diagnosis, age, receipt of systemic targeted therapy following surgery, and increased CCI score were all associated with greater rates of ED visits, greater mortality, and higher costs. For the ED visit models, we tended to find greater odds ratios for the 365-day prediction models than the 30-day prediction models, which suggests many of these underlying factors continue to affect the like-lihood of an ED visit even much after the initial surgery. Furthermore, it Fer-1 suggests these factors explain more of the likelihood of having an ED visit at the later time points, potentially indicative that as both sto-chastic and surgical quality factors fade in importance, these other underlying patient characteristics may be important predictors that clinicians providing longer term follow up care can use to target patients in greater need of monitoring and follow-up care.
    We find little evidence of an association between a hospital’s risk-adjusted ED ratio and its risk-adjusted mortality ratio. This suggests that the 2 measures may be capturing 2 different measures of quality and highlights part of the reason the OCM included this alternate measure. Because mortality rates are relatively low in the kidney cancer population compared to some other cancer sites, this rela-tionship might not hold for other cancer types.
    We also find that a hospital’s 30-day ED visit rate was not significantly associated with 6- or 12-months costs, although the correlation was positive. On the other hand, 365-day ED visits were associated with significantly increased costs even when the costs of the ED visit were excluded. This suggests that ED visits occurring further away from the initial surgery may be capturing a decline in health. For a variety of logistical and organizational reasons it may be difficult for hospitals to continue to provide follow-up care to 
    patients for longer periods after surgery. However, if payers continue to include ED visits as a quality measure to incentivize follow-up with patients and ensure long term care coordination, it will be important for providers to continue to monitor their risk-adjusted ED visit rates longer term.
    Our study had several limitations. First, because kidney cancer is relatively rare many hospitals saw 10 or fewer patients. Because of this, estimates for some hospitals may be more susceptible to out-liers or random fluctuations in ED visits and, in particular, mor-tality. Therefore, the lower mortality rates for kidney cancer relative to other cancer may just reflect an important difference for kidney cancer while limiting generalizability to other cancers. Second, because we have SEEReMedicare data we are limited to elderly patients with Medicare Parts A and B plans.32 However, the CMS OCM focuses on claims so while this may limit predictive power in our study this is also true of the OCM and therefore an important limitation providers need to be aware of. Third, to ensure sufficient sample we focus on all ED visits. However, in terms of a quality metric, providers, patients, and policymakers may be more con-cerned about ED visits that do not lead to admissions. In our data we observed that 53.1% of ED visits within 30 days and 58.2% of ED visits within 365 days did not ultimately lead to a hospital admission. Future work could focus on these visits, specifically. Furthermore, ED visits may be due to a variety of factors including worsening disease that may be less directly related to the initial surgery, particularly for ED visits further from the index surgery. However, these ED visits may still be important to capture since they may be an indicator of greater patient need for follow-up care