Which Predictors Could Identify the Most Costly Patients in Oncology?

June 20, 2019

Comorbidities, toxicities, and certain treatments received, including immunotherapy and bone marrow transplants, were found to be the strongest predictors of high costs among oncology patients.

Comorbidities, toxicities, and certain treatments received, including immunotherapy and bone marrow transplants, were the strongest predictors of high costs among oncology patients, according to an abstract presented at the 2019 American Society of Clinical Oncology Annual Meeting, held May 31-June 4, 2019, in Chicago, Illinois.

“Quality-based payment programs in medicine are currently being introduced nationally, aimed to improve care and reduce costs,” researchers said. “This study aimed to evaluate the top spenders (TS) after cancer diagnosis and predict TS at 2 separate time points using predictive analytics.”

Researchers collected data about patient characteristics, cancer details, treatments, adverse events, and outcomes for patients treated for cancer at the Mayo Clinic from 2007 to 2017. They obtained standardized costs over a 2-year period after first treatment from the Mayo Clinic Cost Data Warehouse. Medicare reimbursements were assigned to all services and adjusted to the 2017 Gross Domestic Product Implicit Price Deflator for inflation. In the study, TS were identified as patients with costs greater than those in the 93rd percentile, which was $113,158 or higher, due to a substantial rise at that level.

Researchers used descriptive statistics and univariate analysis for comparison. They used a prediction model with a training set of 80% and a validation set of 20% using multivariate selection to predict TS. It was repeated using information available at 2 time points, which included time of consultation and time of the last follow-up.

Researchers identified 5626 TS from the 80,385 patients included. The mean overall cost was determined to be $44,953. The prediction models had ROC AUC statistics of 0.82 at the first time point and 0.89 at the second time point in training set, and 0.82 at the first time point and 0.88 at the second time point in the validation set, which indicated good prediction of high costs.

Researcher found the factors most predictive of TS included:

  • the need for blood transfusions within 90 days of treatment, with an odds ratio (OR) of 5.3
  • bone marrow transplant, with an OR of 4.0
  • mild liver disease, with an OR of 3.5
  • hemiplegia, with an OR of 3.4
  • weight loss greater than 10% within 90 days of treatment, with an OR 3.3
  • upper GI cancer, with an OR of 3.0
  • “other” cancer type, with an OR of 2.8
  • immunotherapy use, with an OR of 2.7
  • hospitalizations within 90 days, with an OR of 2.4

Researchers also found the highest costs resulted from hospital services in the TS and non-TS groups. The mean cost of hospital services was $114,258 in the TS group and $13,185 in non-TS groups.

“This is the first study to predict with high accuracy the top spenders in oncology,” researchers said. “Our findings suggest that quality payment programs should adjust for comorbidities, and that reducing toxicity may be an effective method at reducing costs.”

Reference

Waddle MR, Stross WC, Malouff TD, et al. Identifying and predicting the most costly patients in oncology. Presented at: American Society of Clinical Oncology Annual Meeting; May 31-June 4, 2019; Chicago, Illinois. Abstract 6633.