Prime Develops Predictive Model to Identify Savings in Breast Cancer

Prime Therapeutics’ predictive model was able to identify members with high breast cancer pharmacy and medical claims to help clients better manage drug spend.

Among Prime Therapeutics’ commercial insurance plans, one in 20 patients with actively treated breast cancer are predicted to have more than $100,000 in drug therapy costs annually. In a recent analysis, Prime used its predictive modeling program, which integrates medical and pharmacy claims data, to identify patients with higher costs for breast cancer treatment as a way to target those for drug therapy optimization.

Prime’s research found that between 53,510 and 59,911 members in any given month between September 2019 to May 2020 were identified as having active breast cancer treatment, and between 5.1% to 5.4% had more than $100,000 in drug spend.

“There are savings opportunities here; we wanted to identify and manage the treatment of the cancer itself, as well as identify areas where drug therapy biosimilars can be used,” Patrick Gleason, Pharm.D., said in an interview with Formulary Watch. Gleason is Prime’s assistant vice president, health outcomes. “We can now have a more impactful conversation with a provider about a specific patient who is in the beginning of their treatment about biosimilars.”

Prime reviewed medical and pharmacy claims from 16 million commercial members from 2018 to 2020. The company created breast cancer identification rules using diagnosis codes, CPT codes, Optum Symmetry episode treatment grouper, a list of drugs primarily used for breast cancer with significant cost, and gender to specify the breast cancer population. Predictors assessed included member demographics, financial risk scores, diagnosis codes, breast cancer type, previous spending, drug utilization, and timing of surgery relative to drug therapy.

Three years ago, Prime built a predictive model to identify members who were high-cost drug claimants among the 16 million patients covered in the commercial insurance business, excluding Medicare and Medicaid. The model defined a high-cost drug claimant as someone who had more than $250,000 in medical and pharmacy claims.

“We were trying to predict the future on an annual next year spend, as well as identify those patients aren’t there yet at $250,000 to help to inform the focus as to where we can work on ensuring optimal growth,” Gleason said. This is, he said, a small number of Prime’s commercial insurance patients, just 32 patients out of every 100,000, but this has doubled from 16 patients out of 100,000 from five years ago.

During the analysis of breast cancer claims, Prime studied patients who had more than $100,000 in drug spend in order to identify those who could be switched to a biosimilar or other lower cost therapies. Prime found that one patient in 20 with breast cancer had more than $100,000 in drug therapy costs annually in the combined medical and pharmacy benefit.

“Because of our integrated medical and pharmacy data, we can isolate those individuals who are predicted to spend more than $100,000 and then provide actionable information to the managed care pharmacists to have conversations with providers about that given drug therapy regimen to ensure best outcomes,” Gleason said. “We see this as an opportunity to lower costs and to optimize drug therapy with a biosimilar regimen.”

Oncology is Prime’s largest area for medical and pharmacy claims, Peter Bryan, Pharm.D., MedDrive program director, said in an interview with Formulary Watch. Key predictors identified included previous total drug spend in the past 12 months, use of drugs associated with HER2 positive status, metastases indicators, and hormone receptor positive or negative diagnoses.

“Using all of these aspects together, we were able to develop categories to understand where patients are in their treatment journey,” Bryan said. “We wanted to understand the typical treatments patients go through regardless of their biomarker status, beginning from a mammogram, through a mastectomy, surgery, radiation, etc. There may be some nuances within these patient categories based on their tumor type.”

Bryan pointed out this analysis is only possible from a review of both medical and pharmacy claims. “You can’t do this predictive modeling and identify future high-cost drug claimants with pharmacy claims alone,” he said. “You need to be able to integrate medical and pharmacy claims data and some of the factors and indicators within the medical claims in combination with the pharmacy claims to be able to do a forecast.”

Previous Prime analysis identified that 7% of its members with breast cancer contributed 68% of the total drug spend, and this correlated to HER2+ and triple-negative breast cancer status. “Those two specific breast cancers are much higher spend than anything that we’ve seen with hormone sensitive positive breast cancer,” Bryan said.

Gleason said going forward, Prime plans to run the breast cancer predictor model monthly to determine any savings opportunity and to maximize the drug therapy. “We need a better treatment pathway, consistent recommendations, and we’re going to make recommendations and then document and report the savings,” he said.