Financial Analysis/ Cost Estimating

Accountable Care Organization Background:

In 2014 a California-based ACO (Accountable Care Organization) bearing risk for approximately 10,000 covered lives had sought answers for increasing expenditures associated with referrals. As such, Dr. Moneyball^TM examined 6 months of claims experience data for analysis.

Due to the confidential nature of the case under study, preparation of data included sanitation, and the removal of identifying variables from graphs or charts.

Our approach to data mining employed the CRISP-DM. As such, a 6-step process included the following components:

• Developing a Business Understanding
• Developing an Understanding of the Data
• Preparation of the Data
• Statistical Modeling
• Model Evaluation
• Model Deployment

Challenge:

The claims in the database represented expenses for services provided by outside provider specialists to whom the ACO has referred patients. Based on our preliminary understanding of the data, although the diabetic patients represent only 25% of the patient base for the ACO, the costs for these patients are roughly 75% of the claims paid out by the ACO for the 10,000 covered lives. This represents a problem for the ACO, and the business strategy was to preliminarily identify recommendations to reduce those claims.

Methodology for Analysis:

Review and analysis of the data included the application of several modalities. These are as follows:

• By way of a Feature Selection through Statistica, we were able to identify the most relevant variables for the study. The Feature Selection was performed using Approved Claims as the Dependent Continuous Variable. Provider Specialty was the most relevant of the Independent Categorical Variables. It found to have high F-value for Importance.

• With Provider Specialty as the Dependent Categorical Variable, we ran an Importance Plot to identify Provider Number (issued by the ACO) as a relevant variable.

• The positive correlation between the Provider Number and the CPT Code was a clue that not all providers are performing the same procedures even though most providers are billing 90000 codes (office visits).

• To clarify what types of providers are serving these patients we decided to run a 2-dimensional pie chart.

Findings:

The findings were revealing…89% of claims paid for referrals of diabetic patients are for psychiatry!

Preliminary Recommendations:

Based on our work, we concluded that as preliminary insights for beta testing, the ACO should consider one or more of the following options to hedge their financial liability on this patient base. A follow-up to further refine and measure is also recommended.

• Consider the integration of a mental health component with primary care to more aggressively meet patient needs internally to reduce referrals.
• Consider the addition of an “in-house” psychiatrist who may some of the more severe cases on a per diem reimbursement methodology.
• Consider negotiating and/or re-negotiating existing contracts with providers. Consider sub-capitation or package rates for psychiatric services with outside providers.
• Re-examine internal referral patterns of internal primary care providers to more closely examine “medical necessity.”
• Re-evaluate internal referral protocol to ensure proper application of guidelines.
• Refer for an internal second opinion.
• Conduct formal evaluation prior to consideration.