Did you know that approximately one in seven households in the US experience food insecurity or lack of access to an affordable, nutritious diet? The Supplemental Nutrition Assistance Program (SNAP) is a government organization with roots going back to the 1930s and the first food stamp program. It is a vital lifeline in providing food benefits to families so they can afford nutritious food. However, processing errors and system glitches often result in payment inaccuracies. This can cause significant hardships for these families. Recognizing the urgent need for improvements, SAS has developed the SAS Payment Integrity for Food Assistance model to address and mitigate these issues.

SNAP serves an average of 42.1 million people monthly, or 12.6 percent of US residents. State agencies responsible for determining eligibility are constantly performing a balancing act. They need adequate resourcing to meet both the demand for benefits as well as the timely processing of applications. When demand for services outpaces resourcing, essential quality control procedures designed to maintain accurate eligibility decisions and subsequent benefit awards break down. According to the Annual Improper Payments Dashboard, the SNAP program overpaid over $8 billion in 2023. This is up from $4 billion in 2020. The payment error rate also grew in that timeframe, going up from 7.36% in 2020 to 11.55% in 2023. These increases are dramatic yet avoidable. This is where the SAS Payment Integrity for Food Assistance model can help.

What is the SAS Payment Integrity for Food Assistance model?

The increase in payment errors includes deliberate fraud, which usually only accounts for 1-2% of the lost dollars. Most improper payments are due to inadvertent agency and household mistakes. However, the historic process of identifying payment errors is cumbersome and time-intensive. The SAS Payment Integrity for Food Assistance model aids the quality control teams in risk assessment to identify cases most likely to have payment errors.

SAS Payment Integrity for Food Assistance model utilizes minimal data about cases, expenses, and income in combination with AI to identify cases with a high probability of payment errors. The platform is unique in that it does not rely on external data, only the information that the agency already collects. By focusing on cases more likely to have payment errors, this model eases the workload of the Eligibility Unit. It also reduces the processing time to identify payment errors, leading to faster adjudication of these errors.

How does the SAS Payment Integrity for Food Assistance model work?

The model emulates the manual work agencies have historically done to identify payment errors. However, it expands the accuracy and efficiency by incorporating automation and advanced analytics. It begins by aggregating existing cases to create peer groups in the population without relying on third-party data. These peer groups are then used to assess reported income and expenses outliers. Finally, prioritization scores are assigned to cases with significant departures from their relative peer groups.

All replaced values, simulated benefit amounts, and distribution statistics are used to determine the probability that a case has an overpayment over a specified threshold. Based on these probabilities, cases are then assigned to risk levels to prioritize investigation.

What information does this model provide?

The SAS Payment Integrity for Food Assistance model outputs four main components at the individual case level:

1. Estimated payment error
2. Parameter statistics
3. Heuristic scores
4. Reporting tables

In combination, these outputs enable the agency and its caseworkers to identify the cases with the highest risk of payment error. It also enables them to prioritize the investigation of those cases. The parameter statistics provide agencies with transparency and validate that the model correctly identifies cases with errors. The reporting tables track error rates across time so agencies can assess the success of their work.

Summary

The goal of SAS Payment Integrity for Food Assistance model is twofold. The first is to improve agency efficiency. This ultimately reduces the burden on low-income families who are responsible for repaying any debt incurred because of processing errors. The second is that with the SAS Payment Integrity for Food Assistance model, agencies can use AI alongside their existing operations data to streamline their processes and reduce payment errors.

Future enhancements are on the horizon to include AI-driven entity resolution and EBT card fraud detection. The entity resolution component will provide deconfliction across the same identities listed on multiple cases and similar addresses applying for benefits. EBT cards can manifest in various ways, including benefit trafficking and card skimming, leading to significant financial losses to the program.