How AI Is Replacing Manual OIG Screening

For years, OIG screening meant one thing: a compliance officer sitting at a computer, typing names into the OIG's LEIE search page one by one, and copying results into a spreadsheet. It was tedious. It was slow. And it was wrong more often than anyone wanted to admit.

The manual screening era is ending. Here is what is replacing it, and why the shift matters.

The Problem with Manual Screening

Manual OIG screening has two fundamental weaknesses: throughput and accuracy.

Throughput is straightforward. A person can check 50 to 100 names per hour against the LEIE website, depending on how fast they type and how many results they need to review. For a 200-person agency screening monthly, that is two to four hours of work every month dedicated to a single compliance task. Scale that to a multi-site operation with 1,000 employees and the math stops working entirely.

Accuracy is the more dangerous problem. The LEIE's search interface uses exact name matching. You type a first name and last name, and it returns results that match those strings exactly. This means:

• "Michael Johnson" will not match "Mike Johnson"
• "Maria Garcia-Lopez" will not match "Maria Garcia"
• "Jose Hernandez" will not match "José Hernández"
• "Robert Smith Jr" will not match "Robert Smith"
• A data entry error in either your records or the LEIE will cause a miss

Each of these mismatches is a false negative: a result that says "no match found" when a match actually exists. False negatives in exclusion screening are not an inconvenience. They are a compliance violation waiting to happen. The penalties for employing an excluded individual can reach seven figures from a single missed match.

What Fuzzy Matching Actually Does

Fuzzy matching is a family of algorithms that measure how similar two strings are, rather than checking whether they are identical. Instead of asking "are these two names the same?" fuzzy matching asks "how similar are these two names on a scale from 0 to 100?"

There are several approaches to fuzzy matching, but the one that works best for name comparison is called token sort ratio. Here is what it does:

1. Tokenize: Split each name into individual words (tokens). "Robert Allen Smith Jr" becomes ["Robert", "Allen", "Smith", "Jr"].
2. Sort: Arrange the tokens alphabetically. This makes the comparison insensitive to word order—"Smith Robert" will match "Robert Smith."
3. Compare: Calculate the similarity ratio between the two sorted token strings using the Levenshtein distance (the minimum number of single-character edits needed to transform one string into the other).

The result is a score from 0 to 100. A score of 100 means the names are identical. A score of 85 means they are very similar with minor differences. A score of 50 means they share some tokens but are substantially different.

How ClientCare Implements This

ClientCare's screening engine uses a multi-step matching process designed to maximize accuracy while minimizing false positives.

Name normalization is the first step. Before any comparison happens, we normalize both the staff name and the LEIE record name:

• Unicode normalization using NFD decomposition, which strips accenting marks (José becomes Jose, François becomes Francois)
• Case normalization (everything to lowercase)
• Strip non-alphabetic characters except spaces
• Identify and optionally strip name suffixes (Jr, Sr, II, III, IV, V, 2nd, 3rd, 4th)

Index lookup narrows the search space. We do not compare every staff name against every LEIE record—that would be computationally expensive and slow. Instead, we index LEIE records by last name and only compare staff members against records that share a last name or a similar last name.

Dual-threshold matching is where the scoring happens. We screen against both the LEIE and SAM.gov databases and use two thresholds:

• 85 or above: potential match. The names are similar enough to warrant investigation. This threshold catches most name variations while keeping false positives manageable.
• 95 or above: strong match. The names are near-identical. This almost certainly requires immediate action.

The distinction matters for workflow. A 95-score match gets flagged as high priority. An 85-score match gets flagged for review. Your compliance officer investigates both, but the prioritization helps them focus on the most likely real matches first.

What AI Adds Beyond Fuzzy Matching

Fuzzy matching handles name comparison. But modern screening tools add additional intelligence layers:

Contextual validation: When a fuzzy match is found, additional data points can help confirm or rule out the match. Does the state in the LEIE record match the employee's work history? Does the exclusion date make sense relative to the employee's hire date? Does the excluded individual's specialty or role match?

Intelligent scheduling: Not every staff member needs the same screening frequency. New hires need pre-employment screening. Staff who have been clean for years can be screened on the standard monthly cycle. Staff who had a previous potential match (that was investigated and cleared) might warrant closer attention.

Pattern detection: Across an agency's screening history, patterns emerge. If a certain name format consistently causes false positives, the system can learn to flag those differently. If a particular type of name variation is common in a geographic region, the matching engine can be tuned accordingly.

The ROI Calculation

Consider the cost of manual screening for a 100-person agency:

• 2 hours per month of compliance officer time at $35 to $50 per hour: $70 to $100 per month
• Risk of a missed exclusion due to name mismatch: one incident can cost $100,000 to millions
• Risk of inconsistent documentation: potential survey deficiency and corrective action plan

Automated screening eliminates the labor cost, dramatically reduces the accuracy risk, and generates the documentation automatically. The monthly cost of a screening tool is typically a fraction of what you would pay a compliance officer to do it manually—and the tool does not get sick, does not forget, and does not make typos. For a detailed cost breakdown, see The True Cost of Manual OIG Screening.

The real ROI is not in the time saved. It is in the enforcement action that never happens because the screening tool caught what a manual search would have missed.