From 2% to 100%: How One Home Health Agency Closed Its Call Review Gap

A mid-size home health agency in the Southeast was losing accounts it did not know were at risk. Three managed care contracts — representing roughly $2.4M in combined annual revenue — declined to renew within the same quarter. Exit interviews with the payer contacts surfaced a consistent theme: unresolved patient complaints, inconsistent scheduling follow-through, and a general sense that calls into the agency disappeared into a void.

The agency's QA program was not negligent. It was typical. Two dedicated quality analysts reviewed approximately 16 calls per day each, selected through a combination of random sampling and supervisor escalation. That covered about 2% of the agency's daily volume of 800+ inbound and outbound calls. The analysts scored calls against a 14-point rubric, documented findings in a spreadsheet, and routed coaching recommendations to supervisors on a monthly cycle.

By every internal measure, QA was functioning. By every external outcome, it was insufficient.

What 2% Sampling Actually Looked Like

The agency's call volume broke down roughly as follows: 35% scheduling and visit confirmation, 25% clinical intake and triage, 20% billing and authorization inquiries, 15% patient or family complaints and concerns, and 5% miscellaneous. The QA team's random sample pulled disproportionately from scheduling and billing — high-volume, low-complexity calls that were easy to score and rarely surfaced issues.

The calls that mattered most — a patient reporting a fall during a triage call, a family member expressing frustration for the third time in a week, a caregiver concern that matched a pattern across multiple patients — sat in the unreviewed 98%. Not because anyone chose to ignore them, but because there was no mechanism to surface them.

The compliance team discovered this the hard way during a state survey. A surveyor pulled five patient records and cross-referenced them with call recordings. In two cases, patients had reported falls by phone that were never escalated to the clinical team. The calls existed. The documentation trail did not. The resulting deficiency citation required a plan of correction and triggered a follow-up survey six months later.

The Decision to Move to 100% Review

After the survey findings and the lost contracts, the agency's VP of Operations convened a working group: compliance, clinical leadership, IT, and the QA team. The question on the table was not whether the current process was broken — that was established — but what a realistic alternative looked like.

Hiring additional QA analysts was modeled first. Reaching even 10% coverage at the current call volume would require four additional FTEs, roughly $280K in fully loaded annual cost, plus the lead time for hiring, onboarding, and calibration. Getting to 100% through manual review was not on the table. At 12 minutes per call, it would take a team of 67 analysts working full shifts to keep pace with daily volume. That number ended the conversation about scaling with headcount.

The working group evaluated automated call review platforms over a 90-day period. The selection criteria were specific: the system had to process calls in near real-time, support a custom tagging taxonomy rather than relying on generic sentiment scoring, integrate with the agency's existing telephony stack, and produce output that compliance and clinical staff could act on without requiring a data science background.

They deployed SurfacerIQ and began with a 30-day parallel run — automated review running alongside the existing manual QA process on the same call set, comparing outputs.

Building a Taxonomy That Matched the Business

The most consequential decision in the implementation was not technical. It was taxonomic.

The agency worked with its compliance officer, director of nursing, and two senior QA analysts to build a custom taxonomy of 24 tags organized into four domains:

Clinical escalation triggers — fall reported, medication concern, wound status change, hospitalization mention, pain escalation, decline in function. These mapped directly to the agency's clinical escalation policy and state reporting obligations.

Compliance and documentation flags — PHI disclosed to unauthorized party, verbal order not read back, consent language missing, call not documented in EMR, identity verification skipped.

Churn and retention signals — patient expressed intent to switch agencies, repeated complaint (third or more contact on same issue), family dissatisfaction with caregiver, billing dispute unresolved after two contacts, missed visit not proactively addressed.

Operational patterns — hold time exceeding four minutes, call transferred more than twice, agent unable to answer authorization question, scheduler could not offer appointment within 48 hours.

Each tag carried a severity weight and a routing rule. A fall reported during a triage call triggered an alert to the clinical supervisor within two minutes. A churn signal routed to the retention coordinator with the call summary and history attached. A compliance flag landed in the compliance officer's daily digest for review and disposition.

This was not a sentiment analysis dashboard. It was a structured detection system built around the agency's own risk definitions.

What Changed in the First 90 Days

The results from the first quarter of full deployment were granular enough to act on and significant enough to justify the investment.

Falls detection went from incidental to systematic. In the first 90 days, the system flagged 47 calls where a patient or family member reported a fall. Of those, 11 had not been escalated by the receiving agent. Under the prior QA model, the probability of catching any single one of those 11 calls was roughly 2%. The compliance team implemented a same-day review protocol for every fall flag, and the clinical team closed the escalation gap within three weeks through targeted retraining.

Churn signals surfaced before contracts were lost. The system identified 23 patients who had made three or more contacts expressing dissatisfaction within a rolling 14-day window. The retention coordinator reviewed each cluster, and in 19 cases, outreach resolved the underlying issue — typically a scheduling inconsistency or a caregiver mismatch. Four patients still chose to transfer, but in each case the agency had documentation of the issue and the response. When one of those patients' managed care organizations asked for detail on the discharge, the agency had a complete record. That conversation, the VP of Operations noted, was fundamentally different from the ones that preceded the lost contracts.

Compliance findings dropped measurably. The agency's internal pre-survey audit — conducted quarterly — had historically surfaced 8-12 documentation deficiencies related to call-based interactions. In the first post-deployment audit, that number was two. Both were edge cases involving after-hours calls routed to an answering service outside the monitored telephony environment.

The QA team's role shifted, but the headcount did not. The two existing QA analysts moved from random sampling and manual scoring to exception-based review. Their daily workflow shifted to reviewing flagged calls, validating tag accuracy, refining taxonomy rules, and conducting targeted coaching sessions based on specific, documented call examples rather than generic rubric scores. One analyst described the shift as moving from "looking for needles in a haystack" to "reviewing the needles someone already pulled out."

No new FTEs were added. The total cost of the automated review platform was less than the fully loaded cost of a single additional QA analyst.

What This Case Demonstrates

This is not a story about technology replacing people. The QA analysts are still employed, and their work is more impactful than it was before. The compliance officer still reviews findings and makes disposition decisions. The clinical supervisors still manage escalations.

What changed is visibility. At 2% sampling, the agency operated on the assumption that its call handling was generally acceptable because the small slice it reviewed was generally acceptable. That assumption cost them $2.4M in revenue, a deficiency citation, and the operational credibility that takes years to rebuild with payer partners.

At 100% review, every call is categorized, every risk event is surfaced, and every pattern is detectable. The agency does not review every call manually — that was never the goal. The goal was to ensure that no call containing a fall report, a compliance exposure, or a churn signal went undetected. That goal is now met, every day, across every call, without scaling the team.

The gap between 2% and 100% is not a technology gap. It is an information gap. And in healthcare operations, information gaps become compliance gaps, retention gaps, and revenue gaps — usually in that order.