# 50,000 HIV Patients Are Still Seeing Their Doctor. They Haven't Filled ART in Six Months. **Researcher Agent Output · Independent Analysis · April 2026** *ART = antiretroviral therapy · Still in care. Not on treatment.* ### Hero metrics | Value | Label | Detail | |-------|--------|--------| | 49,562 | Seen Not Treated | HIV medical visit in past year, no treatment fill in 6+ months | | 2.8% | Injectable Migration | Share explained by Cabenuva. The remaining 97% stopped filling. | | 3,478 | Irreducible Core | Same prescriber, HIV labs ordered, no ART fill | An analysis of 526,577 patients on oral antiretroviral therapy (ART) in 2022 found that one in three was no longer filling ART two years later. The conventional explanation — patients switching to long-acting injectables — accounts for less than 3%. Approximately 50,000 patients continued to have HIV-coded medical visits without a corresponding ART pharmacy fill. They are not lost to the healthcare system. They are being seen — an average of 20 times per year — without being reconnected to their HIV prescription. ## Where 526,577 Oral ART Patients Went Over Two Years | Status | Patients | Share | |--------|----------|-------| | Retained on Oral ART | 343,803 | 65.3% | | Migrated to Injectable (Cabenuva) | 14,905 | 2.8% | | Seen Not Treated (HIV visit, no treatment fill) | 49,562 | 9.4% | | Active in Healthcare, No HIV Visit | 50,768 | 9.6% | | Dataset Exit (no claims) | 67,539 | 12.8% | --- ## Page 1 — Footnotes ¹ "Seen Not Treated": B20 diagnosis on medical claim Jan 2024–Mar 2025, no oral ART fill in preceding 180 days. Same-provider subset (10,600) and lab-confirmed subset (3,478) control for secondary diagnosis coding. ² Cabenuva via HCPCS procedure codes. Some injectables billed under alternate codes may not be captured; true migration rate may be modestly higher. ³ Same prescriber NPI on last ART fill and recent B20 visit, with viral load (CPT 87536) or CD4 (CPT 86360) lab order at or near the visit. --- ## 45% Were Perfectly Adherent Before They Stopped Among the patients who disengaged, nearly half showed no warning: stable monthly refills averaging 16 per year, then abrupt cessation.⁴ This "sudden stop" pattern occurred across all payer types — 31% commercial, 37% Medicaid, 16% Medicare — ruling out Medicaid coverage disruption as the sole explanation. The remaining 55% split between gradual drifters (widening gaps before cessation), already-intermittent fillers, and single-fill patients. Each pattern implies a different intervention. But the most counterintuitive finding is what happens when these patients return. ### ~7.5 Million Annual healthcare encounters with disengaged patients — 20.5 medical visits + ~44 non-ART pharmacy fills per patient per year. The system sees them. It does not reconnect them to ART.⁵ ## The Patients Most Likely to Return Are Not the Ones You'd Expect | Pattern | Subset | % of lapsers | Ever Return to ART | Return Within 12 Months | |---------|--------|--------------|---------------------|-------------------------| | Sudden Stop | (stable adherence) | 45% | 49% | 26% | | Gradual Drift | (widening gaps) | 13% | 58% | 33% | | Already Intermittent | | 19% | 58% | 31% | | Single Fill Only | | 10% | — | — | Legend: first percentage column is "Ever Return to ART"; second is "Return Within 12 Months." Single Fill Only has no return bars in the interactive chart. The implication: the largest group of lapsers (sudden stoppers, ~24,000 patients) is also the hardest to recover. The most recoverable groups (drifters and intermittent fillers, ~17,000 combined) show detectable warning signs — widening refill gaps — months before they fully stop. Among all lapsers with continued HIV medical visits, 71% eventually returned to treatment, with 38% returning within 12 months. ### Page 2 — Footnotes ⁴ "Sudden stop": ≥12 ART fills in 12 months before last fill, terminal gap ≤45 days. Median terminal gap: 27 days. ⁵ Disengaged patients averaged 20.5 medical visits and 44.3 non-ART pharmacy fills annually — higher than retained patients (19.7 visits, ~40–43 non-ART fills). --- ## 361 Prescribers. One-Quarter of the Problem. Disengagement is concentrated enough for targeted intervention. The top 5% of prescribers (361) account for 24% of all disengaged patients. The top 20% (1,444) account for 57%. These prescribers are largely already within manufacturer field-force target lists. Among the "Seen Not Treated" population, 10,600 patients are still seeing the same provider who last prescribed their ART — and 58% of those lapsed within 18 months, placing them inside the recovery window where return rates are highest. ### Disengagement Is Concentrated Enough to Target Cumulative concentration curve (prescriber percentile vs. cumulative % of patients): reference diagonal = even distribution. Key labeled points on the curve: **361 prescribers → 24%** of cumulative patients; **1,444 prescribers → 57%**. Curve control points (percentile, cumulative %): (0, 0), (1, 7.7), (5, 24.2), (10, 37.9), (20, 56.5), (50, 84.8), (100, 100). ### Entity profile **Community HIV Practice · Internal Medicine · Texas · 664 ART Patients** 32.7% disengagement rate. 122 patients classified as Seen Not Treated, 81% returning to this same provider in planned care settings. Viral load ordering rate: 44.7%. Regimen mix is clinically complex — only 68% on modern single-tablet regimens, with 19.6% on multi-pill combinations — suggesting a clinically complex, treatment-experienced population. Lapse pattern: 53% sudden stops, 21% gradual drift. Among patients who do return, 80% restart on the same regimen. The prescriber is actively monitoring patients whose ART prescriptions have lapsed — the intervention is a conversation, not a referral. *The full analysis profiles 100 prescribers and 50 accounts with behavioral summaries, lapse pattern breakdowns, and regimen-level disengagement rates — available to walk through.* ### Page 3 — Footnotes ⁶ Prescriber identification via last-fill prescriber NPI on pharmacy claims, filtered to exclude pharmacy taxonomy codes. High-volume prescribers with elevated disengagement rates may reflect patient panel complexity at safety-net institutions rather than provider-level quality variation. --- ## Methodology Open and closed medical + pharmacy claims on 330M+ lives (Health Verity), supplemented with MMIT formulary data and provider taxonomy via NPPES. Cohort: B20 diagnosis + 2 or more oral ART pharmacy fills in 2022, tracked through March 2025. Disengagement: no oral ART fill after September 2024. "Seen Not Treated": B20-coded medical visit (Jan 2024–Mar 2025) with no ART fill in prior 180 days. Injectable migration identified via HCPCS procedure codes. ## Sensitivity Results | Challenge | Question | What We Found | |-----------|----------|---------------| | Medicaid coverage loss | Did patients just lose insurance, not stop treatment? | The phenomenon existed at a 5.6% rate before Medicaid disruptions began — roughly 60% of what we see is not coverage-related⁷ | | Invisible dispensing | Could patients be getting medication through channels claims can't see (340B, ADAP)? | Patients whose last fill was at an identified safety-net pharmacy actually had higher retention (73% vs. 65%) — the opposite of what an artifact would show⁸ | | Data quality | Could gaps in the data be creating false positives? | 99.3% of prescription records have valid supply duration; 94% are standard 30-day fills | | Brand-specific effect | Is this a problem with one medication, not the market? | Disengagement rates span only 3.7 percentage points across 11 medications from all three manufacturers when measured over the same time period⁹ | | Diagnosis coding | Are HIV visit codes just automatically carried from old records, not real visits? | 60%+ of these visits were with internists, family doctors, or nurse practitioners — the providers who manage HIV¹⁰ | | Missed medications | Are patients taking HIV drugs we didn't account for? | ~1,300 patients were on unmapped generic formulations — roughly 1% of the disengaged population | | The tightest test | What's the finding when you strip away every possible alternative explanation? | 3,478 patients whose own doctor ordered HIV-specific bloodwork at a recent visit with no treatment filled | ## Limitations 340B in-house dispensing, ADAP direct drug distribution, and cash-pay channels are invisible in commercial claims; an estimated 10–20% of the "Seen Not Treated" population may be receiving ART through these channels. The 340B test used safety-net pharmacy taxonomy as a proxy — a direct overlay of the HRSA 340B registry, Ryan White clinic listings, and contract pharmacy arrangements would further refine this estimate. Observational design — causation cannot be inferred. B20 as a secondary diagnosis may represent carry-forward coding; the same-provider and lab-confirmed subsets control for this. ### Page 4 — Footnotes ⁷ Pre-unwinding: H1 2021→H1 2022. Post-unwinding: H2 2023→H2 2024. 5.6%→9.2% increase implies ~40% attributable to Medicaid unwinding; 60% structural. ⁸ FQHC-taxonomy last-fill patients showed higher retention than standard retail/specialty, directionally inconsistent with 340B artifact. Contract pharmacy arrangements remain a blind spot. ⁹ Duration-matched: patients active H1 2021, retention at H1 2023. Eliminates vintage bias from different approval dates. ¹⁰ Among B20 visits for disengaged patients (Jan 2024–Mar 2025): Internal Medicine 40%, NPs 11%, Family Medicine 10%. Lab/radiology specialties consistent with B20 as secondary diagnosis.