Evidence · idea-011

AliveCor's KardiaK Software Platform, an ECG-based deep-learning hyperkalemia-screening algorithm, received FDA Breakthrough Device Designation in September 2018.

As of the Generator's search, no FDA clearance (510(k)) or De Novo grant for AliveCor's KardiaK hyperkalemia algorithm has been identified — i.e. Breakthrough designation alone has not, as far as the Generator can tell, produced a cleared predicate in this space.

A Mayo Clinic-affiliated deep-learning model trained on paired ECG and serum-potassium data reported an AUC of approximately 0.88-0.89 for hyperkalemia screening from a standard ECG.

Between 20% and 67% of patients with heart failure also have chronic kidney disease, across the range of studies reviewed.

In a multi-country comparison, RAASi prescription prevalence among CKD/HF patients in the United States was reported around 52%, lower than in Germany or France.

The US Renal Data System (USRDS) 2024 Annual Data Report provides Medicare Fee-for-Service CKD burden estimates using a 5% random sample for calendar year 2022; the Generator has not extracted a specific beneficiary count from it.

No HCPCS code, payment rate, or coverage policy specific to a continuous ECG-based hyperkalemia-screening service has been identified by the Generator; existing CMS remote physiologic monitoring codes may be an analog billing structure, but the Generator has not confirmed applicability and is not reproducing any AMA-licensed code descriptor here.

Assumption, not a fact: clinicians managing RAASi titration would be willing to act on an ECG-based hyperkalemia flag without an accompanying lab value, and would order a confirmatory draw promptly when flagged. This is a stated design assumption behind the device's clinical workflow, not a claim any connector in this repo can check.

FDA's Breakthrough Devices Program page states that FDA generally cannot publicly disclose whether a sponsor has submitted a Breakthrough Device designation request, or whether FDA granted or denied it, unless the sponsor discloses it; FDA's own public list of Breakthrough-designated devices includes only devices that have since received marketing authorization. Since no 510(k) or De Novo grant for AliveCor's KardiaK has been located (claim 2), KardiaK's designation status is structurally unconfirmable against any FDA registry at this time, independent of whether the designation itself is real.

No FDA De Novo bulk-download file (analogous to the 510(k) pmn96cur.zip or PMA pma.zip files) exists at accessdata.fda.gov's premarket ftparea; a request for denovo.zip at that location returns HTTP 404, and no bulk-export function was located on the De Novo database search page. This is a source-availability finding distinct from 'not yet fetched': no known stable URL serves this file for engine/scripts/fetch_data_file.py to retrieve.

A real, live CMS URL for the January 2026 HCPCS Level II Alpha-Numeric file exists (confirmed via engine/scripts/fetch_data_file.py --dry-run: HTTP 200, 2,519,574 bytes), but it is a ZIP archive and engine/scripts/fetch_data_file.py validates and writes only CSV/TSV/JSON bodies, so it cannot currently be loaded into data/hcpcs_level_ii.csv without extending the fetch tool to unzip. No specific HCPCS code for remote ECG-based or physiologic monitoring is confirmed by this pass; secondary sources (AAPC, vendor blogs) name candidate code numbers but were not independently verified against this or any primary CMS document.

A Mayo Clinic deep-learning ECG model for hyperkalemia screening (serum K+ greater than or equal to 5.5 mEq/L), trained on 1,576,581 ECGs from 449,380 patients (Mayo Clinic Rochester, 1994-2017) and validated retrospectively on 61,965 patients with stage 3 or greater CKD across three geographically separate Mayo sites, reported AUCs (2-lead, I/II) of 0.883 (95% CI 0.873-0.893) in Minnesota, 0.860 (95% CI 0.837-0.883) in Florida, and 0.853 (95% CI 0.830-0.877) in Arizona -- correcting the prior claim's approximate '0.88-0.89' range, which overstated the upper bound and omitted the two lower-performing sites. At the 90%-sensitivity operating point, specificity was only 54.7%-63.2% across the three sites.

In the same Mayo Clinic study's three validation cohorts, measured hyperkalemia prevalence (the base rate the screening model was validated against) ranged from 2.6% (1,282 of 50,099, Minnesota) to 4.8% (287 of 6,011, Florida) among patients with stage 3 or greater CKD.

The USRDS 2024 Annual Data Report's CKD Reference Tables (B.1-B.6) present point-prevalent (Dec 31) counts of the fee-for-service non-ESRD Medicare population aged 18+, based on the Medicare 5% random sample, with CKD, heart failure, and diabetes each identified from claims in the reported year, and counts multiplied by 20 to represent the full Medicare population meeting the cohort definition -- confirming the methodology cited in the prior claim, though no specific beneficiary count was extracted from this methods document itself (the count lives in a separate reference-table file not retrieved in this pass).

Revealed-behaviour twin for the class-3 question 'would clinicians act on a hyperkalemia flag' (claim 12): in a US outpatient HFrEF registry (CHAMP-HF), over a median 18-month follow-up, RAAS-inhibitor-class drug discontinuation occurred in 12.7% of ACEi/ARB users (444/3,509), 10.4% of ARNI users (140/1,352), and 20.4% of MRA users (435/2,129); chronic kidney disease was the only independent predictor of increased discontinuation risk across all three drug classes. This is a general discontinuation rate, not one isolating hyperkalemia as the stated reason.

In a non-US (Saudi Arabia) retrospective outpatient cohort of 905 RAAS-inhibitor users followed a mean 29.8 months, hyperkalemia (K+ >= 5.1 mmol/L) occurred in 32.8% (295/905), and RAASi therapy was discontinued in 6.2% and down-titrated in 4.5% of patients specifically because of hyperkalemia.

In a non-US (Japan) target-trial-emulation cohort of 2,305 CKD patients who developed incident hyperkalemia (K+ >= 5.5 mEq/L) while on a renin-angiotensin-system inhibitor, 15% (346/2,305) discontinued the drug after the episode; discontinuation was associated with 16% higher all-cause mortality (adjusted HR 1.16, 95% CI 1.02-1.33) than continuation, with no difference in the composite kidney outcome.

In a multinational heart-failure registry across 111 sites in 9 countries (CARE-HK in HF, n=2,558, not US-isolated), 29% of patients had at least one hyperkalemic event over a median 12.3-month follow-up, and RAASi/MRA therapy was down-titrated or discontinued in only 3.6% of cases within 30 days of the first hyperkalemic event.

In a non-US (Manitoba and Ontario, Canada) population-based cohort of CKD patients with RAAS-inhibitor-related hyperkalemia (n=7,200 and n=71,290 respectively), discontinuing the RAAS inhibitor (versus continuing) was associated with higher all-cause mortality (Manitoba HR 1.32, 95% CI 1.22-1.41; Ontario HR 1.47, 95% CI 1.41-1.52) and higher risk of dialysis initiation (Manitoba HR 1.65, 95% CI 1.41-1.85; Ontario HR 1.11, 95% CI 1.08-1.16).

No 510(k)-cleared device anywhere in the openFDA database carries hyperkalemia, electrolyte imbalance, or potassium in its device name (a direct, unrestricted openFDA device/510k query for device_name=hyperkalemia and for unqualified full-text hyperkalemia both return zero results, not merely the single-top-match result the connectors 510k CLI reports); FDA's one classification family built for ECG-based AI 'notification software' (ten product codes under regulation_number 870.2380 in openFDA's device/classification database: SBQ atrial fibrillation risk, SFR hypertension, QXX coronary artery disease, QXO general cardiovascular, SIK cardiac hypertrophy, SIJ valvular regurgitation, SHP cardiac amyloidosis, SII aortic stenosis, QYE reduced ejection fraction, SAT pulmonary hypertension) contains none for hyperkalemia, electrolyte, or potassium. This narrows the regulatory-pathway question specifically to De Novo, with no existing predicate product-code family for a follow-on 510(k) to piggyback against, whereas an analogous condition (reduced ejection fraction) already has five 510(k)s (K233409, K232699, K250119, K250649, K250652) clearing under one De Novo-established code (QYE).

In the US Get With The Guidelines-Heart Failure (GWTG-HF) registry, among 365,494 heart-failure hospitalizations at 418 sites (2014-2019), 234,332 (64%) had discharge eGFR less than 60 mL/min/1.73m2 and 18,869 (5%) were on dialysis; among the 157,439 patients with HFrEF (EF<=40%), use of triple guideline-directed therapy (ACEi/ARB/ARNI + beta-blocker + MRA) fell from 38% at eGFR>=90 to 3% among those on dialysis.

In the DISCOVER CKD multinational prospective cohort (NCT04034992, n=1,052 adults with CKD enrolled September 2019-June 2022), 61.1% (643 patients) were receiving guideline-directed medical therapy (a renin-angiotensin system inhibitor and/or an SGLT2 inhibitor) at baseline, ranging from 42.7% in the USA, the lowest of the countries reported, to 82.2% in Spain and Sweden, the highest.

In a US Medicare fee-for-service claims cohort of 621,171 patients with heart failure with preserved ejection fraction identified 2013-2018 via a validated claims-based phenotyping model, spironolactone (a mineralocorticoid receptor antagonist) initiation rate rose from 16.8 to 19.9 per 100 person-years, while hospitalization for hyperkalemia within 6 months of initiation remained stable at 12.0 to 10.6 per 1,000 person-years among initiators over the same period.

In the DISCOVER CKD retrospective cohort, hyperkalemia prevalence among CKD patients was 20.9% (37,713 of 180,416) in the UK Clinical Practice Research Datalink (2008-2019) versus 7.4% (5,924 of 80,196) in the Japan Medical Data Vision database (2008-2017); following a hyperkalemia episode, RAASi therapy was discontinued during follow-up in 50.6% of UK patients versus 29.2% of Japanese patients, and a potassium binder was initiated within 3 months of the episode in fewer than 0.2% of UK patients versus 18.7% of Japanese patients.