JMIR Form Res. 2026 Aug 18;10:e97202. doi: 10.2196/97202.
ABSTRACT
BACKGROUND: Arteriovenous fistula (AVF) stenosis is an important complication of hemodialysis that may impair dialysis adequacy and patient outcomes. Conventional AVF monitoring relies largely on subjective physical assessment and may delay recognition of abnormal vascular access findings.
OBJECTIVE: This prospective pilot study aimed to assess the feasibility of serial electronic stethoscope monitoring for detecting abnormal AVF sounds before routine nursing assessment, and explore factors associated with a high-risk AVF phenotype, defined as percutaneous transluminal angioplasty within the preceding 3 years.
METHODS: We conducted a 3-month prospective study at the National Taiwan University Hospital Yunlin Branch Hemodialysis Center from March to June 2023. Thirty adults receiving maintenance hemodialysis through a functional AVF were enrolled; 29 completed follow-up and generated 1462 acoustic recordings. AVF sounds were recorded 2-6 times weekly at 2 standardized sites: approximately 3 cm proximal to the anastomosis and the arterial needle insertion site. Signals were analyzed using short-time Fourier transform and expressed as decibels relative to full scale. Clinical, demographic, laboratory, and dialysis-related data were collected. Because recordings were clustered within patients, risk-factor analyses were performed at the patient level. Associations with the prespecified high-risk AVF phenotype were explored using Fisher exact tests and Firth penalized logistic regression because only 9 patients met the high-risk definition.
RESULTS: An exploratory in-sample candidate threshold for abnormal AVF sounds was derived at <-28.59 decibels relative to full scale using a 3SD rule. Because the threshold was developed, refined, and evaluated within the same small dataset, it requires independent external validation. One participant had stenosis confirmed by duplex Doppler before acoustic monitoring began. In this illustrative case, the acoustic signal remained abnormal and was recognized approximately 1 month before routine bedside nursing assessment, although not before imaging. All stenotic recordings originated from this single participant and therefore do not constitute independent evidence of diagnostic performance. In exploratory patient-level analyses, the high-risk phenotype was associated with diabetes mellitus (odds ratio [OR] 12.6, 95% CI 2.2-138.1; P=.005), hypertension (OR 12.9, 95% CI 1.3-31.7; P=.033), and BMI ≤25 kg/m² (OR 19.0, 95% CI 2.0-2568; P=.011). Fresenius FX-series dialyzer use was not significant on exact testing (OR 4.4, 95% CI 0.9-28.5; P=.11). Estimates were imprecise because of the limited number of high-risk patients.
CONCLUSIONS: Serial electronic stethoscope monitoring was feasible. In one illustrative patient with Doppler-confirmed stenosis, persistent acoustic abnormalities were recognized before routine nursing assessment but not before imaging. The candidate acoustic threshold and exploratory correlates of high-risk AVF status require confirmation in larger, externally validated prospective cohorts before clinical implementation.
PMID:42612202 | DOI:10.2196/97202