COST OF FALLS AVERTED AND BURDEN PREVENTED BY CONVERSATIONAL VOICE AI FALL-RISK SCREENING IN OLDER ADULTS: A COST-CONSEQUENCE ANALYSIS
Author(s)
Martha Buckley, MD1, Avik Ray, MD, MS2, Jonathan Agnew, PhD, MBA2, Herprit Mahal, MD, FACP2, Meenesh Bhimani, MD, MHA2.
1OhioHealth, Columbus, OH, USA, 2Hippocratic AI, Menlo Park, CA, USA.
1OhioHealth, Columbus, OH, USA, 2Hippocratic AI, Menlo Park, CA, USA.
OBJECTIVES: Falls are a leading cause of morbidity and mortality in the US, yet Medicare Annual Wellness Visit (AWV) screening relies on self-reported fall history that underestimates the actual incidence. From a US payer perspective, we assessed the estimated medical cost burden prevented by conversational voice AI fall-risk screening.
METHODS: We analyzed a cross-sectional AI telephone health-risk assessment in a US Midwestern Health System’s (OhioHealth, Columbus, OH, USA) Medicare AWV program (511 older adults; mean call 16.0 minutes). The voice AI agent assessed seven fall-risk indicators (e.g., needing hands to rise, unsteady gait, fear of falling); high burden was two or more indicators. AI cost was $2.40 per assessment ($9/hr). Case-finding outcomes used observed data. A population model layered general older-adult fall incidence (27%; CDC/Bergen 2016) and an assumed intervention uptake (50%) with exercise efficacy (23% fall-rate reduction; Sherrington 2019) and inflation-adjusted per-fall cost (~$12,500; Burns 2016). A probabilistic sensitivity analysis (1,000 iterations) varied prevalence, incidence, uptake, efficacy (Beta), and per-fall cost (lognormal), with 95% credible intervals (CrI).
RESULTS: Among prior-year non-fallers (n=378), 32.3% had high fall-risk burden (adjusted OR 8.65, 95% CI 5.26-14.21); number needed to screen 3.1, at $7.43 per hidden case. Scaled to 10,000 beneficiaries, screening ($24,000/year) identified 2,390 hidden cases and averted 74 falls, preventing ~$928,000 in medical cost (95% CrI $282,000-$1,944,000), net $904,000, benefit-cost 39:1 (95%CrI 12-81:1).
CONCLUSIONS: Conversational voice AI identified occult high fall-risk in one in three older adults who denied falling, at ~$7 per case. In this population model, it remained cost-saving in every probabilistic iteration, preventing ~$0.9 million per 10,000 beneficiaries. Voice AI-enabled telephone screening is a low-cost, scalable mechanism for early fall-risk detection in AWV workflows. Prospective validation linking AI-derived indicators to incident falls is warranted.
METHODS: We analyzed a cross-sectional AI telephone health-risk assessment in a US Midwestern Health System’s (OhioHealth, Columbus, OH, USA) Medicare AWV program (511 older adults; mean call 16.0 minutes). The voice AI agent assessed seven fall-risk indicators (e.g., needing hands to rise, unsteady gait, fear of falling); high burden was two or more indicators. AI cost was $2.40 per assessment ($9/hr). Case-finding outcomes used observed data. A population model layered general older-adult fall incidence (27%; CDC/Bergen 2016) and an assumed intervention uptake (50%) with exercise efficacy (23% fall-rate reduction; Sherrington 2019) and inflation-adjusted per-fall cost (~$12,500; Burns 2016). A probabilistic sensitivity analysis (1,000 iterations) varied prevalence, incidence, uptake, efficacy (Beta), and per-fall cost (lognormal), with 95% credible intervals (CrI).
RESULTS: Among prior-year non-fallers (n=378), 32.3% had high fall-risk burden (adjusted OR 8.65, 95% CI 5.26-14.21); number needed to screen 3.1, at $7.43 per hidden case. Scaled to 10,000 beneficiaries, screening ($24,000/year) identified 2,390 hidden cases and averted 74 falls, preventing ~$928,000 in medical cost (95% CrI $282,000-$1,944,000), net $904,000, benefit-cost 39:1 (95%CrI 12-81:1).
CONCLUSIONS: Conversational voice AI identified occult high fall-risk in one in three older adults who denied falling, at ~$7 per case. In this population model, it remained cost-saving in every probabilistic iteration, preventing ~$0.9 million per 10,000 beneficiaries. Voice AI-enabled telephone screening is a low-cost, scalable mechanism for early fall-risk detection in AWV workflows. Prospective validation linking AI-derived indicators to incident falls is warranted.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EE33
Topic
Economic Evaluation, Epidemiology & Public Health, Health Technology Assessment
Disease
Geriatrics