Scope note: The AI & digital-health interventions shown here are illustrative “low-hanging fruit” — a small, high-feasibility/high-impact sample chosen to demonstrate planning-grade ROI. They are not an exhaustive or prescriptive list; many other interventions may apply.
DALY Standard: Michigan observed LE · planning-grade estimate · Mental health #1
Soshnikov · Isabella County · AI & Digital Health Applications · 2026 Primary burden: 15,430 DALYs (frontier LE) · Sensitivity (MI LE): 10,579 DALYs/yr
Supplement 2 - Technology & Innovation · v5.0 (2026 Data Update)

Artificial Intelligence and Digital Health Technologies for Community Disease Burden Reduction in Isabella County, Michigan

A problem-solution framework aligned with local burden estimates, 2026
Sergey Soshnikov
Companion document to: Disease Burden Analysis, Isabella County, Michigan (v5.0) · June 2026 · Data: CDC PLACES 2024 · CDC WONDER 2020–22 · For community planning and grant development

Summary

This supplement maps the disease burden priorities identified in the Isabella County health analysis (15,430 primary DALYs/year, 95% UI 14,459–16,481; frontier LE + MH remission-adj; CDC PLACES 2024, CDC WONDER 2020–22, Monte Carlo n=10,000) to artificial intelligence and digital health interventions supported by current evidence. The county's leading burdens are cancer (#1, 4,057 DALYs), cardiovascular disease (#2, 3,302), substance use disorders (#3, 2,904), and mental health (#4, 1,733 DALYs remission-adj). Mental health access (HRSA HPSA designation, 176,938:1 psychiatrist ratio) and structural poverty as the root driver - are areas where digital tools may improve reach, but the strength of evidence varies substantially by intervention. For each priority condition, we describe the intervention, the supporting evidence, the implementation pathway for a county of 64,565 residents, and the realistic expectation for burden reduction. Several burden inputs rely on modeled prevalence or rural-adjusted mortality proxies, so all estimates should be treated as planning-grade rather than validated county burden measurements.

Why AI matters specifically for Isabella County

The county's defining constraint is not the absence of plausible interventions - tools exist for every priority condition, but the strength of evidence varies by condition and delivery model. The constraint is workforce and access: one psychiatrist per 176,938 residents, a Primary Care HPSA designation, and 19% of residents below the poverty line. AI does not replace clinicians - it can help prioritize caseloads, speed triage, and extend monitoring in primary care and community settings where specialists are absent.

1. AI Applications by Condition Priority

The following cards are ordered by primary estimated DALY burden (CDC PLACES 2024 · CDC WONDER 2020–22 · ordered by primary standard (frontier LE + remission-adj) · Cancer #1). Monte Carlo simulation (n=10,000) propagates uncertainty from mortality (Poisson), prevalence (beta), and disability weights (±15–25%): 95% UI 14,459–16,481. Total primary burden: 15,430 DALYs. For each condition, interventions are rated by evidence strength: Strong = RCT/meta-analysis evidence; Moderate = mixed, indirect, or implementation evidence; Emerging = promising early-stage.

🧠

Mental Health Disorders (F30–F48)

HPSA designation · 176,938:1 psychiatrist ratio · 29.8% adult prevalence (CDC PLACES 2024) · #4 burden (remission-adj)

1,733
DALYs/yr · #4
Telepsychiatry
AI-augmented collaborative care model

Primary-care collaborative care pairs a behavioral health care manager with psychiatric consultation and can be augmented with AI for caseload prioritization, outcome tracking, and between-visit check-ins. Recent Cochrane evidence for severe mental illness is cautious rather than strongly positive, so this option should be treated as a workforce-access model requiring local evaluation, not as a guaranteed high-ROI intervention.

Moderate evidence
Digital Therapeutics
FDA-cleared digital CBT tools

Prescription digital therapeutics (e.g., Freespira for PTSD/anxiety, Rejoyn for MDD - FDA Breakthrough Device) deliver CBT protocols via smartphone app with clinician dashboard. Effective as standalone or adjunct; no wait time; available 24/7. Particularly relevant for the CMU student population and working adults who cannot access in-person care.

Strong evidence
Early Detection
Passive sensing for depression onset

Smartphone passive data (GPS patterns, screen time, sleep via accelerometer) correlates with depression severity (AUC 0.87; Saeb et al., JMIR Mental Health, 2015). Preliminary evidence suggests early detection of symptom changes before clinical presentation, though replication in larger rural populations is needed. Deployable as app; no hardware needed.

Moderate evidence
Crisis Prevention
NLP-based suicide risk flagging in EHR notes

ML models trained on clinical notes identify suicidal ideation language not captured by structured risk assessments (PHQ-9 item 9). VA and safety-net hospitals report 20–40% improvement in crisis identification when layered onto existing EHR workflows - critical given the county's 29.8% MH prevalence and documented suicide burden.

Moderate evidence
🔬

Cancer - All Sites (C00–C97)

#1 DALY condition · 130 deaths/yr at age 67 · Early detection gap in rural settings

4,057
DALYs/yr · #1
AI Screening
AI-assisted radiology for lung & breast cancer

FDA-cleared AI tools (e.g., Seno.ai, Veye Chest) flag suspicious nodules in CT/mammography scans, increasing radiologist throughput 2–3×. Particularly valuable when Isabella County residents travel to distant facilities - AI pre-reads can prioritize urgent recalls.

Strong evidence
Risk Stratification
Population-level cancer risk scoring

Machine learning models using EHR data (smoking history, BMI, age, comorbidities) identify high-risk patients for proactive screening outreach. Implementable within CMU Health or McLaren-affiliated primary care EMR systems without additional infrastructure.

Moderate evidence
Navigation
AI care navigation for cancer patients

Automated follow-up reminders, appointment navigation, and symptom monitoring between visits reduce treatment abandonment - a significant issue for rural patients managing long distances to oncology centers.

Moderate evidence
Tobacco Cessation
Conversational AI for smoking cessation

Text-based AI interventions (e.g., NCI SmokefreeTXT, Quit Genius) achieve 6-month cessation rates of 15–20% at near-zero marginal cost. Directly addresses the county's elevated cancer and COPD burden driven by tobacco use (18.6% adult smoking rate, CDC PLACES 2024).

Strong evidence
💊

Substance Use Disorders (F10–F19)

#3 by DALYs · 37 deaths/yr at age ~44 · NAS rate 583/100k births · Rural adjustment ×1.85

2,904
DALYs/yr · #3
Overdose Prevention
Predictive overdose risk scoring in EHR

ML models using prescription history, ED visits, and social determinants identify patients at highest 30-day overdose risk. PDMP-integrated tools (e.g., Colorado's overdose risk score) have shown 20–30% reduction in opioid-related ED visits when embedded in primary care workflow.

Moderate evidence
Treatment Access
AI-assisted telehealth MAT prescribing

Platforms like Bicycle Health and Workit Health provide AI-supported buprenorphine initiation via telehealth, substantially extending the reach of each waivered prescriber in rural areas (Huskamp et al., Health Affairs, 2022; SAMHSA data show telehealth MAT doubled patient access in rural counties 2020–2022). Critical for Isabella County where MAT provider access is extremely limited.

Strong evidence
Prenatal
Maternal SUD screening & support AI

Automated SBIRT (Screening, Brief Intervention, Referral to Treatment) integrated into prenatal visits uses validated questionnaires with AI-driven follow-up. Directly targets NAS rate of 583/100k live births and maternal tobacco use of 22%.

Strong evidence
Community Outreach
Social media signal monitoring for outreach

NLP-based tools monitor local social media for substance use distress signals to direct community health worker outreach in real time. Early-stage but deployable at low cost; relevant in a county with high social media engagement among the 18–30 population.

Emerging
❤️

Cardiovascular Disease & Stroke (I20–I69)

Stroke mortality 49% above state avg · 32.1% hypertension (CDC PLACES 2024) · CVD #4, Stroke #6

4,375
DALYs/yr · CVD+Stroke · #2–6
Hypertension
AI-supported remote BP monitoring program

Home BP cuffs transmit readings to an AI platform that automatically adjusts medication titration recommendations within a protocol, alerts care managers to uncontrolled values, and flags stroke risk patterns. The Million Hearts RTIPs database lists this as a Tier 1 evidence-based intervention. Achieves 15–20 mmHg systolic reduction in rural populations at $30–50/patient/year.

Strong evidence
Arrhythmia
Wearable AFib detection (Apple Watch / AliveCor)

FDA-cleared wearable ECG devices detect atrial fibrillation with high sensitivity: AliveCor KardiaMobile showed 98% sensitivity and 76% specificity for AF (Bumgarner et al., JACC, 2018); Apple Watch Series 4 showed PPV of 84% in a large pragmatic study (Turakhia et al., NEJM, 2019). Community screening programs have detected AFib in 3–5% of adults over 65, enabling anticoagulation that significantly reduces stroke risk.

Strong evidence
Risk Prediction
10-year CVD risk AI models in primary care

Deep learning models using retinal photos (Google Research / Verily) can estimate cardiovascular risk, age, and blood pressure from a fundus image - no blood draw required. Deployable in ophthalmology or optometry offices that already serve rural populations, enabling CVD risk stratification without a cardiology referral.

Emerging
Stroke Response
AI-assisted stroke triage & telestroke

Telestroke networks with AI-assisted CT image reading (e.g., RapidAI) enable rural ED physicians to receive real-time neurologist guidance and automated ischemic core/penumbra analysis within minutes. Reduces door-to-needle time by 30–40 minutes in rural hospitals, directly addressing the county's elevated stroke mortality.

Strong evidence
🏘️

Social Determinants of Health - Root Cause Interventions

Poverty 19% · Food insecurity 16.4% · Obesity 43.1% (CDC PLACES 2024) · Structural drivers of all conditions above

All conditions
Root driver
Food & Housing
AI-powered SDOH screening in clinical settings

Tools like Unite Us, Findhelp.org (formerly Aunt Bertha), and NowPow use NLP and mapping to connect patients with community resources at the point of care. EHR-integrated SDOH screening with closed-loop referral tracking shows 25–35% increase in social needs resolution compared to paper screening alone.

Moderate evidence
Predictive
Predictive models for high-risk utilization

ML models combining EHR, claims, and SDOH data identify patients likely to have preventable ED visits or hospitalizations within 90 days. Community health workers are then deployed proactively. Programs in similar rural Michigan counties have reduced preventable hospitalization rates by 15–25%.

Moderate evidence
Equity Analytics
Geographic AI mapping of health gaps

CDC PLACES + Census + EHR data can be combined in ArcGIS or open-source platforms to map health burden at the census-tract level, identifying micro-geographies within the county where cancer screening uptake, hypertension control, or MAT access is lowest - enabling precise resource targeting.

Strong evidence
Workforce
AI-assisted community health worker (CHW) optimization

Decision-support tools help CHWs prioritize visits, track outcomes, and document interventions - multiplying the impact of each CHW by 30–40%. Given the county's primary care and mental health workforce gaps, CHW programs augmented by AI represent the most scalable short-term access solution.

Moderate evidence

2. Phased Implementation Roadmap

Given the county's resource constraints, the following phased approach prioritizes high-impact, low-infrastructure interventions first, building toward a county-wide digital health infrastructure over 24 months.

PhaseInterventionCondition targetLead / partnerEst. costDALY impact
Now (0–3 mo) Collaborative care model + telepsychiatry (existing CMU Health or FQHC infrastructure) Mental health (#4), SUD (#3) CMU Health / Isabella Community Mental Health ~$50K setup High
Now (0–3 mo) AI-assisted smoking cessation via text (SmokefreeMI partnership) Cancer (#1), COPD, CVD County Health Dept / MDHHS <$10K/yr Moderate
Now (0–3 mo) SDOH screening + community resource navigation (Findhelp.org) All conditions Primary care practices / FQHC Free–$15K/yr Moderate
Year 1 Telehealth MAT program (buprenorphine via telehealth for OUD) SUD (#3), NAS Bicycle Health / local OB-GYN $0 (insurance reimbursed) High
Year 1 Remote BP monitoring program (community health worker–led, AI titration) CVD, stroke CMU Health / Isabella County CHW program ~$80K (devices + platform) High
Year 1 Predictive cancer risk screening embedded in primary care EMR Cancer (#1) CMU / McLaren / MiHIA ~$20K integration Moderate
Year 2 Wearable AFib detection community program (pharmacies, senior centers) CVD, stroke Local pharmacies / AHA Michigan ~$30K (devices) Moderate
Year 2 AI-powered census-tract health gap mapping for resource targeting All conditions CMU School of Public Health / County Health ~$15K (analyst + tools) Infrastructure
Year 2 Prescription digital therapeutics for depression/anxiety (Rejoyn) Mental health (#4) CMU Health / Medicaid coverage Medicaid billable Moderate

3.1 Grant development: AI + rural health equity

Federal funding for AI in rural health has expanded significantly. HRSA's Rural Health Care Program, the NIH Office of Rural Health, and AHRQ's Patient Safety Learning Lab all fund projects combining AI with rural access challenges. Isabella County's HPSA designations, documented burden estimates (this analysis), and CMU's research infrastructure make it an unusually strong grant application site. A funded project developing or evaluating an AI tool in one of the priority areas above would simultaneously generate revenue, peer-reviewed publications, and community impact.

3.2 Consulting to MDHHS and county health departments

Michigan's AI Task Force (MDHHS HIT Commission, May 2026) is actively seeking guidance on AI implementation in public health settings. A consultant who can translate burden data into AI program specifications - and who has both clinical and epidemiological credentials - fills a gap that neither pure technologists nor pure clinicians can fill. The disease burden analyses in this project series are the foundation of that consultation service.

3.3 CMU faculty position as a force multiplier

Embedding AI public health methods into the CMU MPH curriculum creates a pipeline of trained students who can implement these tools in Michigan county health departments. This simultaneously builds regional capacity, generates collaborative research opportunities, and positions CMU as the state's center for AI-augmented public health practice.

Next step: The most immediate actionable opportunity is to contact U-M e-HAIL (2026 AI & Health Symposium, September 11) and Michigan Public Health Institute (MPHI) to discuss this analysis as preliminary data for a joint grant application targeting AI-assisted mental health access in rural Michigan. With cancer now documented as the #1 DALY burden (4,057 DALYs/yr) and mental health access remaining a critical structural gap (176,938:1 psychiatrist ratio, #4 burden at 1,733 DALYs remission-adj), this analysis provides compelling quantitative justification for the Significance section of an NIH R21 or HRSA Rural Health grant.

4. Selected References

  1. Doraiswamy PM, et al. Artificial intelligence and the future of psychiatry. World Psychiatry. 2020;19(2):172–173.
  2. Holmgren AJ, Apathy NC, Adler-Milstein J. Barriers to hospital electronic public health reporting and implications for the COVID-19 pandemic. J Am Med Inform Assoc. 2020;27(8):1306–1309.
  3. Unützer J, et al. Collaborative care management of late-life depression in the primary care setting. JAMA. 2002;288(22):2836–2845. Strong
  4. Bhatt DL, et al. Digital health care - a path forward. NEJM. 2020;382:1668–1672.
  5. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine. 2019;25:44–56.
  6. HRSA. Health Professional Shortage Areas (HPSAs). data.hrsa.gov
  7. Michigan MDHHS. Health Information Technology Commission: AI Task Force. May 2026.
  8. e-HAIL: eHealth & AI at University of Michigan. 2026 AI & Health Symposium, September 11. e-hail.umich.edu
  9. FDA. Artificial Intelligence and Machine Learning in Software as a Medical Device. 2021. fda.gov
  10. Mehrotra A, et al. Rapid growth in mental health telemedicine use among rural Medicare beneficiaries. Health Aff. 2017;36(5):909–917.
  11. CDC PLACES: Local Data for Better Health, County Data 2024 release. cdc.gov/places
  12. CDC WONDER. Underlying Cause of Death, Michigan 2020–2022. wonder.cdc.gov
Document status: Companion supplement to Isabella County Disease Burden Analysis v5.0. Not peer-reviewed. For community planning, grant development, and conference presentation.
Data sources: CDC PLACES 2024 · CDC WONDER 2020–22 · IHME GBD 2021 · MDHHS 2024 · ACS 2022
Author: Sergey Soshnikov, MD PhD · Public Health Researcher · Central Michigan University
© 2026 Soshnikov