Healthcare Performance Metrics

Most healthcare providers say they care about outcomes. But their systems are still designed around visits, not the patient journey. Real value-based care starts with this mindset shift: You’re not treating a visit. You’re managing a care journey. So, what should providers actually do to make that real? 1. Map the care journey Start with key cohorts—e.g., diabetic seniors or post-acute care patients. Ask: What does a good 6-month journey look like? Then map it backwards. What data, interventions, and check-ins are needed? 2. Expand the data lens Clinical data is just 50% of the story. You need SDOH (housing, food, income), behavior (adherence, mood), and context (caregivers, home support). 3. Stratify risk proactively Don’t wait for ER visits. Build simple models that combine clinical risk + social risk. Then segment patients into high, rising, and stable risk groups. Use AI to predict who's likely to fall through the cracks. 4. Close the loop with AI AI should surface next-best-actions: Who needs a nudge today? What’s changing in their baseline? What care gaps are widening? Think of AI not as a tool, but as a teammate, watching the journey 24/7. 5. Build a longitudinal feedback loop If you don’t measure outcomes across time, you’re blind. Use dashboards that show: Outcome trends per patient cohort, ROI on interventions, Impact of addressing SDOH. At Inferenz, our mission is clear: Help providers operationalize the care journey using data, AI, and human-centered design. Because value-based care isn’t a future model: it’s an execution challenge. And we’re building the rails to make it real.

I track 15+ biomarkers quarterly. But only 3 predict my cognitive health 20 years from now. After reviewing the longevity research and tracking my own data for 5 years, I've learned most biomarkers are noise. These 3 are signal. Peter Attia is right: VO2 max is the strongest predictor of lifespan. But for brain health specifically, these matter more: 1. ApoB (not LDL) Target: Under 60 mg/dL Why it matters: ↳ Vascular dementia accounts for 40% of all dementia ↳ ApoB measures actual atherosclerotic particles ↳ Predicts cognitive decline 15-20 years before symptoms What most docs miss: You can have "normal" LDL and terrible ApoB. LDL is cholesterol content. ApoB is particle number. Particle number drives plaque formation. Plaque drives vascular dementia. How to improve it: ↳ Prioritize fiber (40g daily minimum) ↳ Limit saturated fat ↳ Consider ezetimibe or PCSK9 inhibitors if lifestyle isn't enough 2. Fasting insulin (not fasting glucose) Target: Under 5 uIU/mL Why it matters: ↳ High insulin precedes high glucose by 10+ years ↳ Insulin resistance doubles Alzheimer's risk ↳ Inflammation from insulin resistance damages neurons What most docs miss: Fasting glucose stays normal until insulin can't compensate anymore. By then, you've had insulin resistance for a decade. How to improve it: ↳ Zone 2 cardio (4-5 hours weekly) ↳ Strength training (maintain muscle glucose disposal) ↳ Time-restricted eating (12-14 hour fast minimum) 3. Sleep efficiency (not sleep duration) Target: Above 85% Why it matters: ↳ Brain clears amyloid and tau during deep sleep ↳ Poor sleep quality doubles dementia risk ↳ Sleep fragmentation prevents memory consolidation What most docs miss: 7 hours of broken sleep is not the same as 7 hours of consolidated sleep. You need time in deep and REM stages. How to improve it: ↳ Same bedtime/wake time (even weekends) ↳ No alcohol (destroys REM sleep) ↳ Screen sleep apnea if efficiency stays low My personal data over 5 years: ApoB dropped from >100 to <60 (added rosuvastatin + ezetimibe) Fasting insulin dropped from 8 to 4 (strength training + walking) Sleep efficiency went from <80% to >90% Cognitive testing: Improved across all domains. What I don't track anymore: Total cholesterol (not actionable) BMI (muscle mass matters more) Generic "inflammation markers" Many supplement levels The longevity doctors are right about one thing: What gets measured gets managed. But measuring everything is paralysis. These 3 biomarkers give you the most actionable data for preventing cognitive decline. Start here. ⁉️ What biomarkers do you track for longevity? ♻️ Repost if you believe in data-driven prevention 👉 Follow me (Reza Hosseini Ghomi, MD, MSE) for evidence-based longevity strategies

Can blood tests predict the transition from subjective cognitive decline (SCD) to true dementia? 🩸 A recently published (https://lnkd.in/g2myuf_M) prospective cohort study (SCIENCe) followed 298 individuals with SCD who were evaluated at a memory clinic. Notably, 73.2% of the participants were Amyloid negative (tested with Amyloid-PET or CSF), representing an early, high-risk group. The cohort (mean age 61.55 years, 41.6% female) was tracked for an average of 4.8 years. Researchers measured the rate of change (slope) in four blood-based Alzheimer disease (AD) biomarkers, specifically focusing on phosphorylated tau 217 (pTau217), which reflects tau pathology, and glial fibrillary acidic protein (GFAP), which reflects reactive astrogliosis. Cox proportional hazards models were used to calculate the risk (hazard ratio, HR) of progressing to mild cognitive impairment (MCI) or dementia, and the model's prognostic accuracy was measured by the concordance index, or C-index. The central finding was that the rate of change in biomarkers provided strong, independent prognostic value beyond just the measurement taken at the start of the study. A steeper pTau217 slope was strongly associated with progression risk (HR, 3.61; P<0.001). Crucially, adding this slope improved the predictive accuracy of the model from a C-index of 0.86 to 0.89. Similarly, the GFAP slope was associated with progression (HR, 1.51; P=0.04) and improved accuracy from a C-index of 0.77 to 0.81. Both steeper pTau217 and GFAP slopes were associated with a decline in cognitive function across all domains tested. The Aβ42/40 ratio slope, in contrast, was not associated with progression risk (P=0.43), indicating that longitudinal measurement of this specific biomarker did not improve prognosis. The optimal prognostic model, combining baseline Aβ42/40 and the baseline and slope values for pTau217 and GFAP, achieved an overall accuracy of 0.90. The study also found that approximately 1 in 5 initially biomarker-negative participants converted to a positive status during follow-up. A limitation of the study is that the number of participants who progressed to MCI or dementia was modest (33 participants, 11.1%). Also, the participants were recruited exclusively from a memory clinic, which may introduce selection bias and limit how broadly the findings can be applied. The results suggest that monitoring the longitudinal changes in plasma pTau217 and GFAP may be a promising, non-invasive strategy for identifying and tracking AD pathology in high-risk SCD patients. These biomarkers could hold potential utility for future participant selection and outcome monitoring in preclinical AD intervention trials, but further longitudinal studies across different settings are needed to confirm their reliability for individual patient prediction. #neurology #dementia #alzheimers #prognosis #BloodTest

The Risk Factors for Dementia We Often Miss We talk a lot about memory decline, but we rarely talk about the upstream systems that drive it decades earlier. Most of us still think dementia is a late-life brain problem. In practice, many risk factors begin silently in midlife across metabolic, vascular, immune and lifestyle pathways. Dementia is not one disease but a convergence of metabolic inflammation, vascular dysfunction, lifestyle habits and genetic susceptibility. The good news is that many pathways are modifiable. 1/ NAFLD (Fatty Liver Disease) ↳ Chronic hepatic inflammation accelerates neuroinflammation ↳ Strong association with insulin resistance and cognitive decline Solutions: ↳ Weight reduction of 5–10% ↳ Mediterranean dietary pattern 2/ Gut Microbiome ↳ Dysbiosis increases peripheral inflammation ↳ Alters microglial activation and blood–brain barrier integrity Solutions: ↳ Prebiotic fibre daily ↳ Fermented foods 3 to 5 times weekly 3/ Lipids and Lipoproteins ↳ Elevated LDL and ApoB increase vascular injury ↳ Dyslipidaemia worsens small-vessel brain disease Solutions: ↳ LDL-lowering to guideline targets ↳ Replace saturated fats with unsaturated fats 4/ Hypertension ↳ Chronic high BP damages cerebral perfusion ↳ Strong determinant of vascular dementia Solutions: ↳ Tight BP control ↳ Daily aerobic activity 5/ Genetics ↳ APOE-ε4 increases amyloid accumulation ↳ Interaction with lifestyle amplifies risk Solutions: ↳ Precision lifestyle approach ↳ Aggressive management of all modifiable risks 6/ Diabetes ↳ Insulin resistance impairs neuronal glucose uptake ↳ Higher risk of both vascular and Alzheimer’s dementia Solutions: ↳ HbA1c optimisation ↳ High-intensity lifestyle change 7/ Diet Quality ↳ Ultra-processed foods increase inflammation ↳ Deficiency in omega-3 and polyphenols reduces neuroprotection Solutions: ↳ Mediterranean or MIND diet ↳ Reduce ultra-processed food burden 8/ Physical Inactivity ↳ Reduces BDNF and neurogenesis ↳ Increases vascular and metabolic risk Solutions: ↳ 150 minutes weekly exercise ↳ Strength training twice weekly 9/ Body Mass Index ↳ Midlife obesity predicts late-life cognitive decline ↳ Visceral fat increases systemic inflammation Solutions: ↳ Target 5–10% weight loss ↳ Strength and aerobic training 10/ Smoking ↳ Accelerates oxidative stress ↳ Worsens cerebrovascular injury Solutions: ↳ Smoking cessation support ↳ Nicotine replacement or pharmacotherapy Most dementia pathways begin decades before symptoms. If we intervene early across metabolic, vascular, and lifestyle domains, we can meaningfully shift long-term cognitive outcomes. Cognitive decline is not inevitable. The earlier we intervene, the more we protect the ageing brain. #LifestyleMedicine #BrainHealth #LongevityMedicine #MetabolicHealth #PreventiveMedicine

New paper led by Sadiq Y. Patel: Many state Medicaid officials and Medicaid health plans are concerned with ED visits and hospitalizations for ambulatory care-sensitive conditions (ACSCs, e.g., "avoidable" visits). We studied 48.3 million patients receiving Medicaid across 34 states and the District of Columbia. Nearly 40% of ED visits and hospitalizations were for ACSCs: https://lnkd.in/giVJvHz8 Common risk models to proactively outreach these patients are of poor quality, giving pop health teams limited opportunity to intervene before patients become 'high utilizers'. Using more modern methods and incorporating social risk factors into the models increases their sensitivity 3x without losing specificity: https://lnkd.in/gBj-vuG3 Most importantly, proactive community-based outreach to patients identified by the newer models averted 48% of ACSC hospitalizations and 20% of ACSC ED visits, versus a matched control group: https://lnkd.in/g6FTaqtt

"The gap between a brain’s chronological age and its predicted age on brain imaging appears to be a strong mediator of risk factors for cognitive decline, especially in individuals with cerebrovascular disease (CeVD), new research suggested. In a cross-sectional study of more than 2000 participants, researchers used machine learning and brain scans to create a prediction model. An individual’s “brain age gap” (BAG) was then determined by subtracting the chronological brain age from the predicted brain age. Results showed that a higher number of cognitive impairment risk factors were associated with lower cognitive performance scores — and that a larger BAG may influence this relationship, with the strongest effects observed in participants with CeVD." https://lnkd.in/gke_yPGD

A new Acta Neuropathologica study asks why people with advanced Alzheimer pathology (Braak stage V) show such different cognitive trajectories. Quantitative imaging of frontal and temporal cortex in 61 autopsy cases showed neocortical p-tau burden ranging from 0.2% to 53.7%. In multivariable analysis only p-tau burden and microinfarcts independently predicted cognitive decline; amyloid, LATE-NC, and Lewy body pathology did not. Individuals with low burden (≤13%) maintained markedly better cognitive trajectories over the final 15 years of life than those with high burden (≥23.5%). The work supports neocortical tau quantification as a closer correlate of dementia severity than staging alone. https://lnkd.in/e6MAvKTE

Can a single blood test diagnose multiple dementia-related diseases? A recent Nature Medicine study introduces an AI-driven proteomics model that brings us closer to this possibility. 🙂 Here are some key insights directly from the research: 🔬 Massive dataset & biological depth The model was trained on 17,187 individuals and analysed 7,500 plasma proteins from blood samples - capturing complex biological signals that traditional diagnostics often miss. 🤖 Multi-disease diagnosis (not just one) Unlike conventional approaches, the model simultaneously predicts 6 conditions: -Alzheimer’s disease -Parkinson’s disease -Frontotemporal dementia -ALS -Stroke/TIA -Cognitively normal This is critical because up to 70% of people over 80 have multiple overlapping pathologies. 📊 Strong predictive performance The model achieved: -up to 95% accuracy for ALS -92% for Parkinson’s -81% for Alzheimer’s -with overall AUC values above 0.78 across tasks. 💫 Beyond diagnosis: understanding disease complexity Instead of giving a single label, the model outputs probabilities for each disease, revealing: -hidden co-pathologies -patient subgroups with distinct biological profiles -overlapping disease mechanisms 🆘 Interestingly, some “false positives” may actually reflect early or preclinical disease, not model error. 🧬 Biological interpretability (not a black box) The system identifies specific proteins contributing to predictions, such as: -NEFL (linked to neurodegeneration) -ACHE (associated with Alzheimer’s treatment pathways) -KCNIP3 (linked to ALS biology) The model can even distinguish individuals who will later develop cognitive decline, showing potential for early intervention strategies. 💫 #AIinHealthcare #PrecisionMedicine #Neuroscience #DataScience #MachineLearning #HealthcareInnovation

The Saudi Public Health Agency might take a leading role in adopting the model of community health centres as part of health reform. The community healthcare centre’s core functions should focus on social risk screening, risk stratification, SDoH and how these concepts interrelate in delivering primary healthcare Social Risk Screening (SRS): A systematic process to identify non-medical factors that may affect a patient’s health and ability to access care. Common domains include housing, food, transportation, utilities, safety/violence, social isolation, and education/literacy. Risk Stratification: A method to categorise patients by their overall risk of adverse health outcomes or utilisation (e.g., emergency visits, hospitalisations) based on a combination of medical, social, behavioural, and demographic factors. The goal is to target interventions to those at the highest risk. Social Determinants of Health (SDoH): The broader context and conditions in which people live, learn, work, and age that influence health outcomes. SRS inputs are commonly used in risk stratification models. How a community healthcare centre typically uses these 1) Intake and screening During patient intake or annual visits, staff administer standardised SRS tools (paper-based, EHR-integrated, or patient-reported via portals). Example domains: housing stability, food insecurity, transportation, utilities, safety, social isolation, caregiver burden, literacy/education. 2) Scoring and risk stratification Data from SRS is combined with clinical information (diagnoses, prior utilisation, chronic conditions) to assign a risk level (e.g., low, moderate, high). Tools may include: ZIP-code–level social vulnerability indices Validated screening instruments (e.g., PRAPARE, AHC-HRSN, MHSA social needs screening). Then, Internal scoring algorithms that weigh medical and social factors 3) Care planning and interventions High-risk patients may receive enhanced care coordination, social needs navigation, or community referrals. Interventions might include: -Connecting to food banks -Housing supports or homeless outreach -Transportation assistance for appointments -Utility assistance or energy programs -Domestic violence resources and safety planning -Enrolment in social work, case management, or community health worker (CHW) programs 4) Monitoring and follow-up For example, re-screen communities periodically to detect new or unresolved needs. Then track outcomes such as appointment adherence, medication pickup, emergency department visits, and hospitalisations. Eventually, one can adjust care plans based on changes in social risk and health status. #PublicHealth #Vision2030

Let me tell you a story a large multi-hospital network in the Midwest once faced a harsh reality, despite top-tier specialists and advanced facilities, their readmission rate had climbed to 17%. They realized over 60% of these readmissions were for chronic conditions that could have been predicted and managed months earlier. The missing link - visibility. Their data sat in silos: EMRs, insurance claims, community health databases, lab reports, all disconnected, unable to reveal the full story of patient risk. When they implemented an AI-driven population health platform, everything changed. They began identifying individuals at high risk of readmission 90 days in advance. Nurses were alerted for early outreach. Care plans were personalized. Preventive programs were deployed. Within 18 months, the network saw - ✅ A 27% reduction in preventable hospitalizations ✅ 15% lower cost per patient episode ✅ And most importantly - a rise in patient satisfaction scores by 21% The Shift to Predictive, Value-Based Care. Traditional healthcare economics reward treatment. Value-based care rewards prevention. But the transition is not easy. Healthcare leaders today face three challenges - #Fragmented Data Ecosystems - 80% of healthcare data is unstructured, trapped in incompatible systems (source: HIMSS 2024). #Escalating Chronic Disease Burden - By 2030, chronic diseases will account for 70% of all healthcare costs worldwide (WHO). #Reimbursement Pressure - CMS targets that all Medicare beneficiaries will be in a value-based model by 2030, forcing every hospital to become predictive, not reactive. This is where AI becomes the bridge, turning data chaos into coordinated prevention. How AI Drives Preventive Intelligence - 🔹 Predictive Modeling - Machine learning models analyze patient histories, lifestyle data, and social determinants to predict disease risk trajectories — helping care teams act before a crisis. 🔹 Population Segmentation - AI clusters patients by risk scores and care needs, optimizing outreach and resource allocation across communities. 🔹 Integrated Health Dashboards - Executives gain real-time visibility into care gaps, cost hotspots, and preventive outcomes, shifting decision-making from intuition to evidence. 🔹 Personalized Engagement - AI chatbots, alerts, and digital health tools keep patients engaged in self-care, reducing lapses in medication and follow-up visits. The Result? When prevention is guided by intelligence, healthcare stops being episodic and becomes continuous. As McKinsey reported, AI-enabled preventive models could save the US healthcare system over $200 billion annually by 2030, through early detection and chronic disease management. That’s the kind of transformation that defines leadership in value-based care. AI can empower them, by ensuring no patient becomes invisible in the data. That’s not just technology adoption. That’s humanity, scaled intelligently. #ValueBasedCare #PopulationHealth #PreventiveCare