Using Data to Guide Intervention Strategies

Our new peer-reviewed research on preventing ED visits and hospitalizations among patients receiving Medicaid: https://lnkd.in/g7wrhEhr When patients have multiple conditions, the optimal "next best step" is rarely clear. With 15 minutes left on a Friday night to spend with a patient with uncontrolled schizophrenia and hypertension, do I first spend my time on a prior authorization for a long-acting antipsychotic or on medication adherence counseling for a new blood pressure regimen? Standard guidelines leave these crucial, sequential decisions to varied individual judgement and experience. Our new, peer-reviewed research, published in JMIR AI, shows that Reinforcement Learning (RL) offers powerful guidance for such decisions. Instead of relying on LLMs (which can hallucinate, risking safety), we carefully studied years of intervention data from multidisciplinary population health teams, comparing the outcomes of similar patients who received different interventions or intervention sequences. We used these historical intervention sequences and their outcomes to build a State-Action-Reward-State-Action (SARSA) RL model to recommend the optimal interventions to population health teams. The results: - In this counterfactual causal inference study, the RL-guided approach reduced acute care events (ED visits and hospitalizations) by 12 percentage points compared to the status quo, a 20.7% relative reduction (P=0.02). - It yielded a Number Needed to Treat (NNT) of 8.3 patients getting service to prevent one acute care event. - Crucially, there was no evidence of harm (number needed to harm = infinity). - The RL-guided approach also improved fairness across demographic groups, with a 28.3% reduction in gender-based disparity and a 37.1% reduction in race/ethnicity-based disparity. This work demonstrates that population health teams enabled by RL technologies can outperform those relying on experience- or playbook-based practices alone, particularly when navigating the complex intersections of medical and social needs. This builds on ongoing work at Waymark in RL for Medicaid: Test-Time Learning and Inference-Time Deliberation for Efficiency-First Offline Reinforcement Learning in Care Coordination and Population Health Management: https://lnkd.in/gm8hU7Pd Hybrid Adaptive Conformal Offline Reinforcement Learning for Fair Population Health Management: https://lnkd.in/gVuWmj5p Accepted to Stanford's #Agents4Science2025: Feasibility-Guided Fair Adaptive Offline Reinforcement Learning for Medicaid Care Management: https://lnkd.in/dC48TCdp Andrew Ng James Zou Pranav Rajpurkar Lucas Hopkins Andrea Ramirez Scott Anders Michael Pencina Josh Patten Keith Payet Jerold Mammano Joel Gray Doug McMillen Haroon Hyder Wael Haidar Yasir Tarabichi, Brian Martin, Mohammad Dar Christina Severin Sunita Kasliwal Baligh Yehia, Jeffrey Cohen Paul Testa Suja Mathew Tracy B. Aparna Abburi Erin Nahrgang,Rob Fields Daniel Barchi Shantanu Nundy Vineeta Agarwala, Hui Cheng

Targeting where conservation works best Conservation has long wrestled with a deceptively simple question: not whether to act, but where action will matter most. Restoration, protected areas, corridors and enforcement all compete for limited funding across landscapes that differ widely in ecology, governance & human pressure. Increasingly, research suggests that improving outcomes depends less on new tools than on using existing ones more selectively — directing effort to places where it will make the greatest difference relative to doing nothing. A 2025 paper led by Rebecca Spake described this approach as “precision ecology.” It argued that conservation should move beyond estimating average effects and instead predict site-specific outcomes, tailoring actions to local conditions. The concept draws on precision medicine, which matches treatments to individual patients rather than applying uniform therapies. The logic is straightforward. Conservation operates in heterogeneous systems, where the same intervention can succeed in one place and fail in another. Tree planting may restore ecosystem function where soils, rainfall and protection are adequate, yet fail where drought, fire or grazing dominate. Anti-poaching patrols may deter illegal hunting in accessible reserves but struggle in remote areas. One-size-fits-all strategies are therefore unreliable. The paper highlights statistical methods — drawn from economics & machine learning — that estimate how intervention effects vary with context. Yet conservation has long targeted its efforts. Planning tools design protected-area networks to maximize biodiversity at minimum cost. Restoration programs prioritize areas with high recovery potential, while satellite monitoring directs responses. In practice, managers already concentrate resources where threats or opportunities are greatest. Where precision ecology differs is in emphasis. Traditional targeting often focuses on ecological value or threat. The newer perspective asks about effectiveness: the difference an intervention will make. A site may be biologically rich yet likely to persist unaided, while a less celebrated area might decline rapidly without action. Implementing such approaches depends heavily on data. Advances in remote sensing and environmental monitoring provide unprecedented detail, but gaps remain in many regions, and models built on sparse data can give a false sense of certainty. Practical constraints also matter. Land tenure, community priorities & political feasibility often determine where projects occur as much as ecological potential. Seen this way, precision ecology is a refinement. Conservation has gradually moved toward more evidence-based, context-specific strategies. Perfect prediction is impossible, but better targeting can help ensure scarce resources achieve the greatest impact.  As pressures on ecosystems intensify, the difference between acting everywhere and acting strategically may prove decisive.

Recent population health data suggests that residents in the northern region of Singapore, particularly towns such as Woodlands, Yishun, and Sembawang, have a higher prevalence of diabetes and hypertension compared with national averages. While the numbers are clear, the underlying causes are less certain. Several plausible factors may contribute: • Demographic structure, including an older population profile • Socioeconomic gradients that influence diet, stress, and health behaviours • Differences in physical activity and lifestyle patterns • Ethnic distribution and associated metabolic risk profiles • More active screening and detection through primary care networks However, these remain hypotheses. More rigorous research is needed to understand the drivers behind this geographic clustering of chronic disease. One promising approach is the use of artificial intelligence for population health monitoring. AI can integrate multiple data sources such as electronic medical records, screening programmes, pharmacy data, wearable devices, and socioeconomic indicators to detect emerging patterns of disease. With machine learning and geospatial analytics, health systems could identify high-risk neighbourhoods earlier, monitor behavioural risk factors such as physical activity, and predict which communities are most vulnerable to chronic disease. This would allow health systems to move beyond reactive care toward proactive population health management. Instead of waiting for complications to appear, we can anticipate risk, target prevention programmes, and evaluate whether interventions are working. Understanding why disease burden concentrates in specific communities is essential for designing effective public health strategies. Combining epidemiological research with AI-driven monitoring may help us better understand these patterns and ultimately improve the health of our population. #PopulationHealth #Diabetes #Hypertension #AIinHealthcare #PublicHealth #Singapore

In the current global landscape, where the success of health programs is increasingly contingent on the precise application of data analysis and interpretation, the indispensable role of Monitoring and Evaluation (M&E) has taken center stage. This document offers a well-structured, detailed guide, specifically tailored for M&E professionals and humanitarian workers, highlighting how data-informed decision-making can drive the success of health interventions. With its focus on practical tools and strategies, this guide empowers readers to improve health outcomes through more robust program evaluations and service delivery assessments. By delving into core statistical principles and methodologies, this guide enables M&E workers to refine their ability to monitor program performance and assess the impact of health services. The challenges posed by ongoing global health crises—such as the HIV epidemic, the resurgence of tuberculosis, and the continued prevalence of malaria—underscore the necessity of harnessing accurate, high-quality data for informed decision-making. This guide exemplifies these principles through real-world case studies, such as Thailand’s HIV prevention initiatives and the effective scaling of HIV testing in Nigeria, both of which leveraged data to inform critical policy and operational decisions. Ultimately, this document serves as an invaluable resource for M&E professionals, providing them with the skills and knowledge necessary to improve the quality of their evaluations and the programs they support. By embracing the insights and techniques offered in this guide, readers will be better equipped to influence policies, enhance service delivery, and foster more impactful, data-driven solutions to some of the world’s most pressing health challenges.

Every referral in healthcare is a $10,000 decision. But too often, they’re still made by habit, not data. “We’ve always referred here.” “That’s who we know.” “They’ve always been good to us.” That mindset is common. But it leaves money on the table, and patients at risk. When you redesign referral networks around data instead of anecdotes, 3 things happen: ✅ Patients are guided to specialists with stronger outcomes ✅ Health systems cut avoidable costs at scale ✅ PCPs make decisions with clarity and confidence So how do you put data into action? → 𝗠𝗮𝗽 𝗿𝗲𝗳𝗲𝗿𝗿𝗮𝗹 𝗽𝗮𝘁𝘁𝗲𝗿𝗻𝘀 𝘄𝗶𝘁𝗵 𝘀𝗰𝗼𝗿𝗲𝗰𝗮𝗿𝗱𝘀. Use data to see where patients are actually going, highlight costly leakage, and identify which specialists consistently deliver better outcomes. → 𝗟𝗲𝗮𝗱 𝘄𝗶𝘁𝗵 𝗾𝘂𝗮𝗹𝗶𝘁𝘆 𝘁𝗼 𝘄𝗶𝗻 𝘁𝗿𝘂𝘀𝘁. PCPs care most about patient outcomes. Show them evidence of higher-quality care first, cost savings will follow naturally. → 𝗦𝗲𝘁 𝗰𝗹𝗲𝗮𝗿 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱𝘀 𝘄𝗶𝘁𝗵 𝗱𝗮𝘁𝗮-𝗯𝗮𝗰𝗸𝗲𝗱 𝗮𝗴𝗿𝗲𝗲𝗺𝗲𝗻𝘁𝘀. Define expectations around access, communication, and coordination backed by real data. So accountability is built in from day one. Because the point isn’t to gather more data. It’s to use the right data to guide better action for patients, for systems, for growth. That’s how networks stop being a cost center and start driving sustainable growth. 👉 If you were redesigning your referral network today, which single data point would you put at the center?

A nationwide suicide reattempt prevention program in France that was built on brief contact interventions (BCIs) like crisis cards, phone calls, and handwritten, hand-stamped postcards 𝘳𝘦𝘥𝘶𝘤𝘦𝘥 𝘴𝘶𝘪𝘤𝘪𝘥𝘦 𝘳𝘦𝘢𝘵𝘵𝘦𝘮𝘱𝘵𝘴 𝘣𝘺 38% 𝘰𝘷𝘦𝘳 12 𝘮𝘰𝘯𝘵𝘩𝘴. The program was effective regardless of prior suicide attempt history and showed slightly greater impact among women. With a return on investment of €2.06 per euro spent, it’s not just clinically meaningful; it’s fiscally responsible. This study is a masterclass in pragmatic public health: low-tech, high-touch, and high-impact. For systems grappling with behavioral health crises and budget constraints, this is a rare win-win: better outcomes and lower costs, without needing an app or AI. Suggested action items for healthcare executives: 📬 𝗘𝗺𝗯𝗿𝗮𝗰𝗲 𝗹𝗼𝘄-𝘁𝗲𝗰𝗵 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀: handwritten outreach can outperform digital nudges in behavioral health. 📊 𝗨𝘀𝗲 𝗘𝗛𝗥 𝗱𝗮𝘁𝗮 to identify and stratify patients at risk for reattempts, especially in the first 6 months post-discharge. 📞 𝗙𝘂𝗻𝗱 𝗮𝗻𝗱 𝘀𝘁𝗮𝗳𝗳 𝗰𝗲𝗻𝘁𝗿𝗮𝗹𝗶𝘇𝗲𝗱 𝗳𝗼𝗹𝗹𝗼𝘄-𝘂𝗽 𝘁𝗲𝗮𝗺𝘀 to deliver structured outreach; don’t rely on ad hoc clinician goodwill. 💰 𝗘𝘃𝗮𝗹𝘂𝗮𝘁𝗲 𝗥𝗢𝗜 𝗼𝗳 𝗯𝗲𝗵𝗮𝘃𝗶𝗼𝗿𝗮𝗹 𝗵𝗲𝗮𝗹𝘁𝗵 𝗶𝗻𝘁𝗲𝗿𝘃𝗲𝗻𝘁𝗶𝗼𝗻𝘀 using real-world data, not just RCTs; your CFO will thank you.

It takes, on average, 17 years for new clinical evidence to make its way into routine practice. And even then, only about 14% of published evidence is fully adopted. Why? Because generating evidence isn’t enough—we have to execute on it. Decades of research have explored what actually drives clinicians to change their practice. One of the foundational papers by Grimshaw et al. (Strategies for Changing Clinicians’ Practice Patterns) showed that active implementation strategies—like educational outreach, reminders, and audit/feedback—are consistently more effective than passive approaches. (PubMed: https://lnkd.in/gtAbQutz) And just this year, a systematic review of 204 studies (covering 36,000+ nurses and 340,000+ patient interactions) confirmed this. The interventions that consistently moved the needle? - Individual and group education - Reminders - Tailored interventions - Leveraging local opinion leaders (Source: Implementation Science, 2024: https://lnkd.in/ga8ijMFa) One striking takeaway: more isn’t always better. Adding multiple strategies together didn’t necessarily improve outcomes. Focus and fit mattered more than complexity. Why this matters for Medical Affairs -> Implementation science shows us what works. But it doesn’t tell us where to act first. That’s where internal insights can play a critical role. They help pinpoint the barriers most relevant to your stakeholders, so interventions can be applied with precision. Hypothetical examples might look like this: 1/ If field insights suggest confusion around biomarker testing → this could point toward targeted education as a potential strategy. 2/ If medical information requests highlight repeated dosing questions → this could suggest reminders or quick-reference tools as a useful approach. 3/ If educational outcomes show persistent safety concerns → this might indicate a role for engaging trusted opinion leaders. These are not recommendations—just illustrations of how insights could bridge to the kinds of strategies implementation science has proven effective. In short: -> Insights are the compass (they show where the barriers are). -> Proven strategies are the engine (they drive change forward). By connecting the two, Medical Affairs can move beyond simply “collecting insights” and accelerate the path from evidence → adoption → patient impact. How often do insights directly inform execution—whether that’s education, reminders, or opinion leader engagement? Are we truly closing the loop?

How are you using your program data? If your answer is reporting and then some silence....read on. Your program data should be used for multiple purposes beyond accountability and reporting. Here are some tips to apply to start utilizing your program data for learning and adaptive management; 1) Assess if you are answering the correct questions. ↪ Is the data you are collecting what you need to learn? For example, will the number of farmers trained tell you what you need to do to improve the intervention? 2) Design learning questions with the program decision-makers. ↪ Bring together the critical program decision-makers and go through an exercise to determine what information they rely on to assess if things are working as expected. 3) Review and redesign your data collection and analysis system to address the learning needs. ↪ Think beyond quantitative data collection methods. Incorporate participatory M&E and qualitative inquiry approaches. 4) Provide evidence on time and in the correct format to the different decision-makers. ↪ Have more than 1 format for presenting the evidence gathered and ensure it comes at the right time to influence decision-making. 5) Support evidence translation. ↪ Sharing the evidence in written formats is not enough. Consider evidence synthesis and sensemaking activities that help the team understand what the evidence is 'saying.' 6) Set up follow-up systems ↪ Design systems to track how the evidence is used and how the adaptations affect program outcomes. PS: What would you add to the list? Follow me, Florence Randari, for more tips and resources on learning and adaptive management!

Hey #highered leaders - if you're still using static pivot tables to inform strategy, this post is for you ⤵ Take a peak at the below screenshot. This example, which shows two "paired predictors", is just one way you can turn data into action: 📈 ▶ The top right quadrant are “high achievers”. They have a high GPA + high credit earn ratio. These students might simply receive a message of encouragement. ▶ The top left quadrant are “strivers”. They have lower GPAs, but higher credits earned. These students might receive a nudge related to maximizing their use of available academic resources. ▶ The bottom right quadrant are “setbacks”. They have higher overall GPA, likely from good grades in their early coursework, but are earning fewer credits towards graduation requirements in key courses in their major. These students should probably receive messaging about the need for high-touch interaction with their advisors to stay on track and not lose their early momentum. ▶ The students in the bottom left quadrant are in "survival mode”. They are below average in both areas. These students are probably due for some real human-to-human conversation to better understand their needs. They may need in-depth intervention, with accompanied supports for finding the most successful path towards goals that match the students’ strengths and interests. You may consider nudging and re-nudging them throughout a term. ⤵ There's so many more examples of how Civitas Learning partners are disaggregating data to close equity gaps. If you're curious to learn more, let's connect 💌 #studentsuccessanalytics

Screening tests are often misunderstood as tools to predict injury. In reality, I believe their greatest value lies in guiding interventions at multiple levels: 👉🏼 Group-level interventions – Applied to the entire squad. For example, the entire team undergoes a warm-up to try to prepare them for the demands of the session and reduce the risk of injury. 👉🏼 Cluster-level interventions – Targeted approaches for sub-groups. This might be based on positions, such as specialised shoulder work with Goalkeepers or Quarterbacks, or clustering into groups based on the screening data like ankle, hamstring or hip focus. 👉🏼 Individual-level interventions – Personalised adjustments for athletes based on their injury history, profiling results and/or rehabilitation needs, such as tailored prehab exercises. A single test cannot reliable predict injury for an individual. But using this tiered approach allows practitioners to apply screening test data to reduce risk across the group, individualise preparation and prehab for each athlete, and allocate resources efficiently without overburdening athletes.