Welcome to Population Health, applying health IT to improve population health at the community level. This is lecture b, the objectives for applying health IT to improve population health at the community level, are to compare and contrast traditional public health perspectives with that of the newer and at times controversial public health perspective. Summarize the potential for health information technology to improve the health of populations at the community geographic level. Classic public health IT as run by a state, or a territorial health agency, health department, or a federal agency, for that matter, is really quite important. And by and large, cannot be replaced by informatics or data systems of medical care organizations. It could include vital health statistics that track births and deaths. Increasingly important, is the tracking of surveillance of reportable conditions. It seems like every year a new virus comes or bacteria comes to threaten us globally. So that's all important, as well as longer term challenges to health, such as sexually transmitted diseases, or HIV, or TB, that have been around for a while. But whether or not it is H1N1, or Zika virus, or Ebola, the need to survey all these things is critical, and really can only be done well at the community level. That's not to say that there can't be a role for electronic health records in other medical care. Increasingly, both public health informatics and medical care systems need to be integrated. Another example of that would be the classic role of immunization tracking. Cancer registries and monitoring newborn congenital diseases. Historically these have traditionally been done at the community or state level. But now, increasingly, these types of clinical information can also be captured by a medical care system through electronic health records. Although the role of public health agencies application of HIT goes well beyond that, in general these are fairly high priority. And perhaps the majority of systems in states have these types of HIT applications. There are many examples of HIT supported activities that can be used for population health improvement at the community or geographic level. The first is needs assessment, or prioritization of care, for those individuals with the highest need. A risk pyramid would be an example of that. That is when you array an entire population in a geographic area along a continuum. Some may be at need for early screening, others may be experiencing the early disease stage. And still others may be experiencing the later stages of a disease. Grouping of quote, denominator end quote population into these classes is a classic public health approach in likewise is central to population health. We've already talked about the need to integrate population health at the community level, medical level, and healthcare system level. Therefore, one can use HIT increasingly through electronic medical records, in health information exchanges, where all of the data can be shared within a community. In area yet to be wildly applied, at least in the US at the time of this recording is taking medical and social services data and agencies and combining that on a regular basis. But increasingly, whether or not it's a nutrition program, a housing program and abused partner program those can and should be linked with medical care. The role of the consumer in their own behavior, and the behavior of the community and their family is key. Increasingly, one can use mobile help, or M Help, or other consumer based internet programs to intervene in health and behavior. Certainly activity trackers, for example, a Fitbit, and others are key to changing behavior. And so in population help for many factors that involve the consumer and members of the community that is a very promising area. There is a field of analysis that's been around for awhile called graphic information systems, or GIS. People don't spell that out very often but GIS means, let's look at factors that are link to geographic units. That could me a census tract which could be communities or blocks on the street. That could be zip codes, that could be towns. And let's analyze patterns at the level of geography. That isn't always easy to do it because not all medical care and public health data are available in small units. But where they are, GIS analyses are very promising. We'll share some examples of that Increasingly if one wants to strive for the health of the community given that the greatest preponderance, 97%, of health and public health dollars really go into medical career one must embrace the medical care community hospitals and doctors and insurance companies. Increasingly various communities use pay for performance P for P to incentivize doctors and hospitals to focus on population health. The very influential institute for healthcare improvement, IHI, has come up with something called the triple aim and one of the aims in fact is population health. We will share with you a very exciting a new model here in Maryland where the secondary and tertiary care providers of the community are now being held accountable for population health, and that is done through data systems, through understanding patterns of care, and seeing how well they improve. Now we've already talked about the learning healthcare system or learning healthcare community. The center for Medicare and Medicaid services, CMS, that provides management of the Medicare and Medicaid programs, has just launched an accountable health community program. Trying to integrate social and medical care and insurance plans. And understanding what works which is a very key part of CMS is intervention is key. Again that's part of the big movement toward learning health systems or in this case population health systems and as alluded to in the graphic in lecture a from the Canadian Public Health Association and very important to the future one must evaluate and understand outcomes, and that is key not just to medical care but also in public health and community based care. The next part of this lecture discusses a few case studies that pull together some community level population health informatics applications. The first case study is an interesting one that draw on medical care data from probably the largest electronic health record database in the US. And maybe in the world and that's the department of veterans affairs. The hospital and medical care system across the country that provides care to men and women who have been in the US armed services after they leave the military. They have been leaders in the application of electronic health records, EHRs, sometimes called electronic medical records, EMRs, to medical care and to research. But now, they're also very interested in population health, and they are collaborating with the Johns Hopkins Center for Population Health Information Technology or CPHIT. This is project that tries to blend medical care issues with one that is very amenable, and important at the community level and that is obesity. An epidemic across the US in all settings including former veterans. The VA has probably one of the largest databases in the country for body mass index, BMI, which is the marker of obesity used in the clinical setting. Nearly every time a patient in the VA comes in contact with the system, their height and weight are taken. And one is able to assess their change in weight and potentially obesity level. There is a project linking this clinical information from literally tens of millions of veterans across the country with data available down the latitude and longitude of where the veteran lived at the time they were getting care. So it's possible to link that information with all types of information. With not only that veteran, and the healthcare system, but also with the geographic area where they reside. So one can attempt to link in information on available food, exercise opportunities, whether or not it's an urban or rural setting. And as experts in obesity will say, there's a full array of non-medical as well as medical factors that could potentially impact obesity levels and shifts when interventions are applied. This is at an early stage, but everyone is very excited that this will make it possible to link together public health and socioeconomic factors and community, as well as, one of the better medical databases, arguably in the world that the VA has. CPHIT is currently at an early stage of creating some feasibility assessments of graphing that applied various analytic, sometimes called predictive modeling where they not only identify where individuals or in this case tens of millions of veterans are in terms of their obesity level today, but where they'll likely be in the future. So they are looking at patterns over time and trying to make predictions so that the Department of Veterans Affairs will be able potentially to implement interventions either at the community level, clinical level, or healthcare system level. The map of the United States on the left shows sort of a typical day in the VA. A typical day includes 29,000 BMIs that are represented and available within 24 hours, if not quicker. Showing levels of obesity as represented on this graphic, sometimes this is called a hot spotting chart because it shows spots that are quote, hot, end quote. When it comes to understanding the condition being surveilled or monitored. In the graphic on the right, just to show one of the many dozens of factors being linked in their GIS analysis shows an association between income and obesity, a known factor related to diet and exercise, and other factors that are sometimes linked in geographic areas. In addition to trying to learn about this, and help the VA plan nationally for intervention, CPHIT hopes to move toward a population decision support program. In this case, it would be for the healthcare system serving the veterans in Maryland where Johns Hopkins is located. And this is a model showing data that can't be captured, not in the normal way that a doctor would capture it for his or her patients or a clinic might capture it for the denominator patients, that is a type of population health. But this would allow all VA service providers in a region to link together, and not only look at actual trends in BMI for all the veterans in the state, but also other non-medical factors. There's also, off to the right of this graphic, the various types of data that CPHIT's linking in. It shows that they're bringing in census data, data from the Centers for Disease Control, the USDA which monitors food, even NASA which monitors geography from satellites, that let's you see if there's green space. And we're trying to show how these data can use to understand factors that can improve the health of populations, in this case obesity. It's a very exiting time in the field of population health informatics. Because as more and more data come online, the possibilities for understanding the impact on communities, the health of communities will ever increase. The next series of slides presents a case study of a collaborative effort in the Baltimore and Maryland communities that focuses on community health for the entire state and region. To accomplish this in the US one needs to collaborate across many providers, many government agencies, many private organizations and HMOs. In many state there are active health information exchanges, or HIEs. In Maryland this HIE is called the Chesapeake Regional Information System for Our Patients, or CRISP for short. In Maryland in part because of the unique CMS-sponsored waiver innovation in the state, all hospitals are paid the same amount for all health insurance organizations sometimes called an all payer. This would include Medicare, Medicaid, and all the private insurance companies. All of the organizations likewise share in a non-insured costs. This case study develops metrics and approaches to monitor the health of populations using electronic medical records, administrative records, health insurance records and public health records to develop a sense of the geographic level of some of the healthcare utilization characteristics. Such as, hospitalizations or looking at elders who are likely to fall and injure themselves. A very high priority area in most locations. Or a new very exciting an important initiative concerning the impact of opioid addiction, of both legally prescribed medications and heroin. So there is an array of collaborations, the goal of which is to improve the health of populations in the state and in the region by linking medical and non-medical data, and trying to emphasize to a feasible extent the social and population health factors in addition to the more common medical factors. This is a graphic developed by CPHIT representing the Maryland population health information network, M-PHIN to support the all payer, global based system that every Maryland hospital is using. Again, it's a very interesting model. The hospitals are being paid for what takes place in the hospital, but they will make more or less money based on the health of the population. And increasingly because they're paid at global budget, they do not have incentives necessarily to treat people in the hospital. The goal is to keep them healthy. This graphic describes the various types of data to be collected, following the IHI's triple aim at the top. The goal is to try to monitor both at the patient level and the population level and to understand cause. The network will include all the data for all the hospital pairs including the unified claims data system. The health insurance claims which would be Medicare, Medicaid, HMOs, Blue Cross, or United Healthcare really are the core of information that can be used. Increasingly, electronic healthcare records are available. The majority of hospitals, in Maryland's case 100%, and a very large number of doctors, approximately 80 to 90% in Maryland, have electronic health records. The challenge, however, is that they're not all integrated or inter-operable, but they will be one day. And they will also be pulling together public health, local and state information, as well as national surveys and other sources. It is very important that you don't just have the data, but that you also have a measured framework. And a lot of time is being spent developing measures sensitive to the paradigm presented here in this unit, not just for medical care outcomes, but also for population and social outcomes as well. Maryland's State Health Department is called the Department of Health and Mental Hygiene of the State of Maryland. And they, of course, are key players, as is the Health Services Cost Review Commission, HSCRC, which is the unit that's responsible for developing the payment protocols that are approved by the Medicare program. So it's a very exciting time. Good progress is being made, but there's still much to do. Shown here is an early stage example of a GIS analysis that links in information for the City of Baltimore to understand hospitalizations in various quote, hot-spots, end quote. These are areas that are also lower socioeconomic regions, where patients don't always have access to good primary care or access to other types of services, and need to rely on hospitalization. Other locations in the country are doing this type of research too. But a unique aspect of the Baltimore project is that it's linked across all geographic areas and it's a blending of medical and social. Increasingly, under the umbrella population health, these different data sources are being used to try to address this holistically. The last slide in this case study series is also a Baltimore project. Recently funded by the Robert Wood Johnson Foundation and other sources, it is a collaborative effort between the Johns Hopkins Bloomberg School of Public Health and CPHIT, the City of Baltimore and the CRISP HIE, that attempts to link medical data, social data and even meteorological data to look at factors that are linked to the falls of elders. Many people, particularly older people, who are prone to fall. Yes, ice can do it, but so can not having hand rails or having steps that are in disrepair or being on medications that lead to dizziness. It is not good for the older person to fall, but in a global budgeted healthcare system, every fall also costs the system additional money. So this is an attempt to link data from the Housing Authority, for example, to data from prescribers to get an idea of who's at risk of side effects that could lead to falls and also to link that data to the use of walkers or other equipment. Certainly, understanding social support is something that would be good to do. There have been so many projects, particularly in institutions, trying to look at falls, because patients do fall in hospitals and nursing homes as well. But this is one of the first where there's been an attempt to do it for entire communities, linking medical and social data. HIT will allow an array of things in the future, that couldn't really be done easily before. The first on this list is the way to integrate numerators and denominators to define true populations. Those of you who have taken an epidemiology course understand that one always needs to try to look at who's at risk. Who's the target, is it everybody in the community, for example? And who are the people who have had the incidents or the prevalence? Easy to say theoretically, in the real world it's a challenge to determine who your target population is and who are the individuals that you're most concerned about. But in linking all of these data together, such as CPHIT was able to do with the VA EHRs, and that they hope to do with the HIE data in Maryland, this is increased. And in the domain, sometimes one develops a denominator, a target population from one database and a numerator, those with the disease or at risk, from a separate database altogether. It's getting easier to identify who's at risk by using algorithms and predictive modelling, where it's possible to predict who's going to be at risk, not just today, but in a year or two years. Who is a risk for a fall? Who is at risk for hospitalization? Who is at higher risk than average for opioid abuse, because they're already taking medications that lend themselves to abuse, or they've had past hospitalizations for substance abuse. Who is at risk for hospitalization because their blood sugar is not under control? These types of predictive modeling and risk identification increasingly will and can be used. Most of these new databases being applied today have structured data. That is, there's only a certain number of possibilities to fill out a data item, and it may be one through five. Or it's very clear that anybody can use these structured data. In the electronic medical record and in a lot of other sources of data, people use text, or they speak, or there's just a narrative, much more difficult to use. Increasingly, it's necessary to collaborate with computer scientists to take advantage of some of the techniques used by Google and Microsoft and IBM in business that can be used in healthcare. For example, natural language processing, or text mining, can assist with both quote, reading, end quote, the notes of doctors or the notes of social workers. And this is a domain that increasingly will be important. Most analyses in public health and healthcare have tended to be retrospective. Once in a while, there's an attempt to plan things ahead of time. Call that predictive or forecasting, but increasingly a lot of these things will happen in real time or near real time. Just as there is clinical decision support models for doctors and nurses to help them at the bedside, or when the patient's in the office, increasingly in population health, clinical decision support can be developed to support nurse outreach, to support surveillance or to monitor trends. And so real-time signals and models that are dynamic will increasingly be a future area in population health IT. Also to date many of the medical models have focused on one disease at a time, assessing whether or not clinical guidelines are adhered to, or whether or not an outcome is good or bad. They had not necessarily looked at a balance scorecard of what one achieves, whether or not it's the IHR tripling, or whether or not it's taken a community's priorities into perspective. But increasingly, they are looking to understand the value of a certain intervention or action, not just for the individual patient, but for the entire population. This gets into complex and sometimes philosophical or ethical issues. Something you can't avoid in healthcare or public health. Having decision support that will allow you to take or modify this, depending on the community or payment organization, or the wishes of the person or community will also be a frontier. A challenging frontier but an important frontier. This concludes lecture b of population health applying IT to improve population health at the community level. In summary, in this lecture, you have learned that the domain of population health informatics and population health information technology, HIT, is a composite of various, sometimes independent disciplines. About examples of HIT-supported activities that can be used for population health improvement at the community or geographic level. Additionally, you have learned about case studies that pulled together some community level population health informatics applications. Including a Johns Hopkins CPHIT obesity study, utilizing the Veteran's Affairs, EHR database, and a collaborative effort in Baltimore and Maryland communities that focuses on community health for the entire state and region using the crisp HIE and geo data. In summary, in this unit you've learned, one, how health information technology is being applied to the improvement of population health at the community level. Two, about the relevance of CPHIT's Ecological Framework for Population Health to the concept of population health at the community level. Three, how other types of factors, such as social factors and non-medical factors impact health and wellness. Four, to compare and contrast traditional public health perspectives with that of the newer population health perspective. Five, about the potential for health information technology to improve the health of populations at the community and geographic levels.
