Biomedical Engineering - E-health Methods and Applications

Completed notes of the course

Complete course

1 1 DIGITAL TRANSFORMATION: The first definition of e - health was born around twenty years ago, and it says: “E - health is an emerging field referring to health services and information delivered or enhanced through the internet and related technologies”. Topics and elements that are important in order to define health concept • Empowerment of consumers and patients • Enchanting quality of care • Enabling information exchange • Education • Efficiency • Equity • Extending conceptually and geographically • Ethics • Encouragement • Evidence based This definition was not really restricted to the use of technology and to the use of electronic technologies, in fact Eysenbach hi mself, that was the first coming out with the definition of e - health, stated that the “e - ” of e - health had different meanings, so h e wanted to enlarge the definition since the beginning. Nowadays, after twenty years, we are always speaking about digital transformation. Digital transformation is a term that is u sed every day, especially in this period with Covid19. • Digital transformat ion does not mean digital technologies. à does not mean to have a lot of opportunities that are able to identify a better way to life, but it is changing the type of life but all this element should be connected, integrate things but must consider thi ngs th at are reliable • Digital technologies are only the tools that drive and allow the digital transformation. • D igital transformation is a change in culture and organization, while digital technologies are opportunities to revolutionize the model of care. • Digi tal transformation is not only inside our lives (mobile phones, digital monitoring, etc.), but also inside health care (elect ronic prescription). We should not only provide this technology, but we also should adopt it. • When we are working in the world of digital transformation and when we want to produce something that is an e - health system, we have to be careful because to develop any health system does not mean that this system will be used. • This system is new, we can develop the best system in the worl d, but nobody would not use it. • Necessary to consider monitoring aiging of people or chronic diseas à (confronto tra quell oche è stato creato in un anno durante il period del covid e durante i precedenti 20 anni à quindi se si ha la presenza di una “catalizzatore” come il covid è piu facile che avvengano dei cambiamenti, sono necessari anche dei cambiamenti dal punto di vista politico) • We want to understand the best methodology to build this e - health technology in order to have this technology as a tool for digital transformation, so for a change in our organization. • The two main reasons that call for this digital transformation are ageing and chronic diseases, which are the things that cal led the most for investm ents in the last twenty years. o But the boost towards digital transformation that in one year Covid19 did, was exactly the same that has been done since the beginning of e - health concept. o So, in the end, we can say that Covid19 is a catalyst of digital transformation. This was an optimistic effect of this pandem ic and tragic situation. 1.1 DIGITAL MEDICINE: What is the difference between digital medicine and digital health (sometimes also ca lled digital wellness)? 2 • Digital medicine is a set of software and algorithms (tools) that is used to measure or intervene and to improve human health. We need sensors to do measurements related to health à close loop that is created around the final user à requires to measure some element and to improve human health • Digital medicine application are based on clinical evidence, for this reason it is necessary to consider medical certification (completely change after 2020) à necessary demonstrate clinical validity à compare the steps with something that is clinically used • These measurements produce d ata that flow and inform the intervention, so according to our measurements we can drive the intervention. The result of the intervention, again, is a flow of data that go and create a loop. • So, the intervention is modified producing other data that are s ensed by measurements, and so a loop is created. • This is a technology that can be used not only in digital medicine, but also in digital wellness, for example speaking about an app that monitor our sleep quality. In this case there is a measurement, there is an intervention and so there is a sort of loop. Another exam ple is the health application to cou nt the steps, which is also something that correlates monitoring to intervention. For example I want to do three days a week a running of one hour, though the app I can see if I am able to do it, so I can monitor, check, and correct what I am doing. • There are 2 processing: o Digital medicine is more than this, it is not only a tool that allow me to monitor myself, but it is also something that is b ased on scientific evidence, so it has to be reliable. It is not only the way to measure and intervene, but the intervention should be based on scientific evidence. o Digital medicine requires both verification and validation. What is the difference between verification and validation? Verif ication is mainly done by engineers, while validation is mainly done by clini cians. This already means that there are multiple actors to build any digital medicine tool, so we need a multidisciplinary approach that involves both engineers, that can develop the tool and verify that is right, and clinicians, that has to understand if what is done is needed from a clinical point of view. So, both verification and validation are something that should be processed and understood in order to understand if the digital tool that is developed fits the initial purposes. o Verification à verify what is produce is correct and has the correct function o Validation à includes all the users in order to validate the validity to use this technology • Together are very important 2 E - HEALTH DATA: Role of data in healthcare: All health care processes are based on gathering data and interpreting their meaning in order to take decisions. • Data is the base for the decision make • Gathering data and interpreting their meaning are central to the health care processes • Data are crucial to the process of decision making: o Data are the basis for categorizing the problems a patient may ha ve, for example the first time we go to the doctor to understand the diagnosis associated to some symptoms we may have, or for identifying subgroups within a population of patients. What the doc tor does is using his knowledge and its experience to categori ze people according to symptoms and understand what the diagnosis is, according to data collected by asking questions. o Data help a physician to decide what additional information is needed and what actions should be taken to gain a greater understanding of a patient’s problem or to treat most effectively the problem that has been diagnosed. à treat more efficienty a case but it is necessary to share the data à should be put together data and should be standardized 2.1 Data flow in specialized healthcare • Modern medicine is highly specialized • The achievement of high quality healthcare may imply data flow among physician/providers • The issue is more relevant for the management • These latter also imply higher costs 3 • Consider an example: o Oncology à require the ccoperation of different medical specialistws § Primary care physician o For this reason it is necessary to consider the standardization o Oncological treatments require the cooperation of different medical specialties: § Primary care physician (PCP) § Medical oncologist § Surgical oncologist § Radiation oncologist + nurses and caregivers (w/wo hospitalization) • The important thing is data are not free. If we need additional data to better understand the diagnosis, we need to order and prescribe diagnostic exams, that are associated to a cost. It is always a trade of between the cost of data and the utility that these data can ha ve in understan ding the diagnosis, so their importance and priority. It is difficult to decide which is the best exam to give in a specific situa tion, to go deep into something and acquire additional data. It is a process that is iterative and costs. • Data is also import ant because healthcare is now very specialized and involves different actors also in the medical staff, so data has to be maintained and shared between different people. Also the maintaining and the sharing of these data is another cost. • The achievement o f high - quality healthcare may imply data flow among physician / providers. We should move through the maintenance of electronics records (both data, images etc.). The issue is more relevant for the management of chronic diseases. These latter also imply hi gher costs. • For example, oncological treatments require the cooperation of different medical specialists: primary care physician, medical oncologist, surgical oncologist, radiation oncologist + nurses and caregivers (w/wo hospitalization). • So, if we need the share data between all these people, we need to make these data electronic and available, but also to use standards that are the way technical or semantical to ensure the understanding, the maintenance and the transfer of data around all the people i n volved. • We are not seeing all the technical standards in this course, but we are instead presenting the semantical standards, so the standards that ensure the maintenance of the meaning. 2.2 Interoperability • Interoperability is associated to standards. -- > it is necessary that everyone speak the same language Definition: E - health interoperability means the ability of two or more e - health systems to use and exchange both computer interpretable data and human understandable information and knowl edge. • If we build e - health without interoperability this is what happens, we are not reaching the other people. • There are three interoperability layers: o Organizational interoperability : it usually includes the so - called level interoperability, that someone is addressing as the fourth layer. § Organizational interoperability is the ability to provide and to sign policy and rules to work together. So, before using sta ndards, and before usi ng the other two layers, that are semantic and technical interoperability, the organizational part should be put in place somehow. It consists of all the o Organizational interoperability : it usually includes the so - called level interoperability, that someone is addressing as the fourth layer. § Organizational interoperability is the ability to provide and to sign policy and rules to work together. So, before using standards, and before using the other two layers, that are semantic and technical interoperability, the organizational part should be put in place somehow. § It consists of all the legal framework and workflows. It is the will ability to work together and exchange information. It depends on the overarching environment composed of laws, policies, and cooperation agreements. o Semantic interoperability : when we speak about sematic interoperability, we have standards to maintain the meaning. It allows both sides to understand the meaning of the information exchanged across cultural and linguistic barriers. § This means that the two infrastructures have to understand and associate the same meaning to the s ame data. Is not only the meaning, but also guidelines and protocols that have to be followed. 4 § These general guidelines are recommendation done at international level, but sometimes they do not correspond to the specific application of a hospital. So there are clinical pathways and procedures that follow guidelines and recommendations according to the resources that are available. § Includes different thing à clinical procedure and medical guidelines à meaning in recognizing the situation and treating a situation with recommendation but also include standards and profiles and terminologies So, semantic interoperability is not only the maintenance of the meaning of data, but it is also the clinical procedures and medical guidelines. • Technical interopera bility : when we speak about technical interoperability, we have standards to allow the exchange and the maintenance of data, so the transfer of data between different structures, so it guarantees data security and privacy, data integrity, access to relevan t data associated with an identified patient. 2.3 From data to knowledge systems: • We consider a medical datum to be any single observation of a patient — e.g., a temperature reading, a red blood cell count, a past history of a rush, or a blood pressure reading. The datum itself not always is associated with a meaning. o A single observation like for example the temperature, observation of a patient that can be done in series to monitor somethi ng but to allow single to be im portant for a system it is necessary to put all the data together à structure all the data in order to give to them importance • For example, let us consider the blood pressure reading. We can have 127 mmHg, and we can associate it with a meaning, but th e d atum itself does not mean anything. For example, if we are doing an exercise and we have 160 mmHg, it is a datum that is not associated t o a problem because we are doing exercise, so the datum itself does not have a meaning. • So, medical data are parameter s about a patient. If a medical datum is a single observation about a patient, medical data are multiple observations. Such data may involve several different observations made concurrently, the observation of the same patient par ameter made at several poi nts in time, or both. Thus, a single datum generally can be defined by 4 elements: • The patient in question, so the person associated with that datum. • The parameter being observed (e.g., liver size, urine sugar value, history of rheumatic fever, heart size on chest X - ray film). • The value of the parameter in question (e.g., weight is 70 kg, temperature is 98.6 ̊F). Usually the value is also the unit of measurement. • The time of the observation (e.g., 2:30AM on 14FEB1997), in fact a datum can have a duration, or can evolve in time (for example the HR of children and adults). All the context around the data is important to transform these data in information. To associ ate a meaning to a datum, we need to know all the elements listed above. • It is a matter of perspective whether a single observation is in fact more than one datum. A blood pressure of 120/80 might w ell be recorded as a single datum point in a setting wher e knowledge that a patient’s blood pressure is normal is all that matters. If the difference between diastolic (while the heart cavities are beginning to fill) and systolic (while they are contracting) blood pressures is important for d ecision making or fo r analysis, however, the blood pressure reading is best viewed as two pieces of information (systolic pressure = 120 mmHg, dias tolic pressure = 80 mmHg). • H uman beings can glance at a written blood pressure value and easily make the transition between its unitary view as a single datum point and the decomposed information about systolic and diastolic pressures. 5 2.3.1.1 Modifiers • Another important thing is that we can have modifiers . A modifier is a circumstance under which a datum was obtained. For example, was t he blood pressure taken in the arm or leg? Was the patient lying or standing? Was the pressure obtained just after exercise? Dur ing sleep? What kind of recording device was used? Was the observer reliable? • Such additional information, called modifiers, can be of crucial importance in the proper interpretation of data. A modifier better explains the modality in which we took a specific data. • We do not have to put all modifiers, but only the relevant ones. Th ey can be very important to understand and interpret the meaning of data. So, we are going to go towards a structure of data. We are building around these data a sort of database, because we are asso ciating to the data a lot of characteristics. We are crea ting a sort of object, that is the datum, that has a lot of parameters associated. 2.3.2 Data uncertainty • Another problem is that of uncertainty . We know, as engineers, that each instrument has a specific sensitivity, accuracy, and all these characteristics. W e know that by measuring with a sensor, the sensor itself produces error, so we have error of measurement. So, at each datum corresponds an associated uncertainty. • It is good to know the uncertainty of the data we are looking at. Every instrument introduc e a certain level of uncertainty. It is rare that an observation, even one by a skilled clinician, can be accepted with absolute certainty. • This is associated to a cost. The most relevant is the measurement to understand a certain diagnoses, the less shou ld be the uncertainty related to the data we are producing. • One technique is to collect additional data that will either confirm or eliminate the concern raised by the initial observati on. This solution is not always appropriate, however, because the cost s of data collection must be considered. • The additional observation might be expensive, risky for the patient, or wasteful of time during which treatment could have b een instituted. • Again here, we are coming to the idea of trade - offs in data collection, that thus becomes extremely important in guiding health care decision making. 2.3.3 From data to knowledge based system • The following figure shows a pyramid that represents a sort of hierarchy of meaning: o We have data, we have information, we have knowledge. What is the difference between these three concepts? o Data is a single observation, and we have millions of single observations around; information is the contextualization of a d atum, so we are associating a meaning to a datum, so we are constructing a s ort of information. o To construct this information, we are using knowledge; knowledge are models (probably heuristic models) and guidelines that a re used to construct information form data. o Information and knowledge à information as an interpretation, while knowledge is a model o So, knowledge are guidelines that are used to infer a meaning from a datum. In other words, a datum is a single observational point that characterizes a relationship. It generally can be regarded as the value of a specific parameter for a particular objec t (e.g., a patient) at a given point in time. • Knowledge , then, is derived through the formal or informal analysis (or interpretation) of data. Thus, it includes the results of form al studies and also common - sense facts, assumptions, heuristics (strategic rules of thumb), and models — any of which may reflect the experience or biases of people who interpret the primary data. • The term information encompasses both organized data and knowledge, although data are not information until they have been organized in some way for analysis or display o A datum is a single observational point that characterizes a relationship. It generally can be regarded as the value of a specific parameter for a particular object (e.g., a patient) at a given point in time. 6 o Knowledge , then, is derived through the formal or informal analysis (or interpretation) of data. Thus, it includes the results of formal studies and also common sense facts, assumptions , heuristics (strategic rules of thumb), and models — any of which may reflect the experience or biases of people who interpret the primary data. o The term information encompasses both organized data and knowledge, although data are not information until they have been organized in some way for analysis or display. o A database is a coll ection of individual observations without any summarizing analysis. 2.4 What is a database: • A database is a collection of individual observations without any summarizing analysis. • We have not only database, but we also have knowledge base , (intelligence) which are collection of facts, of information associated to data. • It means that for sure we need a structure, but we also need a structure of the meaning of the data. • Knowledge bas e are the building blocks of systems that takes decisions automatically, because they have data and knowledge, so meaning associated to data, and according to this they are able to analyze and provide suggestions. • So, in this course, we take for granted t he existence of database to which we can associate semantic and meanings, and then we can create new decisions and new knowledge starting from this organized knowledge base. • In other words, a medical record is thus primarily viewed as a database — the place where patient data are stored. • A knowledge base, on the other hand, is a collection of facts, heuristics, and models that can be used for problem solving an d data analysis. If the knowledge base provides sufficient structure, including semantic links amo ng knowledge items, the computer itself may be able to apply that knowledge as an aid to case - based problem solving. Many decision - support systems have been called knowledge - based systems, reflecting this distinction between knowledge bases and databases. o A medical record is thus primarily viewed as a database — the place where patient data are stored. o A knowledge base , on the other hand, is a collection of facts, heuristics, and models that can be used for problem solving and data analysis. If the knowled ge base provides sufficient structure , including semantic links among knowledge items, the computer itself may be able to apply that knowledge as an aid to case - based problem solving . o Many decision - support systems have been called knowledge - based systems , reflecting this distinction between knowledge bases and databases. 2.5 Types of medical data: Some of the data types that we are using in healthcare are the following: • Narrative, textual data. • Numerical measurements. • Recorded signals. • Images. • Drawings. • Photographs. Each of them is associated to a standard, to an uncertainty and to all the mo difiers we were speaking about. Let us see them more in details. 2.5.1 Narrative data: it accounts for a large component of the information that is gathered in the care of patients: • The patient’s description of his / her present illness • Responses to focused qu estions from the physician • The patient’s social and family history 7 • The general review of systems that is part of most evaluations • The clinician’s report of physical examination findings Generally is gathered verbally and is recorded as text in the medical record. The nurses were filling in these diaries with not structu red text and probably, some information in thee diaries that were really important to drive diagnosis were never translated into s omething with a meaning that could be easily accessed by the doctors. This information is fundamental, it somehow helps to make transparent and understandable the situation. There were some attem pts to make these texts structured with standardize question aries, but it is not so easy, because every patient is specific, and the dialogue can evolve in different ways. It is not so easy to generalize, but at least it should be somehow made transparent in the analysis and interpretation. It is also a problem of legality. If something goes wrong, the electronic record is the first thing they are going to check. So, decisions should be taken tran sparently according to the electronic record. The following figure shows an example of narrative data: • No physician disputes the importance of such observations in decision making during patient assessment and manageme nt. The precise role of these data and the corresponding decision criteria are so poorly understood that it is difficult to record them in ways that convey their full meaning, even from one physician to another. Clinicians need to share descriptive informa tion with others. When they cannot interact directly with one another, they often turn to the chart or computer - based record for communication purposes. • In the past 15 years there have been text interpreters that are based on a national processing languag e. As we can see in the following figure, all the words are undergoing a grammar analysis, but also an analysis of meaning. This is done through medical terminologies, in order to translate text into medical “knowledge”. This is not so easy, many acronyms , abbreviations and negations are used. So, it is a process that is not so trivial. So, some narrative data are loosely coded with shorthand conventions (abbreviations) known to health personnel. Examples: Some narrative data are loosely coded with shorthand conventions (abbreviations) known to health personnel • “PERRLA” – Pupils are Equal (in size), Round, and Reactive to Light and Accommodation; “ICU” – intensive Care Unit; COPD – Chronic Obstructive Pulmo nary Disease. • Many are not standard and can have different meanings depending on the context in which they are used: • “MI” – “mitral insufficiency” (leakage in one of the heart’s valves) “myocardial infarction” (heart attack). • Many hospitals try to estab lish a set of “acceptable” abbreviations with meanings, but the enforcement of such standardization is often unsuccessful. 2.5.2 Recorded signals: o (es. ECG) In some fields of medicine, analog data in the form of continuous signals are particularly important. o Example: electrocardiogram (ECG), a tracing of the electrical activity from a patient’s heart. o When such data are stored in medical records, a written interpretation of its meaning is included. § We need to transform this record in something electronic an d associate it to its meaning. It is good to associate the record of a text, a report, to a specific portion of the signals that shows the problem/characteristic described. o This makes the interpretation easy, it also facilitate learning, and it avoids mis understanding when the same data is going to be read from different doctors. 8 o If we have a report, a written description of a specific portion of the signal, there is no misunderstanding, everybody read and see the same. So, there should always be a conne ction between the image of the exam, so the signal, and the text of the report, so its description. If we are working in this direction, we are giving a massive help in terms of transparency, and also in terms of optimizing the care. 2.5.3 Images : the following figure shows different kinds of images with associated information: 2.5.4 Drawings: • In so me electronic reports there are also images that doctors are drawing to let people understand, like shown in the following figure: • Visual images, either acquired from machines (photographs) or sketched by the physician (drawing), are another important category of data. It is common for physicians to draw simple pictures to represent abnormalities that they have observed; example: a sk etch is a concise way of conveying the location and size of a nodule in the prostate gland. 2.6 DATA RECORDING: • As should be clear from these examples, we are in a field in which we should understand what the best way is to associate a m eaning to these data. • The idea of data is inextricably bound to the idea of data recording. Physicians and other health care perso nnel are taught from the outset that it is crucial that they do not trust their memory when caring for patients. They must record their observations, as well as t he actions they have taken and the rationales for those actions, for later communication to th emselves and other people. • The people that collect and handle the data are a lot. Each care is multidisciplinary. Health data on patients and population s are gathered by a variety of health professionals: o Nurses often build relationships with patients th at uncover information and insights that contribute to proper diagnosis, to understanding of pertinent psychosocial issues, or to proper planning of therapy or discharge management. o Office staff and admissions personnel gather demographic and financial inf ormation. o Physical or respiratory therapists record the results of their treatments and often make suggestions for further management. o Laboratory personnel perform tests on biological samples, such as blood or urine, and record the results for later use b y physicians and nurses. o Radiology technicians perform X - ray examinations; radiologists interpret the resulting data and report their findings to the patients’ physicians. o Pharmacists may interview patients about their medications or about drug allergies and then monitor the patients’ use of prescribed drugs. The data are shared among all these people. • Another important thing are the technological devices to generate data, like laboratory instruments, imaging machines, monito ring equipment, measurements able to take a single reading (thermometer, spirometers, ECG...). These measurements can be shown on a display, printed, saved electronically, or can require an expert to interpret the output and provide a referral on it. • Data collection is relevant not only for the information system, but is should be relevant also for the instruments. If we are producin g a medical device, we should know how the data should be collected. For example, if the ECG should be structured in a way that is interoperable, there are rules and standards to collect these data and we should know that also if we produce the device to c ollect these data. • We have to do that also if we are producing an app, because now our healthcare is also done with mobile health (all the applications we may have in mobiles, tablets, etc.). For example, there exist a mobile device 9 that has a digital med ical device that advice the patient when it is close to have an epileptic crisis. This is something that starts form somethin g wearable, that should be validated and verified, and should be reliable. 2.7 The uses of medical data: • Given the increased complexity of modern health care, the broadly trained team of individuals who are involved in a patient’s care, and the need for multiple providers to access a patient’s data and to communicate effectively with one another through the chart, the paper record no longer adequately supports optimal care of individual patients. Another problem occurs because tradi tional paper - based data recording techniques have made clinical research across populations of patients extremely cumbersome. Computer - based record keeping offers major advantages in this regard à Electronic Health Record (EHR). • The main use of medical d ata is their collection for Electronic Health Records (EHR). Then, we also have the patient health record. EHR are knowledge based and they are basic for the care of a patient but also for clinical research. • There are two different types of use of data, w hich are the limited use and the extended use: o S peaking about the limited use of EHR, the doctor records patient information, so EHR, to access information according to whic h, together with other knowledge, decide to provide cares for patients. It is a so rt of loop. It is a use of EHR that provides care to the patient, but it is limited to the specific case. o The extended use adds the so - called secondary use of data, which consists of biomedical and clinical research and clinical trials. It allows to test if a treatment of a patient is better than the usual care. We should test if a treatment is better than wha t is normally done on patients, and we should also inform the national health system about that. So, we should register our trial and update the gu idelines. So, the extended use does not imply only the care of a single patient, but it also implies an exten ded use of data that can increase the knowledge, getting better results. o This means that research is at the base for a learning health care system in which we need to provide and create new protocol s, guidelines, and educational material. o So, to resume, it is not only the care of a single patient, but it is that care that is recorded transparently and that offers data to perfo rm research in a rigorous way that it is used to increase the knowledge around a certain pathology so that the single care is th en based on new knowledge that is more effective. o Only if we use interoperability and standard and rigorous protocols, we can use the extended use, otherwise we cannot provide good results that can be exploited. o Another use of data is that we can anticipate future problems. 2.7.1 Future problems • Screening population and prevent • Data are gathered in order to do prevention. • Data gathered routinely in the ongoing care of a patient may suggest that he/she is at high risk of developing a specific pro blem even though he or she may feel well and be without symptoms at present. Medical data therefore are important in screening for risk factors, following patients’ risk profiles over time, and providing a basis for specific patient education or preventive interventions, such as diet, medi cation, or exercise. • Perhaps the most common examples of such ongoing risk assessment in our soc iety are routine monitoring for excess weight, high blood pressure, and elevated serum cholesterol levels. In these cases, abnormal data may be predictive of later symptomatic disease ; optimal care requires early intervention before the complications have an opportunity to develop fully. • Another use of data is to record standard preventive measures. • The medical record also serves as a source of data on interventions that have been performed to prevent common or serious dis orders. • Other important preventi ve interventions include immunizations : (use data in order to prevent) the vaccinations that begin in early childhood and may continue throughout life, including special treatments administered when a person will be at particularly high risk ( e.g., inject ions of gamma globulin to protect people from hepatitis, administered before travel to areas where hepatitis is endemic). 10 o When a patient comes to his local hospital emergency room with a laceration, the physicians routinely check for an indication of when he most recently had a tetanus immunization. • When easily accessible in the record (or from the patient), such data can prevent unnecessary treatments (in this case, an in jection) that may be associated with risk or significant cost. If we are in the emergency room, it is also a question of time. So, the digitalization of this information really helps. • Another use of data is to identify deviations from expected trends, so the understanding of data in a trend. The following figure, for example shows the curves of growth of children: • Data are often useful in medical care only when viewed as part of a continuum over time. An example is the one shown in the figure above, so the routine monitoring of children for normal growth and development by pediatricians. These curves are used to understand if the growth is inside normality ranges or not. It is important to interpret our data in time, with respect to normality ranges. 2.7.2 Provide a legal record • Another use of medical data, once they are charted and analyzed, is as the foundation for a legal record to which the courts can refer. The medical record is a legal document; the responsible individual must sign most of the clinical information that is re corded. In addition, the chart generally should describe and justify both the presumed diagnosis for a patient and the choice of management. Data do not exi st in a generally useful form unless they are recorded. • The legal system stresses this point as wel l. Providers’ unsubstantiated memories of what they observed or why they took some action is of little value in the courtroom. The medical record is the foundation for determining whether proper care was delivered. Thus, a well - maintained record is a sourc e of protection for both patients and their physicians 2.7.3 Support clinical research • Research is at the base to build evidence and to build guidelines. Support clinical research through the aggregation and stat istical analysis of observations gathered from p opulations of patients • Randomized clinical trial (RCT): RCTs typically involve the random assignment of matched groups of patients to alternate trea tments when there is uncertainty about how best to manage the patients’ problem. The variables that might a ffect a patient’s course (e.g., age, gender, weight, coexisting medical problems) are measured and recorded. • RCT are typically double blind (i.e., neither the researchers nor the subjects know which treatment is being administered) an d multicentric. • Epid emiology: Medical knowledge also can be derived from the analysis of large patient data sets even when the patients were not specifically enrolled in an RCT. Much of the research in the field of epidemiology involves analysis of population - based data of th is type. Our knowledge of the risks associated with cigarette smoking, for example, is based on irrefutable statistics derived from la rge populations of individuals. 2.7.4 Decision making • Studies of medical decision makers have shown that strategies for data c ollection and interpretation are mainly based on an iterative process known as the hypothetico - deductive approach . It is a sort of iterative process, in which we make hypothesis, gather data to understand if the hypothesis are good or not and adjust the p lan. This is exactly how the patient care works. If the data are not enough, we need to collect additional data. It is like a loop: • The starting point is the patient that presents to the doctor with a problem. The doctor asks questions to identify the patie nt (identity, history of the patient, etc.) and makes initial hypothesis, which is called “prior probability”. Then, the doctor ask s additional questions, to refine the hypotheses. Heterogeneous exams can be done, and data can be collected to refine hypotheses. Then, these data have to be sele cted to do a diagnosis. • The diagnosis calls for a treatment. By observing results, we can un derstand if the patient is cure, or if we have to refine the hypotheses and the treatment. In general, it is an iterative process that requires data. 11 • In other words, the central idea is one of sequential, staged data collection, followed by data interpretation and the generation of hypotheses, leading to hypothesis - directed selection of the next most appropriate data to be collected. As data are collect ed at each stage, they are added to the growing database of observations and are used to reformulate or refine the active hypotheses. This process is iterated until one hypothesis reaches a threshold level of certainty (e.g., it is proved to be true, or at least the uncertainty is reduced to a satisfactory level). At that point, a management, disposition, or therapeutic decision can be made. Physicians refer to active hypotheses as the differential diagnosis. 3 TERMINOLOGIES: CLASSIFICATION AND ONTOLOGIES W h at is needed to produce the semantic interoperability in health care, in particular we’re giving an introduction to different kind of terminologies, so classifications and ontologies in particular, and then we’re going to speak more into the details about ... • Different heath care professioals contribuite collection and interpretation • Communication and patient data sharing among authotized health care professionals are centrals points in healthcare o Create a way that (models/meaning) to be sure that who is ready the element can understand everything (speak with an unique language) • Developing coded medical information is crucial to inperpret patient documentation o Associate the code with whatever concept in the medical domain à software can acts • Data should be available for epidemiological studies a nd research o Can do an epidemiologic studies in global way because we have a terminology and each person is associated with a specific epidemiology • Automatic interpretation from CT that is one of the most used terminologies in health care, so: • Computerized information systems in health care should use medical terminologies and classifications to manage clinical data 3.1 Terminology necessity • N eed for different health care professionals to contribute not only to data collection but also into interpretation. What is h appening now in the health care sys tem is that multiple people from different stuff is involved in the care of a single patient, so data sharing and communicati ons are crucial requirements for offering the best health care possible. This is not only a communication as we understood among au thorized health care systems because the system should be authorized to be part of this communication, but there’s a lot of concerns about da ta protection, privacy and integrity but not only speaking about technical interoperability that is ensured by stan dard like HL7, dicom standards but also standard that ensure the maintenance of meaning, the maintenance of the data but also the meaning of the data that are shared among the people. • The other need is the need for developers because there are developers t hat are coding this medical information and these codes are also used for interpretation not only for the final doctor but also for softwares so automatic interpretation of data. • We’re nowadays producing a lot of systems that are the so called clinical dec ision support systems, all these system are an help or a support for the medical doctors but should be built on the interpretation of data preserving their meaning, otherwise if the meaning is n ot somehow standardize is not interpretable by a machine, but is also true that every developer cannot start from the beginning every time, there should be a standard also in this terminologies, and on this and we’re a living in a situation in which the need is really actual, data should be also available for epidemi ological research, when we’re speaking about epidemiological studies we’re speaking, probably since 3 or 4 years ago we were speaking mostly about the health status of a nation, now we’re speaking about how to deal with a pandemic, because epidemiolo gical studies are done nowadays to do statics on how many people are killed by covid or suffering of this situation or other analysis to mo nitor and somehow create counter metrics for the actual situations, but in this case it is not anymore a problem of underst anding the meaning is just a problem of categorizing, for instance in Italy we have the SDO the discharge letter that is given by the hospital, in this letter there are a lot of information but there is just on field that is the reason for the discharge th at is used for founding and DRG analysis in the hospital, there’s just one field cause of that associated to each people that instead that is done for doing epidemiological research and for having all the n umbers that we 12 were looking at last year everyday (not anymore like that) this depends only on specific code, it is not you should reconstruct the full meaning but you have to retrieve the meaning full information with another aim. • So the need for the epidemiological because we should somehow manage medic al data or for analysis or for automatic meaning retrieval and this is why we are speaking about terminologies. • Which is the building block of building terminologies? 3.2 The clinical idea. The clinical idea are the essence of anything we want to record about the health of the person, whatever we want to store, or we want to know about a specific subject is a clinical idea. • So the clinical ideas are the building blocks for all electronical records that are full o f clinical ideas. All the words like otitis media, inflammation, ear infection, middle ear, adenoids, otitis externa ... are all clinical ideas around the concept of the clinical health of the idea, so for a single topic we have millions of different clini cal ideas. • There is a model of concept à clinical idea that allow also software to better understand what we want to explain • If we are just speaking about this little topic, we have clinical ideas of different part of the subject from different field. • Mi ddle ear is an anatomical part, while adenoidectomy is an intervention that is done, inflammation is a clinical status (it sa ys that there’s inflammation in a specific part). It is true that clinical ideas are building blocks to build a terminology, but th ese building blocks alone without any models that links these building blocks one to the other are useless. • Example: o You can anyway create a structure that compute the percentage of a specific area and understands the main sense probably than ks to machine learning techniques, but without a model you can’t reconstruct the real meaning. 3.3 Terminology Is the set of special words of phrases that are used in a particular field • A terminology is not a vocabulary (because a vocabulary is broader areas of words), a terminology is a special set of words u sed in a particular field. So, there’s in this case health care as field of ap plication of the terminology that we’re going to speak today. • Terminologies are also coded and motivated expression. o Where you are associating each concept to an identified to a code that allow retrieval. And terminology is not only based on concept (clini cal ideas) but also the relationship between concepts. • Now you can remind, if you heard about object oriented languages in programming, you can remind of evidence, evidence are wha t links to entity one to the other because is parent of the other, one is th e generalization of the other. • Based on concepts and associations between them • Terminologies are built according to specific rules or procedures (models) à we can see the same architecture of the class, class is the content and the relationship are like the linkage that are present in terminology • Example: o The nose is a part of the body. o Nose is child and body is the parent. o So something that can be associated with a relationship with an association. • But in terminologies there’s not only this kind of associations but there are other kind of associations, for example there are n ot just parent - child relations, so to build a terminology you need more kind of relations between concepts, to create a model for terminology is a bit broader. • Example: o We have four terms, and we want to retrieve automatically the meaning that is ‘’a chest pain is radiating to the back’’, Simona Ferrante, E-HealthMethodsClinicalideas•Clinical ideas are the essence of anything we want to record about the health of the person. They are the building blocks for EHR 13 o If we do not have a model but only a phrase it is difficult to understand what is the concept à better create a model in oder to clarify how the terminology and concept are created the sentence o if you have only these 4 terms without a model connecting an automatic software without experience in health care could easil y mislead the meaning to ‘’a back pain radiating to the chest’’ because in the end the construction of the sentence is exactly correct but it’s the meaning is wrong, for a doctor it’s the opposite, a doctor knows what’s the original sentences mean and he knows that is more likely that a chest pain is radiating to the back rather than the opposite, and the doctor do not make this mistake. • The idea is then providing the model, which is like: o You can define the pain location and the radiation area, and you can associate concept to that sp ecific attribute. So you’re creating a simple model to minimize the errors that an automatic software with any knowledge can easily occur. o Once we have defined terminology let’s do go more into the detail. 3.4 Concept • A concept is a specific clinical idea with a specific meaning and for sure we have simple concept that are defined by single terms and complex concept that needs more concepts to be defined. (unique clinical idea which has a specific meaningful within its scope of use) • Simple concepts can be defin ed thorugh a single term • To rappresente complex(not only a single terms, but it is something more) concepts 2 alternatives are usually followed o P recoordination à already explicit with an ID • Id for every concept • Creating a definition and assigning a unique identification code to single or aggregates of predefined concepts • A terminology is composed by the pre defined list of all the simple and the complex concepts of the domain of interest and their modification • A n example is a classification (class that include also something that is not discover jet) o Post coordination à we have only the single complex and the simple rules in order to create a concept • A code to anatomic concepts • These latter can then be combined to obtaine the concept aggregartes à perforated ulcer, where perforated if the modifier of the original simple concept • In this case aggregates of single concepts can be built when needed • Should allow post coordination on ly if there are rules in order to create rule concept à • When playing with complex concepts, these can be part of precoordinated vocabulary and post - coordindated vocabulary. • Example: o Endoscopic emergency intervention. o I have these 3 terms that forms a comp lex concept. o In a precoordinated vocabulary, the complex concept that is endoscopic emergency intervention, already exist and has a specif ic code. o So I have a series of complex terms, the whole complex concept existent are already coded and existent. o Inste ad, if I have, the same complex concept is a postcoordindated vocabulary, it could be the case in which I have the single sim ple concept, but I can create the complex concept. • So I have for instance: o The definition is: intervention o There is the priority: emergency o The modality: endoscopic o So I’m structuring a little bit the data in a way that is not necessary to know all the possible complex concept, because we are in a world progressing a lot. (Covid19 was new but could be implemented with the postcoordindated vocabulary). o An example of precoordinated technology is classification. • In precoordinated vocabulary is explicitly defined and associated to a code. o For sure we have a lot of univocal identifier code for each explicit concept and in the precoordinated vocabulary you will have all the simple and complex concept defined from the beginning. • In the postcoordindated vocabulary you will have a list of simple concepts (atomic concepts) and then you should combine in c oncept aggregates the co mplex concepts. 14 o This means that you can built new concept using building blocks. o A terminology like that is postcoordindated, for sure you must know how to do this, otherwise you’re creating something that is not consistent or you’re creating combination t hat share similar meaning, and this should be checked in all the terminologies that are postcoordindated. • What are the advantages and the disadvantages of both terminologies? 3.4.1 Precoordinated vocabulary: • Advantages o no ambiguity , as I said everything is expl icit, all the complex concepts are already explicit, so it means there’s the definitions and so there’s no ambiguity and also all the atomic concept there are meaningless are not in the vocabulary, there’s no risk to have meaningless concepts here. o All th ose combination of atomic concepts that are meaningless or not desirable are eliminated as the only permitted combinations of atomic concepts are the default ones o These are big advantages of this kind of terminologies. • disadvantages: o they are so specific and so explicit that are very extensive, this means they’re leaving huge difficulties to handle for you. o It is also impossible to classify new concept aggregates because if you need to create another concept you should somehow hav e al ready defined it, there’s no way to aggregate other concept. o In most of the case concepts are repeated in the vocabulary, just to add an additional meaning but without saying that it’s t he same. o Concepts are repeated many times In the vocabulary 3.4.2 P ostcoo rdination vocabularies • Advantages: o smaller then precoordinated (less identifier, less space occupied on the machine that has to use this vocabulary), they can a ggregate. o New concepts can be created aggregating basic or previously - generated concepts. • Disadvantages o are the rules how to aggregate. In these two slides there is a summary of what we have spoken on today. 3.5 Basic uses of terminologies in general • A bstraction and classification : o when you are, for instance, programming a game in objected oriented languages you’re creating classis, class is an abstractio n of reality because for instance you need a class like ‘’student’’ you’re not describes all the cha racteristic of the students but you’re extracting the student explicitating only the field you need according to the field and according to the scope, because the field student ca n have completely different structure. (abstraction of reality) o An abstracti on is an examination of the recorded data and then selection of items from a terminology with which to label the data. o It’s an abstraction of reality that is linked to the meaning that (slide example: in this case is billing) and according to t his we have one way of putting the data all together, I mean I have a sort of one - dimensional hierarchy of these data, I’m extracting in one dimension and according to this dimension I’m classifying the data. • The representation o is a process a way In which I need mor e details, so I want to create a model that well represent all the terminology, for instance if I create a classification only with IA relationship this IA it is an example of the one dimensional I was mentioning. o Is more close to the ontologies à reality o This means that I cannot represent the whole meanings of the terminology but I’m focusing on this specific dimension and I’m following only this to create a terminology, instead in the representation I try to be more detailed. I’m going to go towards the complete model of meaning. 15 o Another example to be clearer about the structure, if I’m interested in billing and I know that the only scope of my terminol ogy is billing, I don’t need to describe the patient case in a lot of details I need only the diagnosis of his discharge, no more than this. It’s a one way to abstract the patient data, I hope that this example was clearer than the dimension I was doing before. I need just to know th e primary disease that is putted in the discharge letter, this is the only thing I need about the patient, and I need the one for the discharge letter not of the admission, because at the admission I can have diagnosis but then I can do a lot of test and I can refine my diagn osis, the billing is computed according to the discharge diagnosis. • So if I have this need there could be a terminology that helps me to classify all the diagnosis. 3.5.1 Classification purpose and properties • The purpose is the subdivision of phenomenon into classes that are the basis for ordering things. S o it’s an abstraction in this case. • And the properties are: o univocity of concepts and identifier (each concept has its specific unic identifier) à each line is a concept and has one identifier (ID) o being omni comprehensive, that means a priori explicit and precoordinated (they should comprend already all the definition that I need) (it is precoordinated this means that you are listing there also all the complex concept that you want to recreate with all t he detail, so you can answer 10 words after the other to explain better) o classes should be mutually exclusive, I cannot define the same concept with two different things o exhaustive, all the knowledge should be there (not only you have the complexity explicit but also you should be exhaustive wi th respect to the general meaning, I can classify all the virus with a specific origin but also all the other virus without any origins, so that you’re exhaustive because you cannot encounter any other virus that doesn’t belong to any of this classis) • So the example b efore there could be a new virus not present per categories, that includes all the things that was not expected when the classification was created. • The main use of this classification are statistic, epidemiology, audit, planning, financial billing. 3.5.2 Ontolo gy • C lassification is an abstraction O ntology is a representation of reality and it is considered a complete description of a domain of interest by concept and relationship between concepts, but in this case we have not only the hierarchical relationships but we have also all other non hierarchical relati onships such as ‘’it is ca used by’’ ‘’ it is found by’’ and so on. • Hierarchi cal or not rierarchical (association with concept) • The other important thing of ontologies is that they are able to allow semantic interoperability maintaining the meaning, for instance if I’m saying the co ncept student, this concept includes in ontology not only the term student but also the term alumni that is a synonym of that concept but shares with that concept the same meaning, so another example would be myocardial infarction and heart attack, th ese t wo terms share exactly the same meaning, so are inside the single concept, I don’t need two different concepts, because the meaning is the same. • We can also use sinonimous of the same concepts à to retrive the same meaning • We’re going to present two example of ontology that are snomed CT today and umls (universal medical languages system), they both are example of ontologies. 3.6 International classification of disease (ICD) • The international classification of disease (ICD) thi s is exactly what is used by the world (also Italy), to do epidemiological studies. • It is used by health care professional to classify diseas and other healh problems recorded on many types of health and vital records, including death certificates ad deat h records • For instance the ICD was developed by WHO, so was developed at global level, it is used to classify disease as the type but also to classify causes, and also to record health problems in a state in order Simona Ferrante, E-HealthMethodsTre n d s in c a u s e-of deathreporting by ICD revision 16 to decide the quantity of magnitude given to a specific area. So now, more than 100 countries are using the ICD to report mortality, so it’s a unic classification system that is used world wide and it means that is translated also worldwide. • About