|
|
|
Index - Major Sections
Site Map
Product and Services _______________ Index - Same Level Subject
Index - Child Subjects |
Data The USA Institute of Medicine once proposed as essential components of the process of health information the following:
What seems to have been left out is "Usability" While it has been said that the Goal of Information Systems is to consolidate "Silos" of information into useable consolidated database management systems... the fact is...there is no data that can be used no matter how many Silos are combined. Problems There is insufficient data collect to profile the client. There are insufficient variables collected into these silos to be able to evaluate anything. The data that is collected are:
Households/Families are not treated as a group. Almost 75% of Hospitals do not collect information on cultural data (CITATION JCAHO2002 \l 1033 Amy Wilson-Stronks, 2002) Interesting, there is no grouping of individuals in a family/household. Each individual that comes into the healthcare organizations is treated "separately." It makes no difference if everyone else in the household has the same symptoms....that information is not record. Household information is not available.... In less developed countries, individuals may become sick for social and economic reasons. The water/air may be contaminated. There may be no sanitary facilities or the food that is being eaten is contaminated. It is only be chance that any of these problems are recognized. Diagnoses have been reported to be incorrect. In a recent Healthcare conference, the Speaker, the CEO of one of the largest Healthcare organizations in the US...stated that based on their research over 67% of the diagnosis were incorrect. There are also many types of diagnosis such as Primary, Working, Rule Out, Post surgical to name a few. So...which one do we use for analysis. HCP's often do not fully understand the difference between a "diagnosis" and a "symptom". If the "problem" has not been positively identified, then the problem is a symptom and not the diagnosis. Guessing what the diagnosis is can completely mislead any other HCP when reading the patient's record or in treating the patient. Symptoms are not recorded for analysis. It appears that no symptoms are collected or evaluated in order to determine if in fact the diagnosis is correct. The only social data that is collected is birthdates, gender, person responsible for payments, and maybe, where the individual is working. Interestingly, there is no data collected on Families units. The reason that much of this data is stored in silos...is because no one knew what to do with it. Just because you combine the data from all these silos doesn't mean that it is the Data can be used for analysis. Data is summarized (Summary notes). Data is often summarized losing much of the information that the data originally contained. In is the practice that only the "Physicians summary notes" are passed on from Healthcare Professional to Healthcare Professional. What if the particular physicians observations are incorrect? What if he made the wrong diagnoses? What if needed detail information is missing? Summary notes are unsafe and dangerous to the treatment of the individual. Data is not recorded in a manner that can be used for research. Based on the data available, other than very elementary statistics, no research (it will still have to be performed the old fashion way of hunting thought patient notes) can be preformed from this existing data. Standardized data sets are poorly designed and where originally designed years ago for paper based systems. Many of the Standardized data sets have terms that are not in the same domain....mixing diagnoses, problems, symptoms, synonyms in the same data set Data Collection. What data needs to be collected requires that you know how and what you are going to do with the data. Many healthcare professionals have no training in "how to use data." In many cases, data is "not collected" for the simple reason that the healthcare professional says they do not "have the time." This view is an egocentric view where the Healthcare Professional is only thinking of himself. There are a lot more "stakeholders" in the game other than the one Healthcare Professional. Healthcare Professionals may also believe that if data is collected, this data will be used "against" them. More and better data, helps clients participate more effectively in their own care. Data can be used for transparency and accountability. Data Storage. Many so called professionals are still trying to push for a distributed data storage. Real time Data Analysis and Data Mining requires that the data be "Extracted" from the Cleaned", Data Ownership. Data is valuable if you know what to do with it and no one wants to give up their data freely...no matter how many lives it can save. The privatization of healthcare practices is the worst thing that could have happened to healthcare. It is suspected that it was the healthcare organizations themselves that implemented the "privacy laws"...why?...in order to protect their business data. As long as this is the practice, improvement in healthcare will be very slow. InHCc once had a very large healthcare organization tell them that they would never give up their data and if required to do so by law, that they would make the data "unusable"! Data Transfer It has been often promulgated that data must be in a single standard electronic format for all transactions. The only requirement is that the "Database" be in a standard format. The use of tools such as Microsoft's Integrated Services can easily transform any file (text, sql, work, excel) to any other file and reformat any data fields in the record. Data Use Data should be design such that it is collected one and used many times. This means that the original data must be captured in a quality and with enough variable to be able to be used for communication, analysis, data mining, and research. Comparison of data sets There are many studies to see how many terms match up with a different set....This does not work if both the standard sets are "bad sets". Code Sets The ICD-10 codes OPCS-4 are statistical classifications. "These do not have the comprehensiveness, depth or flexibility needed to apply to medical records that are used by clinicians in every day practice" (NHS Clinicians-Guide Part 1, Connecting for Health, October 2008). The US government appears to be making mistakes that will forever keep the healthcare system from being able to do anything about their healthcare costs. In a recent case, the FDA adopted the Veteran Administration and Kaiser Permanente (VA/KP) Problem List Subset of SNOMED as the terminology to represent indications in electronic labels. While a number of evaluations were performed [REFERENCES] reporting that the VA/KP Subset has significant limitations for coding drug indications…the FDA continued to peruse the standardization. According to Humphreys et al adoption of standards by end users requires
It seems that these reasons have become fairly standard for using any particular code set. Let’s examine each proposition.
Atomic Data Sets The problem has been that current data sets were developed when the purpose of coding was to get as much information as possible into the code… An Atomic data set consists of a Lead concept and one or more qualifiers. This is impossible in most of the standards existing today (ICD10, SNOMED) One term having “sufficient coverage” and “precision” is extremely difficult and creates a lot of maintenance on the part of the standard keepers. Example:
Demographics Demographics "...are personal data elements, sufficient to identify the patient, collected from the patient or patient representative and not related to health status or services provided. Some of these elements may require updating at each encounter..." ¨(ASTM E 1384-07) Interestingly, Demographics probably determinate health status more than any other factor (Still being debated) Gender, Age, where you live, languages you speak, education, and occupation are all factors that are a major determinate of your health. It seems that most organizations, including very important standards organizations have missed this completely. The study and research into these relationships are important. Businesses have long understand the power of profiling and the effect that this have on marketing. Instead of implementing a country wide (expensive) no-smoking campaign, it is more effective to profile and target groups of individuals (unless of course, you have unlimited funding!). Why hasn't healthcare used processes that have been used for years in business...Is the Management of Healthcare different than the Management of a business?...as one Director of an International NGO once told me...[and it was also one of the most poorly run organizations I have ever seen or read about]. The determination of which of these factors, over time, can affect the health status of a population, is one of the most important studies that healthcare professionals can perform. "Over time" is also a key ingredient of the study of these relationships. Analysis of these relationships cannot occur unless demographic data on individuals are kept over time. This means that demographic data must be archived and not "updated". When ever demographic data changes on an individual, that date must be saved so it can related back to the health conditions of the client at the time. Does, over time, the increase in the education of the individual produce better health?
Data that "may" be collected but is not in a form that can be used in Data
Mining
Without a sufficiency of variables, there is no way to discover "new" relationships. Using the same "over-worked" data, produces the same results. Exchange of Information First we need to discuss the often quoted practice of "telling the story in the patients own words." While this sounds great, it is not very practical. If the emphasis is placed on the exchange of information, then "Telling the story in the patients own words" may lead to misunderstandings and put the patient at risk. Here is why. Interpretations of meaning is related to the location, who is speaking to whom, their education, their social background, their previous experiences and many other factors. If this "story" is read later from a electronic message, can we be sure that the correct interpretation is made? Likely not. Rather it is more important that the Healthcare Professional that hears this "story" translates this story so other healthcare professionals can understand the problem. As an example, there are many synonyms for the common cold...("common"...now just what does that mean?) and in most cases the patient may only descript the symptoms. It is the healthcare professional task to description these terms correctly (and I think the most important task that they have!) so that others may understand the patient's condition. "Flowery" phases and "their own description" has no place in medical care. There may, and other is, more than one stakeholder in the healthcare process. Yes, the healthcare professional examining the patient may need to have "empathy" with the patient, but more importantly, this healthcare professional must be able to translated the medical data to others. This author has worked in an emergency room where the major part of the population were poor black individuals from the "projects." The physicians that treated them were usually from upper middle class white families (that is changing). They were completely incapable of communicating with each other but the patient still received the best care that I have ever seen from any healthcare organization. Health information exchanges (HIEs) are being developed to allow healthcare data to be shared across organizations within a region, or the community. Really now...is this going to happen with the privacy of information laws that now exist and the competition among private healthcare organizations. And just what is going to be shared?...Is data actually going to be transferred "between" these organizations or is the only purpose in having this data is to be able to "view" the individual's health information? It is not about the Technology...it is about "Why". Think about it! It is not about "sharing of information"...It is not about everyone having everyone else's data. That is NOT going to happen, and there is no reason for it to happen... Except in the processing of data of procedures performed outside the individual healthcare organization, most data is not required to be "exchanged" but rather only viewed. Unless the organization actually need the physical data in their database, then there is no reason to have to "exchange" data. Viewing of the health information of an individual can easily be provided between organizations through an Internet Portal such as that being created by many healthcare organizations over the Internet or through private Portal companies such as Microsoft's "Health Vault"... The questions to ask...Is why would you want to exchange information???
Requirements WHAT IS NEEDED...is a centralized database where information can be used for research and presented to interested organizations. THIS IS THE MAJOR REASON FOR DATA EXCHANGE. This cannot happen in a distributed system where each organization maintains their own data. Google and Microsoft is already producing impressive research with the limited data that is uploaded to them. These private companies use software for data collection and storage from disparate systems into a centralized database. Once the data is stored it is make available to interested healthcare organizations through a Internet Portal. Additional, research against this database can easily be performed and results made available to interested stakeholders. This data is made accessible over the Internet and healthcare professionals can view this information as if it was in their local database by simply clicking an internet link....the Userl (and the Internal IT department) are not able to distinguish the difference...there is no need to "exchange data" and no reason to have single "individual applications". Additional benefits of the centralized database:
A distributed system where computers "exchange" information cannot do this! Organizations that must exchange information already have good systems and do not need the elaborate structure that standard organizations are trying to develop. While they must have means to communication between systems, they also can contribute data to the Centralized Database. Here are the few examples of these types of data:
Standards Coding Systems is one of the greatest problems in Healthcare. In seems that the Standards Organizations of today are either after money, the control, run by physicians, or just haven't a clue to what they are doing...mostly all of the above. This for me is a very lengthy subject and this section will only be an introduction. Standards are a problem. Too many standard organizations all trying to sell their products. It wouldn't be so bad if the data sets were any good, however, Many of these organizations are trying to adapter their old data sets to computer use...it just don't work. Examples: ICD10CM is a joke (it was developed for mortality reporting not as a point of care standard). SNOMED is not a point of care code set. An example of a badly designed HL7 coding system is shown below:
First, if we need an International coding system, the mnemonic of the Codes have no value. Second, the Code is alphanumeric and requires more storage and more CPU power to process. third, with an automatically generated code value from "Lookup tables", there is no need for the check digit, which again, increases the storage space requirements of data and processing time ... you would expect them to know better! Shame on HL7. The real problem is that the standard organizations in order to make it as easily as possible to convert are doing nothing more than "dressing up" (and in some cases nothing) their old codes...As an example SNOMED CT is a combination of two very old paper based standards that in themselves where not developed with HIT systems in mind. It seems that if any one had a new "concept" they would just throw it in the pot. SNOMED was formed in January 2002 by combining SNOMED RT, created by the College of American Pathologists , and the Read Codes of the UK National Health Service. While It benefits are called (by some) "the most comprehensive clinical vocabulary available in any language" it is not useful for a HIS system. Have the most comprehensive vocabulary is not the same thing as being well designed (which it isn't). The Read Codes were introduced in the early 1980`s. HL7 was at least created for the hospital information system, but again, it was founded in 1987 and they are still trying to use the same codes... ASTM E1384-07 "Standard Practice for Content and Structure of the Electronic Record" fails because it incorporates the same coding systems that we have already discussed. It states that "The first part identifies items of information carried in the traditional paper record organized by the source oriented structures common to paper records....[and this is suppose to be designed for the "Electronic Record"...]
Health Problem classification The classification of diseases have been a complete failure in most cases. It has been estimated by one CEO of the largest healthcare organization in the US that over sixty percent of the diagnosis made by physicians are wrong. In one study in Hermosillo, Mexico only 27 percent of the death certification diagnosis were correct. Given this data, we can safety conclude that very few of the statistics we get today are valid. By requiring the physician to make a diagnosis (ICD10) you are forcing him to put "something in the blank space" even if he is not sure. Other than using a valid laboratory test or diagnosis imaging, it is extremely difficult to make a diagnosis. Once the diagnosis is recorded all the "objective" data becomes lost. There is a move in the US away from listing diagnosis to the procedure of listing "Problems". Now these problems can be a diagnosis based on the ICD10 codes but only if the Problem is verified. This we believe is a step in the right direction to improve healthcare. By analyzing the problems (symptoms, findings, and laboratory results) the most likely diagnosis can be determine. If abnormal symptoms/problems appear, then instead of the physician filling in the blank with "Diseases not classified elsewhere" or worse...guessing, new emerging diseases can be recognized sooner. What is really bad, is that many healthcare organizations uses "coders" in the back room to fill in the blank because the "the physician does not have the time to look up the data." The question is...how does the physician know what treatment to perform if he doesn't know what the diagnosis is? And then there is the great feat that the coders must do of reading the physicians notes! Let the healthcare professional list the problems of the patient...if the diagnosis can be made, then well and good...otherwise leave the blank blank. Health Care terms (Entities/Attibutes) A description of a term may have different definitions from one screen input to the other, depending on the context. However, when the term means the same as the term "reflexes", that term will mean the same in a normal examination or in a specialized department such as neurology. A "reflex" is a reflex but the detail of the description may be different and more complete in neurology than for a regular examination. The same goes for Vital Signs. As an example the Blood Pressure may be recorded as 130/80 in one screen. But in another another area where the blood pressure is more important it may be recorded as:
and the Cardiologist will record the "heat rate" in greater detail, such as rate, Location of maximum impulse, murmurs, strength, rhythm, etc. All healthcare terms depend on the Entity/Class of the Concept and the attributes for that Class. As an example:
Symptom is the class/entity. Each of the other terms are the attributes of the Symptom Class. The type, start date, frequency, pain level are attributes of the symptom....the attributes and related to this class and no other. The attributes are created as attribute: value pairs. This structure allows easy communication among care givers, analysis and data mining. Data can be analyzed by type of Symptom, location, frequency, pain level, etc. This data can then be compared to the Symptoms related to a diagnosis (and the probability of each of the symptoms occurring in that diagnosis)...and then compared with that of the patient to see if the diagnosis match with the symptoms. Data Mining can be used to discover new "problems"....i.e. symptoms that do not exist as a combination in any known diagnosis. In the above example, the value "upper" is a modifier of the attribute. It is also a separate field. Now all Symptoms can be analyzed, as an example, by gender, age, previous problems, household/family group, date of occurrence, travel, sexual practices, etc. We can select all symptoms of a certain type, then select by the age and gender of the individual...and so on. This structure allows:
Specialization Data is not duplicated. The same term is used in a normal examination as in cardiology as an example. when a user views the data of an individual they see the "Heart Rate" no matter where it was taken or who took it. While ever effort has been made to include all specialist terms, there has been no effort to "separate" these terms from any other. Specialist terns will be displayed based on the "screens" that they use for their module. Data Transfer Data is often transferred in medical systems between "different parts of the system" by creating messages. In an Integrated system, the application communication directly with the database. Creating "messages" internally is time consuming and very inefficient. Messaging is used in the Internal workings of the computer (networking) but does not have to be used for the normal management of the healthcare organization. Every Input and View comes directly from the database. Every department and individual in the Organization sees exactly the same data and in "real time" Current relational database systems (Microsoft, Oracle) allows the user to import almost any type of data into the Relational database (text, xml, other SQL versions, Microsoft Access and Excel, common delimitated). Coding Every Class, attribute, and/or modifier has a Unique Identifier assigned to its value. There is no requirement that any Healthcare Profession codes...in fact, they do not ever see any codes. Research It appears from reading vendors brochures that the only use of data is communicating between healthcare professionals.. the most import use of data is data analysis and data mining. Communication...you can use a telephone or fax. |
|
|