A Look at Clinical Decision Support Systems

The United States healthcare industry has three major simultaneous developments in knowledge management and the use of information technology (IT). The first is in data collection, i.e., the adoption and “meaningful use” of electronic health records (EHRs). Second is data sharing as exemplified by health information exchanges (HIEs). Finally, there is data analysis as seen in enterprise data warehouse (EDW) and clinical decision support system (CDSS) tools.A CDSS supports clinicians in making evidence-based decisions and diagnoses by providing scenario-pertinent information from patient data in the EHR system. A CDSS can aid the physician in asking questions that are specific to a patient's health, helping avoid errors of omission. A CDSS helps to better utilize the data and intelligence in an EHR system. A CDSS that is configured properly can help to avoid adverse events—for example, by predicting possible harmful drug interactions.With the continuing push toward implementing EHR systems, more CDSS use is sure to follow—some say to the ultimate extent in which a patient will be dealing at home with a super-duper CDSS in the form of a holographic doctor. No waiting! Your virtual doctor would possess all the collated experience and knowledge about your health, prescribing the right treatments and medications. Wait. Didn't they do that already on one of the “Star Trek” TV shows? Although I think that the “Star Trek” holographic doctor was confined to a sick bay, it was always ready to serve. Maybe in the future, modern houses will include a sick bay and virtual medical staff. Future child: “I don't feel so good. I should stay home from school today.” Future parent: “Well why don't you go and log into Sick Bay and see what the doctor says?”While we're not there yet, maybe we are on the path to a home sick bay staffed with holograms. An estimated 41% of hospitals that have an EHR system also have a CDSS in place.There are two general types of CDSSs. One is knowledge-based, much like what I've just described. In essence, such a CDSS is a huge scannable database with a clinical case-based reasoning system. The other type is more of a neural network computing effort. Also known as artificial intelligence, this second type of CDSS involves an iterative “trial and error” learning process similar to one of the ways that humans think and learn.A knowledge-based CDSS uses compiled clinical knowledge to provide expert-level consultation to the clinician for diagnosis and medication selection. Such systems also can have many features, from general medical treatment references to precise information for specific conditions. The suggestions from the CDSS can target a patient's unique clinical and physiological data sets.Data mining may be conducted by the CDSS to examine the patient's medical history in conjunction with relevant clinical research—such as in a succession of IFTHEN queries. This kind of evidence-based analysis also can help predict potential events, such as the previously mentioned drug interactions.An example of the non-knowledge-based approach is a CDSS that uses the Archimedes model in which a series of equations analyze the EHR physiological and clinical data. The results are then fed into a computer model that simulates real healthcare processes and human physiology. This type of CDSS also could pull data from places such as a disease registry and then combine it with data from the EHR system to identify at-risk levels and suggest recommended treatments.EHR adoption is driven mainly by Title XIII of the American Recovery and Reinvestment Act (ARRA) and is expected to continue to accelerate. The Centers for Medicare & Medicaid Services (CMS) published its final rule in 2010 for the EHR Incentive Program. The program sets out requirements for the meaningful use of EHRs as a condition of the inventive payments. According to the Federal Register, “The HITECH Act [enacted as part of ARRA] statutorily requires the use of health information technology in improving the quality of care, reducing medical errors, reducing health disparities, increasing prevention, and improving the continuity of care among health settings.”Also adding to the quantity and quality of data will be ICD-10 when it goes into effect in October 2014. ICD-10 is the 10th revision of the International Statistical Classification of Diseases and Related Health Problems. It will increase the number of codes for identifying diagnoses and procedures nearly tenfold from 17,000 to 155,000, allowing for better classifications. ICD-10 will facilitate the matching of diagnoses with symptoms by increasing the data granularity, in turn producing a higher volume of new data for analysis and evaluation by the CDSS. This could lead to better pattern recognition and outcomes.When there are several disconnected databases and data stores, there is a danger of obtaining only some of the pertinent clinical and health information. Without a process to drive standardization and normalization, pulling data from various disconnected systems will not connect all the dots. Humans will need to be employed to scrub, sort, filter, and combine data manually to get to the same level as what could be achieved with a standardized process. Such a process can get time consuming and costly as it proceeds report by report.Big data as a concept has been around in other industries for a long time. In the IT arena, the term describes a very large data set beyond the ability of common software tools to manage and process within a tolerable elapsed time. Such a data set would be in the range of terabytes (1 terabyte = 1000 gigabytes) and petabytes (1 petabyte = 1000 terabytes), and big data is growing. In light of federally mandated and incentivized EHR adoption and expected changes in ICD-10, big data has become a hot topic in healthcare in the past few years.Adding to the flood of health data is the rise of mHealth and the growth of wellness apps. CDSS promises to make use of all that data on the order of $450 billion in reduced healthcare spending according to a April 2013 McKinsey Report—or about 12 to 17% of the current $2.6 trillion in healthcare costs. To say that this is a huge deal is putting it rather mildly.To help consolidate and control your torrents of data, consider an EDW device. An EDW is a server designed for data storage, access, and use. To get an idea of what a EDW looks like, consider the following specifications for a current top-of-line system: a tower with the dimensions of H: 6½′ × W: 2′ × D: 4′ and weighing in at 1,650 pounds, fully loaded. While it needs 6000 watts of power, it can dynamically support online, real-time use of its 60 petabytes of storage. Now we're talking serious data storage! It uses a concept of data temperature in which highly used data is hot data and tends to reside in the faster solid state drive (SSD). Hard disk drives, on the other hand, have access rates that are 16 to 22 times slower and would store cold to warm data as it's not accessed as much. These systems are also highly fault tolerant and include redundant hardware so that hardware or disk failures are managed seamlessly—transparent to the user. Although that may be an extreme example of an EDW, it is one that points to the future.Physicians, nurses, and other healthcare professionals can use a CDSS to prepare and review a diagnosis as a means of improving patient outcomes. A CDSS needs to be the focal point if meaningful use of EHRs is to realize its potential of better healthcare at lower costs. To achieve this goal, a CDSS needs to deliver relevant medical practice information at the point of care. In some cases, the operational costs of a fully implented CDSS and EDW sytem can be a barrier to adoption. One report indicated that some 74% of those with a CDSS said that the financial viability continues to be a struggle. While that represents a costly learning curve, the expected payoff in healthcare efficiencies and better outcomes would make it worthwhile. All leading indicators and current healthcare industry efforts seem to be pointing in the direction of making use of all the health and clinical data we've been collecting via EHRs. Maybe we're not that far away from the home sick bay idea. Future parent: “Doctor, what's wrong with my child?” Dr. Hologram: “Nothing at all ma'am, all bio indicators and CDSS reports indicate a clean bill of health with no current maladies. Send him to school!”