In mathematics and computer science, an algorithm is a step-by-step procedure for solving a problem. Algorithms are used for calculation, data processing, and automated reasoning. Expressed as a finite list of well-defined instructions and starting from an initial state and initial “input,” an algorithm, when executed, proceeds through a finite number of well-defined successive states, eventually producing an “output” and terminating at a final end state.
Algorithms have been used to develop, describe, and present logical processes of patient care. Since the early 1990s, processes of care designed as computer algorithms have been used to direct the care of critically ill patients. To be logical, a process of care needs to be described as a sequence of measurements, observations, or decisions. To be “computerized,” the process of care must be explicit and comprehensive.
For practical use in clinical medicine, algorithms have been a logical set of rules that precisely defines a sequence of decisions and specifies interventions. Commonly, the term protocol is used to refer to a specific clinical process of care, and a protocol may incorporate multiple algorithms. The concept of computerized algorithms to guide bedside clinical care has been developed and used for critically ill patients. Implemented as computerized protocols, their use has been associated with improved care relative to contemporaneous clinical standards.
Computer technology is now an integral part of the U.S. health-care system. Electronic medical record (EMR) technology is used to document patient measurements and interventions; to record clinician diagnoses, interpretations, and bedside presence; and to associate these data with established codes for billing, diagnosis, and treatment. Direct use by physicians is now mandatory, and government and commercial insurance agencies require computerized record submissions for reimbursement. How well the use of computerized medical record technology improves the efficacy of medical care and the efficiency of its delivery has been a poorly analyzed and unreported aspect of EMR technology.
Apart from the advent of the EMR, computerized medical protocol technology has evolved during the past 25 years. The term computerized protocol implies computer technology providing information to guide patient-specific care at bedside in real time. The problem of variability of patient care among clinicians derives from individualized, subjective decision making in complex clinical circumstances. Computerized protocol technology has been successfully used to implement protocols for complex processes that require standardized decision making to decrease variability in the care of critically ill patients and offers a powerful method for implementing a broad range of evidence-based guidelines. This technology is a desirable option for the intensive care community to establish and maintain the intensivist’s essential role in specifying and implementing best practices.
Computer Technology to Advise Physicians: Medical Informatics and Decision Support
Fifty years ago, Ledley and Lusted hypothesized that medical reasoning could be mathematically modeled. In 1964, the National Library of Medicine created the Medical Literature Analysis and Retrieval System (MEDLARS), and in 1971, Medical Literature Analysis and Retrieval System Online (Medline) was initiated. The National Library of Medicine developed searchable online libraries containing reference information, and the Unified Medical Language System (UMLS) research and development program, initiated in 1986, continues to provide national and international vocabularies and classifications. Since the 1970s, development of government and, more recently, commercial systems has led to computer technologies that provide information to advise physicians. These include online search systems, systems to provide diagnostic assistance, clinical data interpretation, and expert systems to guide patient-specific care. Information search and reference systems are now ubiquitous by means of the Internet.
Development of computer, communication, and network technology and its use in medicine during the 1970s initiated the development of medical informatics as an academic discipline and enabled optimization of acquisition, storage, retrieval, and application of medical information. Medical informatics incorporates computer science, clinical guidelines, medical terminologies, and information and communication systems with an overall goal of promoting patient care that is safe, effective, equitable, efficient, timely, and individualized. Examples of medical informatics include expert systems such as Mycin, a rule based yes/no query system to diagnose bacterial infections and recommend drug therapy, and Internist-I, a ranking algorithm system to diagnose disease; MUMPS (Massachusetts General Hospital Utility Multi Programming System), ( http://en.wikipedia.org/wiki/MUMPS , now also known as M ), a commonly used language specification and programming language for clinical applications that is the basis of the largest enterprise-wide EMR, VistA (Veterans Health Information Systems and Technology Architecture) and its graphic user interface, CPRS (Computerized Patient Record System), which enable health-care providers to review and update a Veterans Administration (VA) patient’s electronic medical record at any VA facility ( http://en.wikipedia.org/wiki/VistA ); and LDS Hospital’s (Intermountain Healthcare Corp, Salt Lake City, Utah) HELP (Health Evaluation through Logical Processing) system, one of the first EMR systems, designed to assist clinician decision making and operational for nearly 40 years.
Decision support tools to advise physicians are not new. For decades, physicians have used pocket editions of texts, antibiotic therapy guides, diagnostic algorithms, and protocol handbooks at the point of care (point of decision making). Computerized decision support tools, such as computerized algorithms and protocols, provide new attributes that include bedside application, incorporation of sufficient detail to be explicit, and reproducible electronic acquisition and storage of time- stamped patient measurements to permit identification of temporal changes, consistency, and reproducibility for use in algorithm logic. When explicit computerized algorithms or protocols are driven by patient measurements, the protocol output (instructions) is patient specific, at once providing individualized treatment and standardizing clinical decisions. This nonintuitive property has proved desirable among clinicians and has improved patient care outcomes.
Currently, the principal role for computer technology in medicine is to record and recall data, including in-hospital patient data, patient-specific medical care payment data, and non–patient-specific publications of medical science and clinical experience. Key to the advance of medical informatics and computer technology is widely available, clinically up-to-date computerized algorithms that can be used to effect immediate decisions for immediate care.
Computerized Algorithms as Models of Medical Reasoning
The concept of models of medical reasoning advanced 50 years ago has been successfully demonstrated in many computerized protocols implemented during the past 25 years. The algorithms to specify clinical care processes were developed by local working groups of clinicians and informaticists to ensure safe, optimized care and to decrease variability of care (see Fig. 25-1 ). The rule-based expert system, comprising explicit rules that facilitate decision making in a logical, workflow-compatible sequence, has been used most extensively to model complex care processes. Rule-based expert systems have been used with long-term success. At patient bedside (point of care; point of decision making), specific, clinically current measurements are used to derive clinical decisions in a sequence that directs incremental interventions as needed to obtain and maintain a specific, measureable effect. This type of expert system has been applicable to many aspects of intensive care. Computerized protocols comprising multiple algorithms driven by readily available, repeatable measurements, which were originally devised and refined by LDS Hospital clinicians and informacists. This technology has been adopted by others to effectively standardize clinical decision making at bedside and provide timely, patient-specific intervention for selected aspects of critical care.
Computerized Algorithms and Protocols: Clinical Experience
The first computerized algorithms and protocol systems to guide complex processes of intensive care were developed in the late 1980s. Computerized protocol technology has since been used to implement bedside protocols to direct care processes for durations of hours to weeks. Protocols are developed by multidisciplinary groups incorporating best available evidence and clinical experience ( Fig. 25-1 ).
Successful implementation of a protocol for mechanical ventilatory support of patients with acute respiratory distress syndrome (ARDS) was reported in 1993. Algorithms were developed through local expert consensus to standardize bedside decision making. Component algorithms provided point-of-care instructions for adjustment of the fraction of inspired oxygen, positive end-expiratory pressure, tidal volume (TV), and respiratory rate in response to threshold rules and measurement of variables directly affecting oxygenation and ventilation. The protocol comprised 30 algorithms and guided the entire process, from intubation through weaning and extubation. Additional algorithms were developed to enable use of pulse oximetry to accurately assess arterial oxygen partial pressure (Pao 2 ) with noninvasive arterial hemoglobin oxygen saturation (SpO 2 ).
Use of this first computerized protocol was associated with a dramatic increase in the survival of patients with ARDS. Bedside clinicians accepted more than 90% of ARDS management protocol generated instructions. The acceptance rate of computer- generated instructions from most other computerized protocol systems implemented since the early 1990s, nearly all in ICUs, has been 90% or greater, indicating detailed understanding of care process and comprehensive design of the care process model. This system was patient dedicated with the protocol logic program continuously “on” and with the user interface at bedside. Explicit criteria based on current measurements to establish a diagnosis of ARDS were required. Another important principle is requirement for the bedside clinician’s judgment to accept or decline all computerized protocol instructions for therapy intervention, referred to as an “open loop” control. When implemented continuously, a computerized protocol guidance system in a medical or surgical trauma intensive care unit (ICU) proved to be practical and safe, providing standardized decision making and individualized interventions for management of mechanical ventilatory support of ARDS.
This protocol system, with algorithms modified to enable 6 mL/kg breaths, was used in a randomized controlled trial (RCT) in which conventional and “small” TVs were compared in the management of ARDS. The RCT, conducted between 1993 and 1998 at 10 different centers, used a bedside program with available desktop computers and user interface. The protocol system was used for 32,055 hours (15 staff person years, 3.7 patient years); was active for 96% of ventilator time; and generated 38,546 instructions, 94% of which were followed. Similar results at a single participating center were documented in patients with trauma-induced ARDS (Shock Trauma ICU, Memorial Hermann Hospital, Houston, Tex.) in which the computerized system was in use 96% of the time with 95% compliance. The trial demonstrated efficacy of computerized algorithms implemented as a protocol that directed permissive hypercapnia with small tidal volume compared with then conventional large tidal volume strategies. Importantly, the trial demonstrated that care that was used with a computerized protocol system for mechanical ventilatory support could be directly transferred to other clinical sites and significantly improve care. This computerized protocol-based prospective RCT provided convincing evidence and generated new knowledge that predated work of the ARDS-Net.
A Houston-based team effort followed to develop a computerized protocol for management of intracranial pressure (ICP) after traumatic brain injury. A cohort study demonstrated that use of the six explicit therapeutic algorithms significantly improved compliance with established guidelines, limiting untoward changes in ICP and cerebral perfusion pressure despite fewer interventions. A prototype computerized protocol system using the algorithms was subsequently developed and successfully tested.
Computerized protocol technology has been developed to guide fluid resuscitation of shocked trauma patients during their first ICU day. The program was based on management principles developed by the LDS hospital group, and used oxygen delivery (D o 2 ) as a quantitative measurement of hemodynamic performance. The oxygen delivery goal was to be greater than or equal to 600 mL/O 2 /min. This “shock resuscitation protocol” was used to guide the treatment of more than 400 patients during 2000-2006 and was implemented with bedside mobile computer workstations. A series of protocol modifications were made through ongoing consensus group review (see Fig. 25-1 ) and accrued data analysis. This demonstrated the “process control” impact of computerized algorithms in a clinical process previously known to be variable and often chaotic. Use of this protocol confirmed the relationship between the volume of resuscitation fluid administered and the development and timing of abdominal compartment syndrome and persistent coagulopathy.
The use of computerized protocol technology improved the implementation of sepsis management guidelines, when compared with conventional guideline approaches. The Houston team constructed a comprehensive sepsis management protocol system, using Surviving Sepsis Campaign, other guidelines, and local expert consensus (see Fig. 25-2 ). The system standardized decision making among surgical intensivists, resident physicians, and nurse practitioners. Antibiotic agents were administered within 1 hour of protocol initiation and moderate volumes of intravenous fluid were administered (2.0 ± 0.2 L during 24-hour protocol). The hospital mortality rate was much lower than that reported by the Surviving Sepsis Campaign guideline initiative (14% vs. 31%). The acceptance rate for computerized protocol-generated instructions was 90%. This compared favorably with the Surviving Sepsis Campaign “bundle” compliance rate of 36%, although it should be considered that the “bundle” at that time contained some highly controversial interventions (subsequently removed) and compliance required administration of the entire bundle. A similar system designed and implemented at a second site (University of Florida Health surgical ICUs, Gainesville, Fla.) had very similar results: 14% hospital mortality rate and 91% acceptance rate of computerized protocol-generated instructions. At the second site, recognition of sepsis occurred earlier after onset of infection with implementation of a computerized sepsis surveillance and diagnosis system, providing a start state for the computerized protocol process, and analogous to specific criteria that had been used to diagnose ARDS or shock due to major torso trauma.
Related posts:
What Lessons Have Intensivists Learned During the Evidence-Based Medicine Era?
What Is the Role of Invasive Hemodynamic Monitoring in Critical Care?
Should Fever Be Treated?
Should Exacerbations of COPD Be Managed in the Intensive Care Unit?
Are Anti-inflammatory Therapies in ARDS Effective?
Has Outcome in Sepsis Improved? What Has Worked? What Has Not Worked?
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
