Children are at significant risk for adverse drug events, and recent studies have begun to describe the frequency and epidemiology of medication errors and adverse events in pediatric inpatients.⁸^–¹⁰ In 2006, The Institute of Medicine released guidelines urging improved surveillance systems to detect adverse events.¹¹ Traditional methods used to detect adverse events in children included manual chart review and voluntary incident reporting. These detection systems are inefficient and significantly underestimate the number and prevalence of adverse events.^12,¹³

Another manual detection strategy relies on trigger methodology where an occurrence, found on manual chart review, triggers further investigation to determine the presence of an adverse event.^11,¹⁴ For example, the administration of flumazenil may trigger the detection of benzodiazepine-induced respiratory depression. Automated adverse-event detection relies on the generation of a trigger report from the EHR, which indicates the possibility that an adverse event has occurred requiring further investigation. This methodology has been proven an efficient and cost effective way to detect adverse events.¹⁵^–¹⁹

Promises and Limitations

Information technology in the form of an EHR promises improved patient care.^20,²¹ Potential benefits of information technology include providing rapid access to integrated clinical data and extant medical knowledge, eliminating illegibility, improving communication, and issuing applicable reminders and checks for appropriate medical actions.²²

A number of studies show that information technology can provide various benefits, including increasing adherence to guidelines (particularly in the outpatient arena) and decreasing some medication errors.^23,²⁴ However, the majority of these studies come from a very small number of institutions with homegrown clinical information systems that were developed by devoted groups of clinicians.²⁵ Very few studies show that the commercially available systems confer similar benefits, and even if they do, it is unclear that their success can be migrated from one implementation to another.²⁶^–²⁸ In fact, any benefit may be outweighed by new problems introduced by the systems themselves. In effect, one set of problems may be traded for another.^29,³⁰

Despite considerable progress, the sentiment expressed by G. Octo Barnett in 1966 is often echoed today, “It is frustrating to meet with repeated disappointments when the objectives are superficially so simple.”³¹ The medical information space is vastly more complicated than it seems at first. EHR software programs are enormously complex, are built by large teams of programmers with input by numerous clinicians, demand high-speed processors and high-bandwidth networks, and rely on often fragile interfaces with other hospital systems. Implementation currently requires tremendous effort by both clinicians and technical specialists to configure these systems according to the specific needs of an institution and in ways that will enhance care rather than impede it. An often unappreciated complicating factor is that the technology does not simply replace paper; it also reengineers care—deliberately or not. (See Unfavorable Alteration of Workflow.)

Errors can and do occur in programming or configuration. Many programming deficiencies can be detected and corrected with thorough testing, preferably in a development environment that does not affect real patients; however, some of these problems will only become apparent under unique circumstances that are presented by patient care. Indefatigable vigilance for these errors is essential.

Numerous other unintended consequences result from implementing an EHR, including the creation of new kinds of errors, an increase in work for clinicians, an untoward alteration of workflow and change in communication patterns, an increase in system demands, a continuation of the persistence of paper use, and the fostering of potential overdependence on the technology.³²^–³⁴

New Kinds of Errors

While some errors can be avoided by using an EHR with computerized physician order entry (CPOE), other errors may be created or propagated.^35,³⁶ Many “new” errors are a result of poorly designed interfaces. For example, clinicians can easily make “juxtaposition errors,” intending to select one item but selecting another close to it on a long, dense pick list in a small font. A similar kind of error is mistaking an open chart of one patient to be that belonging to another or picking the wrong patient from a long list of patients.

Interfaces between electronic systems are particularly vulnerable and can cause various new kinds of errors. Patients who have been physically transferred but remain, disembodied, in their previous electronic location may have all of their care suspended pending completion of the electronic transfer. Worse, should electronic transfers be delayed, medications may be delivered to a patient’s former room and administered to a different patient admitted to that room. Allergies may be entered in the bedside system, but interface problems can prevent that information from reaching the pharmacy or nutrition systems. Occasionally, laboratory results can be inserted into the wrong medical record because of interface issues.

Rigid interpretation of policies and procedures can be configured into the EHR but may lead to difficulties in clinical practice when dealing with ambiguous circumstances and exceptions. Sometimes the process of care is incompletely understood and codification can be disastrous. Policies at most institutions include automatic stop orders that require rewriting medication orders in a specified time frame. Compliance to this rule can be forced with programming, but implementing this rule without safeguards could lead to automatic discontinuation of medications and missed doses.

The benefit of legibility in electronically written notes can be outweighed by novel problems. Overuse of copy-paste functions can result in repetitive, monotonous, and loquacious notes punctuated by the sin of propagating erroneous text verbatim. Automatic transcription of data such as laboratory results or vital signs often bypasses cognition, something that does not happen when data are transcribed by hand.

Increased Work for Clinicians

While transcription errors can be eliminated by computerized order entry, it often falls to clinicians to shoulder the added burden of what might otherwise be considered clerical functions. Documentation in a structured format rather than as free text can enhance completeness and facilitate later data retrieval; however, it can also increase work by forcing the clinician to find ways to fit round pegs into square holes. Similarly, rigidly structured order input can force clinicians to waste time trying different ways to order nonstandard tests or therapies—with little guarantee that these orders will actually be executed if they are routed to electronic limbo.

Clinical alerts can help clinicians make decisions, e.g., when penicillin is mistakenly ordered for an allergic patient, but persistent interruptions of work by alerts can increase the workload of the clinician who must decipher their meaning and assess the risk in each specific circumstance. The frequency of these alerts can become intolerable when they are not delivered to the right clinician with the right information and at the right time and place. When these alerts become too frequent and too predictable, clinicians often adapt by “response chaining”: dismissing the alerts with rote keystrokes much as a pianist plays a familiar tune. Alerts that evoke this response cannot be effective and may be counterproductive.³⁷

Poorly integrated clinical information systems cause clinicians to access many different sources for information to solve a clinical problem, thereby increasing work. Similarly, users should not be required to input the same bit of data in multiple locations in different systems.

Another time-consuming feature of the EHR, and perhaps the most exasperating, is the loss of data, particularly when busy clinicians lose long notes they have just meticulously written. Workstation or interface crashes, network collisions, inopportune time-outs, or system failures of other types can be the culprit. System delays from a wide variety of causes also waste valuable time, as does having to hunt for an available workstation because those installed are insufficient in number or inconveniently placed.

Unfavorable Alteration of Workflow

The introduction of an EHR system significantly alters the sociotechnological milieu. Previously well-functioning medical practices may become entirely dysfunctional. Implementation of an EHR requires modeling of work processes but can sometimes result in ossifying those processes into something too inflexible for efficient and effective patient care.

Patients expected to be emergently admitted to the pediatric intensive care unit (PICU) but still in transit often have medications and urgent therapies ordered and prepared before arrival. A CPOE system may prohibit ordering or dispensing medications for patients who have not yet been admitted. In a paper environment, nurses frequently arrange dosing schedules based on the ordered frequency, the frequency of other ordered medications, and the availability of intravenous access. However, in many CPOE systems, medication orders go direct to the pharmacy and bypass the bedside nurse. Physicians are then saddled with picking the specific times of administration, only to have that schedule revised later by the nurse.

Sometimes, well-defined manual processes can be implemented in more than one way electronically. Without clear delineation of an institutionally sanctioned method, confusion and catastrophe can result. Transferring patients from one unit to another or to the operating room typically requires the discontinuation and reinitiation of all orders, including medications. With implementation of CPOE, clinicians without proper direction could suspend rather than discontinue the old orders. Reactivation of suspended orders could result in duplication and double dosing of medications ordered on transfer.

Redundant orders are sometimes facilitated by CPOE. The gate-keeping function of clerical personnel processing orders for routine radiographs or laboratories is bypassed. Remote access by multiple physicians acting on the same bit of new information can also generate duplicate orders.

Untoward Changes in Communication Patterns

Many care providers blame clinical information systems for unsatisfactory reductions in face-to-face communication. Some users complain that the EHR creates an “illusion of communication,”³⁸ where users believe that information entered into the system will be somehow communicated to the relevant personnel. This assumption can result in missed or delayed execution of orders or failure to appreciate the recommendations of a consultant. Users may erroneously assume that allergies entered into the system will adequately protect patients from receiving offending food or drugs.

Because of the time-consuming nature of CPOE and because workstations are not always available at the bedside, orders are often written after work rounds. In teaching hospitals, order writing is often delegated to the least experienced individual, such as the first-year resident. Residents may not readily appreciate the nuances of an order until they are confronted with an order screen that demands input he or she had not previously considered. By that time, the other members of the care team have disbanded, and clarification of the order requires tracking down and reconsulting the relevant personnel.

Continued Promise

The Institute of Medicine, in its report, “Crossing the quality chasm: a new health system for the 21st century,”² stated that health care should be safe, effective, patient-centered, timely, efficient, and equitable. The Institute further noted that these goals could be more easily reached through judicious application of IT. Automated order entry systems can improve safety. The use of automated reminders based on clinical practice guidelines, computer-assisted diagnosis or management, and evidence-based medicine (EBM) can improve the effectiveness of medical care. IT can enhance patient-centered care that is respectful of and responsive to patient preferences, needs, and values by recording them and appropriately reminding the health care professional. It can facilitate access to clinical knowledge through web sites and online support groups. Clinical decision-support systems can be used to tailor information and disease management messages based on the patient’s individual needs. Timeliness can be improved by e-mail, telemedicine, and direct and immediate access to diagnostic test results and other clinical information. IT can improve efficiency by using clinical decision-support systems to reduce redundant and unnecessary tests and procedures, by improving communication among multiple providers of care to individual patients, and by supplying data for performance and outcome measures. Enhancing equity among patients and across socioeconomic, geographic, race, and ethnic lines can be achieved if IT can improve access to clinicians and clinical knowledge, although it would depend upon equitable access to the technology infrastructure. IT is playing the starring role in the drive to improve the quality of health care today, and the Institute called for a national commitment to build the information infrastructure to support health care.