Brian S. Rothman

Future of electronic health records: implications for decision support.

The potential benefits of the electronic health record over traditional paper are many, including cost containment, reductions in errors, and improved compliance by utilizing real-time data. The highest functional level of the electronic health record (EHR) is clinical decision support (CDS) and process automation, which are expected to enhance patient health and healthcare. The authors provide an overview of the progress in using patient data more efficiently and effectively through clinical decision support to improve health care delivery, how decision support impacts anesthesia practice, and how some are leading the way using these systems to solve need-specific issues. Clinical decision support uses passive or active decision support to modify clinician behavior through recommendations of specific actions. Recommendations may reduce medication errors, which would result in considerable savings by avoiding adverse drug events. In selected studies, clinical decision support has been shown to decrease the time to follow-up actions, and prediction has proved useful in forecasting patient outcomes, avoiding costs, and correctly prompting treatment plan modifications by clinicians before engaging in decision-making. Clinical documentation accuracy and completeness is improved by an electronic health record and greater relevance of care data is delivered. Clinical decision support may increase clinician adherence to clinical guidelines, but educational workshops may be equally effective. Unintentional consequences of clinical decision support, such as alert desensitization, can decrease the effectiveness of a system. Current anesthesia clinical decision support use includes antibiotic administration timing, improved documentation, more timely billing, and postoperative nausea and vomiting prophylaxis. Electronic health record implementation offers data-mining opportunities to improve operational, financial, and clinical processes. Using electronic health record data in real-time for decision support and process automation has the potential to both reduce costs and improve the quality of patient care.

Lack of utility of a decision support system to mitigate delays in admission from the operating room to the postanesthesia care unit.

BACKGROUND: When the phase I postanesthesia care unit (PACU) is at capacity, completed cases need to be held in the operating room (OR), causing a “PACU delay.” Statistical methods based on historical data can optimize PACU staffing to achieve the least possible labor cost at a given service level. A decision support process to alert PACU charge nurses that the PACU is at or near maximum census might be effective in lessening the incidence of delays and reducing over-utilized OR time, but only if alerts are timely (i.e., neither too late nor too early to act upon) and the PACU slot can be cleared quickly. We evaluated the maximum potential benefit of such a system, using assumptions deliberately biased toward showing utility. METHODS: We extracted 3 years of electronic PACU data from a tertiary care medical center. At this hospital, PACU admissions were limited by neither inadequate PACU staffing nor insufficient PACU beds. We developed a model decision support system that simulated alerts to the PACU charge nurse. PACU census levels were reconstructed from the data at a 1-minute level of resolution and used to evaluate if subsequent delays would have been prevented by such alerts. The model assumed there was always a patient ready for discharge and an available hospital bed. The time from each alert until the maximum census was exceeded (“alert lead time”) was determined. Alerts were judged to have utility if the alert lead time fell between various intervals from 15 or 30 minutes to 60, 75, or 90 minutes after triggering. In addition, utility for reducing over-utilized OR time was assessed using the model by determining if 2 patients arrived from 5 to 15 minutes of each other when the PACU census was at 1 patient less than the maximum census. RESULTS: At most, 23% of alerts arrived 30 to 60 minutes prior to the admission that resulted in the PACU exceeding the specified maximum capacity. When the notification window was extended to 15 to 90 minutes, the maximum utility was <50%. At most, 45% of alerts potentially would have resulted in reassigning the last available PACU slot to 1 OR versus another within 15 minutes of the original assignment. CONCLUSIONS: Despite multiple biases that favored effectiveness, the maximum potential benefit of a decision support system to mitigate PACU delays on the day on the surgery was below the 70% minimum threshold for utility of automated decision support messages, previously established via meta-analysis. Neither reduction in PACU delays nor reassigning promised PACU slots based on reducing over-utilized OR time were realized sufficiently to warrant further development of the system. Based on these results, the only evidence-based method of reducing PACU delays is to adjust PACU staffing and staff scheduling using computational algorithms to match the historical workload (e.g., as developed in 2001).

Communication latencies of wireless devices suitable for time-critical messaging to anesthesia providers.

BACKGROUND:Rapid and reliable methods of text communication to mobile anesthesia care providers are important to patient care and to efficient operating room management. Anesthesia departments are implementing automated methods to send text messages to mobile devices for abnormal vital signs, clinical recommendations, quality of care, and compliance or billing issues. The most time-critical communications determine maximum acceptable latencies. We studied the reliability of several alphanumeric messaging systems to identify an appropriate technology for such use. METHODS:Latencies between message initiation and delivery to 3 alphanumeric paging devices were measured over weeks. Two devices used Internet pathways outside the hospital’s local network with an external paging vendor (SkyTel). The third device used only the internal hospital network (Zetron). Sequential cell phone text page latencies were examined for lag-1 autocorrelation using the runs test, with results binned by hour and by day. Message latencies subsequently were batched in successive 1-week bins for calculation of the mean and 99th percentiles of latencies. We defined acceptance criteria as a mean latency <30 seconds and no more than 1 in 200 pages (0.5%) having a latency longer than 100 seconds. Cell phone texting was used as a positive control to assure that the analysis was appropriate, because such devices have (known) poor reliability during high network activity. RESULTS:There was substantial correlation among latencies for sequential cell phone text messages when binned by hours (P < 0.0001), but not by days (P = 0.61). The 2 devices using Internet pathways outside the hospital’s network demonstrated unacceptable performance, with 1.3% and 33% of latencies exceeding 100 seconds, respectively. The device dependent only on the internal network had a mean latency of 8 seconds, with 100% of 40,200 pages having latencies <100 seconds. The findings suggest that the network used was the deciding factor. CONCLUSIONS:Developers of anesthesia communication systems need to measure latencies of proposed communication pathways and devices used to deliver urgent messages to mobile users. Similar evaluation is relevant for text pagers used on an ad hoc basis for delivery of time-critical notifications. Testing over a period of hours to days is adequate only for disqualification of a candidate paging system, because acceptable results are not necessarily indicative of long-term performance. Rather, weeks of testing are required, with appropriate batching of pages for analysis.

Using Information Technology to Improve Quality in the OR

This article summarizes the current state of technology as it pertains to quality in the operating room, ties the current state back to its evolutionary pathway to understand how the current capabilities and their limitations came to pass, and elucidates how the overlay of information technology (IT) as a wrapper around current monitoring and device technology provides a significant advance in the ability of anesthesiologists to use technology to improve quality along many axes. The authors posit that IT will enable all the information about patients, perioperative systems, system capacity, and readiness to follow a development trajectory of increasing usefulness.

Explanation for the Near-Constant Mean Time Remaining in Surgical Cases Exceeding Their Estimated Duration, Necessary for Appropriate Display on Electronic White Boards

BACKGROUND: Consider a case that has been ongoing for longer than the scheduled duration. The anesthesiologist estimates that there is 1 hour remaining. Forty-five minutes later the case has not yet finished, and closure has not yet started. We showed previously that the mean (expected) time remaining is approximately 1 hour, not 15 minutes. The relationship is a direct mathematical consequence of the log-normal probability distributions of operating room (OR) case durations. We test the hypothesis that, with an accurate probabilistic model, until closure begins the estimated mean time remaining would be the mean time from the start of closure to OR exit. METHODS: Among the 311,940 OR cases in a 7-year time series from 1 hospital, there were 3962 cases for which (1) there had been previously at least 30 cases of the same combination of scheduled procedure(s), surgeon, and type of anesthetic and (2) the actual OR time exceeded the 0.9 quantile of case duration before the case started. A Bayesian statistical method was used to calculate the mean (expected) minutes remaining in the case at the 0.9 quantile. The estimate was compared with the actual minutes from the time of the start of closure until the patient exited the OR. RESULTS: The mean ± standard error of the pairwise difference was 0.2 ± 0.4 minutes. The Bayesian estimate for the 0.9 quantile was exceeded by 10.2% ± 0.01% of cases (i.e., very close to the desired 10.0% rate). CONCLUSIONS: If a case is taking longer than the expected (scheduled) duration, closure has not yet started, and someone in the OR is asked how much time the case likely has remaining, the value recorded on a clipboard for viewing later should be the estimated time remaining (e.g., “1 hour”) not an end time (e.g., “5:15 PM”). Electronic whiteboard displays should not show that the estimated time remaining in the case is less than the mean time from start of closure to OR exit. Similarly, if closure has started, the expected time remaining that is displayed should not be longer than the mean time from closure to OR exit. Finally, our results match previous reports that, before a case starts, statistical methods can reliably be used to assist in decisions involving the longest amount of time that cases may take (e.g., conflict checking for resources, filling holes in the OR schedule, and preventing holes in the schedule).

A Perioperative Systems Design to Improve Intraoperative Glucose Monitoring Is Associated with a Reduction in Surgical Site Infections in a Diabetic Patient Population

Background: Diabetic patients receiving insulin should have periodic intraoperative glucose measurement. The authors conducted a care redesign effort to improve intraoperative glucose monitoring. Methods: With approval from Vanderbilt University Human Research Protection Program (Nashville, Tennessee), the authors created an automatic system to identify diabetic patients, detect insulin administration, check for recent glucose measurement, and remind clinicians to check intraoperative glucose. Interrupted time series and propensity score matching were used to quantify pre- and postintervention impact on outcomes. Chi-square/likelihood ratio tests were used to compare surgical site infections at patient follow-up. Results: The authors analyzed 15,895 cases (3,994 preintervention and 11,901 postintervention; similar patient characteristics between groups). Intraoperative glucose monitoring rose from 61.6 to 87.3% in cases after intervention (P = 0.0001). Recovery room entry hyperglycemia (fraction of initial postoperative glucose readings greater than 250) fell from 11.0 to 7.2% after intervention (P = 0.0019), while hypoglycemia (fraction of initial postoperative glucose readings less than 75) was unchanged (0.6 vs. 0.9%; P = 0.2155). Eighty-seven percent of patients had follow-up care. After intervention the unadjusted surgical site infection rate fell from 1.5 to 1.0% (P = 0.0061), a 55.4% relative risk reduction. Interrupted time series analysis confirmed a statistically significant surgical site infection rate reduction (P = 0.01). Propensity score matching to adjust for confounders generated a cohort of 7,604 well-matched patients and confirmed a statistically significant surgical site infection rate reduction (P = 0.02). Conclusions: Anesthesiologists add healthcare value by improving perioperative systems. The authors leveraged the one-time cost of programming to improve reliability of intraoperative glucose management and observed improved glucose monitoring, increased insulin administration, reduced recovery room hyperglycemia, and fewer surgical site infections. Their analysis is limited by its applied quasiexperimental design.

A Narrative Review of Meaningful Use and Anesthesia Information Management Systems.

The US federal government has enacted legislation for a federal incentive program for health care providers and hospitals to implement electronic health records. The primary goal of the Meaningful Use (MU) program is to drive adoption of electronic health records nationwide and set the stage to monitor and guide efforts to improve population health and outcomes. The MU program provides incentives for the adoption and use of electronic health record technology and, in some cases, penalties for hospitals or providers not using the technology. The MU program is administrated by the Department of Health and Human Services and is divided into 3 stages that include specific reporting and compliance metrics. The rationale is that increased use of electronic health records will improve the process of delivering care at the individual level by improving the communication and allow for tracking population health and quality improvement metrics at a national level in the long run. The goal of this narrative review is to describe the MU program as it applies to anesthesiologists in the United States. This narrative review will discuss how anesthesiologists can meet the eligible provider reporting criteria of MU by applying anesthesia information management systems (AIMS) in various contexts in the United States. Subsequently, AIMS will be described in the context of MU criteria. This narrative literature review also will evaluate the evidence supporting the electronic health record technology in the operating room, including AIMS, independent of certification requirements for the electronic health record technology under MU in the United States.

Automated identification of drug and food allergies entered using non-standard terminology

OBJECTIVE An accurate computable representation of food and drug allergy is essential for safe healthcare. Our goal was to develop a high-performance, easily maintained algorithm to identify medication and food allergies and sensitivities from unstructured allergy entries in electronic health record (EHR) systems. MATERIALS AND METHODS An algorithm was developed in Transact-SQL to identify ingredients to which patients had allergies in a perioperative information management system. The algorithm used RxNorm and natural language processing techniques developed on a training set of 24 599 entries from 9445 records. Accuracy, specificity, precision, recall, and F-measure were determined for the training dataset and repeated for the testing dataset (24 857 entries from 9430 records). RESULTS Accuracy, precision, recall, and F-measure for medication allergy matches were all above 98% in the training dataset and above 97% in the testing dataset for all allergy entries. Corresponding values for food allergy matches were above 97% and above 93%, respectively. Specificities of the algorithm were 90.3% and 85.0% for drug matches and 100% and 88.9% for food matches in the training and testing datasets, respectively. DISCUSSION The algorithm had high performance for identification of medication and food allergies. Maintenance is practical, as updates are managed through upload of new RxNorm versions and additions to companion database tables. However, direct entry of codified allergy information by providers (through autocompleters or drop lists) is still preferred to post-hoc encoding of the data. Data tables used in the algorithm are available for download. CONCLUSIONS A high performing, easily maintained algorithm can successfully identify medication and food allergies from free text entries in EHR systems.

Enhancing Point of Care Vigilance Using Computers

Information technology has the potential to provide a tremendous step forward in perioperative patient safety. Through automated delivery of information through fixed and portable computer resources, clinicians may achieve improved situational awareness of the overall operation of the operating room suite and the state of individual patients in various stages of surgical care. Coupling the raw, but integrated, information with decision support and alerting algorithms enables clinicians to achieve high reliability in documentation compliance and response to care protocols. Future studies and outcomes analysis are needed to quantify the degree of benefit of these new components of perioperative information systems.

Case cancellation rates measured by surgical service differ whether based on the number of cases or the number of minutes cancelled.

BACKGROUND:Surgical cancellation rates typically are reported as the number of cancelled cases divided by the number of scheduled cases. However, the total number of cancelled minutes also has financial impact on surgeons’ productivity. Cancellation rates can instead be calculated based on the number of minutes of cancelled cases. Hospitals typically benchmark cancellation rates, since not all cancellations are preventable (e.g., those due to new onset of patient symptoms requiring further workup and treatment before surgery can safely proceed). If the mean estimated duration of cancelled cases were the same as that of scheduled cases, rates would be equivalent whether calculated using the number of cancellations or the minutes of cancellations. It is unknown whether there is a difference between these 2 methods. METHODS:Data for elective, regular workday cases scheduled were obtained from 2 academic hospitals and binned into 8 sequential 13-week periods. Cancellation rates after 7:00 AM or after 7:00 PM on the working day before surgery were calculated by service as (1) the numeric cancellation rate = number cancelled divided by the number scheduled and (2) the duration cancellation rate = minutes cancelled divided by the minutes scheduled. Mean differences (biases) and 95% prediction limits between the numeric and duration cancellation rates were determined. RESULTS:The hospitals’ numeric cancellation rates after 7:00 AM (11.6% and 10.7%) were similar to 12.2% from an academic hospital survey. Bias by individual service ranged from −1.16% ± 1.34% to 1.93% ± 3.01% at one hospital and −1.08% ± 2.76% to 3.05% ± 1.89% at the other. Mean differences between matching services at the hospitals were −0.7% ± 0.6% to 3.3% ± 0.3%. There was considerable variability among services for numeric cancellation rates and the prediction limits of the cancellation rate, calculated using the number of minutes cancelled. CONCLUSIONS:Calculating cancellation rates using case counts can inaccurately represent their impact on surgeon’s productivity compared with using minutes of cancelled cases. Comparing numeric cancellation rates between hospitals or services without checking for bias may lead to inappropriate conclusions. We recommend that hospitals evaluate their data for potential bias to determine whether cancellation rates need to be calculated using scheduled minutes of cases rather than numbers of cancellations.