Clinical decision support alert malfunctions: analysis and empirically derived taxonomy.

Objective: To develop an empirically derived taxonomy of clinical decision support (CDS) alert malfunctions. Materials and Methods: We identified CDS alert malfunctions using a mix of qualitative and quantitative methods: (1) site visits with interviews of chief medical informatics officers, CDS developers, clinical leaders, and CDS end users; (2) surveys of chief medical informatics officers; (3) analysis of CDS firing rates; and (4) analysis of CDS overrides. We used a multi-round, manual, iterative card sort to develop a multi-axial, empirically derived taxonomy of CDS malfunctions. Results: We analyzed 68 CDS alert malfunction cases from 14 sites across the United States with diverse electronic health record systems. Four primary axes emerged: the cause of the malfunction, its mode of discovery, when it began, and how it affected rule firing. Build errors, conceptualization errors, and the introduction of new concepts or terms were the most frequent causes. User reports were the predominant mode of discovery. Many malfunctions within our database caused rules to fire for patients for whom they should not have (false positives), but the reverse (false negatives) was also common. Discussion: Across organizations and electronic health record systems, similar malfunction patterns recurred. Challenges included updates to code sets and values, software issues at the time of system upgrades, difficulties with migration of CDS content between computing environments, and the challenge of correctly conceptualizing and building CDS. Conclusion: CDS alert malfunctions are frequent. The empirically derived taxonomy formalizes the common recurring issues that cause these malfunctions, helping CDS developers anticipate and prevent CDS malfunctions before they occur or detect and resolve them expediently.

Selection of drug-laboratory result pairs for an inpatient asynchronous alert program: results of a Delphi survey.

PURPOSE: Combinations of drugs and laboratory values ("drug-laboratory pairs") that represent practical high-priority targets for potential use in a daily asynchronous inpatient clinical decision-support (CDS) alert report were identified. METHODS: A list of 654 drug-laboratory pairs compiled through a literature review was evaluated by a multidisciplinary expert panel in a modified Delphi procedure. After initial evaluation to narrow the list to 89 drug-laboratory pairs, panelists used Likert scales to rate the remaining pairs on six dimensions (frequency of alert, likelihood of harm, severity of harm, preventability, ameliorability, and global impression of usefulness) in Delphi survey rounds until consensus emerged. Final selection of pairs for potential use as CDS tools was based on global-impression-of-usefulness scores. Correlations between impression of usefulness and other evaluative dimensions were determined. RESULTS: The Delphi process yielded a final list of 24 high-priority drug-laboratory pairs. The highest-ranked pairs were heparin- low platelet count, potassium supplement- high serum potassium, angiotensin-converting-enzyme inhibitor-high serum potassium, and heparin-positive heparin platelet factor 4 antibody test. Medications on the final list included nine anticoagulants, six cardiologic agents, four antimicrobials, and two electrolytes. Most of the selected drug-laboratory pairs related to renal function, serum potassium levels, hematologic results, or pregnancy. Panelists' impression of usefulness was significantly correlated with severity-of-harm ratings (p < 0.0001). CONCLUSION: Expert review of drug- laboratory value pairs for potential inclusion in an asynchronous monitoring program yielded 24 high-priority and practical pairs for monitoring. About 25% of the pairs had not been the focus of previous laboratory-pharmacy CDS at the study panelists' home institutions.

Hand-carried echocardiography by hospitalists: a randomized trial.

BACKGROUND: Hospitalists can use hand-carried echocardiography for accurate point-of-care information, but patient outcome data for its application are sparse. METHODS: We performed an unblinded, parallel-group randomized trial between July 2008 and March 2009 at one teaching hospital in Chicago, Illinois. We randomly assigned adult general medicine inpatients referred for standard echocardiography with indications investigatable by hand-carried echocardiography to care guided by hand-carried echocardiography or usual care. The main outcome measure was length of stay on the referring hospitalist's service. Secondary outcomes included a before-after analysis of reported changes in management due to hand-carried echocardiography and the diagnostic accuracy of hand-carried echocardiography. RESULTS: The difference in length of stay between 226 participants randomized to care guided by hand-carried echocardiography (geometric mean 46.1 hours, interquartile range 29.0-70.9 hours) and 227 participants randomized to usual care (46.9 hours, interquartile range 34.1-68.3 hours) corresponded to a 1.7% reduction in length of stay that was not statistically significant (95% confidence interval, -12.1 to 9.8%). In post hoc subgroup analyses, care guided by hand-carried echocardiography reduced length of stay in participants who were referred for heart failure (P=.0008). Among participants who underwent both hand-carried and standard echocardiography, hospitalists changed management due to hand-carried echocardiography in 37%. Despite the favorable diagnostic accuracy of hand-carried echocardiography, most changes to the timing of hospital discharge occurred after standard echocardiography. CONCLUSION: Hospitalist care guided by hand-carried echocardiography for unselected general medicine patients does not meaningfully affect length of stay. Whether or not it affects care quality remains unstudied.

Computerized physician order entry- based hyperglycemia inpatient protocol and glycemic outcomes: The CPOE-HIP study.

OBJECTIVE: To evaluate the impact of implementing a computerized physician order entry (CPOE)-based hyperglycemia inpatient protocol (HIP) on glycemic outcomes. METHODS: This retrospective, cross-sectional study compared blood glucose values, hemoglobin A(1c) values, diabetes medication profles, and demographic data of diabetic patients admitted to medicine services between March 15, 2006, and April 11, 2006 (before CPOE-HIP protocol was adopted), with data of diabetic patients admitted between October 3, 2007, and October 30, 2007 (1 year after CPOE-HIP protocol was implemented). RESULTS: A total of 241 diabetic patients comprised the pre-CPOE-HIP group and 197 patients comprised the post-CPOE-HIP group. After the protocol was adopted, there was a decrease of 10.8 mg/dL in the mean glucose concentration per patient-day (175.5 +/- 81.2 mg/dL vs 164.7 +/- 82 mg/dL, P<.001). Additional glycemic control improvements included a 5% increase in patient-days with serum glucose concentrations between 70 and 150 mg/dL (41.1% vs 46.1%, P = .008) and a 3.1% decrease in patient-days with glucose concentrations above 299 mg/dL (16.9% vs 13.8%, P = .023). The percentage of patient-days with glucose concentrations less than or equal to 50 mg/dL was not significantly different (0.95% vs 1.27%,P = .15). Compliance with the American Diabetes Association recommendation for hemoglobin A1c inpatient testing frequency increased from 37.3% to 64.5% (P<.001). The length of stay did not differ between the groups. CONCLUSIONS: Implementation of a hospital-wide, CPOE-based, hyperglycemia management protocol had a favorable impact on glucose targets, decreasing excessively high glucose levels without increasing clinically meaningful hypoglycemic events. Compliance with hemoglobin A(1c) testing recommendations also improved.

A hospitalist-run short-stay unit: features that predict length-of-stay and eventual admission to traditional inpatient services.

BACKGROUND: Short-stay units (SSUs) provide an alternative to traditional inpatient services for patients with short anticipated hospital stays. Yet little is known about which patient types predict SSU success. OBJECTIVE: To describe patients admitted to our hospitalist-run SSU and explore predictors of length-of-stay (LOS) and eventual admission to traditional inpatient services. DESIGN: Prospective observational cohort study. SETTING: Large public teaching hospital. PATIENTS: Consecutive admissions (n = 755) to the SSU over 4 months. INTERVENTION: Hospitalist attending physicians prospectively collected data from patients' histories, physical exams, and medical records upon admission and discharge. MEASUREMENTS: Risk assessments were made for patients with our most common provisional diagnoses: possible acute coronary syndrome (ACS) and heart failure. Patient stays were considered successful when LOS was less than 72 hours and eventual admission to traditional inpatient services was not required. RESULTS: Of 738 eligible patients, 79% (n = 582) had successful SSU stays. In a multivariable model, the provisional diagnosis of heart failure predicted stays longer than 72 hours (P = 0.007) but risk assessments were unimportant. Patients who received specialty consultations were most likely to need eventual admission (odds ratio [OR], 13.1; 95% confidence interval [CI], 6.9-24.9), and the likelihood of long stays was inversely proportional to the accessibility of diagnostic tests. CONCLUSIONS: In our hospitalist-run SSU, the inaccessibility of diagnostic tests and the need for specialty consultations were the most important predictors of unsuccessful stays. Designs for other SSUs that care for mostly low-risk patients should focus on matching patients' diagnostic and consultative needs with readily accessible services.

Diagnostic accuracy of hospitalist-performed hand-carried ultrasound echocardiography after a brief training program.

BACKGROUND: The duration of training needed for hospitalists to accurately perform hand-carried ultrasound echocardiography (HCUE) is uncertain. OBJECTIVE: To determine the diagnostic accuracy of HCUE performed by hospitalists after a 27-hour training program. DESIGN: Prospective cohort study. SETTING: Large public teaching hospital. PATIENTS: A total of 322 inpatients referred for standard echocardiography (SE) between March and May 2007. INTERVENTION: Blinded to SE results, attending hospitalist physicians performed HCUE within hours of SE. MEASUREMENTS: Diagnostic characteristics of HCUE as a test for 6 cardiac abnormalities assessed by SE: left ventricular (LV) systolic dysfunction; severe mitral regurgitation (MR); moderate or severe left atrium (LA) enlargement; moderate or severe LV hypertrophy; medium or large pericardial effusion; and dilatation of the inferior vena cava (IVC). RESULTS: A total of 314 patients underwent both SE and HCUE within a median time of 2.8 hours (25th to 75th percentiles, 1.4 to 5.1 hours). Positive and negative likelihood ratios for HCUE increased and decreased, respectively, the prior odds by 5-fold or more for LV systolic dysfunction, severe MR regurgitation, and moderate or large pericardial effusion. Likelihood ratios changed the prior odds by 2-fold or more for moderate or severe LA enlargement, moderate or severe LV hypertrophy, and IVC dilatation. Indeterminate HCUE results occurred in 2% to 6% of assessments. CONCLUSIONS: The diagnostic accuracy of HCUE performed by hospitalists after a brief training program was moderate to excellent for 6 important cardiac abnormalities.

Impact of a bedside procedure service on general medicine inpatients: A firm-based trial.

BACKGROUND: Procedure services may improve the training of bedside procedures. However, little is known about how procedure services may affect the demand for and success of procedures performed on general medicine inpatients. OBJECTIVE: Determine whether a procedure service affects the number and success of 4 bedside procedures (paracentesis, thoracentesis, lumbar puncture, and central venous catheterization) attempted on general medicine inpatients. DESIGN: Prospective cohort study. SETTING: Large public teaching hospital. PATIENTS: Nineteen hundred and forty-one consecutive admissions to the general medicine service. INTERVENTION: A bedside procedure service was offered to physicians from 1 of 3 firms for 4 weeks. This service then crossed over to physicians from the other 2 firms for another 4 weeks. MEASUREMENTS: Data on all procedure attempts were collected daily from physicians. We examined whether the number of attempts and the proportion of successful attempts differed based on whether firms were offered the beside procedure service. RESULTS: The number of procedure attempts was 48% higher in firms offered the service (90 versus 61 per 1000 admissions; RR 1.48, 95% CI 1.06-2.10; P = .030). More than 85% of the observed increase was a result of procedures with therapeutic indications. There were no differences between firms in the proportions of successful attempts or major complications. CONCLUSIONS: The availability of a procedure service may increase the overall demand for bedside procedures. Further studies should refine the indications for and anticipated benefits from these commonly performed invasive procedures.