Identifying barriers to hypertension guideline adherence using clinician feedback at the point of care.

Factors contributing to low adherence to clinical guidelines by clinicians are not well understood. The user interface of ATHENA-HTN, a guideline-based decision support system (DSS) for hypertension, presents a novel opportunity to collect clinician feedback on recommendations displayed at the point of care. We analyzed feedback from 46 clinicians who received ATHENA advisories as part of a 15-month randomized trial to identify potential reasons clinicians may not intensify hypertension therapy when it is recommended. Among the 368 visits for which feedback was provided, clinicians commonly reported they did not follow recommendations because: recorded blood pressure was not representative of the patient's typical blood pressure; hypertension was not a clinical priority for the visit; or patients were nonadherent to medications. For many visits, current quality-assurance algorithms may incorrectly identify clinically appropriate decisions as guideline nonadherent due to incomplete capture of relevant information. We present recommendations for how automated DSSs may help identify "apparent" barriers and better target decision support.

Leveraging point-of-care clinician feedback to study barriers to guideline adherence.

Studies of barriers to guideline adherence have generally surveyed clinicians temporally remote from the clinical scenario in which recommendations were delivered, potentially adversely biasing clinician observations. The user interface of ATHENA DSS, a guideline-based decision support system for hypertension, includes a point-of-care feedback window that accepts clinician-user comments during the display of recommendations. Analysis of this feedback has revealed a number of intriguing patient, provider, and technical barriers to adherence collected during real-time system use.

Evaluating provider adherence in a trial of a guideline-based decision support system for hypertension.

Measurement of provider adherence to a guideline-based decision support system (DSS) presents a number of important challenges. Establishing a causal relationship between the DSS and change in concordance requires consideration of both the primary intention of the guideline and different ways providers attempt to satisfy the guideline. During our work with a guideline-based decision support system for hypertension, ATHENA DSS, we document a number of subtle deviations from the strict hypertension guideline recommendations that ultimately demonstrate provider adherence. We believe that understanding these complexities is crucial to any valid evaluation of provider adherence. We also describe the development of an advisory evaluation engine that automates the interpretation of clinician adherence with the DSS on multiple levels, facilitating the high volume of complex data analysis that is created in a clinical trial of a guideline-based DSS.

Patient safety in guideline-based decision support for hypertension management: ATHENA DSS.

The Institute of Medicine recently issued a landmark report on medical error.1 In the penumbra of this report, every aspect of health care is subject to new scrutiny regarding patient safety. Informatics technology can support patient safety by correcting problems inherent in older technology; however, new information technology can also contribute to new sources of error. We report here a categorization of possible errors that may arise in deploying a system designed to give guideline-based advice on prescribing drugs, an approach to anticipating these errors in an automated guideline system, and design features to minimize errors and thereby maximize patient safety. Our guideline implementation system, based on the EON architecture, provides a framework for a knowledge base that is sufficiently comprehensive to incorporate safety information, and that is easily reviewed and updated by clinician-experts.

Implementing clinical practice guidelines while taking account of changing evidence: ATHENA DSS, an easily modifiable decision-support system for managing hypertension in primary care.

This paper describes the ATHENA Decision Support System (DSS), which operationalizes guidelines for hypertension using the EON architecture. ATHENA DSS encourages blood pressure control and recommends guideline-concordant choice of drug therapy in relation to comorbid diseases. ATHENA DSS has an easily modifiable knowledge base that specifies eligibility criteria, risk stratification, blood pressure targets, relevant comorbid diseases, guideline-recommended drug classes for patients with comorbid disease, preferred drugs within each drug class, and clinical messages. Because evidence for best management of hypertension evolves continually, ATHENA DSS is designed to allow clinical experts to customize the knowledge base to incorporate new evidence or to reflect local interpretations of guideline ambiguities. Together with its database mediator Athenaeum, ATHENA DSS has physical and logical data independence from the legacy Computerized Patient Record System (CPRS) supplying the patient data, so it can be integrated into a variety of electronic medical record systems.

A graphic nomogram for warfarin dosage adjustment.

We assessed the ability of a graphic nomogram to adjust steady-state warfarin dosages and to predict international normalized ratios (INR) after a dosage change, compared with an anticoagulation clinic pharmacist and a Bayesian regression computer program. Study subjects were 108 men and 3 women receiving warfarin anticoagulation. In all patients the median absolute errors in predicted INR values for the nomogram, computer program, and pharmacist were 0.33, 0.46, and 0.48, respectively. The nomogram was significantly more precise than both other methods (p=0.036). In a subset of 50 patients who required dosage reductions, the median absolute INR prediction errors for the nomogram, computer program, and pharmacist were 0.35, 0.54, and 0.48 respectively. The nomogram was significantly more precise than the pharmacist (p=0.005) and computer (p=0.002). The ability to provide more precise dosage reductions of warfarin may be of clinical importance in light of current recommendations for higher-intensity warfarin therapy and maintenance of higher INR values. Prospective validation of the performance of this nomogram in a routine clinical setting is warranted.

Prescribing practice and cost of antibacterial prophylaxis for surgery at a US Veteran Affairs hospital.

This study retrospectively compared the actual drug-related cost of antibacterial prophylaxis for specific operative procedures with the theoretical costs based on recommendations published in Medical Letter on Drugs and Therapeutics, the Surgical Infection Society, and those of the chiefs of each surgical subspecialty at our institution. We identified all patients who received in intravenous bacterial for prophylaxis before a clean or clean-contaminated operation between 1st January and 30th September 1993, using the medical centre's computerised information system. The information included comprehensive surgical case histories, and pharmacy and microbiology records. Only those operations in which recommendations for surgical prophylaxis were present in all 3 guidelines were included. The outcome measures were antibacterial-related costs (drug acquisition and administration costs), the number of antibacterial doses dispensed, and choice of antibacterial agents. During the study period, 3,322 operations were performed, 2,993 of which were excluded. Thus, 329 patients undergoing operations in 6 subspecialties were included in the analysis. The actual mean cost per patient significantly exceeded the projected costs using Medical Letter Consultants' and Surgical Infection Society guidelines for all 6 subspecialties [mean excess cost per patient: $US49.04 and $US34.95, respectively (1994 values)] and institutional guidelines for 4 of the 6 subspecialties (mean excess cost per patient: $US24.36). The actual mean number of doses per patient significantly exceeded those projected using Medical Letter Consultants' and Surgical Infection Society guidelines for all 6 subspecialties (mean excess number of doses per patient: 6.0 and 4.0, respectively) and institutional guidelines for 4 of the 6 subspecialties (mean excess number of doses per patient: 2.9). The choice of antibacterial was appropriate in approximately 90% of cases. We conclude that the practice of antibacterial prophylaxis for specific operative procedures performed by 6 subspecialties is not in accordance with institutional or published guidelines, and the excess cost is primarily a result of prolonged duration of antibacterial prophylaxis.

Lithium and angiotensin-converting enzyme inhibitors: evaluation of a potential interaction.

The potential interaction between lithium and angiotensin-converting enzyme (ACE) inhibitors was investigated in a retrospective, longitudinal, case-control study of 20 hypertensive patients previously stabilized on lithium therapy. The objective of the study was to determine the impact of ACE inhibition on steady-state lithium concentrations and to evaluate the potential association of altered lithium clearance with age, renal function, and electrolyte balance. After initiation of the ACE inhibitor, steady-state lithium concentrations increased by 36.1%, lithium clearance was reduced by 25.5% (p < 0.0001), and four patients presented with symptoms suggestive of lithium toxicity. Significant bivariate correlations were observed for lithium clearance change and age (r = -0.45; p < 0.05) and for lithium clearance change and serum creatinine (r = -0.52; p < 0.02). Multiple regression analysis indicated that 25% of the change in lithium clearance was associated with a change in serum creatinine. This percentage was increased to 35% by the inclusion of age in the regression model. None of the other variables (age, height, weight, or change in serum sodium/potassium) made a significant contribution to this model. The authors concluded that a clinically important increase in lithium concentrations can occur in patients started on ACE inhibitor therapy. As elderly patients may be uniquely predisposed to this interaction, avoidance of this medication combination in older populations should be considered.

Information content of dentists' Yellow Pages advertising.

Content analysis was performed on 1,007 Yellow Pages dental display ads to investigate whether such advertisements include the information important to consumers. The authors found a significant information gap between what consumers considered important when selecting a dentist and the information provided in the advertisements. The research was supported in part by a grant to Caroline Fisher from the American Marketing Association.

Practical computer-assisted dosing for aminoglycoside antibiotics.

In principle, computer-assisted individualization of antibiotic dosing offers the prospect of better patient outcomes through improved dosing precision. In practice, however, the expertise in pharmacokinetics required to operate these programs has precluded their use by most physicians and pharmacists. We developed a computer program for individualization of dosing of aminoglycoside antibiotics under conditions in which access to experts in pharmacokinetics is impractical. The program is accurate, yet it requires less effort for data collection than previous drug dosing programs did. The program generates advice on a broad spectrum of topics, including dose adjustment, interpretation of measured drug concentrations in blood, and recommendations for monitoring drug concentrations. We tested its performance by prospectively comparing it with a clinical pharmacokinetic consultation service in a series of 78 consecutive patients. There were no differences in accuracy or bias in the prediction of drug concentrations. The rate of agreement between the program's dosing recommendations and those of the consultation service was 67 percent. This rate of agreement is typical of interexpert variation. In a stratified set of 24 of the 41 instances with significant disagreement regarding the recommended dose, experts ranked the program's recommendations as highly as those of the consultation service (95% confidence interval for difference in rank, -0.30 less than chi less than 0.47). The results suggest that expert systems can be coupled with pharmacokinetic dosing programs to deliver high-quality clinical recommendations for administration of antimicrobial agents.

Nomogram for dosing warfarin at steady state.

The predictive performance of a nomogram for dosing warfarin was compared with that of a computer program. The nomogram and the computer program were developed from the log-linear model describing warfarin pharmacodynamics at steady state. The nomogram's dose-response curves were generated by using previously reported pharmacodynamic and pharmacokinetic values for an outpatient population receiving warfarin. The series of dose-response curves were plotted by altering the pharmacodynamic values over a range of 3 standard deviations. The ability of the nomogram to predict the steady-state prothrombin time ratio (PTR) after an adjustment in the dosage of warfarin was evaluated, and the results were compared with those of a commercially available program involving Bayesian regression. Data for 65 outpatients were evaluated. The mean +/- S.D. nomogram-predicted, computer-predicted, and measured PTRs were 1.63 +/- 0.27, 1.64 +/- 0.24, and 1.66 +/- 0.23, respectively. The mean prediction errors for the nomogram and the computer program were -0.037 and -0.026, respectively, and the mean percent absolute prediction errors were 11.6% and 11.0%, respectively. Neither method was biased, and differences between the results for the two methods were not significant. The predictive performance of the warfarin dosing nomogram was comparable to that of the computer program.

Cost-effectiveness of prospective and continuous parenteral antibiotic control: experience at the Palo Alto Veterans Affairs Medical Center from 1987 to 1989.

PURPOSE: Controlling inappropriate antibiotic usage is a major focus for hospital quality assurance and cost-containment programs. We assessed the impact of strengthening a parenteral antibiotic control policy and instituting continuous infectious disease service (IDS) reviews of the appropriateness of antimicrobial therapy on cost and patient outcomes. PATIENTS AND METHODS: All patients receiving intravenous antibiotics during a 3.5-year period from 1986 to 1989 were included in either the pre- or post-policy study group. Antibiotic costs 16 months before were compared with antibiotic costs 26 months after implementation of a new policy to restrict inappropriate usage of (1) broad-spectrum antibiotics when not necessary, (2) expensive agents when a less costly agent could be used, and (3) an excessive dosage or interval. Patient subgroups treated 4 months before and 4 months after policy implementation were compared further within diagnosis-related group (DRG) assignments using patient demographic, cost, and outcome measures. RESULTS: The average monthly antibiotic costs during the 26-month post-policy period were $7,600 less than during the 16-month pre-policy period (p less than 0.0001), resulting in an average yearly drug cost reduction of $91,200. The IDS team altered therapy in 611 (34.5%) of 1,769 reviews of antibiotic usage during the 26-month period. The comparisons among similar patient groups by DRG classification revealed the average number of antibiotic doses per study patient admission was decreased 24% (p = 0.005) and drug costs were reduced 32% (p = 0.004) after policy implementation. In two DRG categories (i.e., respiratory infections plus pneumonia), patients in the post-policy group had a 33% decrease in average number of doses (p = 0.05) and 45% decrease in antibiotic costs (p = 0.04) compared with the pre-policy group. Similar trends were observed in most DRG categories. There was an average $70 per admission decrease in drug cost and a reduction of eight antibiotic doses per admission after policy initiation. The overall prevalence of deaths (p = 0.22) and average length of antibiotic therapy (p = 0.29) were less in the post-policy period despite group similarities in patient characteristics and lengths of hospital stay. CONCLUSION: Antibiotic control policies can be developed to ensure quality care and can be designed to select for cost-effective agents. Prospective and continuous monitoring of antibiotic usage by the IDS resulted in a significant and sustained reduction in antibiotic costs without detrimental effect on the length of therapy or deaths.

Effect of using warfarin plasma concentrations in Bayesian forecasting of prothrombin-time response.

The predictive performance of a Bayesian computer program using prothrombin-time (PT) response data with and without warfarin plasma concentrations to forecast patients' PTs at the time of hospital discharge was evaluated. A log-linear pharmacokinetic-pharmacodynamic model was used to describe and predict warfarin dose response in patients recently started on warfarin sodium. Individual patients' pharmacodynamic variables relating warfarin concentration to clotting-factor synthesis were obtained by Bayesian nonlinear regression analysis. Pharmacokinetic values for warfarin clearance and volume of distribution were either calculated using nonlinear regression from measured plasma warfarin concentrations or estimated based on literature-derived population regression equations. Percent mean absolute prediction error (precision) and prediction error (bias) for PT predictions were compared among and between analysis methods that used only literature data to estimate PT response and methods that used zero to five PTs with or without warfarin plasma concentrations. Eleven women and eight men completed the study. Predictions after four days of warfarin therapy using PT measurements beginning after either the first or third warfarin dose were clinically useful regardless of whether warfarin concentrations were used in the predictions. Predictions using fewer than four PT measurements were imprecise and biased. The Bayesian method in this study provided good predictions of PTs immediately before hospital discharge based on warfarin dosing and PT response after either four or five days of therapy. The use of warfarin plasma concentrations in the pharmacokinetic-pharmacodynamic model used here appears unwarranted.

Initiation of warfarin therapy: comparison of physician dosing with computer-assisted dosing.

In a prospective, randomized study at two university hospitals, the authors examined how effectively housestaff physicians (n = 36) managed the initiation of warfarin therapy compared with a computer-assisted dosing regimen (n = 39) using the software program Warfcalc, which was managed by one of the authors. Target prothrombin time ratios were selected by the physicians. Study endpoints included: the time to reach a therapeutic prothrombin ratio, the time to reach a stable therapeutic dose, the number of patients transiently overanticoagulated, the number of bleeding complications, and the accuracy of the predicted maintenance dose, which was assessed at steady-state 10-14 days later. Computer-assisted dosing consistently out-performed the physicians: a stable therapeutic dose was achieved 3.7 days earlier (p = 0.002), fewer patients were overanticoagulated (10% versus 41%), and the predicted maintenance dose was in the therapeutic range in 85% of the computer-dosed patients versus 42% of the physician group (p less than 0.002). For physicians who did not routinely manage warfarin therapy, computer-assisted dosing improved the accuracy of dosing and shortened the time required to achieve a stable therapeutic dose.

Bayesian pharmacokinetic/pharmacodynamic forecasting of prothrombin response to warfarin therapy: preliminary evaluation.

The ability of a pharmacokinetic/pharmacodynamic Bayesian forecasting computer program to predict prothrombin response to warfarin therapy was investigated. The performance of the program was evaluated retrospectively in an inpatient study population of 45 subjects. Predictions of prothrombin response at discharge, based on zero to five serially measured prothrombin ratios, were compared. Precision of prediction was measured by root mean squared error (rmse), bias was measured by average prediction error, and significance (p less than 0.05) was determined by 95% confidence intervals and correlation coefficients. Eleven (3.8%) predictions exceeded established limitations of the pharmacokinetic/pharmacodynamic model and were excluded from data analysis. Correlations between measured and predicted prothrombin ratios for all methods were significant. The five prothrombin ratio feedbacks provided the most accurate predictions (rmse 0.219). These predictions were significantly better than the population parameter (rmse 0.418), one (rmse 0.401), and two (rmse 0.459) prothrombin ratio feedback predictions. The predictions based on population parameters and one prothrombin ratio feedback were significantly biased. When provided with sufficient feedback, the bias was not apparent and the predictive performance improved with each additional prothrombin ratio. The predictive performance of the four and five prothrombin ratio feedbacks is sufficient to provide clinically useful dosage guidelines early in the course of warfarin therapy. The population parameter estimates require further delineation in order to improve the performance of limited prothrombin ratio feedback predictions.