| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the *Department of Anesthesiology, Onze Lieve Vrouwe Gasthuis, Amsterdam, The Netherlands and Department of Anesthesiology, Academic Medical Center, Amsterdam, The Netherlands.
Address correspondence and reprint requests to Markus W. Hollmann, MD, PhD, DEAA, Department of Anesthesiology, Academic Medical Center, PO Box 22660, 1100 DD, Amsterdam, The Netherlands. Address e-mail to m.w.hollmann{at}amc.uva.nl.
Abstract
BACKGROUND: Guidelines for postoperative nausea and vomiting (PONV) prevention are implemented widely but their effectiveness may be limited by poor adherence. We hypothesized that the use of an electronic decision support (DS) system would significantly improve guideline adherence.
METHODS: Medical information of all patients undergoing elective surgery in our regional teaching hospital is routinely entered in an anesthesia information management system at the preoperative screening clinic. Our departmental PONV prevention guidelines identifies patients as "high-risk" and thus eligible for PONV prophylaxis based on the presence of at least three of the following risk factors: female gender, history of PONV or motion sickness, nonsmoker status, and anticipated use of postoperative opioids. Using automated reminders, we studied the effect of DS on guidelines adherence using an off–on–off design. In these three study periods, we queried for all consecutive patients visiting the preoperative screening clinic who were eligible for PONV prophylaxis and studied how often it was prescribed correctly.
RESULTS: Between November 2005 and June 2006, 1340, 2715, and 1035 patients were included in the control, DS and post-DS periods, respectively. As a result of mandatory data entry of risk factors, the percentage of high-risk PONV patients increased from 28% in the control period to 32% and 31% in the DS and post-DS periods, respectively. During the control period, 38% of all high-risk patients were prescribed PONV prophylaxis. This increased to 73% during the DS period and decreased to 37% in the post-DS period.
CONCLUSION: Electronic DS increases guidelines adherence for the prescription of PONV prophylaxis in high-risk PONV patients.
In recent years, several studies have identified risk factors and developed risk models to predict which patients are at risk for postoperative nausea and vomiting (PONV).1–6 Guidelines have been suggested for the prevention and treatment of PONV acknowledging the importance of the problem.7
However, the effectiveness of clinical guidelines may be severely limited by poor adherence.8,9 There may be guidelines-related, physician-related, and/or external reasons for poor guidelines adherence.8 Reasons for poor guidelines adherence might include lack of attention by the attending physician responsible for the preoperative assessment of the patient. This may be overcome with the use of a Decision Support (DS) system that reminds the physician of the action suggested by the guidelines at the time and place of care.10
Recent advances in perioperative care include the introduction of AIMS that may prompt users to take action according to guidelines based on real-time patient data, i.e., DS systems. Relatively few studies focusing on DS in anesthesiology and critical care have been conducted.11,12 We hypothesized that a DS system can significantly improve guideline adherence for the prescription of PONV prophylaxis in a preoperative screening clinic.
METHODS
For identification of patients at high risk for PONV, we used the simplified risk score based on the risk stratification by Apfel and co-workers1,13 This risk score uses four risk factors: female gender, history of PONV or motion sickness, nonsmoking status, and postoperative opioid use. According to this model, 56%–61% of patients are expected to experience PONV in the presence of at least three positive risk factors.1,14 Because it is not clear how to predict postoperative opioid use preoperatively, we have linked this risk factor to the type of admission, assuming that scheduled day care patients were expected not to receive opioids in contrast to clinical patients who were expected to receive opioids. Our departmental guideline dictated that in the presence of three or four positive risk factors, it was indicated to prescribe PONV prophylaxis. PONV prophylaxis comprised dexamethasone 8 mg IV upon induction of general anesthesia and granisetron 1 mg IV shortly before awakening. This guideline was discussed with and accorded by all anesthesiologists. To implement the guideline, meetings were held with all staff (nurses, nurse practitioners, and anesthesiologists) who screened patients preoperatively. These meetings provided information about the guideline, the evidence supporting it, and the expected change in patient care resulting from it. All meetings were held at least 2 wk before the first study period began.
This study was conducted in the preoperative screening clinic of a regional teaching hospital in Amsterdam, The Netherlands. In this outpatient clinic, an anesthesiologist screens all patients scheduled for elective surgery preoperatively before admission for surgery. During this visit, the anesthesiologist plans the anesthetic technique, prescribes the premedication for the day of surgery and decides on the indication for any prophylactic medication, such as PONV prophylaxis. This information is entered in an AIMS (Metavision®; iMDSoft, Tel Aviv, Israel). No patient data are stored primarily on paper. In the preoperative screening department, a workstation is available in all screening offices. In the 12 room operating room complex, all anesthesia machines are also equipped with an AIMS workstation.
The DS system consisted of an automated reminder based on Apfel et al.'s simplified risk score.1 To reliably calculate a simplified risk score for every patient being screened, recording data for all risk factors (gender, smoking status, history of PONV or motion sickness, postoperative opioid use) was made mandatory within the computer program. We programmed the system to remind the anesthesiologist of the indication for PONV prophylaxis if three or four risk factors were positive, general anesthesia was scheduled, and only if PONV prophylaxis had not yet been prescribed. Thus, if all of the conditions mentioned above had been met, the system created a message stating "This patient has at least three positive risk factors for PONV and is eligible for PONV prophylaxis, but is not prescribed prophylaxis yet. Do you want to prescribe PONV prophylaxis?" This message appeared instantaneously after selecting "general anesthesia" as the preferred anesthesia technique. The system prompted the user for a response to the message (either negative or affirmative). Consequently, the anesthesiologist was never forced to adhere to the guideline; they were only reminded of it in eligible patients.
All consecutive patients visiting the preoperative screening department, with the exception of patients scheduled for cardiac surgery, were included.
The study was divided into three study periods, each lasting for 8 or 16 wk, and was set-up according to an off–on–off design. By using this design, both learning effects and bias from conducting a study (as opposed to common practice) can be separated from the effects of a DS system. During the first 8-wk study period (control period), PONV prophylaxis was managed according to the paper version of the guideline. No reminders were used, as our goal was to interfere with normal practice as little as possible. During the 16-wk DS period, the automated reminders, as described above, were activated. After deactivating the automated reminders, the post-DS period was again managed with only the paper guideline for another 8 wk.
After completion of all study periods, the relevant data fields were extracted from the database by prewritten scripts. The primary end-point of our study was guideline adherence. All patients eligible for prophylaxis based on three or more positive risk factors were identified, and the percentage of those that had prophylaxis scheduled was calculated for all three periods. Confidence intervals were calculated and, where appropriate, 
2 tests were used to identify differences among the three time periods.
As a secondary endpoint, we used the percentage of patients that was identified to be at high risk to develop PONV.
To identify learning effects within the three study periods, we stratified the study population into 1-wk time periods. We repeated the analysis for those subgroups.
RESULTS
Between November 2005 and June 2006, 5090 consecutive patients were included in the study: 1340 patients in the control period and 2715 and 1035 patients in the DS and post-DS periods, respectively.
Basic demographics and incidence of risk factors for the three study periods are shown in Table 1. Patient characteristics were distributed equally across the periods.
|
The percentage of patients with three or four positive risk factors increased from 28% (95% CI: 25%–30%) in the control period to 32% (95% CI: 30%–34%) during DS and remained at 31% (95% CI: 28%–34%) in the post-DS period.
Of all patients eligible for PONV prophylaxis, the percentage of patients who were prescribed prophylaxis increased from 38% (95% CI: 33%–42%) in the control period to 73% (95% CI: 70%–76%) in the DS period. After the DS had been deactivated, this percentage decreased to 37% (95% CI: 32%–42%).
The prescription of PONV prophylaxis not according to the guideline, i.e., to patients with two or less positive risk factors, remained unchanged for all study periods. (Table 2).
|
We studied the presence of a possible learning effect within the three study periods by analyzing the 1-wk subgroups. As shown in Figure 1, guideline adherence (prescription behavior) was similar within each study period. The increase in guideline adherence was immediate and complete after introduction of the automated reminders in week 9. After discontinuation in week 24, guideline adherence returned to control levels immediately.
|
To study the effect of DS on prescription behavior of the anesthesiologists, we analyzed the percentage of high-risk patients with prophylaxis prescribed for each anesthesiologist for the three study periods. The number of patients screened by each anesthesiologist was comparable per period. The percentages are shown in Figure 2, which shows that all anesthesiologists except one followed a similar pattern of guideline adherence. In general, guideline adherence increases after introduction of DS and decreases back to control levels after cessation of DS.
|
DISCUSSION
We demonstrated that an electronic DS system significantly improved guideline adherence for the prescription of PONV prophylaxis in the preoperative screening clinic. However, after withdrawal of the DS system, guideline adherence decreased to control levels again. In addition, we demonstrated that identification of patients at high risk for PONV can be increased significantly by making the risk factor data mandatory.
For the current study, we used Apfel and co-workers simplified risk score for PONV, because this score combines adequate discriminative power with ease of use.1,13,14 The incidence of the four risk factors used in this risk score (female gender, nonsmoking status, history of PONV or motion sickness, and postoperative opioid use) has varied considerably among the studies published previously. In the present study, the percentage of identified high-risk PONV patients (three or four positive risk factors) was 31%. This appears to be slightly less than the ±40% incidence of high-risk patients found in the original study by Apfel and co-workers1,14 The smaller percentage of nonsmoking patients and the relatively low incidence of a positive history for PONV or motion sickness in our study presumably accounted for this difference.
The present study is in agreement with earlier reports showing that DS at the time and place of care may constitute an effective option for helping physicians adhere to a certain guideline.8,11,15,16 The increasing use of electronic file systems for medical information, like an AIMS, greatly increases our opportunities to use DS for quality improvement in clinical care. However, for maximum effect, a DS system must "present the right information, in the right format, at the right time, without requiring special effort."17 Our patient-specific DS system is unique within the field of anesthesia to actively select patients who are at high risk for a certain complication (i.e., PONV), to verify the scheduled anesthesia technique (i.e., general anesthesia), and to actively interfere with normal workflow if a guideline violation is impending (i.e., no PONV prophylaxis scheduled in high-risk PONV patients scheduled for general anesthesia). This is in contrast to a system reminding the physician of a guideline in every single case, which may be inappropriate. As an alternative, we could have included the reminder for PONV prophylaxis in a checklist similar to a recently published study by O'Reilly et al. They reported an increase in the timely administration of prophylactic antibiotics as a result of a reminder in a checklist.12 Although the guideline reminder studied by O'Reilly et al. could very well be suited for inclusion in a checklist since it was applicable to the majority of patients, such a checklist may not be the most suitable method for DS on PONV prophylaxis since only 30% of all patients are eligible for PONV prophylaxis. In fact, a guideline reminder for every single case may cause an abundance of unnecessary reminders, which may undermine the confidence of the user in the DS system and may cause unnecessary annoyance. Therefore, the optimal method of DS may depend on the guideline and on the percentage of eligible patients involved.
Even in the presence of DS, a number of high-risk patients remained for whom no PONV prophylaxis was prescribed. The reasons for this residual nonadherence are not known and were beyond the scope of the present study. One might suggest that lack of computer skills or hidden disagreement with the guideline may have played a role in the residual nonadherence. However, the percentage of patients who qualified for PONV prophylaxis, but did not receive a correct prescription, was comparable for all anesthesiologists except one, suggesting that these factors did not play a major role (Fig. 2).
Another explanation for nonadherence may be related to the incidence of the intervention that is dictated by a guideline. For example, billing needs to be done on every patient and antibiotic prophylaxis has to be administered to almost every patient.12,18 Therefore, these actions may become part of the routine for every patient. In contrast, PONV prophylaxis is indicated only in about one third of patients and therefore may not become part of this routine. We speculate that the difference in incidence may have an effect on the maximum level of guideline adherence that can be achieved.
The importance of the multifactorial reasons for nonadherence is illustrated by the fact that the increase in guideline adherence after the implementation of DS has varied considerably among studies reported earlier. In a general population of hospitalized patients, Dexter et al. showed that with the use of automated reminders, the prescription of subcutaneous heparin for prevention of thrombotic complications increased from 19% to 33%.15 Even though this is a significant improvement, it implies that the majority of patients (67%) eligible for preventive subcutaneous heparin did not receive the medication. In the study by O'Reilly et al. mentioned above, correct prophylactic administration of antibiotics within 1 h from incision increased from 62% to 92% over a 1 yr period.12 Residual nonadherence was still 8%.
The above illustrates that the reasons for nonadherence may be more interesting than its mere presence, and stresses the importance for future studies not to focus on guideline adherence only, but also on possible reasons for nonadherence. This may further enhance our understanding of the prerequisites for optimal DS that may help us further improve quality of care.
One might argue that we did not include incidence of PONV as a measure of outcome. However, there were multiple reasons not to study the occurrence of PONV as an outcome measure. First, the effectiveness of PONV prophylaxis as we describe it has been proven repeatedly.19 Second, the application of the simplified risk score to guide prescription and administration of PONV prophylaxis has been proven effective in reducing the incidence of PONV.13 Finally, the incidence of PONV on the day of surgery is influenced by a series of factors other than the correct scheduling of PONV prophylaxis in the preoperative screening clinic. In contrast, the present study was designed to study the effect of DS on the prescription behavior (i.e., guideline adherence) of the anesthesiologist in the preoperative screening clinic. However, it seems logical that improved guideline adherence must finally result in better outcome, since otherwise the validity of the guideline itself may be questioned.
In conclusion, an electronic DS system using patient-specific automated reminders significantly improved the guideline adherence for the prescription of PONV prophylaxis. Moreover, the mandatory data entry of risk factors for PONV improved identification of patients at high risk for PONV. After deactivating the DS system, the effect on guideline adherence disappeared completely.
Footnotes
Accepted for publication November 7, 2007.
This research was exclusively funded using institutional support.
Conflict of interest: Dr. Kooij received travel grants from iMDSoft and ProStrakan Pharma to be able to present research results on international congresses.
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  P. Kranke, N. Roewer, A. F. Smith, S. N. Piper, J. Wallenborn, and L. H. J. Eberhart Postoperative Nausea and Vomiting: What Are We Waiting For? Anesth. Analg., March 1, 2009; 108(3): 1049 - 1050. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
) is the overall average (DS: decision support).
