JSTOR

Methods

At our hospital, internal medicine house staff members are administratively grouped into 3 separate firms; house staff members remain in their assigned firm during their 3 years of training. Each day, a single team from each firm admits patients. Admitted patients are consecutively assigned to firms, resulting in patient populations with similar demographic characteristics, severity of illness, and diagnoses. For the 2005‐2006 influenza season, we assigned an influenza vaccination intervention strategy to each firm: usual care, computer‐generated preselected orders (hereafter, opt‐out orders), or computer‐generated reminders. The computer‐generated messages did not include educational information. The text for the preselected order was as follows: “An order to nursing to administer the influenza vaccine will be generated unless you deselect the order below.” For both computer‐based strategies, the clinical decision support system was triggered for inpatients by the discharge order. Because discharge orders typically are entered once per patient, house staff received a single intervention exposure per patient. There were no additional house staff interventions; in particular, there were no educational sessions.

After the clinical trial, we surveyed the house staff; participation in the survey was voluntary and responses were confidential. The survey was administered after July 1, 2006; therefore, only house staff members who were in their first or second year of training during the influenza season were available to complete the survey. The survey assessed house staff members' knowledge of the patient groups that were considered high‐priority populations for vaccination; this knowledge was evaluated both directly and through clinical vignettes. We also asked about the likelihood that they would recommend influenza vaccine to inpatients and outpatients. The response options for the clinical vignettes and for recognition of high‐priority populations were presented as 4‐point Likert scales. The staff member's likelihood of recommending influenza vaccination to patients was assessed by use of percentiles grouped into quartiles. We obtained institutional review board approval.

We compared responses across the 3 firms. Firms were considered on an ordinal scale, which was based on a prespecified ranking of the intervention’s intensity: usual care was considered to be less intense than the reminder intervention, and the reminder intervention was considered to be less intense than the opt‐out order intervention. The firms were numbered accordingly. We dichotomized Likert scale responses and compared responses between 2 firms by use of the χ²or Fisher exact test. We used the Wilcoxon rank sum test to compare the likelihood of recommending influenza vaccination among firms. We used a nonparametric test for trend to test for the trend across the 3 firms.⁷

Results

Of 70 house staff members, 39 (56%) responded to the survey; the response rate was similar among firms ( ). Most respondents (26 [67%]) reported that they had received influenza vaccination during the current season, and most (28 [72%]) estimated that vaccine efficacy was at least 70% for healthy young adults. House staff recognized long‐term care residents, healthcare workers, patients with cardiovascular disease, and adults older than 65 years as high‐priority populations for vaccination; however, most (22 [56%]) did not indicate that they would vaccinate pregnant women, even after the first trimester (Table). In clinical vignettes, house staff members were most likely to recommend vaccination for patients with chronic obstructive pulmonary disease or pneumonia and least likely to recommend vaccination for patients who were febrile or neutropenic or for pregnant women with asthma (Table). Overall, house staff members were more likely to recommend vaccination for outpatients than for inpatients ( ) (Figure).

Table.  Results of a Survey of House Staff Regarding Influenza Vaccination Practices

Open New Window

(30 KB)

Figure.  Percentage of occasions on which internal medicine house staff recommended influenza vaccination to patients, by house staff group (ie, usual care, reminder, or standing order) and setting. House staff only received the interventions for hospital patients. (Test for trend across house staff groups, for clinic patients and for hospital patients.)

Open New Window

When we evaluated the likelihood that house staff would recommend vaccination to inpatients or outpatients, a significant trend was seen toward increased vaccination across the 3 firms (Figure). House staff members exposed to the opt‐out order intervention were the most likely to recommend vaccination, followed by the staff in the firm exposed to the reminder intervention, and then staff in the firm that performed usual care (Figure). House staff members in the intervention groups were more likely to have received a recommendation from senior physicians to vaccinate patients, compared with the usual care group. ( , test for trend).

In 2‐group comparisons, house staff members in the firm exposed to the opt‐out order intervention were more likely to recommend vaccination for hospital patients and outpatients than staff in the firm that performed usual care ( and , respectively); they were also more likely to recommend vaccination for hospital patients than house staff in the firm exposed to the reminder intervention ( ). House staff in the firm exposed to the reminder intervention were more likely than those in the firm that performed usual care intervention to recommend vaccination for inpatients or outpatients; however, these differences may have been the result of chance ( and , respectively). No differences were found in responses among firms with respect to vaccine efficacy or recognition of high‐priority patients.

Discussion

We found that internal medicine house staff exposed to computer‐generated standing vaccination orders for inpatients reported a higher likelihood of recommending influenza vaccination for both inpatients and outpatients. Also, these staff members were more likely to report that senior physicians recommended administering vaccine. Regardless of firm, there was a failure to recognize pregnant women as a high‐priority population for influenza vaccination.

Other reports have documented the successful use of computer‐based interventions to improve the rate of influenza vaccination among hospitalized patients.^5,8 Although these interventions have successfully changed behavior at the time of the prompt, it is unknown whether the behavioral change would extend to nonintervention settings. We hypothesized that repeated exposure to computer‐generated prompts would change house staff members' attitudes or motivate self education about the vaccine. Our finding that exposure to computer‐based prompts increased the likelihood of recommending vaccination to outpatients suggests that the effect extended beyond the intervention setting. There appeared to be a dose‐response effect (ie, computer‐generated orders influenced self‐reported behavior more than the passive reminder). No differences were found between the intervention firms and the usual‐care firm in their attitudes about vaccine efficacy or knowledge about high‐priority populations targeted for vaccination. It is possible that computer‐generated prompts stimulated house staff to vaccinate outpatients but did not motivate self education about the vaccine.

Because pregnant women have increased morbidity from influenza and because vaccination is safe and effective,⁹ the Advisory Committee on Immunization Practices identifies them as a high‐priority population.¹⁰ Although most house staff correctly identified other high‐priority patient populations, often they did not recognize that pregnant women were a high‐priority group as well. Even after presenting them with a clinical vignette that was framed to encourage vaccination (ie, hospitalization for asthma exacerbation after the first trimester), many house staff members would not recommend vaccination during pregnancy. One explanation is that educational efforts have successfully communicated the message about high‐risk populations for elderly and chronically ill patients, but not for pregnant women. An alternative explanation is that because internal medicine services rarely admit pregnant women, their knowledge deficit resulted from infrequently caring for this population.

Our study had several limitations. There was a small number of potential respondents and a response rate that, although similar to that of other physician surveys, was suboptimal. However, the response rate was similar across firms, suggesting that there was no selection bias across firms. Also, through medical record review we confirmed that the computer‐generated order increased inpatient vaccination (data not shown); however, we did not confirm that for outpatients self‐reported practices represented increased vaccination. Furthermore, we did not have data on the number of discharges per physician; therefore, we could not quantify exposure to the intervention.

House staff members exposed to computer‐generated prompts for influenza vaccination at the time of hospital discharge were more likely to report recommending vaccination to inpatients; this intervention effect extended to outpatients. Focused messages may be required to communicate the importance of vaccinating pregnant women.