Sustaining the Implementation of an Evidence-Based Guideline for Bronchiolitis

Objective  To describe the changes occurring over a 3-year period after implementation of an evidence-based clinical practice guideline for the care of infants with bronchiolitis.

Design  Before and after study.

Setting  Children's Hospital Medical Center, Cincinnati, Ohio.

Patients  Infants 1 year or younger admitted to the hospital with a first-time episode of typical bronchiolitis.

Intervention  The guideline was implemented January 15, 1997. Data on all patients discharged from the hospital with bronchiolitis, from January 15 through March 27, in 1997, 1998, and 1999, were stratified by year and compared with data on similar patients discharged from the hospital in the same periods in the years 1993 through 1996.

Main Outcome Measures  Patient volumes, length of stay for admissions, and use of specific laboratory and therapeutic resources ancillary to bed occupancy.

Results  After implementation of the guideline, admissions decreased 30% and mean length of stay decreased 17% (P<.001). Nasopharyngeal washings for respiratory syncytial virus were obtained in 52% fewer patients (P<.001); 14% fewer chest x-ray films were ordered (P<.001). There were significant reductions in the use of all respiratory therapies, with a 17% decrease in the use of at least 1 β₂-agonist inhalation therapy (P<.001). In addition, 28% fewer repeated inhalations were administered (P<.001); mean costs for all resources ancillary to bed occupancy fell 41% (P<.001); and mean costs for respiratory care services fell 72% (P<.001).

Conclusions  An evidence-based clinical practice guideline for the care of patients encountered in major pediatric care facility has been successfully sustained beyond the initial year of its introduction to practitioners in southwest Ohio.

WE HAVE reported previously on the salutary effect in the first year after implementation of an evidence-based clinical practice guideline describing best practices for inpatient care of infants with a first episode of bronchiolitis.¹ Because first-year successes in adult applications have not guaranteed sustained guideline acceptance by practitioners, in this report we describe some of our experiences with the bronchiolitis guideline during the second and third years of implementation in a major pediatric care hospital in the Midwest.^2-6

Details of the development of the original practice guideline have been published elsewhere.¹ The guideline is limited to the care of infants 1 year or younger admitted to the hospital with a first-time episode of uncomplicated bronchiolitis. Selecting only infants with first-time disease reduced the likelihood of including patients with asthma. Excluded from eligibility were infants with histories of immunodeficiencies, significant congenital heart diseases, bronchopulmonary dysplasia, congenital airway diseases, or any other comorbid condition that might make the effect of the bronchiolitis more severe and, thereby, care more complicated. Patients requiring ventilator or other intensive therapies were also excluded.

The guideline recommends that bronchiolitis, in typical presentation, be viewed as a self-limited disease. It encourages practitioners to assure that patients are well-oxygenated and well-hydrated, while it argues against routine laboratory and radiological studies for patients with typical and uncomplicated bronchiolitis. Respiratory care therapies, such as chest physiotherapy, cool mist, supervised cough and suction, aerosol saline, steroid inhalations, and routine use of bronchodilator aerosol, are discouraged. The guideline also recommends use of a simple respiratory assessment form as a method for helping practitioners judge the efficacy of any β₂-agonist treatment given before deciding on the use of repeated doses of inhaled medications.⁷

Between 1997 and 1998, before the second year of implementation, the guideline was revised to reflect newly published evidence that epinephrine administered in the emergency department was often effective in preventing the need for hospital admission.^8-11 This revised recommendation was accompanied by the insertion of a new algorithm into the guideline that had as part of its structure a connection line that, in retrospect, was interpreted by some practitioners as being somewhat ambiguous and perhaps even as an unintended recommendation encouraging the use of multiple doses of β₂-agonists (Figure 1). For 1999, although the evidence-based epinephrine recommendation was retained, the ambiguous algorithm was removed from the guideline.

During the initial bronchiolitis season after implementation in 1997, the chief resident, a head nurse, and a study coordinator made daily morning rounds on all eligible admissions. During these rounds, guideline principles were reinforced with the ward physicians and nurses, and any problems inhibiting implementation resolved. In 1998 and 1999, only the study coordinator made these daily rounds, communicating with the nurses, ward junior and senior residents, attending physicians, and, when appropriate, the chief resident. The study coordinator was empowered to be eclectic in creating an environment that enhanced practitioner compliance with the guidelines. During various months when the coordinator deemed it appropriate, this included offering simple incentives, such as pizza lunches, as rewards for high compliance.

The assessment of guideline efficacy was structured to test a number of hypotheses: (1) If effective as a management tool, reduction in the use of all resources ancillary to bed occupancy would be sustained and reflected specifically as a reduction in the use of inhalation respiratory therapies and laboratory studies. (2) If acceptable to the practitioners, most infants admitted with bronchiolitis would be started on the guideline, as detected by medical record review of physician orders and management intentions. (3) The numbers of admissions for bronchiolitis and lengths of stays (LOSs) of those admitted would be reduced when practitioners become comfortable with the information reporting on the limited efficacies of inhalations and some other interventions requiring hospitalization.

The bronchiolitis guideline was implemented on January 15, 1997. Data were collected from all guideline-eligible patients discharged from the hospital with a diagnosis of bronchiolitis from January 15 through March 27, 1997, January 15 through March 27, 1998, and January 15 through March 27, 1999. These were compared with each other and to data collected from records on historical control patients discharged from the hospital with a diagnosis of bronchiolitis before guideline implementation between January 15 through March 27 in each of the 4 years, from 1993 through 1996.

After reviewing the guideline, the institutional review board concluded that it was primarily a patient care instrument and, as long as patients were not randomized or identified in publications, consent would not be required to use the guideline recommendations.

Data collected in the 4 years before guideline implementation revealed that 69% of admitted patients with bronchiolitis were given β2-agonist inhalations as part of their therapies and 57% of patients received more than 1 dose. A sample size of 83 patients was deemed necessary to detect a 20% change in the administration of any β₂-agonist with an α of .05 and 90% power. Four hundred patients were required to detect a 20% change in administration of multiple doses of β₂-agonist with 90% power at the P = .05 level.

Patients with bronchiolitis before guideline implementation were identified using the hospital information system's historical database of all patient encounters. For patients admitted after guideline implementation, data were collected concurrently at the time of each encounter. Resource utilization ancillary to bed occupancy, length of hospitalization, and all other information used as a basis for generating hospital charges were obtained retrospectively from the hospital financial and clinical computer systems. Medical record reviews were used selectively to abstract some retrospectively collected data and also to confirm, as a measure of guideline compliance, that a guideline preprinted admission order was signed to initiate the guideline intents for treating infants considered eligible based on a diagnostic impression of bronchiolitis recorded by the admitting physician. The numbers of readmissions within 7 days of a patient's discharge from the hospital were also tracked retrospectively, as well as the reasons for these readmissions. The actual cost of each resource used during hospitalization was extracted based on relative value units computed by the hospital financial system after adjustment of all costs to 1999 values for comparability. Because cost figures for use of resources ancillary to bed occupancy are easier to define and more reliable, hospital charges were not included.¹²

The data were stratified categorically based on whether they represented information about care delivered before or after introduction and implementation of the guideline.

The χ² tests were used for analysis of categorical variables and t tests were performed for normally distributed continuous variables. Wilcoxon rank sum tests were used for nonnormally distributed data. In this report, when expressed as means, results are followed by the SDs. All statistical analyses were performed using PC-SAS software (Release 6.12; SAS Institute Inc, Cary, NC).

Data from 1979 patients were examined; 1300 records from the historical 4-year period before implementation of the guideline were available. These were compared with the data describing 679 patients discharged from the hospital with the diagnosis of bronchiolitis after implementation on January 15, 1997.

Age at admission was not significantly different between years (Table 1). Insurance coverage characteristics for admitted patients with bronchiolitis were different before and after guideline implementation (before, 56% Medicaid/self-pay; after, 50% Medicaid/self-pay) with significant but inconsistent differences in insurance coverage distributions between years (P<.001). There were no important differences in mean age, or insurance characteristics between patients stratified based on whether care was initiated by a physician signing a preprinted guideline-specific admission order sheet.

The homogeneity of the population, as defined by the selection criteria, was reflected in the inclusion of 90% of the population in the single All Patient Refined Diagnosis Related Group (APR-DRG) classification of 096 that is common for patients with asthma or bronchiolitis.¹³ There was no difference between different years of admission in this APR-DRG 096 predominance. Disease severity rating, based on number, type, and mix of comorbid diagnoses, was mildly ill for all but 15% of the patients, who were classified as moderately ill. There was no significant difference in this distribution when tested by year of admission.

During the January 15 through March 27 periods in the 4 years before guideline implementation, yearly admissions for bronchiolitis increased at an average rate of 15% per year for a total increase of 52% (Figure 2). In 1997, after the introduction of the guideline, admissions decreased 29%. Admissions decreased 21% in 1998 and 23% in 1999, for a continuing total drop in admissions of 56% over the 3 years after guideline implementation.

In 1997, a guideline admission order was signed for 67% of eligible infants. In 1998 and 1999, this percentage remained statistically stable, at 71% and 70%, respectively.

The mean ± SD LOS for patients admitted before guideline implementation was 2.9 ± 2.0 days (median, 2 days; range, 1-16 days). In 1997, the mean ± SD LOS was 2.4 ± 1.3 days (median, 2 days; range, 1-9 days). This increased slightly in 1998, to 2.5 ± 1.7 days (median, 2 days; range, 1-11 days), and decreased again in 1999, to 2.1 ± 1.6 days (median, 2 days; range, 1-8 days).

Since there were significantly fewer infants with Medicaid or no insurance coverage in the study group after the guideline implementation, we examined LOS for these groups separately. The LOS decreased significantly after implementation for both groups: those with commercial insurance (from 2.7 ± 1.8 to 2.3 ± 2.7 days) (P<.001) and infants in the Medicaid/self-pay group (from 3.0 ± 2.1 to 2.4 ± 1.4 days) (P<.001).

The Medicaid/self-pay group was composed of infants with Medicaid health maintenance organization (HMO), Medicaid non-HMO, or no insurance coverage. We also looked at LOS for these 3 subgroups separately. For the Medicaid HMO group, LOS decreased from 3.0 ± 1.4 to 2.3 ± 2.1 days (P<.001) after guideline implementation. For the Medicaid non-HMO group, LOS decreased from 3.1 ± 2.1 to 2.6 ± 1.4 days (P<.001). For the noninsured group, LOS decreased from 2.7 ± 1.4 to 2.2 ± 1.2 days (P = .07).

In the years before the guideline, 69% of the admitted infants were given at least 1 β₂-agonist inhalation as part of their therapies. In the 3 years after implementation, this was reduced significantly, to 57% of the infants admitted (P<.001).

Fifty-seven percent of all patients received multiple doses of β₂-agonists in the 4 years before guideline implementation. After implementation, 41% of patients received multiple doses, a reduction of 28% (P<.001).

As illustrated in Figure 3, the greatest decrease in the use of repetitive doses of β2-agonists was from 55% in 1996, to 27% in 1997, which was the first year of implementation. This was followed by a sharp rise in β₂-agonist inhalation therapies to 54% in 1998 after the addition of the algorithm to the guideline. In 1999, subsequent to the removal of the algorithm, the repetitive use of β₂-agonists again began to fall, decreasing to 46%.

The respiratory assessment form was used for fewer than 10% of the cases. This held true even when an admission order sheet prescribing the use of this scoring instrument was signed by the physician.

In the 4 years before implementation of the guideline, 87% to 94% of patients received nasopharyngeal washes to test for respiratory syncytial virus antigen (Figure 4). This fell to between 42% and 44% in the 3 years after implementation (P<.001). Chest x-ray films were ordered in 65% to 77% of infants before implementation (Table 2), and 56% to 62% after the guideline was introduced (P<.001). There was a small, but significant, decrease from the 9% to 13% of infants who had blood gases ordered before the guideline, to 5% to 12% in the years after (P<.001). Similarly, 50% to 65% of children received antibiotics before and 43% to 56% after implementation (P<.001).

Readmissions to the hospital in the 10 days after discharge remained statistically stable, at 2% to 3% of discharge before and after guideline implementation (P = .97). Independent medical record reviews by a community-based practitioner uninvolved in the formulation of the guideline revealed no relationship between guideline recommendations and causes of readmissions.

Total hospital costs for inpatient care fell 15%, from a mean of $3297 during the 4 years before implementation to $2825 for the 3 years after (Table 2). These cost data were not normally distributed. When compared using the natural log of their values, these costs were found to be significantly different (P<.001). Costs for the use of all resources ancillary to bed occupancy fell 41% from a mean of $1028 for the 4 years before guideline implementation to $609 in the 3 years after. The differences in the natural logs of the means were significant (P<.001). Mean respiratory care costs fell 72%, from $293 per patient in the years 1993 to 1996, to $81 in the 1997 through 1999 periods. Because the respiratory care data set included zero values, these costs were tested after transformation to their square roots and the difference was significant (P<.001).

In a prior publication,¹ it was reported that in its first year of implementation this bronchiolitis guideline was associated with a reduction in resource utilization and hospital costs without producing a detectable increase in readmission rates or decreasing family satisfaction. Because the recommendations in the guideline were based on firm scientific evidence, we concluded that the first-year success was not surprising and consistent with another similar and independent report published by Adcock and coworkers.¹⁴

But the first-year success of a guideline does not affirm the efficacy of the instrument beyond its initial season.^2-6 In this report, we detail the results of our efforts to overcome problems reported by others with a demonstration that a bronchiolitis guideline can be sustained for at least 3 years beyond its first introduction. But, as reported herein, our success has not been perfect.

On the continuing success side, we have recorded a persistent drop in hospital admissions for bronchiolitis. This has been sustained sufficiently long that we are confident it cannot be attributed to a diminishing prevalence of the disease in the community. The Centers for Disease Control and Prevention respiratory syncytial virus disease prevalence reports for the nation also support this conclusion.¹⁵

Major components of this guideline are arguments against the routine use of β₂-agonist inhalations for therapy for patients with typical and uncomplicated bronchiolitis. This emphasis was based on numerous randomized clinical trials and meta-analyses of these trials raising serious doubt about the therapeutic efficacy of β₂-agonists in patients with bronchiolitis. Some of these studies have even indicated that treating bronchiolitis with inhalations can cause hypoxia in some infants.^16-19 In spite of the published evidence, these relatively aggressive therapies have remained prevalent.^{10,16,18,20-28}

The persistent use of β₂-agonists in spite of their questionable efficacies has been justified, because 9% to 44% of patients with bronchiolitis patients are reported to subsequently progress to develop asthma^{14,18,19,26-33} and, in the first presentation with wheezing, might be demonstrating this fact. Our experience at Children's Hospital Medical Center (CHMC) supports the lower range of these estimates. Of 3107 infants encountered in the CHMC emergency department between 1996 and 1999 for a first-time episode of bronchiolitis, 501 (16%) were subsequently reencountered in the emergency department and diagnosed as having asthma.

Because of uncertainties about the true risk of misdiagnosing asthma as bronchiolitis in an individual patient, the guideline used in our study does allow that practitioners might appropriately give at least 1 trial dose of a β₂-agonist inhalation to selected patients. But because effectiveness of an inhalation should be clear within an hour of its delivery, the guideline also recommends strongly that repeated doses of inhaled medications be reserved only for those rare patients with a major and sustained clinical improvement in respiratory function after the initial trial. It is interesting that the practitioners apparently rejected the use of the respiratory assessment tool provided to help them make these decisions. Furthermore, even when the use of this tool was prescribed by a physician's order, the nurses and respiratory therapists actually administering the therapies did not include the respiratory score in their documentation of the events. The reasons for this rejection are still unclear.

From anecdotal comments, we know that many practitioners appreciate the fact that the guideline development process included participation by physicians in private practice. They also applaud our explicit acknowledgment in the published recommendations that their use of the guideline was voluntary, and that we expect the recommendations to be applied only when they judged them to be in the best interest of the patient. Although supported again only by anecdotal exchanges with the users, we also interpret the success of this guideline as evidence that the guideline contents are compelling and convincing because the recommendations are supported by referenced evidence that can be reviewed and independently assessed by any practitioner with access to a library. Thus, in our guideline development process, we seem to have addressed, and, in many cases, even overcome, many of the barriers to guideline use identified and convincingly discussed most recently by Cabana et al.³⁴

Although debates continue over the relative strengths of content and implementation methods as influences in guideline success, it is clear, at least to us, that content is not neutral. This is suggested by the recidivism in guideline compliance that occurred in the second year of implementation. Recidivism was reflected as an anomalous increase in the use of β₂-agonist therapies subsequent to adding an algorithm to the document that, by error, introduced an ambiguous connecting arrow that might have been interpreted as a recommendation for a reversal in inhalation therapy practices. This has resolved since the algorithm was removed in the third year.

Finally, it is necessary to acknowledge certain limitations that this study shares with others that report salutary effects of an intervention when the conclusions are based on comparisons of populations that are not concurrent in time. It is simply not possible to control for all potential confounding factors when comparison populations are not concurrent. As a corollary, it is also impossible to prove beyond all reasonable doubt that an intervention alone is the sole cause of any effect measured as an outcome variable. However, to strengthen the link between the bronchiolitis guideline implementation as a likely influence on causing the changes observed and reported in this article, some potential and specific possible confounding factors in this study are worth brief discussion.

We need to certify that, after the guideline was implemented, practitioners did not try to avoid the guideline recommendations by admitting patients with bronchiolitis under a different diagnostic rubric, such as asthma. This is very unlikely in our institution, in which admitting diagnoses are almost invariably assigned by the emergency department physician and not by the referring practitioner. In addition, there were no increases in the number of asthma admissions during the implementation years. Thus, there is no evidence to suggest shifting of diagnoses as an explanation for the fewer bronchiolitis admissions.

The report by Shay et al³⁵ demonstrated that during 1980 through 1996, rates of hospitalization of infants with bronchiolitis increased substantially. Children's Hospital Medical Center experienced the same trend. Children's Hospital Medical Center is the only pediatric hospital in the greater Cincinnati area and serves as a referral center for southwestern Ohio, northern Kentucky, and southeastern Indiana. Ninety-eight percent of children requiring emergency or inpatient hospital services are referred to CHMC. While it has been estimated that the population of the greater Cincinnati area has increased by more than 500,000 over the last 10 years, the birth rate has dropped an estimated 2% yearly. Taking these estimates into account, the admission rate for bronchiolitis at the Children's Hospital was 13 per 1000 population before guideline implementation, and rising steadily, and fell to 8 bronchiolitis admissions per 1000 children after guideline implementation.

Our methods for implementation included daily ward visits by a clinical coordinator. This individual was extremely committed to the task of observing and assuring that barriers to implementation were removed. Our clinical coordinator has high credibility with physicians and ward personnel and these ward visits serve as a daily physical reminder of the guideline as a management tool. The effectiveness and importance of this individual cannot be overstated, but, unfortunately, as an uncontrolled independent variable, this importance cannot be quantitated nor isolated as a measurable variable effecting our success.

It is acknowledged that, coincident with the bronchiolitis implementation, other changes in hospital practices and organization occurred that may have played a role in improving patient care and financial outcomes. In particular, the conversion from a traditional ward attending coverage to more constrained hospitalist coverage, as described by Bellet and Whitaker,³⁶ occurred just before the guideline implementation. This may have contributed to some reduction in hospital days tabulated on 1997. The continuing improvements on subsequent years, however, are not considered associated with the hospitalist intervention that remained stable during the ongoing periods of measurement.

We report herein that an evidence-based clinical practice guideline for the care of patients encountered in a major pediatric care facility has been successfully sustained beyond the initial year of its introduction to practitioners in southwest Ohio.

Accepted for publication May 25, 2000.

We thank all of the additional members of the Bronchiolitis Clinical Effectiveness Committee: Kenneth Zwergal, MD; James Stark, MD, PhD; Donna Casey, RN, MSN; Mary Frey, RN; Tracey Gerstner, RN; Maria Knight, RN; Scott Pettinichi, RRT; Beverly Connelly, MD; Richard Ruddy, MD; Melvin Rutherford, esq; Jennifer Loggie, MD; Irwin Light, MD; Dorine Seaquist, RN; Elisa Immerman, MBA; and Betsy Bushman, MS, for their help with this project; Patti Gubser, for her administrative assistance; and the supportive cooperation of the nurses, physicians, and other practitioners at Children's Hospital Medical Center and in our community.

Corresponding author: Uma R. Kotagal, MBBS, MSc, Division of Health Policy and Clinical Effectiveness and the Institute for Health Policy and Health Services Research, 3333 Burnet Ave, Cincinnati, OH 45229-3039 (e-mail: Uma.Kotagal@UC.edu).