| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Pulmonary (Drs. Dean and Suchyta) and Medical Informatics (Dr. Aronsky) Divisions of LDS Hospital, Intermountain Health Care (Dr. Bateman and Ms. Hadlock), Salt Lake City, UT.
Correspondence to: Nathan Dean, MD, FCCP, Intermountain Health Care, 333 South 9th East, Salt Lake City, UT 84102; e-mail: slndean{at}ihc.com
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Design: Prospective study after implementation vs historic controls.
Setting: Four ambulatory, urgent-care facilities.
Patients: Four hundred sixty-three immunocompetent adults with radiographically confirmed community-acquired pneumonia.
Intervention: A pneumonia practice guideline including decision support logic was implemented for a 12-month period.
Measurements and results: After implementation, physicians used the pneumonia guideline form in 90% of cases. The percentage of patients admitted within 30 days decreased from 13.6% to 6.4% (p = 0.01). Only five patients before (2.5%) and three patients after (1.1%, p = 0.3) guideline implementation required subsequent hospital admission within 30 days after initial outpatient treatment. Only two deaths occurred in the study cohort, both outpatients before implementation. The positive predictive value was 14.4%, and the negative predictive value for admission was 98.8% after guideline implementation. Guideline recommendation for admission was more likely to be followed in patients with more risk factors and hypoxemia.
Conclusions: Decreased admission rate was observed after implementation of admission decision support in combination with specific recommendations for outpatient antibiotic therapy. Favorable outpatient outcomes suggest that implementation of decision support was safe.
Key Words: clinical protocols • patient admission • pneumonia • practice guidelines • severity of illness index • therapy
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Considerable variation in hospital admission rates for patients with community-acquired pneumonia has been demonstrated across geographic regions.8 9 One reason for the variation may be that practitioners overestimate patients’ 30-day mortality risk.10 An objective method of assessing mortality risk (the pneumonia severity index [PSI]) classifies pneumonia patients into five groups, with predicted mortality ranging from 0.1% in class 1 to 31.1% in class 5.11 Observing that mortality in risk classes 4 and 5 was 8.2 to 31.1% among inpatients and 12.2% (5 of 41) among outpatients, Fine et al11 recommended that all patients in classes 4 and 5 be admitted to the hospital. In addition, they proposed that patients in class 3 be considered for brief inpatient observation.
Before publication of the PSI, logic for determining need for admission among patients with community-acquired pneumonia was developed by a committee of physicians and nurses as part of a pneumonia practice guideline at Intermountain Health Care.12 Admission decision support is provided on an 8- by 11-inch form, and requires no calculations beyond summing risk factors. The model also provides specific recommendations for outpatient antibiotic therapy using an oral macrolide with or without daily parenteral ceftriaxone, or an oral ß-lactam. Beginning in 1995, the guideline was implemented at several rural facilities.13 14 After initial studies were completed, the guideline was implemented in four urban, ambulatory, urgent-care clinics (Instacares). We hypothesized that guideline implementation would reduce the rate of admission and decrease failures of outpatient therapy associated with later hospital admission or death.
The primary purpose of this study was to compare the Intermountain Health Care guideline recommendations for inpatient vs outpatient care with actual physician decision making and clinical outcomes at the Instacares before vs after guideline implementation. A secondary purpose was to determine whether the PSI predicts the need for hospital admission in this practice setting.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The Intermountain Health Care pneumonia guideline logic at the time of this study recommended hospital admission for patients with at least two risk factors (adapted from American Thoracic Society guidelines) if they also met an admission criterion (Table 1 ).12 The threshold of two was based on data that patients with fewer than two risk factors (as defined in that paper) do well treated as outpatients.16
|
88% without a room air PaO2 or
saturation ever documented, or
PaO2 < 55 mm Hg measured in
patients transferred to a hospital. Average barometric pressure in Salt
Lake City is 647 mm Hg. We studied all community-acquired pneumonia cases seen at the Instacares during two 12-month periods. The retrospective period was March 1994 to March 1995, before any discussion of the guideline with Instacare staff. Before guideline implementation, Instacare physicians had discussed the management of community-acquired pneumonia in annual team meetings, including review of the American Thoracic Society treatment recommendations. The guideline was first presented to Instacare physicians by a guideline team member in May 1995. Two more physician meetings, distribution of literature and guideline forms, and meetings between a nurse guideline team member and Instacare nurses occurred during the subsequent 8 months. The Instacare physician group determined to routinely follow the guideline in March 1996 by using the form providing decision support for determining site of care and antibiotic selection. We then began prospective data gathering, ending in March 1997. Initial prospective patient identification was performed using the practice guideline forms. Daily logbooks at each Instacare were reviewed weekly to identify pneumonia patients in whom the guideline form was not completed. Our patient list was further compared with a computer-generated list using International Classification of Diseases, Ninth Revision (ICD-9) codes 480 to 486, 487.0, and 511.0 in either primary or secondary position. The same methods were used for retrospective identification with the exception that the practice guideline form was unavailable. We found no cases in the prospective group identified by the practice guideline form who were not identified by ICD-9 coding or logbooks. Four charts of patients identified with pneumonia by ICD-9 coding during the earlier period could not be located; these patients were therefore excluded from analysis.
Instacare staff identified patients presenting with community-acquired pneumonia, and the physician then used the guideline form to support decision making at the time of patient encounter. Form content is summarized in Table 1 ; the actual form has been previously published.13 After completion, the form was faxed to a central office and/or made part of the medical record. No guideline development team member was involved at the time of patient encounter. Physicians were allowed to override guideline recommendations for or against hospital admission, and for antibiotic selection. No incentives or penalties affecting physician determination of site of care were present during either study period. The PSI was not provided to the Instacare staff and was not available as a published article until the last 3 months of the study period.
Included patients had at least two signs and symptoms of pneumonia
(fever, cough, purulent sputum, dyspnea, compatible physical
examination), physician diagnosis of community-acquired pneumonia, and
an infiltrate present on chest radiograph. In addition, all were 18
years old and immunocompetent (
10 mg/d prednisone or equivalent, no
other current chemotherapeutic or immunosuppressant medications, no
known HIV infection). After initial identification of patients with
possible pneumonia, individual chart review was performed by a single
reviewer (M.R.S.) to supplement information present on the guideline
sheet. Both before and after implementation, patient charts and
guideline forms typically recorded the presence of risk factors, but
might not specifically record the absence of other risk factors. If
specific information was not present in the chart or guideline form
(eg, prior history of splenectomy), the risk factor was
assumed not present. Similarly, if measurements of hematocrit, WBC
count, or BUN were not performed, the risk factor was assumed not to be
present.
Hospital admissions occurring within 30 days of the Instacare visit were detected from the Intermountain Health Care inpatient database, by follow-up Instacare visits, and also by phone follow-up performed routinely as part of clinical care 2 to 5 days after initial encounter. We believe it unlikely that undetected hospital admissions occurred in significant numbers using this methodology. Utah vital statistics data were checked against the patient database to decrease the likelihood that deaths occurring outside of an Intermountain Health Care facility would be missed.
This study was approved by the LDS Hospital institutional review board, Salt Lake City. Individual patient consent was not required as a condition of approval.
Data were analyzed by t test for comparing group means. Contingency table analyses were performed using Fisher’s Exact Test for homogeneity of proportions between groups. Risk factors for adverse outcomes were assessed for correlation with actual hospital admission through the use of backward stepwise logistic regression modeling. Only those independent variables significant at the 0.25 level or higher were entered into the multivariate regression model. The SAS (Version 6; SAS Institute; Cary, NC) and S-plus (Mathsoft; Cambridge, MA) statistical packages were used for all analyses.17 18 Results are expressed as mean ± SD. Performance of the decision support systems are reported as the positive and negative predictive values, as well as sensitivity and specificity using standard definitions. All tests are interpreted using a two-tailed significance level of < 0.05.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
After guideline implementation, physicians used the guideline form in 90% of pneumonia cases. The percentage of patients admitted within 30 days of initial Instacare visit decreased from 13.6 to 6.4% after implementation (27 of 199 patients vs 17 of 264 patients, p = 0.01). Only five patients before (2.5%) and three patients after (1.1%; p = 0.3) guideline implementation required subsequent hospital admission within 30 days after initial outpatient treatment. The mean age of admitted patients was 69 ± 16 years after implementation vs 57 ± 22 years before (p = 0.05). Increases in number of risk factors and PSI score after vs before implementation were not statistically significant. The number of patients admitted by number of risk factors is shown in Figure 1 . Twelve of 17 admitted patients after implementation had at least four risk factors. Only two deaths occurred in the study cohort. Both were outpatients before implementation whose death certificates indicated cardiac disease as the cause of death 8 days and 22 days after their initial Instacare visit. Both patients had shown clinical improvement in their pneumonia in outpatient follow-up before death. The patient who died at 8 days had three risk factors, and would have been recommended for admission by the guideline. His PSI score was 71, risk class 3, and he would have been recommended for inpatient observation by the prediction rule of Fine et al.11 The patient who died at 22 days had three risk factors, and would also have been recommended for admission by the guideline. Her PSI score was 81, risk class 3, and she would also have been recommended for inpatient observation by the prediction rule of Fine et al.11
|
|
|
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Only 6.4% of patients were treated as inpatients after guideline implementation despite a recommendation from the Intermountain Health Care pneumonia guideline to hospitalize 39.4% of them. The observed mortality of 1% before implementation and 0% afterward compares favorably with mortality data from other populations.6 It is possible that guideline-directed hospital admissions would have improved outcome in the two patients in the earlier period who died. The observed rate of subsequent hospitalization in this study was 1.7% vs 7.4% in the PORT population. Although the patients were not random-ized between outpatient and inpatient therapy, the low mortality rate and low frequency of patients subsequently requiring admission suggests that decision making by physicians in this study was more appropriate than either the recommendations of the Intermountain Health Care guideline or the prediction rule of Fine et al.11
The observed decrease in admission rate from 13.6 to 6.4% corroborates the decrease in pneumonia admission rate from 39 to 29% observed in an earlier pilot study of the Intermountain Health Care guideline in Sanpete County, UT.13 Although the small sample size in that study did not provide adequate power for detecting a statistically significant difference, the trend was similar to results from the current study. Higher admission rate in Sanpete was caused in part by the inclusion of emergency department patients. Atlas et al19 also reported a decrease in 30-day admission rate among lower-risk pneumonia patients (PSI classes 1 to 3) from 58 to 47% (p = 0.07) in an emergency department after implementation of a practice guideline. Physician decision making in that study was influenced by provision of the PSI, availability of home nurse visits, and free outpatient antibiotics. These results suggest that objective severity assessment tools, even if imperfect, may be helpful in supporting physician decision making regarding the need for hospital admission in patients with community-acquired pneumonia.
The number of risk factors enumerated on the guideline form correlated well with physician and patient decision making. Assessing severity by summing risk factors appears helpful, but the Intermountain Health Care guideline thresholds for recommending admission appear to have been inappropriately low. For example, the threshold oxygen saturation of < 88% was a modification for Salt Lake City’s altitude of a recommendation that patients with PaO2 < 60 mm Hg be admitted to a hospital.15 However, this recommendation was only followed in 41% of cases by Instacare physicians. A threshold oxygen saturation of < 85% for admission fits actual decision making better. Although the logic at the time of study used a threshold of two risk factors for admission, Figure 2 indicates that three or even four risk factors is a more appropriate admission threshold for Instacare patients. In addition, 60% of patients with uncontrolled coexisting illnesses were treated as outpatients. We have modified the decision support logic in our guideline as a result of these data, increasing the threshold number of risk factors to three, decreasing the threshold oxygen saturation to 85%, and changing the coexisting illness criterion to only those patients in whom coexisting illness separately requires admission. Applying the modified guideline logic to the Instacare database after implementation, the number of patients recommended for admission is reduced to 36, including 13 of 17 actually admitted. The positive predictive value increases to 36.1%, although the negative predictive value remains high at 98.2%. Instacare data support raising admission thresholds even higher, but at the risk of the guideline working less well in emergency department settings. The modified admission logic maintains our conservative intent that guideline recommendation of outpatient treatment for an individual patient be safe and reliable.
Although advanced age, underlying medical problems, and moderate disease severity were common in these patients, we suspect that self-selection occurs among patients with community-acquired pneumonia who seek medical care at an ambulatory clinic miles from a hospital. It has been suggested that patient preference regarding admission should influence the decision between inpatient and outpatient care.7 Although we did not investigate this question, it is possible that Instacare patients have lower assessments of their disease severity and need for hospital admission compared with hospital emergency department patients. If so, patient preference could influence Instacare physicians toward outpatient care.
Another possible explanation for the predominant outpatient treatment of pneumonia we observed is the orientation and practice style of the physicians staffing Instacares. The physicians are mostly younger internists who practice exclusively outpatient medicine, even though their training as students and house officers was predominantly inpatient. The outpatient orientation of these physicians, plus the capability of Instacares to administer parenteral antibiotics and provide daily outpatient follow-up care increases the likelihood of patients not being admitted. The differences in patients, physicians, and outpatient medical care delivery capability in the Instacares suggest the need for caution in generalizing the predominantly outpatient care of pneumonia to other settings.
Comparison of the Instacare population to the PORT study cohort indicates that age was the predominant factor in determining risk class in our patients. However, the link between age and mortality is partially explained by severity of disease and underlying illness.20 21 22 A potential limitation of the PSI is that despite the derivation and validation cohorts of > 50,000 patients, only 944 outpatients (41 in risk classes 4 and 5) were studied. We observed increasing likelihood of admission with increasing PSI score or risk class. However, the positive predictive value for admission of 0.28 for class 4 and 5 patients after guideline implementation seems too low to validate the PSI as an arbiter of the need for hospital admission in this practice setting. In addition, 4.9% of patients in risk classes 1 and 2 required inpatient hospital treatment despite a low risk of mortality, making up 36% of all patients admitted from the Instacares. Caution is warranted in interpreting the utility of the PSI inasmuch as physicians were not given the patients’ PSI scores during the study. In addition, the low pneumonia admission rate in the Instacares challenges the ability of any decision support logic to reliably identify patients requiring hospitalization.
The American Thoracic Society guidelines for the treatment of community-acquired pneumonia were based in part on the data from which the PSI was derived. The PSI was derived and validated as a mortality prediction model, although it has been proposed to be helpful in making more rational decisions about hospitalization.11 The Intermountain Health Care guideline differs from the PSI in not weighting risk factors, and thereby being simpler to use. It was developed using principles of continuous quality improvement and continues to evolve. In addition, it provides a specific recommendation for site of care on the basis of factors other than severity of illness, and includes recommendations for outpatient antibiotic therapy (Table 1) . Our study design and differences in approach between the two decision support systems prevent us from comparing their performance quantitatively.
A limitation of this study is the partially retrospective method of data collection. Incomplete documentation could underestimate the true number of risk factors and PSI in these patients. The guideline form was filled out prospectively in 90% of patients after implementation, making data collection potentially more complete than in patients before implementation. However, the equivalent number of risk factors and PSI scores before vs after implementation (Table 3) suggests that the accuracy of risk factor assessment was similar in both periods. Furthermore, the absence of on-site blood gas and blood chemistry analytic capability could be associated with failure to detect low blood pH, hepatic or renal disease, and electrolyte abnormalities. Limited laboratory capability is common in outpatient settings except emergency departments. However, if the patients had more risk factors and greater severity than data reflect, the observations of low admission rates compared with decision support recommendations, and excellent clinical outcomes would only be strengthened.
In conclusion, decreased admission rate for community-acquired pneumonia patients was observed after implementation of admission decision support in combination with specific outpatient antibiotic recommendations. However, physicians appropriately overrode guideline recommendations frequently, and thereby decreased the rate of admission. We have revised the admission thresholds of the Intermountain Health Care pneumonia guideline logic on the basis of this study. Additional study is needed to further refine the logic and to determine its applicability among community-acquired pneumonia patients in outpatient settings such as emergency departments and private offices.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Supported, in part, by an unrestricted educational grant from Pfizer USPG to the Deseret Foundation, Salt Lake City, UT. Pfizer had no role in study design, data collection, analysis, or interpretation, and no right to approve or disapprove preparation and publication of this article.
Received for publication March 24, 1999. Accepted for publication December 16, 1999.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Spindler and A. Ortqvist Prognostic score systems and community-acquired bacteraemic pneumococcal pneumonia Eur. Respir. J., October 1, 2006; 28(4): 816 - 823. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P B O'Donohoe, R Kessler, and T F Beattie Exploring the clinical utility of blood ketone levels in the emergency department assessment of paediatric patients. Emerg. Med. J., October 1, 2006; 23(10): 783 - 787. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. T. Rosenbloom, A. J. Geissbuhler, W. D. Dupont, D. A. Giuse, D. A. Talbert, W. M. Tierney, W. D. Plummer, W. W. Stead, and R. A. Miller Effect of CPOE User Interface Design on User-Initiated Access to Educational and Patient Information during Clinical Care J. Am. Med. Inform. Assoc., July 1, 2005; 12(4): 458 - 473. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Ramsdell, G. L. Narsavage, J. B. Fink, and for the American College of Chest Physicians' Home Management of Community-Acquired Pneumonia in the Home: An American College of Chest Physicians Clinical Position Statement Chest, May 1, 2005; 127(5): 1752 - 1763. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. A. Halm, C. Horowitz, A. Silver, A. Fein, Y. D. Dlugacz, B. Hirsch, and M. R. Chassin Limited Impact of a Multicenter Intervention To Improve the Quality and Efficiency of Pneumonia Care Chest, July 1, 2004; 126(1): 100 - 107. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. A. Halm and A. S. Teirstein Management of Community-Acquired Pneumonia N. Engl. J. Med., December 19, 2002; 347(25): 2039 - 2045. [Full Text] [PDF]
|
|
|
|
 |
|
 M. S. Niederman, L. A. Mandell, A. Anzueto, J. B. Bass, W. A. Broughton, G. D. Campbell, N. Dean, T. File, M. J. Fine, P. A. Gross, et al. Guidelines for the Management of Adults with Community-acquired Pneumonia . Diagnosis, Assessment of Severity, Antimicrobial Therapy, and Prevention Am. J. Respir. Crit. Care Med., June 1, 2001; 163(7): 1730 - 1754. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
