Objective: To assess the effectiveness and safety of implementingan inpatient management and discharge strategy based onempiric antibiotic therapy with ceftriaxone sodium and a guidelineto promote timely discharge for clinically stable patients hospitalizedwith community-acquired pneumonia.
Study Design: A cluster randomized controlled clinical trialwith 30 days of patient follow-up at 8 teaching hospitals and 17nonteaching hospitals nationwide.
Methods: Participants included 240 intervention patients and209 control patients admitted by 85 physician groups betweenDecember 1998 and December 1999. Within each hospital,defined physician practice groups were randomized to the interventionarm (physician notification coupled with ceftriaxone sodiumas empiric therapy) or control arm (neither component).Physicians in the intervention arm were notified when theirpatients met guideline criteria for clinical stability; physicians inthe control arm were not contacted.
Results: The median length of stay was 4 days in both studyarms. The observed reduction in costs associated with the interventionwas not statistically significant when cost outliers wereexcluded. Mortality, serious adverse event, and rehospitalizationrates did not differ significantly across study arms.
Conclusions: Implementation of an inpatient management strategybased on physician reminders coupled with empiric use of ceftriaxonesodium did not reduce length of stay or associatedmedical care costs for patients hospitalized with community-acquiredpneumonia. These negative findings are most likely dueto insufficient potency of the intervention, inadequate guidelineimplementation, or imbalances in baseline patient characteristics.
(Am J Manag Care. 2005;11:491-499)
Community-acquired pneumonia (hereafter referredto as pneumonia) results in more than 1million hospitalizations annually in the UnitedStates, at a cost of $9 billion in 1994 dollars.1-3 Two keycomponents of the inpatient management of patientswith pneumonia are the initiation of appropriate empiricantibiotic therapy and the discharge decision.Observational studies4,5 demonstrate that appropriateempiric antibiotic therapy is associated with improvedsurvival for this illness. In addition, prior studies6-8demonstrate variation in length of stay that is unrelatedto severity of illness at presentation, suggesting thatpatients with pneumonia could be managed more efficientlywithout compromising their safety.
Observational data from the Pneumonia PatientOutcomes Research Team cohort study9 demonstratedthat most inpatients with pneumonia reached stabilityby hospital day 3 based on an objective definition ofclinical stability and that 77% stayed in the hospital atleast 1 day after reaching stability. Managing physiciansbelieved that only about one fifth of patients remainedhospitalized 1 or more days after reaching clinical stability.10 Therefore, physicians may not accurately determinewhen patients reach clinical stability and, despitethe high costs of hospital care, may not use this informationto guide the hospital discharge decision. A 1-dayreduction in length of stay is associated with an estimatedmean savings of $680 per patient.11
In prior studies12-23 to assess the effectiveness ofvarious strategies to reduce the length of stay of adultshospitalized for pneumonia, the clinical criteria usedto guide hospital discharge varied widely, and thereduction in length of stay ranged from 0 to 3 or moredays. Limitations of many of these studies include nonrandomizedstudy designs12,13,15-17,20-22 and the enrollmentof homogeneously low-risk patients15,21 or smallnumbers of patients, often from a single studysite.13,14,16,17,19,21,22 The Community-Acquired PneumoniaIntervention Trial Assessing Levofloxacin (CAPITAL)study,18 conducted in Canada, demonstrated thata critical pathway addressing the admission decision,empiric antibiotic therapy, and antibiotic conversionand discharge decisions could safely improve the efficiencyof care for patients with pneumonia. A prior trialof a multifaceted guideline implementation strategyshowed modest reductions in the duration of parenteral antibiotic use and nonsignificant reductions inlength of stay for patients with pneumonia managed at7 Pittsburgh hospitals.23
The primary aim of this multicenter randomizedcontrolled trial was to assess the effectiveness of implementingan inpatient management strategy involvingempiric antibiotic therapy with ceftriaxone sodium andphysician notification to promote timely discharge forclinically stable patients hospitalized with pneumonia.We also compared the associated total medical carecosts and the medical outcomes of patients managed byphysicians in the 2 study arms.
Inpatient Management Strategy
Empiric Antibiotic Therapy.
Patients in the interventionarm received parenteral ceftriaxone sodium astheir initial antibiotic therapy (1 g/d administered intravenouslyor intramuscularly), with or without concomitanttherapy with a macrolide. Ceftriaxone sodium is awidely used agent for the empiric treatment of pneumonia,because of its broad antimicrobial coverage ofgram-positive and gram-negative bacteria, once-dailydosing, and favorable safety profile.24 During the timeframe of this study, sole use of this agent was recommendedas appropriate empiric therapy by subspecialtypractice guidelines.25 All approved medications for thetreatment of pneumonia, except for ceftriaxone sodiumas empiric therapy, were allowed in the control arm. Allnecessary medications to treat medical comorbiditywere allowed in both arms.
Guideline for the Discharge of Clinically Stable
The discharge guideline was based on a reviewof the medical literature and empiric evidence on thetime to reach clinical stability.9 Each component of theguideline was discussed by a national panel of experts inpulmonary medicine, infectious diseases, and internalmedicine until consensus was reached. In contrast tomany prior guidelines,14,15,19,21,22 the guideline shown inTable 1 is based on the premise that patients with pneumoniareach clinical stability at different times duringthe course of treatment, rather than on a fixed day ofhospitalization.
Implementation of the Discharge Guideline
A case report form operationalized the daily assessmentof patient stability, which began on hospital day 2(the day after admission). Patients in both study armswere to be assessed daily by on-site medical personnelbetween 6:30 and 9:00 AM and again between 2:00 and4:30 PM if the morning assessment indicated potentialfor stability later that day. Each patient's stability statuswas recorded on the case report form but not in themedical chart. When a patient in the intervention armmet all of the guideline criteria for clinical stability, themanaging physician was notified that the patient wasstable for discharge. At all times, managing physiciansretained autonomy for the discharge decision.Physicians who were managing patients in the controlarm were not contacted regarding their patients' stabilitystatus.
To ensure that all providers began with the sameknowledge base, participating physicians in both studyarms attended an initial educational session in whichthe empiric evidence base for the discharge guidelinewas presented. Although it is possible that this 1-timeprestudy session influenced provider behavior duringthe trial, prior work has shown that education alone isnot an effective means of changing provider behavior.26
Study Sites and Identification of Physician andPatient Participants
A national sample of study hospitals was selectedbased on the interest of site investigators, an adequatevolume of inpatients with a discharge diagnosis of pneumonia,and site reviews performed by Medical andTechnical Research Associates, Inc. The principalinvestigator at each site identified participating physiciansand their associated medical groups. Usually,groups were defined as existing practices, individualphysicians, or teams of physicians who formally orinformally provided patient care coverage for eachother. Less frequently, groups were created for the purposeof this study. Before randomization, participatingphysicians agreed to be randomized to either study arm.The project statistician (RAS) randomized the physiciangroups within each study site, using generatedpseudorandom numbers balanced in blocks of size 2.Randomization of physician groups within each siteminimized contamination of the physician-level interventionbetween physicians caring for the same patientand provided a control for hospital-level characteristics.Because of the nature of the intervention, physicianscould not be blinded to their study arm assignments.Although patients were not aware of their physician'sassigned study arm, the antibiotics were open-label.Screening of potentially eligible patients occurred at 31study sites in 17 states in the United States betweenDecember 1998 and December 1999. Patients wererecruited at the time of presentation to a study siteemergency department and followed up for 30 days.
Patient Eligibility Criteria
Patients with a working diagnosis of pneumonia anda chest radiograph positive for a new pulmonary infiltrate consistent with pneumonia were eligible for thestudy if they were at least 18 years of age and admittedfor care by a participating physician. At presentation,patients were classified according to their estimated riskof mortality using the Pneumonia Severity Index (PSI).27
We excluded patients who (1) were in PSI risk classV; (2) required mechanical ventilation or had activeunderlying pulmonary disease; (3) had serious comorbidillness; (4) required admission to a critical care unit;(5) were immunocompromised; (6) had a metastaticconcomitant infection; (7) were hospitalized for palliativecare only; (8) resided in a skilled nursing facility orwere homeless; (9) were pregnant, nursing, or of childbearingpotential and not using reliable contraception;(10) currently used illicit drugs; (11) had been in anacute care hospital within the past 10 days or had beenhospitalized for an established diagnosis of pneumoniawithin the past 30 days; or (12) had a known or suspectedhypersensitivity to ceftriaxone sodium,cephalosporins, or penicillins.
For patients who met all eligibility criteria, thepatient or legal representative was approached forinformed consent to participate in the study. This studywas conducted in full compliance with the principles ofthe Declaration of Helsinki28 or with the laws and regulationsof the United States, whichever afforded thegreater protection of the patient, and was approved byall relevant institutional review boards.
The study population was defined as all enrolledpatients except (1) inadvertently enrolled patientswhose physicians had an unconfirmed study armassignment at the time of patient enrollment and (2)patients who had at least 1 exclusion criterion presentat the time of enrollment that was identified after enrollment.Patients who did not receive a parenteral dose ofan antibiotic or who withdrew consent during the indexhospitalization were excluded after enrollment.
Study Assessments and Procedures
Baseline characteristics, processes of care, and medicaloutcomes were assessed for patients in both arms.Within 24 hours of admission, the medical record wasreviewed for the patient's medical history, use of previousantibiotics or concomitant medications, and resultsof laboratory tests used in the PSI. Demographic datawere obtained by patient interview. Vital signs wereassessed daily, and antibiotic treatment was recordedthroughout the hospital stay. Guideline discharge criteriawere monitored from hospital day 2 through the dayof discharge, and serious adverse events and mortalitywere monitored throughout the 30-day study period.Length of stay for the index hospitalization was definedas the day of discharge minus the day of admission. Inaddition, the 2 components of length of stay wereassessed separately, namely, time to stability (the day onwhich clinical stability was first documented minus theday of admission) and poststability length of stay (theday of discharge minus the day on which stability wasfirst reached). Through a retrospective review of thecase report forms, we ascertained whether the empiricantibiotic therapy, daily assessment, and physiciannotification components of the protocol appeared to beappropriately applied for each patient. Because pneumoniaguidelines have been revised since this study wasconducted, with the most recent guidelines recommendinga second- or third-generation cephalosporinplus a macrolide or a fluoroquinolone alone for wardpatients similar to the participants in this study,29,30 wealso assessed whether antibiotic therapy within 24hours was consistent with the most recent InfectiousDiseases Society of America (IDSA) guidelines.29
Hospital bills for all hospitalizations at study siteswere collected for the 30-day study period. Index hospitalizationcosts were estimated by adjusting the chargeson the UB-92 bills by the Medicare hospital cost-tochargeratios.31 The total cost of medical care for 30days included costs for the index hospital admissionand the postdischarge events of rehospitalization, outpatientphysician visits, emergency department visits,chest radiographs, antibiotic therapy, home health visits,and postacute care. The cost of the intervention wasnot quantified in this study, given prior data demonstratingthat the cost of such an intervention is low($3.20 per patient assessment) relative to the total costof inpatient care.23
This study was designed with 80% power to detect a0.7-day decrease in length of stay from an assumedbaseline of 5 days, using a 1-sided .05-level test. Thiscorresponds to a 14% to 15% decrement in total medicalcare costs. The sample size was adjusted for the clusteringof patients within physician groups,32 assuming amean of 2 patients per cluster and an intraclass correlationof 0.1. The target sample size was increased by 5%to allow for potential loss to follow-up.
Length of stay was considered as a time-to-event outcome,in which the event was hospital discharge.Patients still hospitalized for the index hospitalizationwere censored at hospital day 30. The primary samplingunit was the physician group. To account for the manytied event times and the correlation between patientswithin the same physician group, discrete proportionalhazards models were estimated using SUDAAN, release8.0.33 Preliminary analysis indicated significant nonproportionalityof the hazards with respect to PSI riskclass (< .001) but not treatment arm (= .14) or site(= .75). Subsequent analyses were stratified on PSIrisk class, with covariate adjustment for the site effects.
The same approach was used to analyze time to stabilityand poststability length of stay. Patients who weredischarged before reaching stability were censored on theday of discharge in the analysis of time to stability andwere excluded from the analysis of poststability lengthof stay. All of these time-to-event analyses were rerunto take into account compliance of the empiric antibioticregimen with the most recent IDSA guidelines.29
Except for descriptive purposes, costs were naturallog transformed; log costs were analyzed using a generalizedestimating equation approach with a robustvariance estimate.33,34 These analyses were adjustedfor PSI risk class and site. To assess sensitivity toextremely high costs for some patients, this analysiswas repeated with cost outliers excluded. Additionalsensitivity analyses included adjustment for thosepatient-level baseline characteristics that were imbalancedbetween the study arms at the .10 level basedon c2 statistics.
All-cause mortality was compared for patients in the2 treatment arms using an exact permutation test that acommon odds ratio is 1 across the 4 PSI risk class strata.35 Rehospitalization rates were compared using theSTRATE command in Stata.36 For the length of stay andcost outcomes, treatment effects were assessed using 1-sided .05-level tests. Two-sided tests were used for allother comparisons.
Five hundred thirty-six patients were enrolled at 28study sites. Of these, 20 patients were excluded becauseof unconfirmed randomization assignments of theenrolling physicians; 33 intervention patients and 30control patients were excluded because of discovery ofan exclusion criterion after enrollment; and 2 patientsin each arm withdrew their consent during the indexhospitalization. Therefore, the final study populationcomprised 449 patients from 25 sites in 11 states, ofwhom 53.5% were in the intervention arm (Table 2).Imbalances in patient enrollment by study armoccurred at some sites.
These 449 patients were enrolled by 201 physiciansin 85 physician groups (Table 2). More than 90% ofphysicians in each arm were attending physicians ingeneral internal medicine (55% overall), an internalmedicine specialty (27% overall), or family practice(11% overall) (data not shown). All but 2 of the participatinghospitals were designated as urban, and 17 werenonteaching community hospitals.
Baseline Patient Characteristics
Of the 240 patients in the intervention arm, 54.6%were 65 years or older, 46.7% were male, and 83.8%were white (Table 3). More intervention patients thancontrol patients were admitted from emergency departments.Except for current smoking, each comorbid conditionlisted in Table 3 was more prevalent amongintervention patients, particularly diabetes mellitus andthe presence of 2 or more comorbid conditions. The PSIrisk class distributions of the intervention and controlpatients were similar.
Almost all (99.6%) of the patients in the interventionarm received parenteral ceftriaxone sodium, and45.0% received a macrolide within the first 24 hours;4.2% started macrolide therapy between 24 and 48hours after admission. In the control arm, the most frequentlyused parenteral agents were cephalosporinsother than ceftriaxone sodium (56.0%), fluoroquinolones(31.1%), and penicillins (24.9%); 5.7%received ceftriaxone sodium. Among control patients,58.4% received a macrolide within the first 24 hours,and 1.4% started macrolide therapy between 24 and 48hours after admission. The median time from presentationto initiation of parenteral antibiotic therapy was3.8 hours for intervention patients and 4.3 hours forcontrol patients (= .005). In this study, 46.7% of interventionpatients and 79.4% of control patients receivedempiric therapy within 24 hours of presentation thatwas compliant with the most recent IDSA guidelines.29
Length of Stay, Time to Reach Stability, andPoststability Length of Stay
The median length of stay was 4 days for patients inboth study arms (Table 4). Compared with controlpatients of the same PSI risk class at the same studysite, intervention patients were less likely to be dischargedon a given day of hospitalization (relative risk ofdischarge, 0.88; = .90). Overall, 88patients were discharged before reachingstability (approximately 20% in eacharm), and 1 patient was dischargedbefore any stability assessment. Amongthe 353 remaining patients, 101 interventionpatients (54.0%) and 73 controlpatients (44.0%) were dischargedwithin 1 day of reaching stability.Compared with control patients,intervention patients took longer toreach clinical stability (relative risk ofreaching stability on a given day, 0.79;= .02) and were discharged morequickly once they became stable (relativerisk of discharge, 1.26; = .006).
We reran these analyses with adjustmentfor whether or not patientsreceived antibiotic therapy within 24hours that was compliant with the mostrecent IDSA guidelines.29 The hazardratios associated with the interventionwere virtually unchanged for length ofstay, time to reach stability, and poststabilitylength of stay. For each ofthese outcomes, the hazard ratio forpatients who received IDSA-complianttherapy within 24 hours rangedbetween 0.94 and 0.96 (> .61 foreach), indicating that within eachstudy arm the time to these events wasonly slightly shorter for patients whoreceived the currently recommendedempiric antibiotic therapy.
Total Costs of Medical Care
The estimated median total costs ofmedical care were $4686 per patient inthe intervention arm and $4517 perpatient in the control arm (Table 5).Log costs were lower for patients in theintervention arm whenthe analysis was adjustedfor PSI risk class and studysite (= .09) and were significantlylower whenalso adjusted for observedbaseline imbalances(= .03) (Table 6).Five of the 6 identifiedcost outliers were in thecontrol arm, including 4low-risk control patients.Log costs were not significantlylower in the interventionarm when thecost outliers were excluded(> .64).
Five intervention patientsand 3 control patientsdied within 30 days.No statistically significant differences were observedbetween intervention and control patients with respectto 30-day mortality (2.1% vs 1.4%, = .73), occurrenceof serious adverse events (15.8% vs 11.0%, = .10), orrates of rehospitalization (8.5% vs 7.6%, = .21).
Study Protocol Implementation
The protocol was implemented as planned for 184intervention patients (76.7%) and 181 control patients(86.6%). In the intervention arm, 9 patients received anempiric antibiotic agent other than ceftriaxone sodium;the discharge strategy was not implemented as intendedfor 38 patients (including no physician notificationfor 17 stable patients and incorrect notification regarding20 patients who were not stable); and neither componentwas implemented for 9 patients. In the controlarm, 6 patients received ceftriaxone sodium as empiricantibiotic therapy; the physician was notified incorrectlyof stability regarding 18 patients; and both ofthese events occurred for 4 patients.
In this large multicenter randomized controlled trialof patients hospitalized for pneumonia, implementationof an inpatient management strategy (empiricantibiotic therapy with ceftriaxone sodium and physiciannotification when patients met guideline criteriafor discharge) did not reduce overall length of stay. Weraise the following 4 possible explanations for thesenegative findings: (1) insufficient potency of the interventionto modify provider behavior, (2) inadequateguideline implementation, (3) imbalances in baselinepatient characteristics, and (4) systematic differencesin the empiric antibiotic therapies that affected thetime to reach clinical stability. Prior research conductedas part of the Cochrane collaborative review ofprovider behavior change has demonstrated thatreminders, including electronic and paper formats, areeffective means to change provider behavior.37 Priorwork also shows that telephone delivery of suchreminders by nursing staff, similar to the interventionused in this study, is an effective means of changingprovider behavior.38 However, such an approach byitself may not have been sufficient to promote earlierhospital discharge. A more potent multifaceted implementationstrategy may be necessary to ensure timelyphysician notification that a patient has reached clinicalstability. The fact that the study protocol was notimplemented as planned for almost one quarter ofintervention patients, primarily because of errors inthe physician notification process, suggests that therigor of this guideline implementation may have contributedto the observed null effect of the intervention.In addition, randomizing physician groups does notguarantee balance at the patient level, even theoretically.In this study, intervention patients were morelikely to have multiple comorbid illnesses. Althoughlength of stay and cost results were somewhat morefavorable to the intervention when the treatment comparisonswere adjusted for baseline imbalances, theseobserved imbalances did not account fully for thelonger time to stability observed in the interventionarm. Finally, the choice of empiric antibiotic may havedifferentially affected the time to clinical stability in the2 study arms. Although the empiric antibiotic regimenswere consistent with existing pneumonia guidelinesat the time this study was conducted,25,39 the use of ab-lactam alone in the intervention arm is not consistentwith current pneumonia guidelines.29,30 In this study,almost one third of control patients received fluoroquinolonesas empiric therapy; therefore, more controlpatients than intervention patients received empiricantibiotic therapy that was consistent with currentIDSA pneumonia guideline recommendations29; however,time to clinical stability and length of stay were notassociated with thereceipt of IDSA-recommendedtherapy. Oneprior study40 demonstratedthat treatmentwith empiric antibiotictherapy within 4 hours ofpresentation was associatedwith a reducedlength of hospital stay,with no clear associationbetween guideline-recommendedantibiotic therapyand length of stay.
In this study, clinicallystable patients in theintervention arm weresignificantly more likelyto be discharged on agiven day than comparablestable patients in thecontrol arm. The hospitaldischarge criteria inthe present study are virtuallyidentical to thosein a 2003 study of a multifacetedguideline implementationstrategy.23Although poststabilitylength of stay was notassessed in the 2003study, the relative risk ofdischarge for stableintervention patients inthe present study issomewhat higher thanthat of interventionpatients in the 2003study. Together, thesestudies provide some evidence that interventions tomonitor hospitalized patients with pneumonia for stabilityand notify their physicians when they are clinicallystable can facilitate discharge withoutcompromising patient outcomes. These findings areconsistent with a recent meta-analysis41 showing that,although modest reductions in length of stay are possible,the overall reductions are not statistically significantacross well-designed studies.
This study has limitations that are important toacknowledge. First, the exclusion of patients who wereseverely ill (PSI risk class V), were admitted fromnursing homes, or had severe comorbid illnesses limitsthe generalizability of the study findings. Second,we estimated costs indirectly, by adjusting charges onthe hospital bills by the Medicare hospital-specific cost-to-charge ratios. This imperfect approach is used toadjust for between-hospital differences in the markupover costs in hospitals that do not have state-of-the-artaccounting systems31 and should not bias our within-hospitalcomparisons. Third, like the CAPITAL study,18this study was not designed to assess the separateeffects of the empiric therapy and discharge guidelinecomponents of the proposed inpatient managementstrategy; unlike other studies,18,23 the conversion decisionwas not explicitly addressed. Fourth, we do notknow why some intervention physicians were not notifiedwhen their patients reached clinical stability;unfortunately, we did not collect data on the barriersand facilitators of guideline implementation.
In conclusion, our implementation of an inpatientmanagement strategy based on physician reminderscoupled with empiric use of ceftriaxone sodium did notreduce overall length of stay or associated medical carecosts for patients hospitalized with pneumonia. Thesenegative findings are most likely due to insufficientpotency of the intervention, inadequate guidelineimplementation, or imbalances in baseline patientcharacteristics.
We acknowledge the contributions of the study site investigators, whowere responsible for facilitating the initiation and conduct of this trial atthe participating sites. We also acknowledge the assistance of AntoinetteKlein, MS, and D. Scott Obrosky, MS, in the preparation of the manuscript.
From the Departments of Biostatistics (RAS, MKM) and Health Policy and Management(JRL), Graduate School of Public Health, Division of General Medicine, Department ofMedicine (LJH, MJF), and Center for Research on Health Care (RAS, JRL, LJH, MJF),University of Pittsburgh; and VA Center for Health Equity Research and Promotion and VAPittsburgh Healthcare System (RAS, MKM, JRL, MJF); Pittsburgh, Pa.
This research was funded by protocol NR15534/M44119 from Roche Laboratories,Inc., Nutley, NJ. The investigators from the University of Pittsburgh were responsible for thedesign, analysis, and reporting of this study. Under a separate contract with RocheLaboratories, Inc, an external clinical research organization, Medical and TechnicalResearch Associates, Inc, Natick, Mass, was responsible for the recruitment of sites andinvestigators, data collection, oversight of the conduct of the study at the sites, and provisionof a final database to the University of Pittsburgh investigators.
Address correspondence to: Roslyn A. Stone, PhD, Department of Biostatistics,Graduate School of Public Health, University of Pittsburgh, 304 Parran Hall, Pittsburgh, PA15261. E-mail: email@example.com.
1. Niederman MS, McCombs JI, Unger AN, Kumar A, Popovian R. The cost oftreating community-acquired pneumonia. 1998;20:820-837.
Vital Health Stat 13.
2. Kozak LJ, Lawrence L. National Hospital Discharge Survey: annual summary,1997. 1999;No. 144:i-iv, 1-46.
Semin Respir Crit Care Med.
3. Lave JR, Lin CC, Hughes-Cromick P, Fine MJ. The cost of treating patients withcommunity-acquired pneumonia. 1999;20:189-198.
Arch Intern Med.
4. Gleason PP, Meehan TP, Fine JM, Galusha DH, Fine MJ. Associations betweeninitial antimicrobial therapy and medical outcomes for hospitalized elderlypatients with pneumonia. 1999;159:2562-2572.
5. Houck PM, MacLehose RF, Niederman MS, Lowery JK. Empiric antibiotic therapyand mortality among Medicare pneumonia inpatients in 10 western states:1993, 1995, and 1997. 2001;119:1420-1426.
Am J Med.
6. McCormick D, Fine MJ, Coley CM, et al. Variation in length of hospital stay inpatients with community-acquired pneumonia: are shorter stays associated withworse medical outcomes? 1999;107:5-12.
J Gen Intern Med.
7. Lave JR, Fine MJ, Sankey SS, Hanusa BH, Weissfeld LA, Kapoor WN.Hospitalized pneumonia cases in Pennsylvania: outcomes, treatment patterns, andcosts in urban and rural areas. 1996;11:415-421.
8. Fine MJ, Singer DE, Phelps AL, Hanusa BH, Kapoor WN. Differences in lengthof hospital stay in patients with community-acquired pneumonia: a prospectivefour-hospital study. 1993;31:371-380.
9. Halm EA, Fine MJ, Marrie TJ, et al. Time to clinical stability in patients hospitalizedwith community-acquired pneumonia: implications for practice guidelines.1998;279:1452-1457.
Arch Intern Med.
10. Fine MJ, Medsger AR, Stone RA, et al. The hospital discharge decision forpatients with community-acquired pneumonia: results from the Pneumonia PatientOutcomes Research Team cohort study. 1997;157:47-56.
11. Fine MJ, Pratt HM, Obrosky DS, et al. Relationship of length of hospital stayand the costs of care for patients with community-acquired pneumonia. 2000;109:378-385.
Infect Control Hosp Epidemiol.
12. Ehrenkranz NJ, Nerenberg DE, Shultz JM, Slater KC. Intervention to discontinueparenteral antimicrobial therapy in patients hospitalized with pulmonaryinfections: effect on shortening patient stay. 1992;13:21-32.
Arch Intern Med.
13. Ramirez JA, Srinath L, Ahkee S, Huang A, Raff M. Early switch from intravenousto oral cephalosporins in the treatment of hospitalized patients with pneumonia.1995;155:1273-1276.
14. Siegel RE, Halpern NA, Almenoff PL, Lee A, Cashin R, Greene JG. Aprospective randomized study of inpatient IV antibiotics for community-acquiredpneumonia: the optimal duration of therapy. 1996;10:965-971.
15. Rhew DC, Riedinger MS, Sandhu M, Bowers C, Greengold N, WeingartenSR. A prospective, multicenter study of a pneumonia practice guideline. 1998;114:115-119.
Arch Intern Med.
16. Ramirez JA, Vargas S, Ritter GW, et al. Early switch from intravenous to oralantibiotics and early hospital discharge: a prospective observational study of 200consecutive patients with community-acquired pneumonia. 1999;159:2449-2454.
17. Ramirez JA. Switch therapy with β-lactam/β-lactamase inhibitors in patientswith community-acquired pneumonia. 1998;32:S22-S26.
18. Marrie TJ, Lau CY, Wheeler SL, Wong CJ, Vandervoort MK, Feagan BG;CAPITAL Study Investigators. A controlled trial of a critical pathway for treatmentof community-acquired pneumonia: Community-Acquired PneumoniaIntervention Trial Assessing Levofloxacin. 2000;283:749-755.
19. Omidvari K, de Boisblanc BP, Karam G, Nelson S, Haponik E, Summer W.Early transition to oral antibiotic therapy for community-acquired pneumonia:duration of therapy, clinical outcomes, and cost analysis. 1998;92:1032-1039.
J Healthc Qual.
20. Ross G, Johnson D, Kobernick M, Pokriefka R. Evaluation of a critical pathwayfor pneumonia. 1997;19:22-29, 36.
Am J Respir Crit Care Med.
21. Weingarten SR, Riedinger MS, Hobson P, et al. Evaluation of a pneumoniapractice guideline in an interventional trial. 1996;152:1110-1115.
22. Hendrickson JR, North DS. Pharmacoeconomic benefit of antibiotic stepdowntherapy: converting patients from intravenous ceftriaxone to oral cefpodoximeproxetil. 1995;29:561-565.
Am J Med.
23. Fine MJ, Stone RA, Lave JR, Obrosky DS, Hough LJ, Kapoor WN.Randomized trial of a practice guideline to reduce length of stay for patients withpneumonia. 2003;115:343-351.
Infect Dis Clin
24. Brown RB. Ceftriaxone intervention in an inpatient setting. 1996;9(suppl):S1-S6.
Am Rev Respir Dis.
25. Niederman MS, Bass JB Jr, Campbell GD, et al. Guidelines for the initialmanagement of adults with community-acquired pneumonia: diagnosis, assessmentof severity, and initial antimicrobial therapy: American Thoracic Society:Medical Section of the American Lung Association. 1993;148:1418-1426.
26. Davis D, O'Brien MAT, Freemantle N, Wolf FM, Mazmanian P, Taylor-Vaisey A. Impact of formal continuing medical education: do conferences, workshops,rounds, and other traditional continuing education activities changephysician behavior or health care outcomes? 1999;282:867-874.
N Engl J Med.
27. Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify low-riskpatients with community-acquired pneumonia. 1997;336:243-250.
Declaration of Helsinki.
28. World Medical Association. Rev ed. Hong Kong:41st World Medical Assembly; 1989.
Clin Infect Dis.
29. Mandell LA, Bartlett JG, Dowell SF, File TM Jr, Musher DM, Whitney C;Infectious Diseases Society of America. Update of practice guidelines for themanagement of community-acquired pneumonia in immunocompetent adults.2003;37:1405-1433.
Respir Crit Care Med.
30. Niederman MS, Mandell LA, Anzueto A; American Thoracic Society.Guidelines for the management of adults with community-acquired pneumonia:diagnosis, assessment of severity, antimicrobial therapy, and prevention. 2001;163:1730-1754.
31. Lave JR, Pashos CL, Anderson GF, et al. Costing medical care: usingMedicare administrative data. 1994;32(suppl):JS77-JS89.
Am J Epidemiol.
32. Donner A, Birkett N, Buck C. Randomization by cluster: sample size requirementsand analysis. 1981;114:906-914.
SUDAAN User's Manual, Release 8.0.
33. Research Triangle Institute. ResearchTriangle Park, NC: Research Triangle Institute; 2001.
34. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuousoutcomes. 1986;42:121-130.
StatXact 4 Windows.
35. Mehta C, Patel N. Cambridge, Mass: Cytel SoftwareCorp; 1998.
Stata Statistical Software, Release 7.0.
36. StataCorp LP. College Station, Tex:StataCorp LP; 2001.
37. Bero LA, Grilli R, Grimshaw JM, Harvey E, Oxman AD, Thomson MA;Cochrane Effective Practice and Organization of Care Review Group. Closing thegap between research and practice: an overview of systematic reviews of interventionsto promote the implementation of research findings. 1998;317:465-468.
Arch Intern Med.
38. Davidson R, Fletcher S, Retchen S, Duh S. A nurse-initiated reminder systemfor the periodic health examination. 1984;144:2167-2170.
Clin Infect Dis.
39. Bartlett JG, Breiman RF, Mandell LA, File TM Jr. Community-acquired pneumoniain adults: guidelines for management. The Infectious Diseases Society ofAmerica. 1998;26:811-838.
Arch Intern Med.
40. Houck PM, Bratzler DW, Nsa W, Ma A, Bartlett JG. The timing of antibioticadministration and outcomes for Medicare patients hospitalized with community-acquiredpneumonia. 2004;164:637-644.
41. Rhew DC, Tu GS, Ofman J, et al. Early switch and early discharge strategiesin patients with community-acquired pneumonia: a meta-analysis. 2001;161:722-727.