Original Research INFECTION Pneumonia* Criteria for Patient Instability on Hospital Discharge Albedo Capelastegui, MD, PhD; Pedro P. Espafia, MD; Anu...

659KB Sizes 0 Downloads 3 Views

Recommend Documents

No documents
Original Research INFECTION

Pneumonia* Criteria for Patient Instability on Hospital Discharge Albedo Capelastegui, MD, PhD; Pedro P. Espafia, MD; Anulia Bilbao, MSc; Marimar Martinez-Vaxquez, MD; lnmaculadu Gorordo, MD; Mike1 Orihe, MD; Isabel U m t i a , MD, PhD; andJose' M. Quintana, MD, PhD

Background: A study was undertaken to identify and weigh at the time of hospital discharge simple clinical variables that could predict short-term outcomes in patients with pneumonia. Methods: In a prospective observational cohort study of 870 patients discharged alive after hospitalization for pneumonia, we collected oxygenation and vital signs on discharge and assessed mortality and readmission within 30 days. From the P-parameter obtained in a multivariate Cox proportional hazard regression model, a score was assigned to each predictive variable. The effects of instability at discharge on outcomes within 30 days thereafter were examined by adjusted models with use of the pneumonia severity index at hospital admission, the length of stay, the Charlson comorbidity index, or the preillness functional status. Results: Four variables related to a 30-day mortality rate from all causes were identified in the multivariate model; these included one major criterion (temperature > 37.5"C) and three minor criteria (systolic BP < 90 mm Hg or diastolic BP < 60 mm Hg, respiratory rate > 24 breathslmin, and oxygen saturation < 90%). The developed score remained significantly associated with a higher risk-adjusted rate of death. Patients with a score 2 2 (one major criterion or two minor criteria) had a sixfold-greater risk-adjusted hazard ratio (HR) of death (HR, 5.8; 95% confidence interval, 2.5 to 13.1). Conclusions: Four criteria of instability on discharge seem to be related to the mortality rate after discharge, but each of the factors must be weighed differently. The resulting score is a simple alternative that can be used by clinicians in the discharge process. (CHEST 2008; 134:595-600) Key words: instability on discharge; longitudinal study; pneumonia; predictive models Abbreviations: AUC = area under the receiver operating characteristic curve; CI = confklence interval; CURB-65 = confusion, urea nitrogen, respiratory rate, BP, age 2 65 years; HR = hazard rabR PSI = pneumonia severity index

ractice guidelines for management of pneumonia aim to reduce the variation in key aspects of care thus, to improve the efficiency and effectiveness of health care.1,2 Although the criteria for clinical stability that must be met for hospital discharge have been considered key factors in the care of patients admitted with pneumonia,3.4 there have been few clinical indicators proposed to assess readiness for hospital discharge.5-8 The aims of our study were to analyze the basic indicators of clinical instability at discharge proposed by Halm et al,6.7 and to evaluate their relationship

P and,

and importance to mortality rates and readmission. We first hypothesized that each of the key variables to measure clinical stability may have a different weight and significance in predicting short-term outcomes. Our second hypothesis was that death and readmission in the short term after discharge are such different outcomes that they may not fulfill the three requirements for a composite end point: a similar relative risk reduction, similar frequency, and similar importance to patients.9 Hence, an observational study involving hospitalized patients with pneumonia was performed. CHEST I 134 1 3 1 SEPTEMBER, 2008


MATERIALSAND METHODS Setting of Stzrdy This study was performed at Caldakao Hospital (Spain), a 400-bed, nonurban teaching general hospital serving a population of 300,000 inhabitants that provides free unrestricted care to nearly 100% of the population. The project was approved by the hospital ethics review board. Study Sample

All patients 2 18 years old who were hospitalized with pneumonia consecutively between July 15, 2003, and June 30, 2006, were prospectively enrolled in an observational cohort study. Pneumonia was defined by clinician judgment in combination with a new infiltrate on chest radiograph. Patients were excluded if they were known to have a positive test result for HIV, were chmnicdy immunosuppressed,or were hospitalized for the previous 14 days.A total of 945 patients were admitted to the hospital for pneumonia; 75 of these patients (7.9%)died in the hospital. This study sample was restricted to the total of 870 patients who survived the index hospitalization. Dato Collection

During hospitidization, patient care was managed according to a clinicill guideline."' The in-hospital assessment included all the variables of the pneumonia severity index (PSI)lI and the variables included in the CURB-65 (confusion, urea nitrogen, respiratory rate, BP, age 2 65 years) scorekf recorded within 24 h of hospital admission, and an assessment of preadmission functional status. After discharge, the care of all patients was managed by their family physicians, and a control visit at our center at 30 days.

The stability on discharge criteria were obtained two times in the last 24 h before hospital discharge, and the worst was taken for final decision. These data were obtained by previously trained study personnel. A patient was in stable condition if the temperature was 5 375°C (we also assessed a cut-off point of 37.8"C. as used by Halm et al7), heart rate was 5 100 beatdmin, respiratory rate was 5 24 breathdmin, and systolic BP was 2 90 mni Hg and/or diastolic BP was 2 60 mni Hg. Oxygenation was considered stable if the oxygen saturation rate was z 90% or the Pao, was 2 60 mm Hg. Patients whose oxygenation was measured while they still were receiving supplemental oxygen during hospital stay, with a fraction of inspired oxygen 5 24% or no more than oxygen at 1Wmin via nasal cannula, were considered to be in stable condition at discharge if they had an oxygen saturation rate z 95%. Patients considered to have unstable oxygenation on discharge were sent to their homes with supplemental oxygen. Patients who had used supplemental oxygen at home before hospital admission were not considered to have unstable oxygenation on discharge. All patients at discharge were able to eat (or resume long-term tube feeding) and to receive oral medication. outcofrw~s

The outcomes for this study were death from all causes or hospital readmission within 30 days and 45 days after discharge. Vital status and readmission information for all patients were determined initially by telephone interviews up to 90 days after discharge. All reported deaths and dates of deaths were confirmed by a review of medical reports, public, death registries, or both. All discharge diagnoses were determined for each readmission. Readmission was classified as pneumonia related if pneumonia was an immediate or underlying cause of readmission or if' it played a major role in the readmission. None of the patients were readmitted to other hospitals.


We assessed pre-illness functional status from 2 weeks before hospital admission by inquiring about the performance of 15 daily activities, which were an expanded version of the activities of daily living index published by Katz et al.13 Previous studies'4-" have demonstrated the validity of retrospective reports for assessing functional status prior to hospitalization in acutely ill patients. The activities were graded according to a 4-point system. A snmmary score was obtained by the sum of the scores across all 15 activities (range, 15 to 52: with 15 being autonomous fiinction in all recorded activities).18 Informed consent was ohtained, and trained clinicians conducted structured interviews with patients and family members within 72 h of hospital admission. 'From the Pneumology Service (Drs. Capelastegui, Espaiia, Gororclo, Oribe, and Urrutia), Department of Emergency Medicine (Dr. Martinez-Vazquez).oid Research Unit (Dr. Quintana), f 10s ital de Galdakao-Usansolo-CIBER Epidemio~ogia Salud P d i c a , Galdakao; and the Basque Foundation for Healti Innovation ant1 Research-CIBER Epidemiologfa y Salud Phbhca (Mrs. Bilhao), Sondika. Bizkaia, Spain. The tiuthors have no conflicts of interest to disclose. Manuscript received December 21, 2007; revision accepted March 25, 2008. Reprodiiction of this article is rolubited without written pemGssion from the American College ofchest Physicians (www.chestjournd. ordmisdre rintr shtinl) Correspon(fe w e- to. ' . Alherto ' Ca ielastegui, MD, PhD, Service of Piieiiinology, Hospital [le Gal[lui&o-Usansolo, e-48960 Galdakao, Bizkaia, Spain; e - m i l : [email protected] DOI: 10.1378/chest.07-3039 596

Statistical Analysis

Descriptive statistics included frequency tables, mean, SDs, and median. Sociodemograpliic and clinical characteristics of patients responding to the Katz uestionnaire were compared to . those of the nonresponders. x and Fisher exact tests were performed for categorical variables, and the Student t test and nonparainetric Wilcoxon test were used for continuous varialh. To identlfy which instability criteria were associated with death or readmission within 30 days, univariate and multivariate Cox proportional hazard regression models were used. We assigned H weight to each instability criterion in relation to each P-parmeter. To obtain the total instability score, we added the weights of each of the selected variables. We performed the same analysis with logistic regression models. Effects of the instability score on unadjusted and risk-adjusted 30-day inortality were examined by logistic and Cox regression models. We fitted the first adjusted model with the PSI and the history of COPD, which was the same as Hdm et d,7and the second with the CURB-65 score, Katz index, Charlson coinorbidity index, and length of stay. Kaplan-Meier graphs were constructed for the instability score categories, and comparisons were performed by the log-rank test. Finally, we estimated the sensitivity,specificity,positive and negative predictive \ d u e s , aid the area under the receiver operating characteristic curve (AUC) for different cut-off points of the instability score. All ef'fects were considered significant at p < 0.05. All statistical analyses were performed using SAS for Windows statistical software (version 8.0; SAS Institute: Cary, NC) and S-Plus 2000 (MathSoft; Seattle, WA).

2 '

Original Research

Table lxharacteristics and Outcornes of 870 Patients Hospitalized With Pneumonia Who SuroiVed to Hovital Discharge*

Table %Frequency of Zmtability Variables on Discharge (n = 870) Variables



Host related Age, r Age 65 Nursing home resident Female gender Charlson comorbidity index 0 1-2 2 3 Baseline functional statust 15 1629 2 30 Severity of illness on admission4 PSI risk class 1-111 IV-v PSI CURB-65 0+1 2 >2

Length of stay. d Median 1 3 d Mortality 30 d after discharge 45 d after discharge Readmission for all causes 30 d after discharge 45 d after discharge

69.9 5 16.1 618 (71) 64 (7.4) 309 (35.5) 281 (32.4) 478 (55.1) 108 (12.5) 300 (38.4) 257 (32.9) 224 (28.7)

423 (48.6) 447 (51.4) 91.4 ? 31.8 358 (41.2) 317 (36.4) 195 (22.4) 4.4 5 4.2 3 445 (51.2) 29 (3.3) 35 (4) 72 (8.3) 94 (10.8)

*Data are presented as No. (%) or mean ? SD unless otherwise indicated. tMeasured by Katz Data refer to the suminary score (n = 781). The characteristics of those patients in the sample for whom preihess functional status was not assessed did not vary significantly from those who had undergone assessment of preillness functional status (functional status range, 15 to 52; excellent Functional status, 15). IAssessed with the PSI and CURB-65 score.

RESULTS Characteristics of the study subjects are provided in Table 1. Of the 870 patients discharged alive, pneumonia-related causes accounted for 27 of the total readmissions (37.5%)within 30 days, and for 34 of the total readmissions (36.2%)within 45 days. The most frequent causes of pneumonia-related readmission within 30 days were a new rise in temperature and increased respiratory symptoms (18 patients, 66.7%)and pleural effusion (6 patients, 22.2%). Of those readmitted for pneumonia-related causes, two patients died and one patient needed vasoactive treatment after being readmitted. The rest of patients who died without being readmitted showed severe comorbidities, with an elevated index of dependency, and old age.

Temperature, "C > 37.5 > 37.8 Systolic BP < 90 mm Hg and/or diastolic BP < 60 mm Hg Respiratory rate > 24 breathdmin Heart rate > 100 beatdinin Oxygen saturation < 90% Number of instability factorst 0 1 >1

No. (%)* 14 (1.6)

7 (0.8) 120 (13.8) 139 (16.6) 60 (7) 138 (15.9) 485 (58) 255 (30.5) 96 (11.5)

*Percentages exclude patients with missing data. tThe criterion for temperature instability is > 37.5%

Instability Variables on Discharge Associated With outcom4?s Instability variables on discharge are provided in Table 2. Table 3 shows the variables associated with outcomes in the univariate and multivariate analyses, We found that the relationship between the instability variables on discharge and the risk of death or readmission within 30 days was similar to the results within 45 days. At discharge, only an oxygen saturation rate < 90% was associated with a higher likelihood of readmission within 30 days for pneumoniarelated causes in the univariate analysis (hazard ratio [HR], 3.2; 95% confidence interval [CI], 1.5 to 6.9) and in the multivariate analysis (HR, 3.1;95% CI, 1.4 to 7.2). Points were assigned to each predictive variable from the P-parameter obtained in the 30-day mortality multivariate model. Variables were grouped into major (temperature > 37.5"C, 2 points) and minor (systolic BP < 90 mm Hg and/or diastolic BP < 60 mm Hg, respiratory rate > 24 breathdmin, and oxygen saturation < 90%, 1 point respectively) criteria for instability. We totaled the points assigned to each variable and determined a score for each patient. Patients with a score of 0 (no instability criteria) or with one minor criterion showed low risk of death after discharge (death rate, 2.3%). If the score was 2 2, the prediction of 30-day mortality could be based on the presence of one major criterion or two ininor criteria. The logistic regression model for the continuous score had an AUC of 0.69, and for a score 2 2 the AUC was 0.66. All the analyses were repeated with a temperature variable of > 37.8"C, and very similar results were obtained. Logistic regression models confirmed all previous findings. The score established with the variables for instability at discharge remained significantly associated CHEST I 134 I 3 I SEPTEMBER, 2008


Table &Relation of h t a b i l i t y Variables on Discharge by Univariate and Multivariate Cox Proportional Hazard Regression Analysh With Mortality and Readmission for dl Causes (n = 870)* Univariate

' Variables Mortality Temperature > 37.5"C Systolic UP < 90 min Hg and/or diastolic BP < 60 mim Hg Respiratory rate > 24 breathshin Oxygen saturation < 90% Heart rate > 100 beatshin Readmission Temperature > 37.5"C Systolic BP < Yo inin Hg and/or cliastolic BP < 60 inin Hg Respiratory rate > 24 hreaths/min Oxygen saturation < 90% Heart rate > 100 heatshnin




Estimated p Coefficient

30 d After

4Fj d After



4.8 (1.1-20.3) 2.9 (1.3-6.4)

4 (1-16.6) 3.4 (1.7-6.8)

3.2 (1&6.7) 2.9 (1.3-6.2) 1.6 (0.5-5.2)

2.7 (1.e3-5.4) 2.5 (l.LL5.1) 1.3 (0.44.2)

0.9 (0.1-6.2) 0.7 (0.3-1.4)

1.3 (0.35.4) 0.7 (0.4-1.4)


1.4 (0.8-2.6) 2.1 (1.3-3.5) 0.4 (0.1-1.5)

1.6 (1-2.6) 2 (1.3-3.2) 0.4 (0.1-1.3)


30 d After


p Value


1.49 0.97

4.5 (1-19.2) 2.6 (1.25.8)

0.0450 0.0174

2 1

0.87 0.85

2.4 (1.15.2) 2.4 (1.1-5.2) 0.9 (0.2-3.6)t

0.0294 0.0342

1 1

- 0.43

0.9 (0.1-6.2) 0.7 (0.3-1.4)

0.8812 0.2876

0.30 0.61 - 1.07

1.4 (0.8-2.4) 1.8 (1.1-3.2) 0.3 (0.1-1.4)

0.3128 0.0301 0.1355



*Data we presented as HR (95%CI) unless otherwise stated. Each instability criterion was examined individually in the univariate analysis, and all inst'ibility criteria were examined jointly in the multivariate analysis t Data forced the heart rate variable to multivariate analysis.

charge, but with significantlydifferent weights for each variable. At discharge, only one variable, oxygen saturation < 90%,was related to readmission for all causes and to pneumonia-related readmission within 30 days. This study confirms previous findings7.8 but highlights the following relevant information: (1)not all the previously proposed instability variables on discharge were associated with short-term mortality. Those that had an association differed in the magnitude of their predictive capacity, allowing us to develop an instability score; (2) short-term outcomes of readmission had very little association with the instability criteria on discharge; and (3) from a clinical perspective, the prognostic value of the selected instatdity variables is limited (AUC, 0.66). New studies are needed to [email protected] other important factors related to death besides instability at discharge.

with posthospital 30-day mortality after controlling for other important potential confounders (Table 4). When a score 2 2 was applied, it showed a positive predictive value of 13.5%and a negative predictive value of 97.7% (Table S).The survival plot of death for the instability score categories on discharge are shown in Figure 1.

DISCUSSION Our study showed that four easily obtainable measures of patient instability on discharge-one major criterion (temperature > 37.5"C) and three minor criteria (systolic BP < 90 min Hg and/or diastolic BP < 60 mm Hg, respiratory rate > 24 hreathdmin, and oxygen saturation < 9O%)-were independently related to 30-day mortality after dis-

Table 4-Efiects of Proposed Score for lrsatability at Discharge on Unadjusted and Risk-Adjusted 30-Day Mortality Unadjusted Variables


HR (95% CI)

First Adjusted Model* p Value


'HR (95% CI)

p Value

Second Adjiisted Modelf



'HR (9s%C;I)



2.8 (1.74.5) 5.8 (2,%13.1)

< 0.000 I < O.OO()l


2.4 (1-5.9)


Definition of instability on discharge Score4 Score 2 2$ No. of instability factors 2 I vs 0

2.3 (1.7-3.2) 6.4 (3-13.3)

< 0.0001 < 0.0001

3.2 (1.5-6.9)


2.2 (1.5-3.1) 4.2 (2-9) 2.3 (1-4.9)

< 0.0001

0.05 15

"First model adjusted by PSI and history of COPD (n = 839). tSecond model adjusted by CURB-65 score, Katz index, Charlson comorbidity index, and length of stay in = 759). $The total score for instability is obtained by adding the weights of each of the variables selected: temperature > 37.S0C, 2 points: systolic BP < 90 rnm Hg and/or diastolic BP < 60 mm Hg, respiratory rate > 24 breathshin, and oxygen saturation < 90%, 1 point respectively. $A total score 2 2 points is the level at which a patient is considered unstable. 598

Original Research

Table 5-Cut-off Point8 of Znstabicfty at Discharge Score for Death Within 30 Day8 Score of Instability at Discharge* 2 2 2 2

1 2 3 4

No. of DeathsRotd (%)

Sensitivity, %

Specificity, %

Positive Predictive Value, %

Negative Predictive Value, %


19/317(6) 12/89(13.5) 1/6 (16.7) 1/2 (50)

65.5 41.4 3.5 3.5

63.2 90.5 99.4 99.9

6 13.5 16.7 50

98.1 97.7 96.6 96.7

0.64 0.66 0.51 0.52

*The total score was obtained by adding the weights of each of the selected variables.

In contrast with previous studies,7VRwe have not used any composite end point, such as death and/or readmission, to evaluate instability variables on discharge. Use of composite end points usually assumes that the effect on each of the components will be similar and will have a similar level of importance for the patient.9 Our findings indicate that the relation between instability variables on discharge and postdischarge mortality differs substantially from the relation between these variables and their weights with readmission. We did not consider all instability factors to be of equal weight, and we believe that this issue is clinically important for making discharge decisions. We have observed that temperature (major criteria) was a stronger indicator than hypotension, oxygenation, and respiratory rate (minor criteria), and that heart rate had no value. Why might an elevated temperature increase the risk of death? We hypothesized that the criteria of instability on discharge may be surrogate markers of systemic inflammation that can consistently predict a poor prognosis. Indeed, the levels of inflammatory cytokines induced during an episode of pneumonia










1 .







B I_

2 -.






have been correlated with the severity of pulmonary infection and with prognosis.19,20 It is relevant that our data reflect the medical practice from 2003 to 2006. Compared with an earlier study,7 our analysis shows an important reduction in the length of stay and a higher rate of instability on discharge. However, our outcomes (mortality or readmission) up to 30 days were similar to those in the study by Halm et al.7 Although our length of stay has progressively been reduced since 2000,21,22the rates of death and readmission within 30 days after discharge were similar.22 A recent study showed that most rehospitalization cases following pneumonia are comorbidity related and are the result of underlying cardiopulmonary and/or neurologic diseases.2" The authors observed no association between readmission for all causes and instability at discharge. Our study underlines this issue: we did not find that instability on discharge had much correlation with readmission for all cause or with pneumonia-related readmission. This raises the issue of whether readmission rates may not be a good determinant of quality of care at the hospital. Some other factors such as influenza vaccination before hospitalization,24 quality of communication at discharge, or home care intervention25 may play a more important role. The strengths of this study are its prospective design, the relatively large sample of unselected patients, comprehensive assessment of outcomes, standardized assessment of physical function, detailed collection of clinical data, and use of a robust risk-adjustment model that included preillness functional status. Our study also had some limitations. First, because it was conducted in a single geographic area, it may reflect a single standard of practice. However, the prognostic variables are similar to those in previously published models, which suggest that these findings were not population specific. Likewise, the clinical characteristics of patients admitted to our hospital did not differ from those in studies in the United States" and Europe.12 Second, the precise cause of death in our subjects was not obtained. Although potentially available from death certificates, the deficiencies and lack of CHEST/ 134 f 3 f SEPTEMBER, 2008


reliability in this approach are well docurnented.z6In our study, most patients died at home, which implies that reliable information for cause of death was unavailable. Third, we did not include mental condition as a stability criterion. However, at discharge our patients could rise from bed, walk (except for previous incapacity), take oral medication, and eat (or resume long-term tube feeding). Fourth, we used the same data set to derive the prediction model and to test the model. Under these circumstances, the performance of the model is often overestimated. Finally, following Halm et al,7 patients were defined as having stable oxygenation if they were receiving supplemental oxygen and oxygen saturation was 2 95%. Some of these patients may have been hypoxic if on room air but this is unlikely with a fraction of inspired oxygen no more than 24% or oxygen no more than 1 Umin via nasal cannula, as it was in our case. In conclusion, our study confirms instability on discharge as a marker of posthospital mortality. Additionally we have described which of these criteria are significant, together with their weights. However, we did not find that instability on discharge had much correlation with readmission. From a clinical perspective, patients with a score of 0 (no instability criteria) or with one minor criterion seem to have a low risk of death after discharge; however, our study implies that patients with a score 2 2 (one major criterion or two minor criteria) should be monitored closely because their risk seems to be high. Such a proposal should be validated in other studies and settings. ACKNOWLEDGMENT: We thank the staff members of the different services for their support. We also wish to thank Ms. Sally Ebeling for her assistance editing the manuscript.

REFERENCES 1 Miller J, Petrie J. Development of practice guideline. Lancet 2000; 35582-83 2 Marrie TJ, Lau CY, Wheeler SI, et al. A controlled trial of a critical pathway for treatment of community-acquired pneumonia. JAMA 2000; 283:749-755 3 Macfidrlane JT, Boldy D. 2004 update of BTS pneumonia guidelines: what’s new? Thorax 2004; 59:364366 4 Kosecoff J, Kahn KL, Rogers WH, et 0.1. Prospective payment system and impairment at discharge: the “quicker-and-sicker” story revisited. JAMA 1990; 264:1980-1983 5 Halin EA, Mittman BS, Walsh MB, et al. What factors influence physicians’ decisions to switch from intravenous to oral antibiotics for community-acquired pneumonia? J Gen Intern M d 2001; 16:599-605 6 Halm EA, Fine MJ, Marrie TJ, et al. Time to clinical stability in patients hospitalized with community-acquired pneumonia: implications for practice guidelines. JAMA 1998; 279: 1452-1457 7 Halm EA, Fine MJ, Kapoor WN. et al. Instability on hospital discharge and the risk of adverse outcomes in patients with pneumonia. Arch Intern Med 2002; 162:1278-1284


8 Dagan E, Novack V, Porath A. Adverse outcomes in patients with community acquired pneumonia discharged with clinical instability from internal medicine department. Scand J Infect Dis 2006; 38:860-866 9 Montori V, Permanyer-Miralda G, Ferreire-Gonzalez I, et d. Validity of comuosite end Doints in clinical trials. BMT 2005: 330:594-596 10 Esparia PP, Capelastegui A, Quintana JM, et a]. A prediction rule to identify allocation of inpatient care in communityacquired pneumonia. Eur Respir J 2003; 21:695-701 11 Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identi5 low-risk patients with community-acquiredpneumonia. N Engl J Med 1997; 336:243-2.50 12 Lim WS, Van der Eerden MM, Laing R, et al. Defining community-acquiredpneumonia severity presentation to hospital: an international derivation and validation study. Thorax 2003; 58:377382 13 Kati S, Downs TD, Cash HR, et al. Progress in development of the index of ADL. Gerontologist 1970; 1:20-30 14 Sager MA, Franke T, Inouye SK, et A. Functional outcomes of acute medical illness and hospitalization in older persons. Arch Intern Med 1996; 156:645-652 15 Inouye SK, Peduzzi PN, Robison JT, et al. Importance of functional measures in predicting mortality among older hospitalized patients. JAMA 1998; 279:1187-1193 16 Metlay JP, Fine MJ, Schulz R, et al. Measuring symptomatic and functional recovery in patients with community-acquired pneumonia. J Gen Intern Med 1997; 12:423-430 17 Covinsky KE, Palmer RM, Counseu SR, et al. Functional status before hospitalization in acutely ill older adults: validity and clinical importance of retrospective reports. J Am Geriatr SOC2000; 48:164-169 18 Clermont G, Angus DC, Linde-Zwirble WT,et al. Does acute organ dysfunction predict patient-centered outcomes? Chest 2002; 121~1963-1971 19 G l p n P, Coddey R, Kilgallen I, et al. Circulating interleukin 6 and interleukin 10 in community acquired pneumonia. Thoraw 1999; 54:5155 20 Taniguchi T, Koido Y, Aiboshi J, et al. Change in the ratio of interleukin-6 to interleukin-10 predicts a poor outcome in patients with systemic inflammatory response syndrome. Crit Care Med 1999; 271262-1264 21 Capelastegui A, Espaila PP, Quintana JM, et a\. Improvement of process-of-care and outcomes after implementing a guideline for management of community-acquired pneumonia: ‘I controlled before-and-after study. Clin Infect Dis 2004; 39:95,5-963 22 Capelastegui A, Espaiia PP, Quintana JM, et d. Evaluation of clinical practice in patients admitted with communityacquired pneumonia over a 4-year period. Arch Bronconeumol2006; 42:283-289 23 Jasti H, Mortensen EM, Obroslcy DS, et al. Causes and risk factors for rehospitalization of patients hospitalized with community-acquired pneumonia. Clin Infect Dis 2008, 46: 5505 5 6 24 Herzog NS, Bratzler DW, Houck PM, et d. Effects of previous influenza vaccination on subsequent readmission and mortality in elderly patients hospitalized with pneu1noni.i. Am J Med 2003; 115:454-461 25 Naylor MD, Brooten D, Camphell R. et d. Comprehensice discharge planning and home follow-up of hospitalized elders: a randomized clinical trial. JAMA 1999; 281:613-620 26 Sington JD, CotreU BJ. Analysis of the sensitivity of death certificates in 440 hospital deaths: a comparison with necropy findings. J Clin Pathol 2002; 55:499502

Original Research