Severity Assessment of Lower Respiratory Tract Infection in Elderly Patients in Primary Care FREE

LOWER RESPIRATORY tract infections (LRTI) include infections of the tracheobronchial tree (bronchitis) and the lung parenchyma (pneumonia). It has been estimated, based on the data published from the United Kingdom, that a third of the adult patients with LRTI in community settings seek help for their symptoms. Antibiotic agents are prescribed to a fourth of these patients, and community-acquired pneumonia (CAP) is diagnosed in less than 2% of the cases (less than 0.5% of the whole adult population with LRTI in community settings).¹ However, the incidence of CAP is higher than this in the elderly population.^2- 3

Community-acquired pneumonia is the fifth most common reason for hospitalization among patients aged 65 to 74 years and the leading cause for hospitalization among those 85 years or older.⁴ The cost of treating CAP depends crucially on the place of treatment.⁵ Safe strategies to minimize hospitalization would lead to marked cost savings.⁶ Moreover, most patients with CAP with little risk of death prefer treatment outside the hospital,⁷ and, presently, most patients with CAP are actually treated as outpatients.^1,8 Several models have been developed for assessing the severity of CAP.^9- 18 However, these models have been typically developed among hospitalized patients with CAP and are not necessarily applicable to outpatients with CAP or, more generally, to patients with LRTI in primary care because the scoring systems in the models often require both chest radiography and laboratory analyses, which are not necessarily available in primary care units.

The purpose of our prospective study was to find out whether the information available in the primary care consulting office can be used to assess the severity of LRTI and the need for hospital treatment. Toward this end, we investigated the outcomes of 950 home-living immunocompetent elderly patients (age, ≥65 years) with LRTI who had signs suggestive of pneumonia.

The study was carried out in association with a pneumococcal and influenza trial among persons 65 years or older in northern Finland.¹⁹ During September and October 1992, before the baseline of the study, one of the researchers (P.O.H.) visited all the municipal health centers and municipal district hospitals in the follow-up area to inform the physicians about the research design and the way of reporting patients with LRTI. The follow-up visits to the participating centers were made in the autumns of 1993 and 1994. The study population consisted of all persons 65 years or older living in 55 municipalities in northern Finland. The census on December 31, 1992, included 59 790 people, 38.8% of whom were male. The study protocol was approved by the ethical review committee of the medical faculty of the University of Oulu, Oulu, Finland. Enrollment in the follow-up study was voluntary. The physician made the diagnosis of LRTI on the basis of the patient's history and physical examination findings. During the whole study period, the primary care physicians made their decisions concerning the prescription of antibiotics and the possible referral of their patients independently.

During the 2-year study period, 1743 cases of LRTI with signs suggestive of pneumonia were reported among the study population of 59 790 persons 65 years or older. The number of patients visiting health centers with signs suggestive of pneumonia is not known, nor is the proportion of patients with LRTI but without signs suggestive of pneumonia. Of the cases, 22 were excluded: 21 because of missing information on the day of the initial visit and 1 because of age (<60 years). One episode of suspected LRTI per patient was reported in 1424 cases. Only the first episode was included in the study for 133 patients with more than 1 episode of LRTI. Only home-living patients were included in the further analysis (1072 cases). Patients reported to be in a terminal state (10 cases) or with dementia (89 cases) were excluded because our questionnaire did not include information about whether their treatment was active or not. Of the remaining patients, those reported to be bedridden (14 cases) were excluded because our primary goal was to investigate an elderly population capable of normal daily activities. Nine patients receiving immunosuppressive medication (>5 mg/d of cytostatics or prednisolone) were excluded.

The following patient information was recorded on the case report form by the attending physician to describe the patient's condition during the first visit: presence or absence of respiratory and other symptoms (cough, dyspnea at rest, pleuritic chest pain, acute confusion, acute deterioration of general condition, and/or acute aggravation of a coexisting chronic disease [eg, the impairment of glucose balance in diabetes or the deterioration of congestive heart failure]); duration of symptoms; date of examination; residential information (home, nursing home, municipal health center, or hospital ward); initial place of treatment (home, nursing home, municipal health center ward, or hospital ward); results of the initial physical examination (body temperature, respiratory rate, systolic and diastolic blood pressure, heart rate, and abnormal finding on chest auscultation, ie, rales); and whether the patient was in a terminal state. The patient data were complemented during the follow-up visit (or after death, if the patient died before the scheduled follow-up visit) with the basic laboratory data recorded at the initial visit: hemoglobin and C-reactive protein (CRP) levels; platelet and white blood cell (WBC) counts; erythrocyte sedimentation rate; concurrent illnesses or disabilities (eg, congestive heart failure, asthma, chronic obstructive pulmonary disease, dementia, chronic pyelonephritis, and/or type of diabetes); residence on a long-term ward; bedridden state; immunosuppressive treatment (ie, >5 mg/d of cytostatic medication or prednisolone); previous and current smoking habits; consumption of alcohol; information about possible travel abroad within the past month; final diagnosis; and, in fatal cases, whether the death was due to LRTI. To obtain valid and accurate information on the date of death and to confirm the registered coexisting illnesses, data concerning the study population were also drawn from the national register of the Finnish Social Insurance Institution, Kela. Information on dementia, dietary diabetes, alcohol abuse, and bedridden and terminal state was not available in the national register and was therefore obtained by a questionnaire. All data were stored in a computer database.

The end point of the severity of LRTI was defined as mortality due to LRTI within 30 days after the first visit to a primary care physician.¹⁶ The statistical analysis was performed using the SPSS software (SPSS Inc, Chicago, Ill). Survival after the initial visit was calculated with the Kaplan-Meier method. For the analysis of categorical variables, the χ² test (or the Fisher 2-tailed exact test when appropriate) was used. The continuous variables were skewed, and median values and interquartile ranges (25th to 75th percentile) were therefore used. The association between the continuous variables and survival was analyzed using the Mann-Whitney test. For further analysis, the statistically significant continuous variables were dichotomized by selecting clinically relevant cutoff points. To identify the independent risk factors among the variables that showed statistically significant associations (P<.01) with mortality in univariate analyses, the relative risk of death was estimated using hazard ratios calculated by Cox forward stepwise regression analysis. The P value for entry into the model was .01 and for removal, .05. The patients having none or only 1 missing value of predictor variables were included for Cox regression analysis, and missing data were replaced by the geometric means of the study population: 51 mg/L for CRP level, 22/min for respiratory rate, and 9.3 × 10³/µL for WBC count.

The final study population consisted of 950 home-living elderly patients with LRTI, 38 (4.1%) of whom died within 30 days after the initial visit. Of these deaths, 37% occurred within the first 7 days (Figure 1). Their deaths were not dependent on the initial place of treatment (Table 1). The mortality rates were similar for both sexes (Table 2). In this population with LRTI, dyspnea at rest tended to be more often observed in the patients who died within 30 days (63%) than in those who survived (48%) (P = .07; Table 2), while the presence of rales did not statistically differ between the groups (81% vs 85%). Most of our patients with LRTI actually had real pneumonia; 83% of them had rales on chest auscultation, and 48% had dyspnea at rest during the first visit to a primary care physician. Moreover, results of a retrospective analysis of the chest roentgenograms (CRXs) showed definite pneumonia in 61% and probable pneumonia in 13%. However, we excluded the CRX findings from this study to mimic the real conditions because CRX was only available in 44% of the cases at the time of the first visit to a primary care physician in our series.

Of the categorical variables, only acute aggravation of a coexisting illness had a statistically significant association with survival (Table 2). However, the comorbidity was not associated with mortality in univariate analysis (data not shown). Of the continuous variables, the patient's age, respiratory rate, WBC count, and CRP values were statistically associated with death within 30 days (Table 3). When the importance of these variables was evaluated in the Cox forward stepwise regression model that included 719 cases, acute aggravation of a coexisting illness, respiratory rate (≥25/min), and CRP level (≥100 mg/L) were identified (in this order of magnitude of hazard ratios) as independent relative risk factors of death within 30 days (Table 4). The mortality rate of these patients with LRTI was 2.2% within 30 days if they had none or only 1 independent risk factor. The corresponding mortality rate was 20% if they had all 3 risk factors.

In this study, 2 parameters immediately available in the consulting office (preceding aggravation of a coexisting illness and respiratory rate of ≥25/min) together with an elevated serum CRP level (≥100 mg/L) at the initial visit to a primary care unit, were independently associated with the risk of death within 30 days among elderly patients with LRTI and can be used to assess the severity of LRTI.

As our results show, in most of the patients with LRTI in whom the attending physicians had primarily suspected to have pneumonia, the diagnosis of CAP was later verified by CXR. Thus, a comparison of our results with earlier reports concerning mainly CAP seemed justifiable.

It is well known that elderly patients may lack the typical symptoms of pneumonia; the earliest clues may be unspecific symptoms, such as lethargy, mental confusion, and "failure to thrive," as well as the deterioration of a preexisting disease (eg, congestive heart failure).²⁰ In our series, acute aggravation of a coexisting illness, such as the impairment of glucose balance in diabetes or the deterioration of congestive heart failure, was independently associated with mortality within 30 days. This acute aggravation of a chronic illness was judged by the criteria of the attending physicians.

The measurement of respiratory rate is a useful marker of acute respiratory dysfunction.²¹ Its prognostic importance in CAP has been shown in several investigations.^{9- 12,14- 16,18} The cutoff value of respiratory rate has varied from 20/min or higher²² to 30/min or higher.^{9- 12,14- 16,18} In our study, tachypnea (with respiratory rates of ≥25/min used as a cutoff point) remained a significant predictor of mortality even in multivariate analysis. Our results stress once again the importance of respiratory rate—a simple measurement—in the clinical assessment of respiratory infections. On the other hand, heart rate, systolic and diastolic blood pressure, and body temperature, which have previously been reported as important risk factors associated with death in CAP, did not have similar predictive value in the present series.

Because CRP level is more sensitive, decreases faster after a favorable treatment response, and is independent of age, its use as a marker instead of erythrocyte sedimentation rate has been recommended for the diagnosis and follow-up of infections.^23- 24 In adults with acute LRTI, a CRP level of 50 to 75 mg/L has been considered suggestive of CAP.^1,23 Our results showed that a CRP level 100 mg/L or higher was independently associated with mortality in elderly patients with LRTI in primary care. Thus, our study emphasizes the importance of CRP in assessing the severity of LRTI among elderly patients in primary care. Our results also suggest that the use of CRP as a marker is recommendable in the severity assessment of respiratory tract infections.

In our series, the overall rate of mortality of elderly home-living patients with LRTI within 30 days was 4%, which was lower than the percentage published in, for example, a recent meta-analysis in which the mortality rate of elderly hospitalized patients with CAP varied from 5.7% to 32.9%.²² In our primary care population, the mortality rate was even lower (2%) if the patients had no or only 1 of the independent risk factors associated with mortality. However, the mortality rate was 20% among our patients with 3 positive risk factors. The latter figure agrees with an earlier report on mortality (21%) from pneumococcal bacteremia in Finland.²⁵

In earlier studies, controversial observations on the influence of age on the mortality in CAP have been reported, with some studies supporting its importance^{11,17,26- 27} and others not.^28- 30 In our series, age was a significant factor in univariate analysis but did not remain significantly associated with mortality in multivariate analysis. In many series, comorbidity has been shown to be an independent risk factor of mortality.^16- 17,22 No such association was seen in our series. In the earlier studies in which comorbidity has been recognized as a risk factor, patients with CAP have been typically treated in the hospital. In earlier studies, leukopenia and leukocytosis have also been shown to be independent prognostic markers in CAP.^{9,11,18,22,27} In our series, WBC count was significantly associated with death in univariate analysis, but when contrasting the patients with a WBC count of less than 4 × 10³/µL or higher than 11 × 10³/µL with those with a WBC count within the normal range, WBC count was no longer independently associated with death in Cox regression analysis. Whether the use of cytostatics or steroids could be an independent prognostic variable even in primary care patients should be investigated in a population with a larger proportion of patients undergoing such treatments.

Recently, a prognostic model for the evaluation of the severity of CAP has been described¹⁵ and validated in elderly hospitalized patients with CAP.³¹ We were unable to use this prognostic system, however, as some of the laboratory measurements essential for the calculation of the risk scores of individual patients, such as arterial blood gas analysis and serum urea nitrogen, were not available in the consulting offices of our primary care physicians.

We tried to find markers for the assessment of the severity of LRTI among elderly patients in primary care. These 950 home-living LRTI patients with a clinical suspicion of pneumonia were encountered by primary care physicians in their everyday practice, and the ongoing study did not influence on their treatment decisions. At least 3 parameters independently associated with mortality within 30 days are easily available: patient history (preceding aggravation of a coexisting illness); physical findings (respiratory rate ≥25/min); and laboratory measurement (CRP level ≥100 mg/L). Whether these markers are generally applicable to the evaluation of the severity and treatment decisions of LRTI among elderly patients in primary care should be ascertained and validated in a new prospective study.

Accepted for publication March 29, 2001.

Corresponding author and reprints: Hannu Syrjälä, MD, PhD, Department of Infection Control, Oulu University Hospital, FIN-90220 Oulu, Finland (e-mail: hannu.syrjala@ppshp.fi).