The objective of this study was to determine if elderly ED patients admitted with pneumonia were less likely to receive recommended care interventions than younger patients. We found no relative deficiencies in care received, and actually found that elders were more likely to receive antibiotics and to have oxygenation assessed in the ED. The rate of process implementation did not vary by decade over age 65.
The Pneumonia National Quality Measures are used by CMS and JCAHO as well as private entities to assess quality of care for patients with pneumonia [7]. The requirement for rapid administration of antibiotics is supported by several observational studies which demonstrated decreased mortality in patients with rapid (<4–8 hours) time to antibiotics [11, 12, 15]. Elders may benefit from timely antibiotics even more than other populations [16]. Recommendations for oxygenation assessment are based on the observation that hypoxemia is associated with increased mortality [11].
However, controversy has arisen regarding the usefulness of individual measures, particular time to antibiotics and need for blood cultures [10, 11, 16, 17]. The blood culture quality measure has been recently modified to include cultures only for those patients going to an intensive care unit within 24 hours of arrival, and to only require blood cultures prior to antibiotics if drawn in the ED [7]. It is based on specialty society recommendations [18] as well studies that have demonstrated an association between blood culture collection and decreased mortality [11]. This measure has been criticized by other studies demonstrating low rates of true positive cultures, equivalent false-positive rates, and rare changes in management [17].
Despite these concerns, monitoring of adherence to Pneumonia Quality Measures is now widespread. However, these recommended processes of care have not been uniformly applied to all patient populations. For example, a 2004 study by Mortensen et al found that black patients were less likely to receive antibiotics within eight hours of presentation than whites, but were just as likely as whites to have blood cultures obtained, to have oxygenation assessed, and to receive guideline-concordant antibiotics [8]. Similar disparities have also been documented in inpatients with pneumonia [5, 19, 20]. Of note, we did not any significant management differences based on race within our sample.
The elderly are another group at potential risk of disparities in care. Studies of other disease processes in the ED have found that age can influence the quality of care a patient receives. In 2005, Magid et al found that elderly patients presenting to the ED with myocardial infarction were significantly less likely than younger patients to receive aspirin, beta-blockers, and reperfusion therapy. This occurred even when there was no contraindication to the therapy [9]. This work confirmed studies with similar results in elder inpatients [21–23].
In elders with pneumonia, Fine et al in 2002 found that a significant percentage of patients do not receive optimal care [5]. They studied patient and hospital characteristics associated with care processes in elderly Medicare patients hospitalized with pneumonia. Several factors were associated with failure of process including nonwhite race, hospital teaching status and size, and hospital location in the South. Presence of fever was positively associated with process performance, as we found with certain measures (antibiotics within 4 hours and blood cultures) in this study. However, the applicability of Fine's study to current ED practice is unclear as the data was obtained in 1994–1995, only 57% of study patients were admitted through the ED, and guidelines at that time did not emphasize prompt antibiotic therapy. Furthermore, it did not investigate potential disparities between elders and younger patients.
Waterer et al described delayed administration of antibiotics in 451 patients with community-acquired pneumonia [10]. In this study, age when assessed as a continuous variable was minimally associated with delay in antibiotic administration (OR 1.01; 95% CI, 1.00–1.06). Delay in treatment was more strongly associated with a nonspecific clinical presentation including presence of altered mental state (OR 2.89; 1.53–5.45), absence of hypoxia (OR 1.82; 1.09–3.04), and absence of fever (OR 1.59; 1.06–2.40). All are factors associated with the presentation of pneumonia in elders.
This previous work demonstrates several potential reasons that diagnostic and therapeutic interventions may be delayed in elders. These include preconceived physician biases towards the group, concerns over side effects or complications of therapy, and delays in diagnosis due to nonspecific presentation [9]. The results of this study which reveal that elders do not suffer from inadequate process of care administration in the ED alleviate these concerns. In our institution, elders received equal or more aggressive care as younger patients.
We suspect that concern over potential side effects of therapy would most commonly be found in patients undergoing invasive procedures or receiving medications with potentially severe side effects (e.g., thrombolytics). We speculate that in the case of pneumonia, the non-invasive nature of obtaining blood cultures and assessing oxygenation as well as the relative lack of direct contraindication to most antibiotic therapies prevented concerns over side effects from contributing to a difference in care patterns.
Several studies have demonstrated that elders with pneumonia are less likely to develop common clinical symptoms, including a >50% reduction in the rate of fever as compared to patients <65 years of age [24, 25]. In our database, elders were less likely to present with classic symptoms of pneumonia such as cough, fever, and shortness of breath. They were more likely to present with the non-specific complaint of confusion. We identified an independent association between antibiotic administration and confusion as antibiotics were less likely to be given in patients with confusion. This finding is consistent with that of Waterer et al who also found that an altered mental state was an independent predictor of failure to receive antibiotics within 4 hours (OR 3.2; 95% CI, 1.4 to 6.1). However, in aggregate their non-specific clinical presentation did not prevent elders from receiving appropriate processes of care as compared to the younger population. Similarly in elders with myocardial infarction, Magid et al found that differences in presentation alone did not explain process of care differences in elderly versus younger patients.
Although as a group elders were actually more likely to receive antibiotics, the data from the regression analysis provides information on areas to target for further improvement. Patients (both elder and non-elder) who are confused, have a normal lung exam, have HIV, or complain of shortness of breath are less likely to receive antibiotics in the ED. Those with lower temperature, lower respiratory rate, or normal lung exam are less likely to receive antibiotics within 4 hours. Greater sensitivity on the part of the physician to the diagnostic difficulties inherent in non-specific clinical presentations should result in performance improvement on the quality measures, particularly early antibiotic administration.
The patient's PSI score was not included in the regression models for two reasons. First, it covaries strongly with age (r = 0.60). Second, it has been suggested to be inaccurate in the elderly and thus evaluation of individual variables was warranted[26, 27]. As an alternative, we included most of the components of the PSI as individual components in the regression.
A primary limitation of the study was its retrospective chart review nature. However, steps were taken to ensure accuracy of data abstraction, a large percentage of data elements were completed, and agreement of the abstractors was good for multiple variable types (process, history, symptoms, laboratory, vital signs). Data was collected from only one center and results may not be easily generalized to all EDs, particularly as individual hospital characteristics have been shown to influence performance of these measures [5]. However, our findings demonstrate equivalent or better process implementation as compared to previous studies [5]. To provide conservative estimates, missing data for process measurements was coded as not performed. We are confident that there is good concordance for each of the primary study outcomes. It is likely that the antibiotics were actually administered in each case as it would be rare for a nurse to chart a specific administration time without administering the medication. Likewise, actual values are recorded for oxygen assessment. It is unlikely that such values were fictitious. For blood cultures, we attempted to ensure that they were actually obtained by requiring not just a physician order but also either documentation by the nurse or documentation of receipt in the microbiology lab.
We chose to include all patients who received a diagnosis of pneumonia in the ED [28]. Strict interpretation of the JCAHO criteria would result in exclusion of patients who do not have an abnormal chest x-ray or CT scan during the hospitalization. We found that 15% of our study patients did not have an abnormal chest x-ray in the ED, but due to database limitations we were unable to account for rates of chest x-ray negative, CT-scan positive pneumonia. Such rates have generally been approximately 7% in ED patients with pneumonia [28, 29]. We are confident that the inclusion of patients without chest x-ray abnormality but with ED diagnosis of pneumonia did not introduce substantial error into our findings for several reasons. First, such an approach has previously been used in the literature, and rates of patients failing to meet strict JCAHO inclusion criteria are no higher in our study [28]. Also, in our study the rates of antibiotic administration were the same between those with and without abnormal chest x-rays. Based on past work, we expect that approximately 7% of ED patients will have negative chest x-rays but positive CT scans [28, 29]. In addition, patients with abnormal chest imaging during their admission also meet JCAHO criteria. As a result, the true rate of patients without any abnormality on imaging study should be much lower than 15%. Given that ED physician behaviour is based on clinical diagnosis at the time of patient interaction rather than post-hoc radiologic interpretation, we feel that this methodology provides the most accurate measure of behaviour.
Among the recommended process measures, we chose not to assess influenza vaccination, pneumonia vaccination, or smoking counseling as these are generally performed on the inpatient wards. We also did not address antibiotic appropriateness as recommended antibiotics vary depending on several factors including presence of healthcare-associated pneumonia, structural lung disease, recent antibiotic use, or an immunocompromised state [18, 30]. Due to the nature of our database, we were unable to guarantee accurate determination of several of these variables and therefore could not judge appropriateness of antibiotic selection. The study also only examined ED course and did not account for meeting the process of care requirements after admission. We believe this would likely increase compliance with the blood culture and oxygen measurement guidelines. Combining the time spent in the ED with the delays in initiating antibiotics after admission, we do not believe it would change significantly the proportion of patients receiving antibiotics within 4 hours.
The study may suffer from incorporation bias, which in this case could occur when the variable being measured, advanced age, is one that could also affect both the diagnostic and treatment decisions leading to enrollment in the study. The presence of such bias could affect the final results and effect measures of the study.
In this study, such bias could occur due to differences in admission decision and diagnostic evaluation between elders and non-elders. The potential for and magnitude of admission bias can be seen in the absolute rates of admission, 60.1% for younger patients and 88.0% for older patients. Additional potential incorporation biases could be due to age-related differences in pursuing and making the diagnosis itself. Such an effect is difficult to measure for several reasons. For example, as elderly patients often present with non-specific signs and symptoms of pneumonia, the use of symptom-based inclusion criteria would result in significant spectrum bias. Thus to identify diagnostic accuracy in patients would require examination of the entire population of ED elderly patients, regardless of presenting symptom. We chose to use the previous literature as a guide and chose methods similar to other studies for inclusion criteria. The fact that rates of radiographic evidence of pneumonia were similar between groups does provide some assurance regarding the specificity of the diagnosis, although it does not provide information about diagnostic sensitivity. As a result of these issues, in this retrospective study we were unable to confirm accuracy of diagnosis, adequacy of diagnostic workup, or reasoning behind the admission decision. All of these factors affected patient entry into the study and potentially contributed to incorporation bias.
The effect of incorporation bias is somewhat mitigated by the fact that our process outcome measures are not directly related study entry. That is, these process measures are not directly related to the decision to make an initial diagnosis of pneumonia or admit the patient. The most egregious example of incorporation bias would be using clinical manifestations to make a diagnosis and then examining the frequency of those clinical findings in the patients so diagnosed. In this study, the mere act of administering antibiotics, obtaining blood cultures, or checking pulse oximetry is not generally a component of making a pneumonia diagnosis or admission decision and thus not part of triggering study entry criteria. Although obtaining a pulse oximetry may more often be done in patients suspected of pneumonia, its ubiquitous use means that the obtaining of pulse oximetry (as distinct from the value obtained) is not a criterion for pneumonia diagnosis. However, this does not mitigate the bias discussed above which may be introduced due to difference in admission and diagnosis decisions between patients of different ages and so it must be stressed that our study only applies to a patient population constituted after a diagnosis and admission decision is made.
There is also a possibility that the differential admission rates between elders and younger patients may have arisen due to different admitting paradigms on the part of ED physicians with the result that comparison between the two groups is meaningless. For example, younger patients may have been admitted because they appeared to the ED physician to be more ill, while elder patients may have had a greater weight given to age alone. Such imbalance likely represents real-world practice, as even the PSI places a large weight on age and co-morbidities. Although we cannot specifically measure the effect that such differences would have, we do provide information regarding differences in populations in Table 1. The populations do differ in symptoms at presentation and co-morbidities. However, there are no differences in rates of abnormal vital signs or chest x-rays. We then performed a regression analysis to determine if age was truly a factor independently affecting process (Table 2). It should be cautioned that, although these methods do provide a comparison between groups and examine age as an independent variable, they were only measured in patients meeting study entry criteria so that any interpretation must still consider potential incorporation bias as discussed above.
The result of this heterogeneity between populations is that the study is limited in several ways. We are unable to determine why a particular process measure was or was not followed. We can only provide the rate of process conformity. We can only conclude that in the population of patients diagnosed and admitted with pneumonia, the rates of compliance with recommended processes of care are as noted and are affected by the covariates noted in the regression model. This ultimately limits the generalizability of our study results as we cannot make inferences on the data we did not analyze. The results apply only to those patients diagnosed with pneumonia and in whom the decision is made to admit. They do not apply to the infected elderly population in general or to patients prior to the diagnosis or admission decision.