Is antibacterial treatment intensity lower in elderly patients? A retrospective cohort study in a German surgical intensive care unit

Severe infectious complications such as sepsis are common and increasing in critically ill patients in intensive care units (ICUs) [1, 2]. Concurrently, the rising prevalence of multi-resistant organisms in ICUs forces clinicians to use antibacterial combinations and agents of last resort for management of such infections, thereby further fostering antimicrobial resistance development [3‐5]. Antibacterial therapy thus places intensive care physicians in a worsening dilemma of providing adequate therapy for the individual while considering not only the good of the single patient but of the whole ICU [6]. This conflict especially applies to older patients who are on the one hand at greater risk for developing infections [7] and on the other hand may only experience limited benefit from aggressive therapy, especially at the end of life [6, 8]. Elderly patients may consequently receive a disproportionately large share of antibacterials and thus are a group of interest upon which to focus antibacterial utilization improvement activities, such as antibiotic stewardship programs. By contrast, general treatment intensity in ICUs declines with advancing age [9‐13], although it is acknowledged that age alone should not be decisive for this [7, 9, 14]. These conflicting treatment tendencies make it unclear if current utilization of antibacterials in elderly ICU patients differs from that of younger ones. This question is of increasing importance given that the proportion of elderly patients requiring complex care is predicted to rise due to the demographic change [11, 15‐18]. To the best of our knowledge, however, no studies have yet investigated differences in antibacterial therapy in older compared to younger ICU patients. Moreover, previous antibacterial drug utilization studies in ICUs have focused on total antibacterial consumption or impacts of antibiotic stewardship programs, while data quantifying antibacterial therapy based on individual patient data are scarce.

In this study we therefore compared antibacterial therapy in adult ICU patients above and below 80 years, a cut-off applied previously for studies in this setting [8‐10, 13, 15, 16, 18]. We hypothesized that in our ICU elderly compared to younger patients more often require antibacterial treatment. In addition to all antibacterials we also separately assessed different antibacterial drug classes to detect potential differences in local therapy patterns between the two age groups, and to reveal differential utilization of antibacterials in our ICU associated with high resistance risk. Because in our ICU both patients with urgent and no need for antibacterial treatment are treated concomitantly, we accounted for this situation with adequate analytical methods.

Of 17,464 valid admissions to our ICU between April 2006 and October 2013 the majority were male (57.1%) and 2640 (15.1%) belonged to the age group of 80 and above (Table 1). Mean age was 63.8 ± 16.3 years and the median length of stay 0.9 days (interquartile range (IQR): 0.8–2.6), with a higher median of 2.8 (IQR: 0.8–8.3) in the 1448 (8.3%) non-survivors. Slightly more than one third had at least one valid SAPS II- and TISS-10-score (36.3%). The median of the means was 20.1 (IQR: 13.3–28.9) for SAPS II without age-points and 9.7 (IQR: 5.0–12.5) for TISS-10. Patients with a SAPS II- and TISS-score contributed to 72.2% of total patient days. Most admissions of the cohort and in young patients were neurosurgical, whereas older patients were most often transferred from orthopaedics (Additional file 1: Table S1). For all ICU patients the median Charlson score was 2, (IQR: 0–3) with higher scores in old compared to young patients (median 2, IQR: 1–4 vs. median 2, IQR: 0–3, p < 0.0001).

About one third of patients were treated with antibacterials during their ICU stay (Table 1). Of these, an even larger proportion than in the full cohort was male (64.6%) and younger than 80 years (87.5%). Mean age was thus slightly lower compared to the whole cohort (62.4 ± 16.2 years), whereas length of stay (median 2.2 days (IQR: 0.9–8.7) was greater. While in the full cohort the proportion of females in patients aged 80 or above was larger than that of males, only about half of the elderly patients treated with antibacterials were female.

More than 190,000 antibacterial prescriptions were recorded for all patients during the study period (ca. 140,000 when excluding erythromycin), for which penicillins accounted for the most (Table 2). This is likely due to predominant administration of the penicillin class member piperacillin with syringe pumps, for which frequent records are created. In contrast, the largest number of exposed patients received cephalosporins, while most exposed patient days were observed for macrolides (predominantly erythromycin).

In patients 80 years and older the largest proportion was exposed to cephalosporins, imidazoles and penicillins, in the younger group the largest fraction of patients was exposed to cephalosporins, macrolides and imidazoles. Antibacterial prescriptions, however, were mostly recorded for penicillins and macrolides in both age groups. Penicillins, followed by imidazoles and macrolides, accounted for the most exposed patient days in elderly patients, whereas younger patients were by far most often exposed to macrolides, cephalosporins and carbapenems.

Overall, in the cohort there were 38.70 (95%CI: 38.26–39.15) exposed patient days per 100 patient days (global rate), and 32.81 (95%CI: 32.41–33.22) exposed patient days per 100 patient days when excluding erythromycin (Table 3). The rate for patients aged < 80 years was higher (39.73 (95%CI: 39.25–40.21)) than the rate for older patients (30.97 (95%CI: 29.84–32.15)). These observed rates with confidence intervals from an unadjusted standard Poisson model reflect the actual status quo of antibacterial exposure in our ICU for all patients and both age groups. In contrast, the rate ratios of the zero-inflated Poisson-model can be interpreted as the intensity of treatment in patients above 80 compared to those below 80 years for those patients who were likely to be treated. The exposure rate ratio (RR) adjusted for sex, mean SAPS II- and TISS-scores, and year of treatment for all antibacterials was 1.02 (95%CI: 0.98–1.07) indicating a comparable treatment intensity in both age groups. For all antibacterials except erythromycin (adjusted RR 1.07 (95%CI: 1.02–1.12) and all individual antibacterial drug classes except for macrolides, aminoglycosides and glycopeptides adjusted exposure rate ratios reflected significantly higher rates in older patients or a greater treatment intensity compared to patients younger than 80 years. An analysis restricted to patients with more than 48 h of stay in our ICU yielded highly robust results (Additional file 1: Tables S3 and S4). In this group, only 59 or 1.2% had neither SAPS II- nor TISS-scores.

For all antibacterials there was no overall time trend in treatment intensity and no difference in comparative treatment intensity for the age groups over time as both the year and interaction of year and age did not reach statistical significance in the zero-inflated Poisson model. This is congruent with positive microbiological test results from blood cultures and bronchioalveolar lavage. Their number increased steadily during the study period, but the most common infectious agents were consistently identified as Staphylococci, E. coli, and Candida albicans (Additional file 1: Table S2). Treatment intensity was slightly lower, however, for female compared to male patients (adjusted RR 0.96 (95%CI: 0.93–0.98)).

While the proportion of patients aged ≥80 years remained stable during the study period at around 15% in our ICU (Fig. 1), in both age groups the number of exposed patient days increased relative to 100 patient days. There was an even greater increase in antibacterial treatment days relative to patient stay, reflecting increasing complexity of treatment. Except for the years 2007 and 2008 the same trends were observed for both age groups, although the rates for patients < 80 years were consistently higher.

During the study period about one in three admissions to our ICU received antibacterials and one in eight was 80 years or older. Antibacterial prescriptions, antibacterial therapy days and exposure rate to antibacterials were all lower in older compared to younger patients. However, alongside high-risk patients our ICU admits patients for routine post-surgical surveillance with low infection risk who are predominantly elderly and contribute numerous person days free of antibacterial exposure. We therefore accounted for the baseline probability of antibacterial treatment with zero inflated Poisson regression. The resulting adjusted exposure rate ratios indicated no difference for all antibacterials, i.e., global treatment intensity was comparable, for most antibacterial classes there was even evidence for higher treatment intensity in elderly patients.

Based on exposure rates, the most often used classes in our ICU during the study period were macrolides, cephalosporins, penicillins and carbapenems. Ignoring macrolides - predominantly erythromycin used off-label as prokinetic agent - our findings are consistent with other studies reporting β-lactams as most often used, with penicillins in a French [27] and a German ICU [28], and highest consumption of cephalosporins followed by penicillins, carbapenems, and fluoroquinolones in Swedish ICUs, [29]. In a Spanish multi-centre study also β-lactams as a whole, carbapenems, and fluoroquinolones were most widely used [30].

In contrast to our findings, previous studies on mechanical ventilation have revealed less aggressive therapy and less ICU admissions for the elderly even after adjustment for disease severity [9, 10, 12, 13].

The discrepancy between observed and adjusted comparative antibacterial exposure can be reconciled as both results representing different aspects of the antibacterial treatment process, with the former reflecting treatment decision and the latter treatment implementation. Our results thus suggest there is no differential treatment of elderly patients once the treatment decision is made. The decision for treatment initiation, however, appears to be dependent on age or factors strongly associated with age. This includes the association of age with both a low treatment need, particularly in patients admitted for short-term routine post-surgical recovery, as well as with a decision against treatment. The greater difference between observed exposure rates in old and young patients regardless of treatment duration compared to rates from those with > 48 h of treatment supports that the first scenario was probably true for a significant proportion of old patients. Unfortunately, information on life-sustaining treatment limitations was not at our disposal to distinguish both scenarios definitively. We expect that younger and older patients with a priori explicit strong wish not to be aggressively treated are unlikely to be admitted to our surgical ICU. However, information on treatment preferences is often not sought before ICU admission [31], but, as varying by centre [32], may concern a significant proportion of elderly [33], potentially leading to a lower treatment probability for elderly in our study. Indeed, given their higher mortality in our ICU, for elderly probably more often the question arose whether life-prolonging therapy was still useful and in the patient’s interest [6, 8].

The lower treatment intensity in female patients may reflect sex-specific differences in infectious disease incidence but also differences in treatment implementation due to other causes. While some studies report contradictory results on differences in general treatment intensity in ICUs [34‐37], Nachtigall et al. [37] did not find sex-based differences in antibacterial therapy in a sepsis cohort.

To our knowledge no previous study has explicitly focused on age-dependent differences in antibacterial therapy in ICU patients. Instead of aggregated data often used for antibacterial drug utilization studies we analysed patient-level data. We could thereby account for potential confounding in the age-group comparison by considering individual disease severity using the SAPS II-score together with therapeutic intensity reflected by TISS-10. Our findings of an at least comparable antibacterial treatment intensity in older ICU patients were confirmed in analyses restricted to patients with > 48 h of ICU treatment, who can be considered as high-risk for infection. This supports use of the zero-inflated Poisson model for all patients which facilitates a comparison to ICU-wide descriptive antibacterial utilization data. Moreover, this suggests general utility of this method for situations when a two-stage treatment process is likely but restriction of analyses to a subgroup with high treatment probability is not desirable or possible.

Our results provide a snapshot of antibacterial utilization in the ICU setting, treatment indication, however, was unknown. A distinction between ICU-imported and ICU-acquired infections was also not possible and the fraction of empirical antibacterial treatments is unknown. However, as the vast majority (83.3%) of patients treated for > 48 h received antibacterials within their first 48 h, a significant proportion of infections present on admission or soon thereafter is likely. Moreover, as our results reflect a single surgical ICU in Germany their generalizability is limited. Infectious disease incidence and infectious agents expectedly vary in other intensive care settings, such as neurological or cardiac surgery ICUs, and the difference between age groups may vary accordingly. In addition, antibacterial prescription practice can already differ between ICUs of the same country [24, 29]. Conclusions beyond German ICUs should be drawn even more cautiously since health care systems and services vary substantially across Western Europe and North America [38]. This variation also affects the ICU admission process [39]. Triage decisions lead to selective admission of elderly expected to benefit from admission who are likely more often treated than elderly refused ICU admission. It is unknown if our local triage was more or less restrictive than elsewhere and information on refused patients was not available. Thus, interpretation of our results requires consideration of the fact that they are only based on patients admitted according to local decision practice. Nevertheless, in addition to the cohort characteristics, the information on referring departments and microbiological isolates from our ICU provides some context for external comparisons.

While a binary age classification facilitated a simple interpretation of the exposure rate ratios, in further studies a greater number of age subgroups should be investigated, e.g., as previously defined as young-old (65–74 years), old-old (75–84 years), and very-old (> 84 years) [40]. Moreover, future investigations should include information on potential confounders such as antibacterial treatment indication and individual comorbidity, particularly diabetes and malignant diseases, immunosuppressant medication, as well as body weight, smoking and alcohol intake. A possible time-dependency of the treatment difference in the age groups based on length of stay should be considered. Future studies should also distinguish ICU-imported and ICU-acquired infections.

In our surgical ICU elderly patients were less often treated with antibacterials than patients younger than 80 years. In those likely to be treated, however, intensity was at least similar in older compared to younger patients. These findings support differentiation between age as a factor for treatment decision and treatment implementation. This requires confirmation in studies from other surgical and non-surgical ICUs, which ideally should include data on treatment indication.