Page 237 - EJMO-9-3

Eurasian Journal of
Medicine and Oncology ICU pharmacists and clinical outcomes

Methodological quality was evaluated using the reactions, or other preventable injuries, according to
Cochrane Risk of Bias Tool for randomized controlled the criteria specified in each individual study.
trials, and an adapted version of the Newcastle-Ottawa All outcomes were reported as aggregate effect sizes,
Scale was employed for observational studies. Risk of and subgroup comparisons were performed where data
bias assessments were independently performed by two were available.
reviewers to ensure objectivity and consistency. Any
22
disagreements were resolved through consensus between 3. Results
the reviewers, and, when necessary, a third reviewer was
consulted to reach a final decision. 3.1. Study selection and characteristics
Bias levels were rated as low, moderate, or high, based Out of a total of 3,301 unique records identified through
on how adequately studies addressed core methodological the comprehensive database search, 16 studies met the
elements, including participant selection, intervention predefined eligibility criteria and were included in the
25-40
implementation, outcome measurement, attrition final meta-analysis (Figure 1). These studies were
handling, and reporting transparency. published between 1999 and 2022 and collectively enrolled
37,925 patients admitted to ICUs. Of these, 23,060 patients
2.5. Statistical analysis received care involving CCPs, while 14,865 patients served
as controls (i.e., managed without the involvement of
The meta-analysis was performed using Review Manager CCPs).
version 5.3, provided by the Nordic Cochrane Centre. For
binary outcomes such as mortality and ADEs, odds ratios The included studies varied in design, comprising both
(ORs) with corresponding 95% CIs were calculated using prospective and retrospective observational cohorts, as well
either fixed-effects or random-effects models, depending as interventional trials, and were conducted across multiple
on the degree of heterogeneity among the included studies. countries, including the United States, China, Belgium,
Thailand, Egypt, and India. Sample sizes of individual
Heterogeneity was quantified using the I statistic,
2
interpreted as follows: 23 studies ranged widely, from as few as 70 patients to as many
as 30,032 (Table 2). The interventions performed by CCPs
• I value of 0% indicates no observed heterogeneity included direct participation in ICU rounds, antimicrobial
2
• I value of 25% reflects low heterogeneity stewardship, pain and sedation management, prevention of
2
• I value of 50% suggests a moderate level of drug interactions, and optimization of pharmacotherapy
2
heterogeneity in critical illness.
• I value of 75% or higher represents substantial or high
2
heterogeneity. 3.2. Impact on mortality
When the I² statistic exceeded 50%, indicating moderate A pooled analysis of all 16 studies demonstrated a significant
to high heterogeneity, a random-effects model was reduction in mortality associated with the inclusion of
employed to account for inter-study variability. For lower CCPs in ICU teams. The overall OR for mortality was
levels of heterogeneity, a fixed-effects model was applied. 0.72, with a 95% CI of 0.56 – 0.92 (p=0.01), favoring the
Where data permitted, subgroup analyses were planned to intervention group (Figure 2). Patients receiving care
investigate potential sources of variability across studies. with CCP involvement had a 28% lower likelihood of
To evaluate the robustness of the results, sensitivity mortality compared to those managed without pharmacist
analyses were performed by excluding studies deemed to involvement.
have a high risk of bias or those with small sample sizes Nonetheless, the analysis demonstrated substantial
(n ≤ 100). Potential publication bias was assessed visually heterogeneity across the included studies, as reflected by
through funnel plot inspection and statistically using an I statistic of 78% (Figure 2). This considerable variation
2
Egger’s regression test. A p=0.05 or higher was interpreted in effect sizes is likely attributable to differences in study
24
as no significant publication bias. methodologies, patient populations, and the nature of
pharmacist-led interventions. As a result, the use of a
2.6. Outcome measures random-effects model was deemed appropriate to account
The primary outcomes assessed were: for this variability.
(i) Mortality: Defined as all-cause in-hospital or ICU
death, as reported in each study. 3.3. Impact on ADEs
(ii) ADEs: Defined as any harm associated with medication The meta-analysis also revealed a substantial reduction in
use, including medication errors, adverse drug the incidence of ADEs among patients managed with CCP

Volume 9 Issue 3 (2025) 229 doi: 10.36922/EJMO025150116