This is the first study to exclusively evaluate the critically ill population for risk factors for ADEs. Approximately one-half of the potential patient-related ADE risk factors were significant and all of the drug-related risk factors were significant in the univariate analysis. Based on the multiple variable analysis, identifying critically ill patients who are at risk for ADEs include those who are admitted emergently, have acute kidney injury, have thrombocytopenia, and receive a substantial number of medications administered intravenously. These medication regimens are influenced by pharmacokinetic alterations that occur in critically ill patients, which can further increase the risk for ADEs.
The risk factors evaluated in this study can be categorized according to the opportunity for an intervention during in-patient care to potentially prevent an ADE, also known as modifiable risk factors or no potential for intervention described as fixed (nonmodifiable) risk factors. Many of the risk factors evaluated would be considered modifiable, including all the drug-related risk factors and length of stay. Risk factors such as number of drug allergies, hepatic injury, acute kidney injury, low albumin, and low platelet count may be more challenging to intervene on but could still be considered modifiable with the assistance of an advanced computer monitoring system that assists with the detection of drug-related hazardous conditions, allowing for the prevention of ADEs (4
). The fixed risk factors would be age, gender, severity of illness, type of insurance, and substance abuse. The majority of the risk factors determined significant by the univariate analyses are considered modifiable and present an opportunity for preventing ADEs in the ICU. It is difficult to determine which is more important for ADE surveillance tracking the drug-related risk factors or the patient-related risk factors. Drug-related risk factors could be easier to modify but patient-related risk factors have greater odds ratios. Ideally, an institution would track both to optimize ADE prevention; however, if this is not feasible because of a lack of computerized monitoring systems to assist with the patient-related factors, then the drug-related risk factors may be an easier target. The following discussion emphasizes the modifiable risk factors.
Evan et al (11
) performed a large retrospective analysis of risk factors for hospitalized patients. During the 10-yr study, 4,376 ADEs were identified in 4,140 patients, with gender, age weight, creatinine clearance, number of comorbidities, and drug dosage found to be important ADE risk factors. This study did not examine organ dysfunction as a potential risk factor, nor were critically ill patients analyzed as a subset for these risk factors. Advanced age has been described as a potential risk factor differentiating ambulatory and hospital patients (23
). However, in our critically ill population, age and gender were not significant risk factors. We were unable to assess weight because of the lack of documented information in the chart. Although drug dosage was not assessed, we did assess many other drug-related risk factors that were significant. In our critically ill population, the number of patients with acute kidney injury was significantly greater in patients having an ADE, but Charlson comorbidity index was not. Other studies have indicated that liver failure is a risk factor for ADEs, but this was not apparent in our study (28
). These findings indicate that the same risk factors that apply to a non-ICU population may not similarly apply to critically ill patients (11
Drug classes evaluated in our study demonstrated that anti-infectives, autonomic, cardiovascular, electrolyte, gastrointestinal, and drugs affecting the central nervous system were more commonly administered in the patients having an ADE. Patients having an ADE received more high-risk medications than the control group. The drug classes identified continue to demonstrate that targeting high-risk medications for ADE prevention is appropriate.
A prospective study investigated patient risk factors for ADEs in hospitalized patients, evaluating 113 ADEs from a 6-month study period (8
). The authors concluded that sicker patients who are in the hospital longer are more apt to have an ADE. There were only eight risk factors analyzed in this previous evaluation and the question remains if these risk factors are applicable to the critically ill population. In our analysis, the critically ill patients in both the case and control groups had high SAPS II scores, so there was not much variability to allow for comparison. In fact, patients without an ADE had a higher SAPS II score, although the Charlson comorbidity index was similar between groups. Also, patients without an ADE had a lower mortality, shorter ICU length of stay, and received less medication. A higher mortality and longer ICU length of stay for critically ill patients having an ADE have been demonstrated previously (2
). Patients with an ADE in our study had a longer length of stay before ICU admission, presenting another potential risk factor that could be associated with the increased number of medications administered.
Critically ill patients experience many physiologic changes that can influence drug metabolism and excretion. Organ dysfunction, in particular renal insufficiency, can lead to the potential for ADEs. In our study, acute kidney injury was a significant risk factor. This emphasizes the need for constant monitoring of drug clearance throughout a patient’s ICU stay. Computer-based decision support often assists with the initial dosing based on glomerular filtration rate, but more advanced alerting systems are needed for continuous monitoring. The pharmacist can be an asset to the patient care team, with a primary role in monitoring the patient’s pharmacokinetic and pharmacodynamic responses. Surprisingly, low albumin was not a risk factor for ADE occurrence because patients with ADEs received more protein-bound drugs. Patients having an ADE did receive more CYP P450 enzyme-inhibiting and enzyme-inducing drugs potentially contributing to alterations in the pharmacokinetic properties of the patient’s drug regimen.
Intravenous therapy is considered a high-risk activity because of the potential for errors and resulting harm (11
). Previously, intravenous administration has been shown to be a risk for ADEs in hospitalized patients (11
). Our study confirms that the intravenous route of administration is a risk factor for ADEs in the ICU. ADEs related to intravenous medications in academic centers is costly, resulting in an additional $6,647 in costs and a 4.8-day longer stay (31
). Intravenous administration of medications should be minimized when possible by encouraging switches from intravenous to oral routes or initiating oral therapy when possible. Other preventive measures include the use of advanced infusion pumps, standardization of infusion concentrations and rates, development of intravenous administration policies, and promoting best practices.
In general, drug regimens complicated by the high number of medications consistently appear to be a risk for ADEs (9
) and, from the results of our study, this remains true for the ICU environment. Diligent monitoring of medication regimens in the ICU is essential in an effort to discontinue medications when no longer necessary, such as drugs with prophylactic indications. Pharmacist participation in patient care rounds can be helpful with this monitoring (3
). Clinical decision support to monitor for drug–drug interactions can be useful as well. Patients with longer ICU stays should have their medication regimens evaluated frequently for opportunities to discontinue medications when appropriate.
The combination of ADE detection methods used in our study did not capture all ADEs that could be identified with a comprehensive chart review (6
). Different ADE detection methods are more or less likely to find different types of ADEs; spontaneous reporting systems contain more serious ADEs than chart review, chart review misses ADEs related to administration errors that are more likely caught by spontaneous reporting if severe and direct observation is more likely to detect administration related ADEs. The use of spontaneous reporting may explain why 90% of the ADEs in our study were category II and III. A comprehensive chart review can detect the most ADEs; however, because of the time constraints and substantial resources associated with this approach, most surveillance systems use targeted chart reviews provoked by triggers, as our institution does. In fact, the Institute for Healthcare Improvement recommends the targeted chart review using triggers as a part of an institution’s surveillance systems (34
It is possible that patients in the control group may have had an ADE. We intended to minimize this possibility by using a selection approach of the next two patients who met the matching criteria after admission of a case patient and by excluding patients with an E-code. The E-code is assigned by the patient reimbursement department to double-check for the occurrence of an ADE. The best design to avert this limitation is to proceed with a prospective study. We chose to match controls on admission date, admission service, and type of ICU in sequential enrollment rather than matching with random selection, because in the 7.5-yr study we wanted to assure that patient care was as similar as possible and felt that relatively close admission times were important. Matching did not occur based on gender, age, and severity of illness because these were risk factors of interest in the analysis. The multiple variable analyses did control for severity of illness.
The retrospective study design was selected to obtain data over a long period of time (7.5 yrs) because it is an efficient way of tracking presaging factors to verified outcomes. Although a prospective study offers the opportunity to obtain all ADEs occurring within the environment, it is a resource-intensive approach to obtain an adequate sample size. For example, the study by Bates et al (8
) occurred during a 6-month period and yielded a sample of 113 ADEs in 2019 admissions, which is a statistically inadequate sample for our analysis. Data collection in our study was optimized using a combination of electronic chart review of medical notes and administrative data such as International Classification of Disease revision nine codes. Also, a prospective study could provide a better understanding of the higher SAPS II score and the low albumin appearing mildly protective in our study.
Additional limitations include the drug administration analysis that was completed for the entire ICU stay and did not reflect before or after the ADE because this could not be calculated for the controls. In addition, the oral and intravenous drug administration analysis was based on drug charges, therefore necessitating the assumption that a charge equaled administration to the patient. The definition for an ADE included the likelihood ranking of possibly or higher is a common approach used in other ADE assessment studies (19
); however, this approach is likely to overestimate the true ADE rate because it includes borderline events. Although the type of ICU and admitting service related to the ADE were previously identified, these could not be evaluated as risk factors because they were also used for the selection of the control group. Type of ICU and service have been identified previously as risk factors for ADEs (2