Clinical Decision Support System and Incidence of Delirium in Cognitively Impaired Older Adults Transferred to Intensive Care

Abstract

Background

Elderly patients with cognitive impairment are at increased risk of developing delirium, especially in the intensive care unit.

Objective

To evaluate the efficacy of a computer-based clinical decision support system that recommends consulting a geriatrician and discontinuing use of urinary catheters, physical restraints, and unnecessary anticholinergic drugs in reducing the incidence of delirium.

Methods

Data for a subgroup of patients enrolled in a large clinical trial who were transferred to the intensive care units of a tertiary-care, urban public hospital in Indianapolis were analyzed. Data were collected on frequency of orders for consultation with a geriatrician; discontinuation of urinary catheterization, physical restraints, or anticholinergic drugs; and the incidence of delirium.

Results

The sample consisted of 60 adults with cognitive impairment. Mean age was 74.6 years; 45% were African American, and 52% were women. No differences were detected between the intervention and the control groups in orders for consultation with a geriatrician (33% vs 40%; P = .79) or for discontinuation of urinary catheters (72% vs 76%; P = .99), physical restraints (12% vs 0%; P=.47), or anticholinergic drugs (67% vs 36%; P=.37). The 2 groups did not differ in the incidence of delirium (27% vs 29%; P = .85).

Conclusion

Use of a computer-based clinical decision support system may not be effective in changing prescribing patterns or in decreasing the incidence of delirium.

More than 50% of hospitalized adults 65 years and older experience cognitive impairment during their hospital stay.¹ Compared with other patients, these patients are more prone to falls, injuries, pressure ulcers, restraints, use of urinary catheters, and inadvertent exposure to agents with anticholinergic properties, all of which can lead to delirium,^2–6 especially in the intensive care unit (ICU).⁷ Delirium in the ICU is an independent predictor of longer stays in the hospital and ICU, increased health care costs, and higher mortality rates.^8,9

The Institute of Medicine¹⁰ has recommended integrating information systems into health care as a way to improve the safety and quality of care of older patients. Medical informatics refers to the acquisition, storage, retrieval, management, and optimal use of medical information, data, and knowledge.¹¹ Informatics can be used to customize clinical decision support systems (CDSSs) to accommodate patients’ and providers’ needs. A CDSS can be used to retrieve relevant, individualized, and updated information from a health system’s data repository and then feed these data directly to clinicians at the time of decision making.¹² The efficacy and feasibility of these systems have been evaluated in multiple clinical trials; the results have indicated that use of a CDSS can improve the process of care, lead to better patient outcomes, reduce medical errors, and decrease health care expenditures.^13–15

We tested the efficacy of a CDSS-based intervention that consisted of alerting physicians to the presence of cognitive impairment, recommending early referral to a geriatrician, and suggesting discontinuation of the use of urinary catheters, physical restraints, and anticholinergic drugs in the large randomized Enhancing Care for Hospitalized Older Adults With Cognitive Impairment (e-CHAMP) trial.¹⁶ The results of the original e-CHAMP trial generally indicated that use of a CDSS intervention did not change the ordering behavior of physicians providing care to patients with cognitive impairment. We now sought to explore the results of the same intervention tested in the e-CHAMP trial for a subgroup of patients enrolled in the larger trial who were transferred to the ICU at any point during the patients’ hospital stay. We hypothesized that the patients randomized to the CDSS intervention would have a reduced exposure to inappropriate anticholinergic medications, urinary catheters, and physical restraints and would have more frequent referrals to the inpatient geriatric consultation service, resulting in a decreased incidence of delirium among cognitively impaired elderly patients transferred to the ICU.

Methods

The study was approved by the institutional review board at Indiana University School of Medicine, Indianapolis, Indiana, and informed consent was obtained from patients or their legally authorized representatives.

Study Setting and Patients

Patients were included in this study if they were transferred to the ICU services from a general medical care area of Wishard Memorial Hospital and were already enrolled in the e-CHAMP trial between July 1, 2006, and March 30, 2008. Patients transferred to an ICU were identified via the charges accrued by them during their ICU stay. The parent e-CHAMP study was a randomized clinical trial funded by the National Institutes of Health (clinicaltrials.gov identifier: NCT00182832) to determine the efficacy of using a CDSS to alert physicians to the presence of cognitive impairment, recommend early consultation with a geriatrician, and suggest discontinuation of the use of urinary catheters, physical restraints, and anticholinergic drugs. Details of the e-CHAMP trial have been published elsewhere.¹⁶

Wishard Memorial Hospital is a 457-bed, university-affiliated, urban public hospital staffed by Indiana University School of Medicine faculty and house staff. It serves a population of approximately 750 000 residents in Marion County, Indiana. It has a 22-bed ICU with a nurse to patient ratio of 1 to 2 most times and of 1 to 1 when necessary. Patients admitted to the ICU are cared for by clinical teams consisting of 1 critical care faculty member, 1 critical care fellow, 2 senior residents, 2 junior residents, 1 nurse practitioner, and 1 critical care clinical pharmacist. Ancillary services such as geriatrics provide consultation.

Patients were eligible for the study if they were at least 65 years old, were transferred to the ICU at any point during their hospital stay, spoke English, and had cognitive impairment at the time of admission to the hospital. Patients were excluded if they had previously been enrolled in any other study, were aphasic, or were unresponsive at the time of screening.

Procedures and Data Collection

The computerized Regenstrief Medical Record System¹⁷ is the primary instrument for processing data and monitoring patients’ and physicians’ activity for the hospital. The system consists of registration and scheduling, laboratory, pharmacy, and database modules and maintains a number of other databases, including vital signs, results of laboratory tests and diagnostic tests, discharge summaries, and inpatient and outpatient charges. The hospital uses the GOPHER Physician Order-Entry System,¹⁷ a computerized system linked to the Regenstrief system that records all physician orders. As orders are entered, the system sends them electronically to the nurses’ workstation on the patient’s unit, and requisitions are printed at appropriate locations.

Trained research personnel used the Short Portable Mental Status Questionnaire¹⁸ to determine each patient’s cognitive status. The questionnaire is a 10-item test with a sensitivity of 86% and specificity of 99% for dementia that has been validated in both community-dwelling and hospitalized older adults. Patients with a score of 8 or less on the questionnaire were considered to have cognitive impairment. Delirium was assessed by using the Confusion Assessment Method (CAM)¹⁹ at the time of enrollment and every weekday. The CAM score was considered indicative of delirium if the patient displayed both acute and fluctuating changes in mental status and inattention, plus either disorganized or incoherent thinking and/or an altered level of consciousness. The CAM has a sensitivity of 97% and a specificity of 92% for diagnosis of delirium.¹⁹ The Regenstrief Medical Record System was used to determine patient demographics such as age, sex, race, education level, length of hospital stay, mortality, mortality 30 days after hospitalization, 30-day readmission rates, discharge placement, and hospital-acquired complications. Each patient’s level of comorbid conditions was assessed by using the validated Charlson Comorbidity Index.²⁰ Severity of illness was calculated by using the Acute Physiology Score derived from the Acute Physiology and Chronic Health Evaluation III score.²¹ Once research personnel determined a patient’s cognitive status and entered the results into GOPHER, eligible patients were automatically randomized in a 1 to 1 ratio to a CDSS intervention group or to usual care via a computer-generated process.

Intervention

The CDSS intervention was derived from 2 published systematic evidence reviews^6,22 and through discussions among an interdisciplinary team that met monthly for 1 year. Initially, the recommendations of the interdisciplinary team were to decrease patients’ exposure to 18 mutually agreed-upon anticholinergic medications,¹⁶ but ultimately, through consensus, the team decided to expand the intervention to include minimizing the use of urinary catheters and physical restraints. Complete details of the intervention have been described previously.¹⁶ The intervention included the following:

• Each time a physician entered an order in GOPHER for a patient randomized to the intervention group, the physician received noninterruptive alerts of the presence of cognitive impairment, urinary catheter, physical restraints, and anticholinergic drugs or the need for consultation with a geriatrician.

• If the physician ordered use of a urinary catheter, he or she received interruptive alerts recommending that use of the catheter be discontinued.

• If the physician ordered physical restraints, he or she received interruptive alerts recommending that physical restraints be replaced by use of a professional sitter or low-dose trazodone.

• If the physician ordered any of the 18 inappropriate anticholinergics,¹⁶ he or she received interruptive alerts recommending that the drug be stopped, suggesting an alternative medication, or recommending dose modification.

• The physician was required to make a decision to accept, reject, or modify any of the interruptive alerts. However, when noninterruptive alerts were presented, physicians were able to quickly exit the screen by using the F8 key. Physicians providing care to patients randomized to usual care did not receive the CDSS alerts but were able to review the results of the cognitive screening.

Outcome Measures

Outcomes of interest were orders for consultation with a geriatrician; orders for discontinuation of the 18 potentially inappropriate anticholinergic medications, urinary catheters, or physical restraints; and occurrence of delirium as indicated by CAM scores among patients who had no CAM indication of delirium at the time of admission to the hospital.

Statistical Analysis

Baseline demographic and clinical variables were analyzed by calculating mean and standard deviations for continuous variables and percentages for categorical variables. Intention-to-treat analysis was used to compare the intervention and control groups: Fisher exact tests for binary categorical variables and t tests for continuous variables. Skewed variables were analyzed by using Wilcoxon rank sum tests. All data analyses were performed by using SAS 9.3 software (SAS Institute). A power analysis for 80% power and 2-sided α = .05 disclosed that 242 patients would be needed to detect a 15% difference in the incidence of delirium between the intervention group and the control group.

Results

The study sample consisted of a cohort of 60 patients who were transferred to the ICU for at least 1 day from an original population of 424 patients with cognitive impairment enrolled in the e-CHAMP randomized controlled trial. The mean age was 74.6 (SD, 8.4) years; 45% were African American, and 52% were women. Table 1 gives the baseline characteristics of the study population. The 2 groups (intervention and control) did not differ significantly in age, sex, ethnicity, education, chronic comorbid conditions, preexisting cognitive impairment, or severity of illness. The mean length of stay before transfer to the ICU was available for 40 patients; it was slightly higher in the CDSS group (5.5 days; SD, 4.5) than in the control group (3.4 days; SD, 3.7) but was not significant (P = .11). These 40 patients stayed in the ICU for 3 or more days. A total of 5 patients in the CDSS group and 7 in the control group had orders for consultation with a geriatrician before they were transferred to the ICU. Before transfer to the ICU, the CDSS group had 9 orders for a urinary catheter and 5 orders for anticholinergic drugs; the control group had 7 orders for a urinary catheter and 7 orders for anticholinergic drugs. The anticholinergic drugs ordered included promethazine (9 orders); diphenhydramine (6 orders); and nortriptyline, meclizine, and hydroxyzine (1 order each).

Table 1.

Baseline characteristics of the clinical decision support system (CDSS) and the usual care groups

Characteristic	CDSS (n = 30)	Usual care (n = 30)	P
Age, mean (SD), y	74.2 (7.6)	75.1 (9.1)	.70
Women, %	57	47	.61
African American, % of patients	47	43	>.99
Education, mean (SD), y	10.2 (2.3)	10.0 (3.4)	.90
Mechanical ventilation, % of patients	17	17	>.99
Charlson Comorbidity Index, mean (SD)	2.1 (2.0)	2.6 (1.7)	.30
Score on Short Portable Mental Status Questionnaire, mean (SD)	5.3 (2.6)	4.7 (3.2)	.40
Severity of illness score,a mean (SD)	32.1 (16.0)	32.7 (19.1)	.98

^aMeasured by using the Acute Physiology Score.

Intention-to-treat analysis (Table 2) revealed no differences between the intervention and control groups in the occurrence of physician’s orders for consultation with a geriatrician or discontinuation of orders for urinary catheters, physical restraints, and anticholinergic drugs. The 2 groups also did not differ in mean length of hospital stay, survival rate 30 days after discharge, discharge to home, incidence of overall delirium, and incidence of delirium in the ICU.

Table 2.

Differences between the outcome measures in the clinical decision support system (CDSS) and the usual care groups

Outcome	CDSS (n = 30)	Usual care (n = 30)	P
Order for consultation with a geriatrician, % of patients	33	40	.79
Order to discontinue use of restraints,a % (No.)	12 (1/8)	0 (0/9)	.47
Order to discontinue use of urinary catheter,a % (No.)	72 (13/18)	76 (13/17)	.99
Order to discontinue use of anticholinergics,a % (No.)	67 (6/9)	36 (4/11)	.37
Incidence of delirium overall, %	27	29	.85
Incidence of delirium in the intensive care unit, %	18	12	.64
Length of stay in intensive care unit, mean (SD), d	7.4 (11.3)	5.7 (6.6)	.71
Length of stay in hospital, mean (SD), d	14.5 (13.1)	12.2 (10.3)	.60
In-hospital mortality, %	7	17	.42
Survived at 30 days after discharge, %	87	83	.99
Discharged home, %	37	30	.78

^aDenominator was the number of orders eligible for discontinuation.

Discussion

Our results indicate that using a CDSS to influence the behavior of health care providers in entering orders for patients with cognitive impairment neither increased orders for early consultation with a geriatrician nor decreased the use of urinary catheters, physical restraints, and anticholinergic agents. Use of the CDSS also did not reduce the incidence of delirium among older adults with cognitive impairment who were transferred to the ICU. In other studies,^13,14 use of a CDSS resulted in improvement in patients’ outcomes and a reduction in medical errors, but the patients in the studies were not elderly patients with cognitive impairment. The e-CHAMP trial¹⁶ was the first randomized trial to study the effect of a CDSS on the care of cognitively impaired elderly patients, and, to the best of our knowledge, our study was the first to evaluate the impact of a CDSS in reducing delirium among cognitively impaired elderly patients transferred to the ICU.

According to studies by Inouye et al²³ and Marcantonio et al²⁴ and a systematic review,²⁵ multicomponent interventions have been more effective than single-component interventions in preventing delirium. On the basis of these studies, a multicomponent intervention such as ours should have had a good chance to prevent delirium, but the results of both the original e-CHAMP trial and of our analysis of the e-CHAMP-ICU subgroup differed from the results of other multicomponent interventions. Multiple factors could explain our inability to reduce delirium in critically ill elderly patients with cognitive impairment. Earlier studies^26,27 on risk factors for delirium in the ICU indicated that numerous predisposing and precipitating factors contribute to the occurrence of delirium. Because our intervention was limited to 3 risk factors, its effect might have been overwhelmed by the presence of other risk factors responsible for delirium. In addition, in critically ill patients, use of physical restraints and urinary catheters generally cannot be avoided, a situation that could have further reduced the effect of our intervention. We might have had better results if we had directed the CDSS to reduce the use of restraints and urinary catheters to nurses rather than to physicians. The e-CHAMP intervention was focused on reducing use of anticholinergic agents, but other pharmacological risk factors are associated with delirium in the ICU, such as use of sedative-hypnotics in the form of benzodiazepines and pain control with morphine.^7,9,28 Reducing delirium without modifying exposure to benzodiazepines and opioids may not be possible. Use of an intervention to decrease exposure to a greater number of ICU risk factors might be more successful in decreasing the incidence of delirium in critically ill patients. The rates of use of anticholinergic drugs, urinary catheters, and physical restraints were much lower than we initially expected. The lower rates compromised the power of our study, which was already limited because of the small sample size, and could have contributed to the negative results. A CDSS-related factor that might have reduced the impact of our intervention is the possibility that the physicians were overwhelmed by repeated computer alerts. Within the current clinical environment, several alerts are used to highlight formulary modifications in prescribing, notices on hospital bed occupancy, and laboratory updates that ICU staff consider unnecessary “noise.” In such instances, the providers tend to ignore the alerts, and subsequently the reputation of these computer-based alerts, warnings, and reminders is compromised and compliance with them decreases progressively. This decrease in compliance occurred with use of other CDSS^29,30 as well as in our study. “Alert fatigue” among physicians was indicated by an increase in use of the F8 key to ignore the alert and exit the screen. Supplementing the CDSS alerts with a human element may be necessary to achieve compliance with the intervention. To test the hypothesis that supplementing the alerts with a human element will increase compliance, we have designed a trial³¹ of pharmacological management of delirium that is under way at the medical center. We are testing a multicomponent intervention that uses both a CDSS and human intelligence to reduce the duration and severity of delirium in the ICU.

Strengths and Limitations

The strengths of our study include the diverse population of patients (inclusion of women and African Americans), use of reliable and validated assessment tools for cognitive impairment and delirium, and daily delirium assessment except on weekends. The limitations include the small sample size selected from participants in a larger randomized trial, limited scope of the intervention (targeting only 3 risk factors predisposing to delirium), inclusion of only 18 medications with anticholinergic properties, inability to calculate the overall use of anticholinergics, initiation of the intervention with patients on medical units before the patients’ transfer to the ICUs (thus, findings are not applicable to patients directly admitted to the ICU), once-daily delirium assessment and no CAM screenings on weekends (possible loss of some data because of the fluctuating nature of delirium), and lack of a human element to reinforce CDSS-recommended intervention.

In conclusion, use of a computer-based physicians’ CDSS did not change providers’ practices in the ICU and did not decrease the incidence of delirium in cognitively impaired elderly patients. Future studies with interventions targeting multiple risk factors and using an advanced CDSS with integration of human intelligence may be required to establish a role for computer-based alerts in preventing delirium in the ICU.