Delirium: Still Searching for Risk Factors and Effective Preventive Measures

BACKGROUND: Noting the outcomes of delirium reveals the importance of recognizing patients who are at the highest risk for delirium and developing prevention‐oriented systems of care. Outcomes found in prospective studies are falls, pressure ulcers, longer length of stay, long‐term cognitive deficits, and functional impairment. Many of these are used as indicators of quality geriatric care. The actual incidence rates of delirium depend on the characteristics of the cohort studied and whether it included all consecutive geriatric patients, emergency surgical patients, and elective surgery patients or specifically excluded patients with terminal illness or advanced dementia. Even for a given institution to determine its incidence rate is difficult. The frequency of use of the International Classification of Diseases, Ninth Revision, code for delirium at discharge is not an accurate measure because hospitals and physicians are not likely to use a code that does not add to hospital reimbursement under Medicare. Notably, there has never been a study of the cost of either the intensity of nursing care or additional diagnostic testing involved in the care of the delirious patient. Successful prevention depends on our ability to accurately identify those specific patients at risk for delirium and measure the relative contribution of the individual risk factors. This literature review addresses two systematic reviews by Cole et al., one on risk factors for delirium and the other on the effectiveness of preventive strategies. QUESTION: What are the risk factors for incident delirium in hospitalized geriatric patients? DATA SOURCES: A Medline search was performed using the terms “delirium,”“risk factors,” and “aged” for studies published in either French or English from 1966 through 1995. Current Contents and the bibliographies were also searched for relevant articles. STUDY SELECTION CRITERIA: One hundred one studies were identified in the search, of which 27 met the following inclusion criteria: prospective enrollment of patients age 50 and older, use of an accepted diagnostic criteria for delirium, and identification of at least one risk factor. No unpublished studies were included. DATA EXTRACTION: Each article was scored with an instrument specifically developed to assess validity of a study of risk factors. The criteria used were (1) adequate description of patient characteristics and recruitment process, (2) description of selection process of cases and controls, (3) adequate measure of and adjustment for confounding variables, (4) use of a systematic measure of risk factors, (5) blinded assessment of risk factors, (6) use of an operational definition of delirium, (7) adequate sample size, and (8) report of a measure of association with confidence intervals (CI). The mean quality score of the 27 studies was 9.3 out of total possible score of 16. A checklist was used to extract data from the articles. When original data were available in two or more studies, the unadjusted odds ratio (OR) for each risk factor was calculated for the individual studies. The adjusted OR was not used because each study had adjusted for different variables. A test of heterogeneity was done of the log OR for each risk factor. A combined Mantel‐Haenszel OR was calculated for risk factors that showed an association with delirium in at least one study. MAIN RESULTS: The review included 1,365 subjects with either incident or prevalent delirium and 5,513 controls. The mean age of the patients was 75.7 (range 50–97). The types of cohorts were medical (11), surgical (9), psychiatric (5), and mixed (2). Most studies used Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria to establish a diagnosis of delirium. Sixty‐one different risk factors were included in the studies, and 27 risk factors were evaluated in two or more studies There were four risk factors significantly associated with delirium in more than one study without significant heterogeneity. Dementia, positively associated in 12 of 15 studies, was the strongest risk factor (combined OR 5.2 (95% CI = 4.2–6.3), homogeneity P = .01). Depression was found to be a risk factor in two of the five studies in which it was included (combined OR 1.9 (95% CI = 1.3–2.6), homogeneity P = .01). Visual impairment was significant in two of four studies (OR 1.7 (95% CI = 1.2–2.3), homogeneity P = .05). An association between delirium and “abnormality” of serum sodium concentration was found in 4 of 5 studies (combined OR 2.2 (95% CI = 1.3–4.0), homogeneity P = .03). There were five risk factors found significant in more than one study, but with significant heterogeneity between studies. These were: alcohol abuse in three of five studies (combined OR 3.3 (95% CI = 1.9–5.5), homogeneity P = .90); impairment in the activities of daily living in two of four studies, (combined OR 2.5 (95% CI = 1.4–4.2), homogeneity P = .60); auditory impairment in three of five studies, (combined OR 1.9 (95% CI = 1.4–2.6), homogeneity P = .17); and male gender in four of nine cohorts (combined OR 1.9 (95% CI = 1.4–2.6), homogeneity P = .32). The authors calculated an OR of 3.8 (95% CI = 2.2–6.6) for “severe medical illness” as a risk factor but did not further operationalize this term. Variables presented as qualitative risk factors because data from the studies did not permit calculation of OR were age in eight of 12 studies, “structural brain abnormality” in both psychiatric cohorts, and azotemia in two of four studies. Hypoxia and “abnormal” cerebrospinal fluid were found to be significantly associated with delirium in one study each. Specific to surgical cohorts, the only identified qualitative risk factors were emergency surgery and impaired postoperative mobility in one study each. In single studies, the following were not found to be associated with delirium: race, marital status, education, first language other than English, history of nocturnal confusion, pain, dyspnea, elevated blood pressure, incontinence, abnormal serum calcium, abnormal liver tests, and pyuria. In two studies each, a history of delirium and history of falls showed no measurable association with delirium. An “abnormal” hemoglobin concentration was not a risk factor in three studies. Variables not associated with delirium for surgical patients were the type of anesthesia and duration of surgery, in one study each. Specific to the psychiatric patients, neither bilateral electroconvulsive therapy nor the number of treatments was found to be associated with delirium in one study each. Narcotics were specifically included in two of the 13 trials that evaluated medications, and neither showed a significant independent association, (combined OR 1.5 (95% CI = 0.9–2.3), homogeneity P = .96). CONCLUSIONS: The authors were concerned about the limitations of their review and suggested specific improvements in methodology for future studies of risk factors for delirium. The four main concerns were possible underreporting of nonsignificant risk factors, lack of the use of reproducible definition of “abnormality,” heterogeneity of the studies included, and inclusion of studies with low validity scores. The contributing factors to heterogeneity were diversity of enrollment criteria between the individual cohorts and different methods used for assessing risk factors. This was especially true for risk factors such as medications that were classified as a specific drug, a medication class, or a dose. Even a quantitative variable such as age was most often presented in qualitative terms. Contributing to lower validity ratings were inadequate sample size, lack of blinding for risk‐factor assessment, and poor control of confounding variables. The following recommendations were made for methodological improvement of future studies. The validity criteria for a study of risk factors developed by these authors should be included; risk factors should be studied within and not between medical, surgical, and psychiatric cohorts; and only incident cases of delirium should be included, to distinguish risk factors from confounders and precipitating causes.