Given the prevalence, costs, and morbidity of cognitive impairment, and the availability of several brief screening instruments [11-13], screening for cognitive impairment among elderly patients seen in primary care has been recommended [14-20]. Early recognition and diagnosis of cognitive impairment are believed to be beneficial for at least three reasons. First, a diagnosis provides some comfort to the patient and family by explaining the changes in the patient's behavior and also allows the practitioner to counsel the patient and family about prognosis. Second, an accurate diagnosis of cognitive impairment and assessment of its functional and social effects may facilitate access to rehabilitative, social, and financial services, as well as help inform decisions about competency and guardianship [21-23].

Third, early recognition may allow an opportunity to alter the course of the cognitive impairment. Physicians may be able to do so if diagnostic evaluations show a reversible or partially reversible cause [24-26] or if the rate of progression of a dementing illness or a comorbid illness can be slowed [27-29]. Indeed, routine diagnostic investigations for reversible causes of dementia have been recommended [19, 20, 24]. However, the prevalence of reversible causes of dementia and the clinical utility of routine diagnostic investigations are still being debated [13, 30-35]. Thus, although early documentation of cognitive impairment offers many other benefits, clinicians might appropriately choose either an extensive or a limited diagnostic evaluation for any given patient.

Our study had four objectives. First, we determined the prevalence of cognitive impairment by screening a group of urban, elderly, primary care patients using the Short Portable Mental Status Questionnaire (SPMQ). Second, we describe their primary care physicians' rates of documentation and evaluation of cognitive impairment. Third, we compare use of health services in the year after the patient's screening examination. Fourth, we compare practice patterns and use of health services among patients with documented and those with undocumented cognitive impairment.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Our study was done in the General Medicine Practice of the Regenstrief Health Center, a multispecialty ambulatory care clinic associated with an urban county hospital [36]. The General Medicine Practice is a university-affiliated primary care practice staffed by 35 general internists on the faculty and 118 internal medicine residents. Beginning in January 1991, all patients aged 60 years and older were screened for cognitive impairment, depression, and alcoholism during their regularly scheduled visits. We excluded prisoners, patients residing in a nursing home, patients unable to speak English, and patients who had hearing impairment. The screening instruments included the Short Portable Mental Status Questionnaire [37], the Center for Epidemiologic Studies Depression scale (CES-D) [38], and the CAGE alcoholism questionnaire [39]. Professional research assistants administered the screening interview to patients at the time of scheduled office visits. Screening scores on the SPMQ completed for the purposes of our study were not revealed to the primary care physicians.

Patients with 3 to 4 errors on the SPMQ were considered to have mild cognitive impairment, and patients with 5 or more errors were considered to have moderate to severe cognitive impairment. We corrected the SPMQ score for years of education by allowing an additional error for patients with 8 or fewer years of education [37], but we did not correct scores for race. When administered to community-dwelling elderly persons, the SPMQ has a specificity of greater than 90% and a sensitivity of 50% to 82% (cutoff, 3 or more errors) [12, 40]. The positive predictive value of the test in a community-dwelling patient sample was 87%, and the validity was tested against structured psychiatric assessments [37]. We used the standard cutoff score of 16 or greater on the CES-D to indicate significant symptoms of depression [38]. Patients with two or more positive responses on the CAGE questionnaire were considered to have evidence of alcoholism [39].

Interview data were merged with information routinely collected and stored in the patients' electronic medical records [41]. These data included all outpatient diagnoses as recorded by the patients' physicians. The ICD-9-CM codes for the individual diagnoses for each patient were collapsed into 1 of 40 unique categories. We report only the 20 most common diagnoses. We also determined whether each patient had ever had any of the following: tests for hemoglobin concentrations, serum chemistries, thyroid function, erythrocyte sedimentation rate, syphilis (VDRL or fluorescent treponemal antibody absorption or Treponema pallidum microhemagglutination assay), serum cobalamin levels, and serum folate levels; radiograph of the chest; neuroimaging (computed tomography or magnetic resonance imaging); lumbar puncture; or electroencephalography. We determined whether patients had had neurologic evaluation in the year before or in the year after the screening date and whether they had been prescribed any of the major classes of central nervous system-active medications (anticholinergics, narcotics, antidepressants, benzodiazepines, or major tranquilizers) in the year after the screening date. We determined whether patients had ever received pneumococcal vaccination and whether they had received influenza vaccination or colon cancer screening (hemoccult test, barium enema, sigmoidoscopy, or colonoscopy) in the year after the screening date. We determined whether female patients had had mammography and Papanicolaou smears in the year after the screening date. We also determined whether patients had been prescribed aspirin, a nonsteroidal anti-inflammatory drug, or estrogen replacement in each of the 2 years before the screening date.

For use of health services in the year after the screening date, we determined the following: whether patients had been hospitalized or had visited the emergency department and the mean number of these episodes among patients with at least one of these visits; the mean number of outpatient visits, including those to medical and surgical subspecialty clinics; and total outpatient charges, including charges for office visits, outpatient diagnostic testing, and pharmacy services. Mortality was determined as of December 1992 using hospital discharge condition, death summaries, autopsy reports, and the Indiana State Board of Health's death certificate files.

Statistical Analysis

For dichotomous variables, we used chi-square tests to compare frequencies among patients with no cognitive impairment, mild cognitive impairment, or moderate to severe cognitive impairment. We used analysis of variance to compare the mean log-transformed values for continuous and count variables among these three groups. We also looked for significant differences between those with "documented" cognitive impairment and those with "undocumented" impairment. A patient with documented cognitive impairment had an SPMQ score of 5 or greater and had "dementia" (or a synonymous term such as the organic brain syndrome or Alzheimer disease) listed in their outpatient medical record.

With moderate to severe cognitive impairment as the dependent variable, we next used logistic regression to determine which variables were independently associated with cognitive impairment. Because the data used in these models are cross-sectional, these analyses cannot ascribe causality. The multivariable regression was done to determine whether age, education, and race were independently correlated with cognitive impairment even when we controlled for comorbid conditions. In the first model, we included all comorbid medical illnesses as independent variables. In the second model, we included the significant variables from the first model in addition to the demographic variables, with significant bivariate relations including age, sex, years of education, and low serum albumin level. The calculation of patients' SPMQ scores for these models did not include corrections for race or education. For these analyses, we randomly divided the data set in half for derivation and validation sets. We also did logistic regression analyses to determine if cognitive impairment was independently correlated with hospitalization, emergency department visits, outpatient visits, use of preventive health care services, or mortality after we controlled for age, race, and education.

Results

Top Methods Results Discussion Author & Article Info References

From January 1991 to May 1993, we contacted 4413 patients aged 60 years and older during routine visits to the primary care practice; 3954 (90%) of these patients completed the SPMQ. One hundred fifteen patients refused the screening examination, 284 were ineligible, and 60 could not complete the SPMQ. Among all patients screened, the mean age was 68 years (range, 60 to 102 years), 68.8% were women, 63.4% were black, 43.6% had 8 or fewer years of education, 10.5% had evidence of alcoholism, and 16.2% had symptoms of depression. Six hundred twenty patients (15.7%) had evidence of cognitive impairment; 414 (10.5%) patients had 3 to 4 errors on the SPMQ, and 206 (5.2%) had 5 or more errors on the SPMQ. Prevalence of cognitive impairment by 5-year age intervals and race is shown in Table 1. The prevalence of moderate to severe cognitive impairment progressively increased across age groups, ranging from a prevalence of 2.0% in patients aged 60 to 65 years to 25.6% in patients older than 85 years.

As shown in Table 3, patients with cognitive impairment were more likely to have ever had one or more of the diagnostic tests that have been recommended in the medical literature for the evaluation of dementia [19, 20, 24]. As would be expected, the tests done most frequently were those with the broadest range of indications. For example, more than 80% of patients had hemoglobin and serum chemistry assessments and chest radiographs, with no differences observed among groups. Thyroid-stimulating hormone assessment, serologic testing for syphilis, determination of serum cobalamin or folate levels and erythrocyte sedimentation rate, neuroimaging, and lumbar puncture were done less frequently. However, these tests were done more often in the patients with cognitive impairment. Electroencephalography was done is less than 1% of patients.

Patients with cognitive impairment were more likely to have been hospitalized and more likely to have visited an emergency department in the year after the screening date (Table 3). In contrast, patients with moderate to severe cognitive impairment had fewer outpatient clinic visits on average. These differences remained significant after we controlled for age, race, and education. No significant differences were seen among the three groups in the mean number of hospitalizations, the mean length of hospital stay, or mean outpatient charges, but the mean number of emergency department visits was significantly higher among patients with mild cognitive impairment. One-year mortality was significantly greater among patients with moderate to severe impairment, even when we controlled for age, sex, years of education, and race. Among patients with moderate to severe impairment who died, 53% had been hospitalized in the year after the screening date.

Among the 206 patients with moderate to severe impairment, 156 (76%) had undocumented impairment as defined by the absence of a "dementia" or similar diagnosis listed in the outpatient medical record. Documentation had important practice pattern implications because patients with undocumented impairment were significantly less likely to receive diagnostic evaluations (Table 4). Patients with undocumented cognitive impairment had slightly lower mean SPMQ scores (6.2 compared with 7.1; P < 0.05), but no other significant differences were seen between these two groups in any of the demographic or clinical characteristics listed in Table 2. These two groups did not differ significantly in use of health services or in mortality Table 4, but the sample size limited the power to detect clinically significant differences.

Discussion

Top Methods Results Discussion Author & Article Info References

We report the findings of a systematic screening effort for cognitive impairment among an elderly primary care population and describe physician practice patterns and patient use of health services. Cognitive impairment was relatively uncommon among elderly primary care patients aged 60 to 70 years (2%) but was extremely common among patients older than 75 years (25.6%). The prevalence of cognitive impairment identified in these patients is remarkably similar to that found in previous studies [1-4], despite differences in patient populations and criteria for identifying cases. Our patient sample was derived from a primary care clinic rather than from a community-based sample and includes more urban black patients (63%) than previously reported and many patients with 8 or fewer years of education (45%). All of these patients had comorbid medical illness, and all had been evaluated by a primary care physician at least once during the observation window.

Patients with moderate to severe cognitive impairment in our study were older, had fewer years of education, and were more likely to be black. These data cannot be used to assess a causal relation between demographic variables and the prevalence of cognitive impairment. Among these patients, older patients were more likely to be black, and older patients had fewer years of education on average. However, race and education remained significantly independently correlated with cognitive impairment, even when we controlled for age and comorbid conditions. In the original scoring instructions for the SPMQ, one additional error was allowed for persons with 8 or fewer years of education or for those who were black. When we repeated our analyses and allowed one more error for black race or low education, the relations between race and education and cognitive impairment remained significant. Using data from the Duke Established Populations for Epidemiologic Studies of the Elderly (EPESE) project, Heyman and colleagues [46] also reported a higher estimated prevalence of cognitive impairment among black patients. In their study, the SPMQ was used to identify patients with cognitive impairment, and a sample of 164 patients with high and low scores also completed a formal neurologic assessment. Although our findings are consistent with those of other studies, both education and race have important consequences for socioeconomic status, occupational exposures, comorbid illness, and access to health care that would be expected to confound their relation with cognitive impairment. In a study specifically designed "to determine whether limited educational level and occupational attainment are risk factors for incident dementia," Stern and colleagues [47] concluded that more years of education or higher level of occupation reduce the risk for Alzheimer disease. However, these investigators could not ascertain whether this reduction in risk was caused by a clinical reserve in brain capacity or by simply a difference in the ease of detecting dementia. Thus, the issues of whether and how education reduces risk [48] and whether and how race confers risk for cognitive impairment remain controversial.

In our study, all of the patients had comorbid medical illness, and the mean number of conditions was similar among groups. However, mortality in the year after the screening date was significantly greater among patients with moderate to severe cognitive impairment. These data cannot address whether this increase in mortality is directly caused by cognitive impairment, less aggressive therapy among patients with cognitive impairment, or confounding medical or social conditions. Several interesting findings emerge from the comparison of the prevalence of the most common diagnoses. Patients with cognitive impairment were more likely to have congestive heart failure, anemia, and cerebrovascular disease. The simplest explanation for these relations is that these illnesses cause or contribute to cognitive impairment. Indeed, each of these conditions would be expected to compromise the delivery of oxygen to the brain. Only cerebrovascular disease was independently correlated with moderate to severe cognitive impairment after we controlled for age, race, and education. The most intriguing question about these relations is whether more aggressive treatment of comorbid conditions would result in improved cognitive function or other patient outcomes. The finding that upper-airway disease, obstructive lung disease, and gastrointestinal disorders were less common among patients with cognitive impairment may be explained by a higher early mortality rate for cognitively impaired patients with these illnesses. Also, patients with cognitive impairment in our study were less likely to have a history of smoking, although this also may be due to early death among cognitively impaired patients who smoke. The differences in the prevalence of these three conditions were not significant when we controlled for age, race, and education.

Patients with cognitive impairment were also more likely to have evidence of malnutrition. This was not simply a manifestation of ascertainment bias because the three groups did not differ in the rate of testing for albumin or of documentation of weight. The associations among malnutrition, cognitive impairment, and death have been described in previous studies [49-54]. The potential problem of poor eating habits, inability to eat, or lack of access to food among cognitively impaired patients suggests that the provision of social supports such as Meals-on-Wheels, occupational therapy, or home health aids could improve patient outcomes. Unfortunately, malnutrition is frequently not documented and is often untreated in older adults [55].

Patients with cognitive impairment were more likely than patients without impairment to have been evaluated for reversible causes of cognitive impairment, but documentation of cognitive impairment was clearly associated with a higher use of these diagnostic tests. Because of the current debate on the clinical utility of these tests [13], it is not possible to contend that the observed rate of testing is inappropriate. However, recognition and accurate diagnosis of dementia in the elderly has important benefits that are independent of the evaluation for reversible causes, and impairment was documented in only 23.5% of patients with moderate to severe cognitive impairment. It is possible that a diagnosis was made for most of these patients and that counseling was initiated but not recorded in the chart either because the physician did not plan to otherwise act on the diagnosis or because of patient or family preferences. Patients with cognitive impairment were less likely to receive the common cancer screening tests recommended for adults, but the appropriate use of these screening tests in cognitively impaired older adults is also controversial [56]. Thus, although these data clearly document variable rates of diagnostic evaluations and preventive health care based on cognitive impairment, they cannot address issues of appropriateness.

Finally, patients with cognitive impairment were more likely to be hospitalized, more likely to visit the emergency department in the year after the screening date, and less likely to visit the ambulatory clinics. Because we could not monitor use of health services outside of the facilities associated with Indiana University, it is possible that differential rates of health services use occur because patients without dementia are more likely to obtain care from other providers. Hospitalized patients with cognitive impairment did not have more hospitalizations or significantly longer lengths of stay. This finding contradicts those of Torian and colleagues [57], but the patient sample for their study was limited to a subset of elderly patients hospitalized on a specialty geriatric unit.

One limitation of our study is the reliance on the SPMQ to identify patients with cognitive impairment. Although this instrument has been shown to be valid and reliable for determining cognitive impairment [37], all patients with cognitive impairment do not have dementia. For example, some patients who have positive screening results on the SPMQ could have delirium or psychiatric illness. The SPMQ is also less sensitive in screening for cognitive impairment than longer instruments or structured interviews [12, 40], although it has a relatively high specificity for cognitive impairment. At a sensitivity of 70% and a specificity of 95% in a population with a 15% prevalence of cognitive impairment, the positive predictive value of the SPMQ at a cutoff of 3 or more errors would be approximately 74%. We did find prevalence rates that closely mirror those of studies that used more detailed measures of cognitive impairment, and the SPMQ did lead to the separation of this patient population into three groups on the basis of important differences in clinical characteristics, health services use, and mortality. The advantage of the SPMQ is its brevity both in the number of items [10] and in the ease in tabulating scores. Also, administration of the SPMQ does not require additional props. These characteristics make SPMQ particularly useful for screening in large epidemiologic studies and for busy clinicians. White and Davis [12] have recently published an excellent review of other available cognitive screening tests.

In conclusion, cognitive impairment was common among these elderly primary care patients and was associated with greater use of health services and greater mortality. Cognitive impairment was not documented in more than 75% of patients with moderate to severe impairment, and absence of documentation had important implications for practice patterns. These findings suggest a need for systematic screening for cognitive impairment among older primary care patients in an effort to improve not only diagnosis and evaluation but also long-term patient management and counseling [58].