In this study we used two cognitive tests to show that cognitive impairment was common in medical inpatients, with 73% of patients having at least one abnormal test result during hospitalisation. By using a multivariable Cox proportional hazards model, we showed that cognitive impairment was associated with a three-fold increase in 12-month mortality.
We divided the patients into three groups according to their results on cognitive tests (with 0, 1 and 2 abnormal tests). Demographic data, presenting complaints and comorbidity showed only subtle differences between the three groups. In terms of recognition, the hospital staff did not recognise cognitive impairment in the group with 1 abnormal test any more than in the group without impairment. This is alarming, as the increase in mortality risk of this group was substantial and similar to that of the group with 2 abnormal tests. To improve detection of cognitive impairment, our study suggests that subjective memory complaints are too unspecific to be used effectively, but that informants should be considered as valuable assets.
Our regression model did not retain variables that one would expect to be significant predictors of mortality, such as age or certain comorbidities, e.g. cancer. Regarding age, this could be due to the designation of patients to different departments at the hospital. At general internal medicine wards, patients tended to be either very old or slightly younger but with multiple comorbidities, for example, a 65-year old with a heart attack would typically be transferred from the ER to the cardiology department but a 65-year old with diabetes, hepatic failure, obesity, a neurologic disease and an infection would more likely be transferred to a general internal medicine ward. This selection could have affected the impact of age on mortality in our study population. Regarding comorbidities, these were coded merely as absent/present, there was no severity ranking or temporal perspective. Therefore, a patient who had had surgery 30
years ago for cancer of the colon was not distinguished from a patient with present metastatic disease. This is of course not a proper representation of the clinical situation. However, the inclusion of Charlson comorbidity index, a very stable and highly significant comorbidity variable, in our model did not affect the finding that cognitive impairment independently predicted mortality.
Studying nurses’ and physicians’ recognition at admission only may seem unfair to regular staff, as impairment could have been discovered later on. However, for ethical reasons, regular staff was informed promptly when cognitive impairment was disclosed in a patient, thus prohibiting studies of the staff’s recognition further along the admission. Changes in cognition could have occurred between admission and administration of cognitive tests, for example delirium resolving quickly or incident delirium occurring between admissions and testing. However, our cognitive tests were performed when laboratory values were within an acceptable range and patients were generally in a better state when tested than upon admission. This indicates that changes of cognition were likely to have been to the better, thus favouring the staff’s recognition on admission. The physicians and the nurses were aware of the study and it is likely that they were more vigilant towards cognitive symptoms than in regular conditions. Furthermore, we used the most generous cut-offs possible to represent recognition. Taken together, it is unlikely that the recognition rates by hospital staff have been underrated.
Representativity also needs to be addressed, as many patients were not included. Patients excluded for hospital-related reasons (with unknown cognitive status) were fewer (n
89) than patients excluded due to disease-related conditions (n =121). The latter could be assumed more cognitively impaired, given the exclusion criteria (severe delirium, terminal disease, aphasia etc.). This would give a bias towards including the healthier part of the population. This notion is also supported by the fact that all included patients managed to fulfil the interview and the MMSE. Thus, it is unlikely that the prevalence of cognitive impairment is overrated.
We did not aim to diagnose dementia or delirium but rather to study cognitive impairment in a broader sense including its recognition and consequences in terms of mortality. To some extent, we tried to exclude patients with delirium but most likely patients fulfilling delirium criteria were included in the study. However, our findings imply that acknowledging cognitive impairment is important in medical inpatients regardless of its duration.
The main strengths of our study are its simplicity and the possibility to apply the findings in a clinical setting. We used two widely employed cognitive tests, taking approximately 15
minutes to administer. These were performed in a standardised way, much similar to that of clinical routine. We used an easily applicable approach with 0, 1 or 2 abnormal test results to make a crude estimation of cognitive impairment. Despite this simple approach, our estimate of cognitive impairment was a significant predictor of mortality in a clinical material of 200 patients with multiple diseases. Furthermore, we used simple questions and measures already applied in hospital routine to study the recognition of cognitive impairment from the patients’ perspective as well as from informants and different healthcare professionals.