In developing countries, such as Brazil, the impact of aging on health services is considerable. As the Brazilian population is aging at a faster pace compared with other countries [14], it is important to identify high-risk patients who would benefit from an intensive approach to address their individual needs.
In this study, longer hospital stays among elderly patients were significantly associated with several variables, including age, stroke, congestive heart failure, diabetes mellitus, dementia, delirium, incontinence, difficulty swallowing, nutritional risk, decreased level of consciousness, pressure ulcers, anemia, number of medications (more than 5 medications increased the risk of longer hospitalization), and cognitive and functional abilities (all p values < 0.05; Table 2). In our study, we observed that only diabetes and an inability to perform bed/chair transfers (assessed with the Barthel Index) were independently associated with higher risk. Therefore, we suggest that these two variables may provide the best initial screening to identify patients to receive a later, more comprehensive assessment, which will help them get the proper care needed.
Other studies also reported that diabetes is associated with an increased risk of hospitalization and longer hospital stays [15,16,17]. In our study, 29% of patients had a diagnosis of diabetes, and of these, 20% had hospital stays longer than 10 days furthermore, diabetic patients had approximately twice the risk of long-term stays compared to nondiabetic patients (odds ratio - OR 1.30–3.65).
Some authors have also implicated functional capacity as a strong predictor of long hospital stays as well as institutionalization and death [18, 19]. For example, frailty markers are associated with adverse health outcomes, both within the hospital and in the community at large. Gait speed could be used as an initial screening for risk of long hospital stays and for home discharge. However, its applicability is limited to patients with testable clinical, physical, and cognitive conditions. Therefore, we believe that assessing bed/chair transfer is more useful as a risk detector in the hospital setting, as it would not exclude a significant proportion of the population [20].
The Barthel Index is an internationally used instrument for functional assessment, and some authors have also found it useful to detect adverse events during hospitalization [10, 12, 21]. In our study, patients who obtained a score of 0 or 5 on the chair/bed transfer item of the Barthel Index had eight times the risk of remaining in the hospital for a longer period (OR 4.61–13.76).
It should be noted that in some studies cited, the average length of stay in the hospital for elderly patients was 10 days (7–14 days) [22], and 28 days was considered the cutoff point for long hospital stays. In contrast, in our study, the average hospital stay was six days, and only 6.3% of our population stayed in the hospital for longer than 10 days (as opposed to approximately 54% in other studies).
The limitations of our study were the relatively small number of elderly patients who were hospitalized for more than 10 days (n = 70). This reflects our hospital’s constant concern with the clinical consequences of long hospitalization times and the efforts to reduce them. Additionally, in contrast to other studies, we excluded hemodynamically unstable patients under intensive care and semi-intensive units, because they are prone to longer stays by definition. Furthermore, we hoped to identify other risk factors in our sample, and we considered critically ill patients to be a confounding factor. However, an important feature of our work is that we included elderly patients from all clinical and surgical units of our large general hospital, regardless of diagnosis at admission. The nutritional evaluation was not performed using a standardized tool for elders. However, an experienced clinical nutritionist assessed all elders during the hospital admissions; this would have a good sensitivity to identify elders at risk of undernutrition. Additionally, the number of elders enrolled in this study was slightly lower than estimated number determined by the sample size calculation. However, on ROC analysis, we observed good accuracy (AUC = 0.773) demonstrating good quality of the classification model.
These findings reinforce the need for multidimensional evaluations for hospitalized elderly patients to prevent serious adverse events, as suggested by Ellis et al. [23]. However, recognizing a subset of questions that could potentially identify patients at risk for longer hospitalizations seems important, as those elders would have multidimensional needs for intensive rehabilitation and clinical surveillance. The utilization of geriatric wards would help meet those needs. However, we must identify the patients who would most benefit from this still limited and costly resource.
The goal of this study was to construct an instrument that would predict a specific outcome important for daily practice. In fact, we believe that there might be an overlap of high-risk patients identified by this model and elders with frailty or disability. However, the selection of patients according to simple characteristics, such as diabetes and the inability to transfer, followed by referral to an appropriate model of care, would be necessary to reduce hospital length of stay. Specific rehabilitation and geriatric assessment wards for the population at risk would be necessary to be able to rehabilitate patients at functional/clinical risk. Certainly, a comprehensive geriatric assessment would provide a better estimate of patients’ needs, but it would require time and the training of a team.