A 52 month follow-up of functional decline in nursing home residents – degree of dementia contributes

This was a 52 months prospective study with four assessments. Baseline assessment took place between November 2004 and January 2005. The follow-up assessments took place after 12, 31 and 52 months.

All residents in 26 nursing homes in 18 municipalities and four counties in Norway with a stay of a minimum of 14 days were asked to participate in the study. The municipalities were chosen to make the sample representative in terms of municipality size [8]. In all, 1165 residents were eligible. Two residents or their next of kin refused participation, four had important missing information, and 233 had no dementia according to the Clinical Dementia Rating (CDR) scale, i.e. CDR < 1 [41]. Thus, 932 residents were included in the longitudinal study of P-ADL.

The Demographic information age, level of education, marital status and length of stay at inclusion, was collected from the medical records.

The Medical health information was collected from the medical records with the use of diagnoses in the International Statistical Classification of Diseases and Related Health problems 10^th Revision (ICD-10). The diagnoses were sorted into diagnostic groups, i.e. stroke-, cardiovascular-, pulmonary-, musculoskeletal-, digestive-, endocrine-, neurological-, genito-urinal- and neoplasm diagnoses [42]. A co-morbidity index was calculated by summing up the occurrence of diagnoses (yes/no) in each of the nine groups of diagnoses.

Sensory loss was registered by two single items for vision and hearing impairment, respectively, i.e. perceived severely impaired or not. No specific measurement was used and the occurrence of vision and hearing impairment are based on the primary professional caregiver’s clinical judgement.

Dementia was assessed using the Clinical Dementia Rating (CDR) scale which covers six domains (memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care) with five response categories (0, 0.5, 1, 2, 3) [41, 43]. The total score is calculated by means of an algorithm giving priority to memory [41]. Residents with a total score of one or higher were regarded as having dementia [44‐46]. In the analysis we used the CDR sum score (sum of boxes), which ranges from 0 to 18 with a higher score indicating greater dementia severity. The correlation between the CDR score and the CDR sum of boxes is high. However, due to an increased range of values, the CDR sum of boxes offers important advantages when analyzing the data. [47].

Neuropsychiatric symptoms were assessed using the Neuropsychiatric Inventory (12-item NPI) [48] in a translated and validated Norwegian version [49]. The 12-item inventory covers the following symptoms: delusion, hallucination, agitation/aggression, disinhibition, irritability/lability, depression/dysphoria, anxiety, apathy/indifference, aberrant motor behaviour, sleep and night-time behaviour disorder and appetite/eating changes. For each symptom, severity (score 1–3) multiplied by frequency (score 1–4) gave a score ranging from one to twelve. Based on a previous principal component analysis, psychosis (delusions, hallucination), agitation (agitation/aggression, disinhibition, irritability) and affective (depression, anxiety) sub-syndrome scores were formed by summing the score of the included items [50‐53].

Psychotropic medications were grouped according to the ATC code, into antipsychotics (N05A except lithium), antidepressants (N06A), anxiolytics (N05B), hypnotics/sedatives (N05C) and antidementia medication (N06D) (yes versus no). The information was collected from the medical record of each resident.

Level of functioning in personal Activities of Daily Living (P-ADL), the dependent variable, was measured by the Physical Self-Maintenance Scale (six items, score range 6–30, i.e. P-ADL scale score) [54]. High scores indicate a lower level of functioning.

Prior to the study, 16 research assistants (all registered nurses) attended a two-day training program on how to conduct the interview. A one-day training program was carried out prior to each follow-up assessment. The project leader (GS) was available for consultation while the data were collected. The research assistants collected the data by means of a standardized interview with the primary professional caregiver and by extracting information from the medical records.

Study information was given to the patient and their family members. An explicit consent was not required for enrollment, but the patients and their next of kin were informed in writing that they could refuse to participate at any stage of the study. The study, and the procedure for information and the right to decline participation were approved by the Regional Ethics Committee of Eastern Norway in 2004, the Data Inspectorate and the Directorate for Health and Social Affairs.

Continuous socio-demographic and clinical characteristics were presented as means and standard deviations (SD), while frequencies and proportions were used for categorical characteristics. A linear regression model was estimated to assess the development in P-ADL throughout the follow-up period of 52 months. As P-ADL consists of repeated measurements, the assumption of independent observations required for ordinary linear regression was violated. Therefore, a linear mixed model containing both linear and second–order time components as fixed effects (growth model) as well as random effects for intercepts and slopes for time was estimated first. Such a model accounts for correlations due to repeated measurements and accommodates any degree of imbalance in longitudinal data. That is, the model is particularly suited for analysing data with missing values and drop-outs due to death, for instance. The SAS MIXED procedure was used to fit the model. The main independent variable, degree of dementia assessed with CDR sum of boxes, and interaction term between degree of dementia and time were then entered into the model as fixed effects. Two different models with respect to adjustment were further estimated. In both models, adjustment variables were demographic characteristics (length of stay prior to baseline, age, gender, education, and marital status), physical health information (co-morbidity index, severe sensory loss), neuropsychiatric symptoms and use of psychotropic medications. Length of stay, age, and gender were defined to be the confounders, while the other adjustment variables were defined as secondary independent variables. The first model contained adjustment variables all measured at baseline. The second model included demographic characteristics, somatic health difficulties at baseline, while neuropsychiatric symptoms of dementia and use of psychotropic medications were entered as longitudinal variables. The following modelling strategy was employed. Firstly, a model containing all defined secondary independent variables was reduced by applying Akaike’s Information Criteria (AIC). The least significant secondary independent variables were removed from the model one at a time and AIC was calculated at each step. The model with the smallest value of AIC was chosen. Secondly, the CDR sum of boxes, interaction term between the CDR sum of boxes and time, and the three confounders were entered into the reduced model and a final model was estimated. Lastly, to assess eventual gender-specific differences, interactions between gender and CDR sum of boxes, age, and education were entered into the final model. However, none of them were significant and therefore excluded. The results of the regression analysis were tabulated as coefficients with the corresponding 95% confidence intervals (CI). Regression coefficients were used to estimate the P-ADL score for each time point from an unadjusted model as well as from both adjusted models at the median baseline CDR sum of boxes value and were presented graphically.

Analyses were performed in SAS v9.2 and SPSS v20. P-values below 0.05 were considered statistically significant. All tests were two-sided.

At baseline, the mean (SD) age and length of stay for the participants were 84.5 (7.5) years and 932 (910.1) days, respectively (see Table 1). In all, 246 (26.4%) of the residents were men and the mean score of the co-morbidity index was 1.8 (SD 1.3, ranging from 0 to 6 groups of diagnoses). The mean (SD) baseline CDR sum of boxes was 13.1 (4.0). Of the 932 participants at baseline, 160 (17.2%) participants were available at the fourth follow-up (see Figure 1). The individual development of P-ADL throughout the follow-up period and the mean P-ADL at four time points (see Table 2) indicated worsening in P-ADL functioning, which was confirmed by the second-order growth model. The positive linear time component (95% CI) of 0.14 (0.12; 0.16) shows that the P-ADL functioning declined over time, and negative second-order time component of -0.001 (-0.001: -0.0005) shows that the mean rate of decline decreased with time.

The importance of degree of dementia for P-ADL functioning over time was studied in two adjusted models. In the first adjusted model (Table 3), the association between degree of dementia at baseline and P-ADL functioning over time was assessed. There was a significant second-order time trend in P-ADL functioning, and more severe dementia at baseline was associated with declined P-ADL functioning over time. The association between degree of dementia and P-ADL functioning was strongest at baseline, the development in time flattened and the association between the baseline degree of dementia and the decline in P-ADL functioning decreased with time.

Table 4 presents the results from the second model finding degree of dementia at each time point being associated with poorer P-ADL functioning at the same time points. As in the first model, there was a decline in P-ADL functioning throughout the follow-up period. Furthermore, as in the first model, the association between degree of dementia and P-ADL functioning was strongest at baseline, the development in time flattened and the association between baseline degree of dementia on decline in P-ADL functioning became weaker with time.

In addition, longer length of stay, being married, having a higher comorbidity index score and having severely impaired vision at baseline were associated with worse P-ADL functioning through the observation period. Among the independent variables with assessments at all four time points included in the analysis, a higher agitation sub-syndrome score and apathy score, as well as not using anxiolytics and antidementia medication were associated with lower P-ADL functioning at the same time points.

Figure 2 illustrates the unadjusted P-ADL development over the follow-up period as well as P-ADL development adjusted in two different ways (two models). Independently of the adjustments made, there is an upward trend in the P-ADL values over time, i.e. the P-ADL functioning declines, but the rate of decline flattened during follow-up. The degree of decline in P-ADL functioning during follow-up was moderated using the adjusted models.