We systematically searched electronic databases for articles published from as early as 1974 through May 2008, including MEDLINE, CINAHL, Psyc-INFO, EMBASE, Dissertations International, and the Cochrane Database of Systematic Reviews. In addition, we manually searched the reference lists of systematic reviews and continued to manually search databases up through September 2008. For each database, we used master heading terms relevant to dementia and stage (advanced, severe, or late stage) and treatments (treatments, therapies, interventions, psychotherapy). We limited the searches to research, systematic reviews, meta-analyses and practice guidelines. This broad search strategy identified interventions studies related to all types of treatments (pharmacological and non-pharmacological) for NPS in dementia. We saved all identified articles to an online research management database and then excluded all studies that did not meet inclusion criteria.
English-language intervention studies published in peer-reviewed journals were reviewed that included: (i) all participants meeting criteria for severe or very severe dementia documented by the use of a validated cognitive or functional measurement instrument, and (ii) measures of NPS as the primary outcome variable(s). As shown in , severe dementia was defined by the: (a) Mini-mental State Examination (MMSE; Folstein et al., 1975
) with scores ranging from 0 to 10 out of a possible 30; (b) the Clinical Dementia Rating (CDR; Hughes et al., 1982
) stage 3 (Perneczky et al., 2006
); (c) the Global Deterioration Scale (GDS; Reisberg et al., 1982
) stages 6 (severe) and 7 (very severe); (d) the Functional Assessment Staging (FAST; Sclan and Reisberg, 1992
) stages 6 (severe) and 7 (very severe); and (e) the International Classification of Diseases (ICD-10; World Health Organization, 2007
) stage severe. By definition, cognitive impairment worsens with each increasing stage of dementia. For individuals who fall below a score of 6 on the MMSE, the Severe Impairment Rating Scale (Rabins and Steele, 1996
) provides a measure of preserved cognitive function for very basic tasks.
Staging of advanced dementia
In addition, studies that stratified participant responses by severity of cognitive impairment were included when they met all other criteria. When the range of severity scores was not reported, the determination of severity for inclusion in the review was based upon the mean cognitive impairment scores plus two standard deviations from the mean falling between 0 and 10 on the MMSE or an equivalent measure. (The cognitive impairment criterion was later broadened to include 0–17 on the MMSE or the equivalent.)
Studies were excluded that: (i) had a primarily pharmacological treatment focus, or a combined pharmacological and non-pharmacological focus, including the use of herbal or dietary treatments; (ii) examined the effects of electrical stimulation; (iii) focused on palliative or end-of-life care (unless the focus was specifically on relieving NPS); (iv) focused on sleep dysfunction or sleep architecture in dementia (unless other NPS such as agitation were also outcomes of interest); or (v) were single case reports, observational studies, or qualitative studies.
All studies meeting inclusion criteria were appraised for design strength and quality of evidence using an adapted version of the Johns Hopkins Nursing Evidence-Based Practice (JHNEBP) model (Newhouse et al., 2007
). The JHNEBP model includes three levels of strength referring to the rigor of the study design, and three broadly defined levels of quality that address sample size, experimental control, and the definitiveness of the conclusions. The same quality guidelines were used with the addition of a fourth component, the consistency of the direction of effects across studies, to evaluate groups of studies within intervention categories. The JHNEBP model was operationalized using criteria from the Forbes (1998)
Validity Rating Tool.
Using the adapted rating tool (), the studies were first rated for strength of design. Experimental designs received a strength rating of Level I and quasi-experimental designs received a rating of Level II. Non-experimental Level III designs were not included in the review. Because of the difficulties inherent in doing clinical research with this frail population, nearing the end-of-life, we included in the Level I strength category the factorial or crossover designs in which individuals were randomly assigned to groups and participated in all conditions. Studies that lacked experimental control or the ability to randomize were included in the Level II category. The quality ratings were determined as follows: scoring proceeded top down, from high to low. For a study to have an evidence rating as high, all of the quality ratings had to be in the high category, with one exception. The only allowable diversion from the high grade for studies that qualified as high was the absence of a power analysis, as long as the criteria for sample selection were clear and the sample size was sufficiently large. Studies rated as having a moderate quality of evidence met all criteria from the moderate category or a combination of moderate and high categories. Finally, a low rating was given to studies that had any of the low quality of evidence indicators.
Strength and quality of research evidence rating scheme for individual studies