Cognitive decline is the cardinal symptom in AD; however, noncognitive symptoms as well as patient quality of life and caregiver burden are also clinically relevant targets for treatment of AD . Presumably, a set of rating scales would be necessary for use in daily medical practice, as in clinical trials of pharmacological and nonpharmacological interventions, to assess multiple AD-relevant domains. Assessment is therefore not only time consuming but current scales are not always applicable to all dementia stages and fail to measure therapy effects reliably . A need for more sensitive instruments to detect AD symptom progression over time has been acknowledged [14, 15].
The ROSA is a novel and unique multidomain assessment scale in AD, which has been designed for the need of daily clinical practice to assess disease progression over time and to measure dementia treatment effects. In the present article, we report the ROSA factorial structure based on clinical study data and discuss the psychometric characteristics of the scale, demonstrating that the ROSA is a valid, reliable and time-efficient observer-rating instrument to aid medical practitioners in sensitively assessing changes in AD symptoms over time.
The ROSA items cover a broad spectrum of AD symptoms, including cognitive impairment (Items 1 to 3), communication and social interaction abilities (Items 4 to 6), behavioral symptoms (Items 7 to 11), and symptoms relevant to activities of daily living (Items 12 to 14). Items 15 and 16 provide an overall evaluation of the patient's quality of life and the caregiver's burden, respectively. As demonstrated by the factor analysis, all items initially included in the ROSA were retained in the final ROSA clustered into two factors. The item assessing patient's quality of life (Item 15) showed the lowest loading within Factor 1, which indicates a rather weak relationship with the items comprising cognitive and functional AD symptoms in the ROSA. On the other hand, Item 15 showed a factor loading higher than the defined threshold of 0.4 also within Factor 2, implying that patient quality of life shows a similar relationship with all behavior items within the ROSA. This confirms the complexity of a patient's quality of life rating and the influence of a number of factors on the quality of life, including cognitive dysfunction, functional disabilities, and neuropsychiatric symptoms, as recently shown in a longitudinal study investigating the change in proxy-rated quality of life in a large cohort of home-living patients with AD . The authors clearly demonstrated that measures of quality of life seem to comprise different functions than typical clinical variables such as cognitive and noncognitive dysfunction and activity of daily living . It is also interesting that Item 13, describing the interest of the patient for his/her surroundings, was included in the ROSA factor comprising cognitive and functional/activity of daily living items (Factor 1). Given that this symptom is usually considered a core feature of apathy, its non-inclusion in the ROSA behavioral factor allows it to be focused only on the most disruptive symptoms.
High internal consistency and test-retest reliability of the final ROSA were shown, pointing to a high reproducibility of the ROSA results. The content validity analyses demonstrated a good correlation between the ROSA and each of the validated scales (ADAS-cog, SIB, NPI, and DAD) used as standard measures for the assessment of AD-relevant symptoms evaluated with the ROSA. In the current study, the ROSA was always applied prior to any other scale used in order to prevent a possible influence on the ROSA by the other scales. A putative limitation of the inter-rater reliability analysis applied in this study was that ROSA assessment was restricted to assessing the patient on video and to having clinical information on the patient only available in a written report. The results, however, showed good inter-rater reliability for the ROSA. A more direct approach for assessing inter-rater reliability required personal interviews by 61 raters in Germany and Austria, which was not feasible.
The ROSA responsiveness - that is, the ability of the ROSA to detect changes over time due to an intervention - was tested by two widely used criteria: the RI and Cohen's d coefficient. The RI provided an estimate of the ability of the ROSA to discriminate between patients who have changed due to treatment and those who have not benefited by the treatment. In turn, Cohen's d coefficient, which provides an estimate of the effect size due to an intervention, was analyzed to demonstrate the ability of the ROSA to detect any change over time due to memantine treatment. Given the differences in the responsiveness statistics of these two parameters, it is not surprising that the values of RI and Cohen's d differ greatly. One could expect that the RI values are higher than Cohen's d values because the RI numerator increases with exclusion of change scores from nonimproved patients, and the respective denominator becomes smaller with the inclusion of the standard deviation of change only for stable patients.
Altogether, the results indicate that the ROSA can sensitively measure relevant changes in all disease severity stages (RI ≥0.8 in all stages) and can reliably analyze the effect size of drug treatments. The latter was supported by the data for the ROSA estimate of the effect size due to treatment with memantine, which was shown to be within the known range of the memantine efficacy in moderate and severe AD stages . Subsequent clinical experience with the ROSA, as well as its further use in clinical studies, would be necessary to permit an accurate assessment of the change in the total score that constitutes a clinically important difference, and thus to enable more accurate assessment of the ROSA responsiveness.
All clinicians taking part in the study were experienced professionals in AD diagnosis and treatment. Prior to initiating the study, all raters (physicians, psychologists and nurses) were trained in the application of the ROSA and the other AD scales.
With the ROSA training, the raters were instructed how to estimate a patient's disease severity prior to the ROSA administration and how to use the relevant examples of the scenarios within the ROSA. The raters were informed that the initial staging is an essential prerequisite for the ROSA administration; it establishes the assessment framework in the ROSA - that is, which example of a scenario (early, middle, or late) should be used for the assessment - and it ensures that, when patients' clinical symptoms are reassessed over time, any changes in the severity stage could be detected by the rater prior to the reassessment with the ROSA. Also, raters were instructed to be careful when applying modified scenarios; that is, to document any change of an item scenario and use the modified scenario again at the ROSA follow-up assessment. A video example of the ROSA administration was shown to demonstrate the use of the scenarios within the ROSA and the possibility to adjust an item scenario to the individual patient when needed.
For 95% of the study patients, the same clinician performed the assessments with the ROSA. The ROSA was scored on the basis of information obtained from the clinician interview and information available on the patient's medical history, clinical data, and previous encounters of the clinician. The interview partner was in most cases the primary caregiver who regularly sees the patient, thus being able to answer the questions on the basis of personal interaction with the patient during the preceding week.
The possibility to cover broad AD severity stages is a substantial advantage of the ROSA in daily clinical practice. One should, however, consider that the ratings are done for each stage using the same scoring range, while the scenarios used for each disease stage represent increasing levels of impairment. The ROSA total scores can thus be directly compared only within a patient or between patient groups of one and the same disease severity stage. This is a certain limitation in using the ROSA assessment in clinical trials for comparison of patient groups with different disease severity stages. Patients who change their severity stage over time may have to be excluded from statistical analyses in clinical trials; however, the individual results of the ROSA survey can still be compared relative to one another. Any change in the severity stage of a patient could be used as a global value of the AD progression and effects of an intervention. Besides, the same scoring range for all disease stages does not limit the use of the ROSA in routine medical practice where a single patient is assessed and comparison between patients is not needed. If a change in the severity stage of a patient occurs since the last assessment with the ROSA, the clinician can perform the new assessment in the same way, using the same scenario for each item but with the new examples provided for the respective new severity stage. In such cases, the ROSA total score could not be used for a direct comparison between the new and the last assessments of the patient but the change in the severity stage itself is a global estimate for the progression of the patient disease over the past time.
Altogether, the study results reported here demonstrated that AD-relevant symptoms and their severity can be reliably assessed with the ROSA in patients at all disease severity stages. In addition to the ROSA total score, a graphical presentation of the ROSA single-item scores could easily be displayed on the ROSA sheet by connecting the ticked fields on the numerical scale. Given the possibility to use the same ROSA sheet for following assessments of a single patient, clinicians could directly compare two or more repeat evaluations based on the graphical display of the single-item scores as well. The ROSA may thus contribute to identifying particular symptoms assessed with single ROSA items that necessitate treatment and counseling. The item-specific ratings may have the advantage that, despite worsening in the ROSA total score, item scores may indicate the AD-relevant domains within the ROSA responsible for the worsening and also the domains that remain unchanged or even improve. This may be of particular benefit for studies on pharmacological and nonpharmacological interventions where effects on specific domains are expected to occur. The use of the ROSA item scores in clinical practice and research, however, needs to be appropriately validated in the future.