Identification of community-dwelling older adults at risk of frailty using the PERSSILAA screening pathway: a methodological guide and results of a large-scale deployment in the Netherlands

The PERSSILAA screening pathway was a two-step annual screening programme to identify older adults at risk for developing frailty (i.e. pre-frailty). First, older adults were asked to complete a self-screening questionnaire to assess their general health status and their level of decline on physical, cognitive and nutritional frailty domains. Second, older adults who, according to step one, were at risk of developing frailty were invited for a face-to-face assessment that focuses on the physical, cognitive and nutritional domains in more depth.

During the first step, older adults with an age between 65 and 75 years old were asked by their General Practitioner (GP) to complete a comprehensive set of questionnaires including questions on health status and demographic characteristics. This was done via a mass mailing (postal survey) to ensure that all older adults within defined GP practices were reached. These GPs signed up to participate after a general meeting of GPs organized by the municipality. Prior to posting the PERSSILAA screening questionnaire to patients, GPs identified those patients who were, in their opinion, too frail to participate. These patients included those with multiple serious illnesses or those with limited life expectancy monitored extensively by their GP; these did not receive an invitation. Based on the preferences of the older adult, questionnaires can be completed on paper or online. Previous research has shown that offering different types of administration (online or on paper) does not affect screening results [17, 18]. When the questionnaire was completed online, older adults could receive their results immediately. When the questionnaire was completed on paper, adults could hand over the questionnaire at their GP office or send it by regular mail. Afterwards they received their results via email or by regular mail (when their email address was unknown). Older adults screening as robust were not contacted further; older adults screening positive for pre-frailty were invited for a face-to-face screening (Step two of the PERSSILAA screening); older adults classified as frail, were contacted by their GP [15, 19].

During step two of the PERSSILAA screening, a face-to-face assessment was conducted aiming to gain more information on the level of functional decline on physical, cognitive and nutritional domains. During the PERSSILAA project the face-to-face assessments were performed by trained volunteers or students at a location in the older adult’s neighbourhood (e.g., a local retirement home). After the face-to-face meeting, older adults received the outcomes of their assessments, information about healthy ageing and advice about the various health programs available in their community. Older adults that were identified as pre-frail on one of the three domains were offered existing training services targeting physical and cognitive domains available in their community organized by their municipality or local welfare organisation, and education about healthy eating for older adults. The service model that accompanies the PERSSILAA screening model was the result of a participatory design process, in which all stakeholders were actively involved. Among other things, it was decided to limit the age span of the screening program from 65 to 75 years of age, in order to not interfere with existing screening programs for those over 75 years old and to take advantage of the technology skills of the less old cohort. For more information about the development of the model we refer to Van Velsen et al., 2015 [15].

The initial instrument scored was the Dutch version of the Groningen Frailty Indicator (GFI) [20, 21] administered to assess the older adult’s general level of frailty. The GFI is a validated 15-item screening instrument that measures loss of function and resources across physical, cognitive, social and psychological domains. Each item is rated on a 3-point scale (Yes, No, Sometimes). The total score of the Groningen Frailty Indicator has a range from 0 to 15. A score of 4 or higher represents moderate to severe frailty. The SF-36 [22] physical functioning subscale (PF-10) was then administered to examine perceived physical functioning. The PF-10 consists of 10 items, each rated on a 3-point scale (yes, limited a lot; yes, limited a little; and no, not limited at all). In order to calculate an overall score, all answers are summed and then transformed to a 0–100 scale. Higher scores represent better health status. Older adults are classified as limited in physical functioning if they score below a cut-off score of 61 (pooled mean score in a general older Dutch population) [23]. The SF-36 PF-10 is validated in Dutch [24]. To screen for cognitive impairment, the self-administered version of the AD8 Dementia Screening Interview (AD8) [25, 26] was used to detect early cognitive changes associated with many common dementing illnesses. Older adults were asked to rate changes in their ability for each of the 8 items, without attributing causality. Each item is rated on a 3-point scale (Yes, a change; No, no change; and Don’t know). The final score is a sum of the number items marked “Yes, a change”. Based on clinical research findings and validation samples, the following cut points are provided: 0–1: Normal cognition and 2 or greater: cognitive impairment is likely to be present. The AD8 is not validated in Dutch. Finally, the Mini Nutrition Assessment Short-Form (MNA-SF) [27] was administered to identify those who were malnourished or at risk of malnutrition. The MNA-SF consist of 6 items, focusing on food intake, weight loss, mobility, psychological stress, neuropsychological problems and Body Mass Index (BMI). By summing up the scores on all 6 items, a total score is derived. A score of 12 to 14 points indicates a normal nutritional status, a score of 8 to 11 points indicates risk of malnutrition and a score of 0 to 7 points indicates that the participant is malnourished. The MNA-SF is validated [28], but not in Dutch.

Based on the GFI, older adults were initially stratified as either robust, pre-frail or frail. GFI scores higher than 4, were classified as frail; scores of 4 were classified as pre-frail [20]. Older adults were also screened and categorised based on deficits in physical, cognitive and/or nutritional domains identified from the other screening instruments. Older adult were also assigned to the pre-frail group irrespective of their GFI score when they scored 60 or below on the PF-10 items of the SF-36 and / or when they scored 2 or higher on the AD8 and / or if their score on the MNA short form was 7 or lower. Finally, older adults that did not meet at least one of the conditions for being classified as pre-frail or frail were classified as robust. Table 1 provides an overview of the screening questionnaires used during the first step of the PERSSILAA screening procedure and the triage to the three groups.

To assess the physical status of older adults four tests were used. First, the timed up and go (TUG) test [29] was administered. The TUG test is a simple office-based test, used to identify persons at risk of falling because of balance or gait problems. The TUG test is performed 3 times and an average score is calculated. Second, the chair stand test (CST) [30] was administered. The CST tests the level of lower extremity strength. Third, the chair sit and reach test (CSRT) [31] was administered. The CSRT assess the level of physical flexibility. Fourth and finally, the two-minute step test (2MST) [32] was administered. The 2MST tests the level of aerobic endurance. All four objective tests have normal range scores related to gender and age based on previous research [32]. Scoring for TUG test was adapted to the range scores of Rikli & Jones [32], where an 8-ft TUG test was performed instead of 3 m. Scoring of the CSRT is indicated in inches to compare the score with the normal range. Scoring below the pre-defined cut-off scores indicates “functional decline”.

For the assessment of the cognitive status, the Quick Mild Cognitive Impairment (Qmci) screen was used. The Qmci screen is a short screening test for cognitive impairment, developed as a rapid, valid and reliable tool for the early detection and differential diagnosis of mild cognitive impairment (MCI) and dementia [33]. The Qmci screen has six subtests covering the following cognitive domains: orientation, working memory (registration), visuospatial/executive function (clock drawing), semantic memory (verbal fluency), and two episodic memory tests (delay recall and logical memory) [34]. Validated in Dutch [35], the Qmci screen can be completed in 3 to 5 min [34]. The overall Qmci screen score ranges from 0 to 100, with lower scores indicating cognitive impairment [36].

For the assessment of the nutritional status, the extended form of the Mini Nutrition Assessment (MNA) [27] and waist circumference were used. The extended version of the MNA is a validated screening tool that identifies older adults who are malnourished or at risk for malnutrition. In addition the items of the MNA-SF, the full MNA consists of an extra 12 items to provide additional information. The overall score of the full MNA is calculated by summing the scores on all 18 items.

Considering to Dutch law (Medical Research Involving Human Subjects Act), the nature of this research (general screening of older adults by questionnaire and face-to-face screening) and that public screening initiated by GPs for different health conditions is usual care in the east of the Netherlands, this study did not require formal ethical approval. The appropriate ethics committee (METC Twente) ruled that no formal ethics approval was required in this particular study (K14–42). At the end of the questionnaire, older adults were asked whether they agreed to the use of their data for research purposes, when they agreed they had to tick a checkbox. Only the data from those consenting were included in the PERSSILAA screening database. All data were anonymized.

Statistical analyses were performed with IBM SPSS Statistics 19 for Windows. All outcome measures were visually inspected for normal distribution using histogram and probability plots, prior to the selection of appropriate statistical tests. Descriptive statistical methods were applied for each of the outcome measures. Normal data were presented as mean ± standard deviation (SD), non-normal as median with range. As the PERSSILAA screening was an annual screening during the length of the PERSSILAA project (three years), only the baseline (first) complete two-step screening data were analyzed and presented in this paper. For statistical analyses, the level of significance was set at α < 0.05.

The PERSSILAA project stated in November 2013 and ended November 2016 and this study was carried out in year two and three of this project. In total, 32 GP offices participated and 10,331 community-dwelling older adults were invited to participate in the first step of the PERSSILAA screening. With a response rate of 36.6%, the data of 3777 of the older adults were stored in the PERSSILAA screening database. The demographic characteristics of the 3777 older adults included in this analysis are presented in Table 2. The mean age of the older adults was 69.9 (SD ± 3.8). The majority of those participating had access to the internet, and 30% of the older adults completed the questionnaire online.

Based on the first step of the PERSSILAA screening, 16.8% of the older adults were classified as frail (n = 634), 20.6% of the older adults were classified as pre-frail (n = 777), and 62.3% of the older adults were classified as robust (n = 2353). The results of 13 participants were incomplete. Table 3 provides an overview of the average score on each measurement instrument according to the classification of the participants. Among the older adults identified as frail, 11.3% showed impairments in all three domains, 24.6% had decline on two of the three domains (16.9% in the physical and cognitive domains, 4.8% in the physical and nutritional domains and 2.9% on the cognitive and nutritional domains), and 36.8% experienced problems in only one domain (21.7% in the physical domain, 11.1% in the cognitive domain, 4.0% in the nutritional domain). Among those identified as pre-frail, 0.8% showed impairment in all three domains, 10.5% in two of the three (7.8% in the physical and cognitive domains, 2.2% in the physical and nutritional domains, and 0.5% in the cognitive and nutritional domains), and 68.3% experienced decline in only one domain (39.7% in the physical domain, 25.6% in the cognitive domain, 3.0% in the nutritional domain).

In total, 623 older adults marked as pre-frail accepted the invitation for step two of the PERSSILAA screening. Table 4 presents the outcomes of the second set of assessments for the physical domain. Based on these outcomes, 69.7% of the pre-frail older adults were considered to have functional decline in the physical domain. The average score on the Qmci screen was 65.1 (SD ± 11.3). Based on their Qmci screen scores (Table 4), 31.0% of the participants were identified as having probable cognitive decline. Those participants experience especially problems with the task focusing on verbal fluency and logical memory. Finally, two instruments provided extra information on the nutritional domain. The average score on the extended version of the MNA was 26.7 (SD ± 2.1) and none of the participants were classified as having malnutrition. However, only 32% of participants had a normal or healthy weight and 67% were overweight or obese. Table 5 provides an overview of the BMI and waist circumference scores. The average waist circumference was 101.3 (SD ± 14.8). Based on the outcome of step two of the PERSSILAA screening 82.0% of pre-frail participants showed functional decline on at least one of the assessed domains.

To investigate the merit of the PERSSILAA screening procedure over the use of a frailty scale in isolation, the classification (frail, pre-frail, robust) of the older adults based on the GFI alone and based on the PERSSILAA screening procedure were compared. Based on the GFI alone, only 8.2% (n = 207) of the older adults were classified as pre-frail and 75% (n = 2.823) as robust. This compares to the proportion identified after step one of the PERSSILAA screening (frail = 16.8%; pre-frail = 20.6%; robust = 62.3%) which differs significantly (Χ² = 5003.6; p < 0.05).