Assessing life-space mobility

A detailed description of these questionnaires is provided in Table S2. The majority of them consider not only spatial (e.g., distance) but also contextual aspects (e.g., mobility aids, personal assistance) of mobility. Nine questionnaires have initially been developed for community-dwelling older persons (LSD, LSQ, UAB-LSA, LSA‑F, LSA-CI and MBA) [12, 14‐16, 18‐21, 24], with three of them specifically focusing on persons having difficulties in going outdoors (HBMA, LSH and Hb-LSA) [18‐20], one for power mobility device users (LSA companion) [16], and three for institutionalized older persons (NHLSD, LSA-IS self-report and proxy versions) [13, 22, 23]. The majority of questionnaires are based on self-reports by structured interviews, with one being a web-based survey (MBA) [24]. The NHLSD is the only one specifically designed for proxy documentation of nurses [13]. Four self-report questionnaires are available as proxy reports (LSA companion, LSA-IS) [17, 23] or have used proxy reports (HBMA, LSH) [18, 19]. All self-report questionnaires assume full mental abilities of the respondent, except for two (LSA-CI, LSA-IS) [21, 22]. Most questionnaires are conducted via face-to-face interview, two can also be performed via phone (HBMA, LSA-F) [16, 18]. Specific equipment is only needed for the web-based MBA, while all other questionnaires were designed as paper and pencil tests. The observation period ranges between 24 h [22, 23] and 4 weeks or 1 month [12, 15‐17, 20]. Rating for the different questionnaires varied with total scores ranging from 0–8 points [18] to 0–120 points [15‐17, 22, 23], with 2 questionnaires having an open-ended score (LSH, MBA) [19, 24]. Higher scores consistently indicate higher mobility levels. Some questionnaires provide options to analyze only spatial mobility aspects [15, 16, 21‐23], rate specific sub-groups depending on frequency of being outdoors [12] or differentiate between low or high [25], or low, intermediate, and high LSM [18].

Figure 2 summarizes the results on the psychometric properties and recommendations for each questionnaire based on the 37 included validation studies. Most studies (n = 19) were identified for evaluating the psychometric properties of UAB-LSA [15] in different languages or populations [26‐45], additionally the phone-based [16, 46, 47] and proxy-reported questionnaires [17] are also based on the UAB-LSA. Detailed information on the individual studies, questionnaires, and COSMIN quality ratings is provided in Table S3.

Sample sizes of the validation studies ranged from 5 [16] to 2147 [32] participants, with 16 studies having less than 50 participants. Mean age of participants was 75 years, ranging from 43 years to 85 years. Studies included community-dwelling older persons without [12, 14, 15, 17‐20, 24, 26‐28, 31‐36, 38, 39, 47, 48] and with cognitive impairment [21], persons with power mobility devices [16], persons with chronic obstructive pulmonary disease [43‐45], persons with stroke [40, 42], persons with vestibular disorders [37], patients with Parkinson disease [49], persons with spinal cord injury [46], persons with critical illness [29], persons in palliative care [30], geriatric inpatients [22, 23, 50], and nursing home residents [13]. Construct and concurrent validity, and test-retest reliability were the most frequently evaluated psychometric properties.

Floor or ceiling effects were analyzed in 16 studies, with all reporting absence of such effects [12, 13, 15, 16, 20‐23, 31, 36, 42‐44, 46‐48, 50]. Completion rates ranged from 82.6% to 100% but were only rarely reported [12, 16, 21‐23, 47, 48]. Completion times were provided even less frequently, averaging between 3 min and 9 min across 5 studies [16, 21, 22, 27, 48], with phone-based questionnaires taking longer.

Almost all studies included a validity analysis. Construct validity was assessed in 30 studies [12‐15, 18‐23, 26, 27, 29, 30, 32‐38, 40, 42‐48, 50], using 1–20 construct variables of different domains (e.g., physical, cognitive, psychosocial, financial, environmental, and/or sociodemographic status). Four studies used a conceptual framework for mobility [4] to select construct variables [21‐23, 48]. Overall, results on construct validity yielded expected directions and magnitudes of correlations with cut-offs of r > 0.5 for similar (e.g., physical variables) and r > 0.30–0.50 for related constructs (e.g., financial, psychosocial variables) for evaluation. Concurrent validity was assessed in seven studies [17, 19, 20, 23, 24, 49, 50]. Apart from two studies [19, 24], concurrent validity results ranged from acceptable to good, depending on the closeness of the comparison instrument to the concept of the LSM questionnaires. Content validity was assessed with positive results in four studies (all for the UAB-LSA) [16, 27, 28, 36], with two focusing on the translation process [16, 28].

Reliability results were evaluated in terms of test-retest reliability in 22 studies, showing predominantly acceptable to good results with intraclass correlation coefficients or Kappas of > 0.7 [11]. Poor test-retest reliability was reported in only one study for the MBA [24]. Acceptable interrater reliability was observed in three studies [13, 19, 27].

Only eight studies analyzed sensitivity to change of the LSM questionnaires [15, 21‐23, 31, 35, 47, 48]. Most of them evaluated the ability to detect intervention-induced changes [21‐23, 35, 47, 48]. Changes over time were analyzed in only two studies [15, 31]. For the analysis, most studies used distribution-based methods [21‐23, 35, 47, 48], with one of them also using an anchor-based method [31]. Studies calculating standardized response means reported small (< 0.5) [23, 35], moderate (> 0.5) [21, 47], and large (> 0.8) [22, 48] sensitivity to intervention-induced changes.

The methodological quality of the included validation studies assessing the different psychometric properties was predominantly moderate to high across studies (Table S3). Overall, methodological quality of validity studies was predominantly negatively affected by low sample sizes that were not statistically justified [12, 17, 19, 40, 42, 50] or low number of construct variables [12, 18, 36, 43, 44, 46, 47]. Reliability analyses were limited due to very long time period between test and retest (1 year) [14], different test conditions used for test and retest (face-to-face vs. phone) [15], analysis of only relative agreement [13], and small sample sizes [13, 19, 26, 31, 34, 37, 40, 42]. For sensitivity to change, no adequate statistical analysis method was observed in one study [15], and low sample size in another [48]. Five studies could not be rated due to insufficient information as only the abstract was available [20, 29, 32, 35, 45]. A comprehensive validation approach to assess several psychometric properties with consistently high methodological quality was conducted in two studies [21, 22], and in four studies with moderate to high quality [13, 16, 23, 48]. For the UAB-LSA [15] as the most frequently evaluated LSM questionnaire, methodological quality of the studies was heterogeneous and ranged from low to high, with most of them showing, however, higher methodological quality.

The LSA-CI and both versions of the LSA-IS were comprehensively validated in the German language, with overall good results on feasibility, validity, reliability, and sensitivity to change [21‐23, 48]. A German version is also available for the UAB-LSA, whose construct validity has recently been documented in German community-dwelling older adults [33].