ARTICLE IN PRESS

International Journal of Nursing Studies 43 (2006) 357365 www.elsevier.com/locate/ijnurstu

Geriatric fear of falling measure: Development and psychometric testing

Tzu-Ting Huang
Chang Gung University, School of Nursing, 259 Wen-Hwa 1st Road, Kwei-Shan, Tao-Yuan, Taiwan, ROC Received 16 November 2004; received in revised form 25 March 2005; accepted 14 April 2005

Abstract Development of the geriatric fear of falling measure (GFFM) was based on a previous qualitative study of Taiwanese community-dwelling elders. Relevant parameters or items for assessing subscales were identied and tested in a threestage process: item development, content validity testing and reliability testing. The GFFM was tested on two groups of community-dwelling elders in Taiwan (N 100 and 384) to examine validity and reliability. Good testretest, internal consistency and inter-rater reliability were found. Conrmatory factor analysis and good overall model ts supported construct validity of the GFFM. Although these data are preliminary, the GFFM could be used as a quick screening instrument to evaluate fear of falling and an outcome indicator of nursing interventions. r 2005 Elsevier Ltd. All rights reserved.
Keywords: Fear of falling; Taiwanese elders; Instrument development; Psychometric evaluation

What this paper adds to the literature What is already known about this topic?

 The scale goes beyond a focus on activity to include

subscales on: psychosomatic symptoms (PS), adopting a risk prevention attitude, and modifying behavior.

 Fear 

of falling amongst older people has been investigated using a variety of measurement scales. Most of these scales have focused primarily on activity of older people, and few are grounded in older peoples experiences and perceptions.

1. Introduction Fear of falling is a common experience among the elderly population. Previous studies have estimated that between 25% and 55% of community-dwelling elders are afraid of falling (Arfken et al., 1994; Howland et al., 1993, 1998; Suzuki et al, 2002; Tinetti et al., 1994). Fear of falling can lead to deconditioning, thereby increasing the risk for falling (Friedman et al., 2002), compromising social interaction, and increasing the risk of isolation (Clague et al., 2000), depression, anxiety, and also impacts on the quality of life (Suzuki et al., 2002) of elderly people.

What do we now know as a result of this study?

 The

geriatric fear of falling measure (GFFM) is grounded in the perceptions of older people living in Taiwan.

Fax: +011 886 3 211 8800x5326.

E-mail address: thuang@mail.cgu.edu.tw.

0020-7489/$ - see front matter r 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijnurstu.2005.04.006

ARTICLE IN PRESS
358 T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365

Several approaches have emerged for measuring selfreported fear of falling. The most common approach has been to ask subjects directly. The subjects answers can be dichotomous (yes or no) or ranked by level (very much, somewhat, and not at all) (Arfken et al., 1994; Franzoni et al., 1994; Maki, 1997; Lawrence et al., 1998). Although these methods yield are informative, they cannot detect possible variation in levels of fear of falling and responses associated with other fears. Another approach to assess fear of falling utilizes a self-efcacy framework (Bandura, 1977). Tinetti et al. (1990) developed the falls efcacy scale (FES), based on the operational denition of this fear as low perceived self-efcacy at avoiding falls during essential, nonhazardous activities of daily living. The FES, a 10-item scale on a 10-point scale, identies how condent subjects feel about performing each activity of daily living without falling. Tennstedt et al. (1998) modied the FES by adding two items: carrying bundles from the store and exercising. They also changed the 10-point rating scale to a four-point scale (1 not at all sure, 4 very sure). Powell and Myers (1995) developed the activitiesspecic balance condence (ABC) Scale that included a wider continuum of activity difculty and more detailed item descriptors. Lachman et al. (1998) developed a new instrument (the Survey of Activities and Fear of falling in the Elderly, SAFE) to assess the role of fear of falling in activity restriction. The SAFE assesses fear of falling during performance of 11 activities, and gathers information about participation in these activities as well as the extent to which fear is a source of activity restriction. Velozo and Peterson (2001) developed a fear of falling measures (FFM) for community-dwelling elderly and consists of 19 common activities that intend to indicate how worried they would be if they were to perform the activities. It is a four-point rating scale, from very worried (4) to not at all worried (1). Although the above scales, FES, ABC, SAFE and FFM, demonstrated good reliability and validity and may allow for a more differentiated assessment than a direct one-item question about fear of falling, they have some shortcomings. First, problems arise with the FES when the subject does not engage in a given activity included in the scale (Lachman et al., 1998). Second, the response format can be too complex for some subjects. The FES (Tinetti et al., 1990) uses a 10-point rating scale from 0 (no condence) to 10 (complete condence). The ABC (Powell and Myers, 1995) uses a more precise assessment, with a scale from 0% (no condence) to 100% (complete condence). In our experience and that of Lachman et al. (1998), older adults, especially those with limited education, had difculty responding to these complex ratings. The third problem is that the content of these four instruments is concerned only with

restriction of activity; fear of falling may involve other issues such as psychological or social concerns. Lastly, no measure of fear of falling has been developed from the perspective of elders. Such a scale would measure the true value of fear of falling. Elderly people (65 years or older) are the fastest growing segment of the population in Taiwan. Because of changes in Taiwanese families leading to a decrease in adults co-residing with elderly parents and an increase in two-income families (Huang and Acton, 2004), elders are often left home alone. However, increased frailty related to aging raises the likelihood of accidents, including falls. Research has indicated that falls and fear of falling can be a major threat to independence for the elderly (Tinetti and Powell, 1993), impacting their quality of life. Finding ways to decrease the risk of falls would benet elders. Screening for fear of falling is the rst step to preventing falls among elders. Even though fear of falling is a signicant health problem among the elderly, few research studies to date have investigated this issue in Taiwan. Therefore, it is important to develop a culturally relevant instrument for measuring fear of falling in Taiwan. Since salient variables about fear of falling in Taiwanese elders have not been identied, the authors used grounded theory to gain a fresh perspective (Huang, 2005). This approach generated a pool of factors important to elders, which were then used to develop items for measuring fear of falling among community-dwelling elders.

2. Initial instrument development 2.1. Stage one To develop the geriatric fear of falling measure (GFFM), relevant items for the assessment subscales were identied and tested in a three-stage process. The rst stage, item development, established a pool of 46 items. This initial pool of items was drawn from items and descriptive examples in a previous qualitative study by the author (2005), which developed a model for understanding fear of falling from the perspective of Taiwanese elders. This model suggests that the strategy of managing fear of falling has four broad themes: psychosomatic symptoms (PS), adopting an attitude of risk prevention (RP), paying attention to environmental safety (ES), and modifying behavior (MB). PS in elderly people are associated with physical symptoms and emotional reactions stimulated by their fears. Adopting an attitude of RP reects elderly participants thoughts or behaviors about preventing falls and includes the categories of increased vigilance and readiness for emergencies. Paying attention to ES includes both environmental modication and use of

ARTICLE IN PRESS
T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 359

safety devices to eliminate possible risk factors for falls. MB refers to elderly peoples conscious adjustment of actions and limiting social activities to eliminate potential risk factors for falls. 2.2. Stage two The second stage tested the content validity of items using the index of content validity (CVI) (Waltz et al., 1997). The items were reviewed by a panel of experts, including two psychologists, one psychiatrist, two physical therapists, and two nurses. All panel members held either a Ph.D. or an M.D.. The CVI was dened, on a scale of 14, as the proportion of items in each subscale given a rating of 3 or above by 80% of the experts (Waltz et al., 1997). Five items did not meet these criteria, and were thus modied according to the panel members suggestions. The nal list of items had a CVI of 86%. 2.3. Stage three Prior to testing the GFFM in a larger sample, its clarity was tested by administering it to ve elders. No item was deleted or added, but three items were reworded. At this stage, the GFFM comprised 41 items to be used in a study with a larger sample.

providers, who would rate PS, paying attention to ES, adopting an attitude of RP, and MB. The FES (Tinetti et al., 1990) has been the tool most commonly used in the literature for measuring fear of falling. The 10-item FES measures fear of falling associated with 10 activities, on a scale of 110, to identify how condent elders feel about performing each activity without falling. A lower score reects lower efcacy or condence, and a higher score corresponds to higher condence. The testretest reliability of the FES was 0.71 (Tinetti et al., 1990). The testretest reliability of a Chinese version of the FES during a 2-month period was 0.90 and its internal consistency was 0.83 (Huang and Acton, 2004). 3.4. Data analysis Construct validity of the GFFM was evaluated via conrmatory factor analysis (CFA), using SPSS AMOS 3.6 (Arbuckle, 1995). As all items of the GFFM were ordinal, asymptotically distribution free test statistics were used (Yuan, Bentler, 1998). The root mean square error of approximation (RMSEA) expresses the lack of t due to reliability and also model (mis)specication (Browne and Cudeck, 1993). RMSEA expresses t with the model per degree of freedom and should be o0.08 for an acceptable t, with 0.05 or lower indicating a very good t to the model. The goodness of t index (GFI) and adjusted goodness of t index (AGFI), which adjusts for the number of parameters estimated, range from 0 to 1, with values of 0.9 or greater indicating a good tting model (Jo reskong and So rbom, 1996). These last two indices are analogous to R2 in multiple regression. The comparative t index (CFI) assesses t relative to a null model using non-centrality parameters (Bentler, 1988). The CFI also ranges from 0 to 1, with values of 0.9 or greater indicative of good t to the model. To assess the homogeneity of items within each subscale, one-factor models for each subscale were evaluated. Items were only loaded on presumed latent factors, thus all error terms were independent. In addition, a two-factor model was tested in both study groups, and a three-factor model was tested in Study 2. Concurrent validity of the GFFM was analyzed by comparing results using the GFFM with those from the FES (Tinetti et al., 1990) using Pearsons r correlations between total scores. Stability of the GFFM over a 2week period was determined by testretest using Pearsons correlation coefcients. Spearman rank correlations were used to examine the GFFMs inter-rater reliability between two trained nurses. 3.5. Human subjects protection

3. Methods 3.1. Design Two descriptive, cross-sectional studies were conducted in 2002. After the Medical Research Ethics board of Chang Gung University approved the study protocol. Eligible elders were approached by one of two trained data collectors, who briey introduced the study purposes and procedures. After informed consent was obtained, a face-to-face interview was arranged at the subjects convenience. 3.2. Participants Study 1, a pilot study, comprised a systematically randomized sample of 100 eligible community-dwelling elders (X65 years) in Taipei City. Study 2, which further examined factorial validity, comprised a randomized sample of 384 eligible community-dwelling elders (X65 years) in Taipei County and Tao-Yuan County. 3.3. Measurement The GFFM included 41 items in four subscales, with item scores ranging from 1 (never) to 5 (always). The GFFM was designed for completion by health care

Elders willing to participate in the study were informed of the studys goals and methods, their right

ARTICLE IN PRESS
360 T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 Table 1 Demographic characteristics of elderly participants Variable Study 1 N 100) n (%) 75.3576.71 57 (57.0) 43 (43.0) 45 (45.0) 55 (55.0) 64 (64.0) 36 (36.0) 66 (66.0) 22 (22.0) 12 (12.0) 5 16 75 3 18 18 61 3 (5.0) (16.0) (75.0) (3.0) (18.0) (18.0) (61.0) (3.0) Study 2 (N 384) n (%) 74.3876.75 190 (49.5) 194 (50.5) 189 (49.2) 195 (50.8) 241 (62.8) 143 (37.2) 252 (65.6) 103 (26.8) 29 (7.6) 30 75 275 4 48 75 249 12 (7.8) (19.5) (71.7) (1.0) (12.5) (19.5) (64.8) (3.1)

to withdraw from the study at any time and to refuse to answer questions, and the strategies used to protect condentiality. The participants were then asked if they agreed to participate in the study and their verbal responses were recorded. To ensure condentiality, all personally identifying information was eliminated from participant response records. Condentiality was kept by all research personnel with access to the database.

Age (years) 6574 ^75 Gender Male Female Marital status Married Single Education Illiterate %6 years 46 years Religion None Buddhism Taoism Christianity Living status Alone With spouse only With family With others Chronic condition ^1 0 History of falls (o65 years) ^1 0 History of falls during previous year ^1 0

4. Results 4.1. Study 1: preliminary reliability testing The pilot study (Study 1) focused on item reduction, testing of internal consistency, testretest reliability, inter-rater reliability, concurrent validity, and construct validity. This study applied CFA to data collected from a systematically randomized sample of 100 eligible community-dwelling elders (465 years) in Taipei City. The 100 participants in Study 1 had a mean age of 75.3576.71 years, 55% were female, 64% were married, 66% were illiterate, 75% were Taoists, 18% lived alone, 55% had at least one chronic condition, and 16% had at least one fall during the previous year (Table 1). An item reduction procedure was employed to screen items, thus producing a more compact form of the GFFM. Only items with an item-scale correlation coefcient 40.3 and inter-item correlation coefcient o0.6 were retained (Strickland, 2000). The internal consistency of each subscale was examined using Cronbachs a; a reliability of 0.70 or above was considered acceptable (Nunnally, 1978). Item-total correlations and Cronbachs a were recalculated after each item was removed. This process was repeated until a satisfactory compact instrument was obtained. After eliminating items with an item-scale correlation coefcient 40.30, 30 items were retained (all seven items on the ES scale were deleted in this step). After eliminating items with an inter-item correlation coefcient o0.60, 18 items remained. One-factor models were tested for all three subscales. One item in each scale (three items in total) was then deleted because the modication indices were greater than 10. In the nal set of items, which were retained after CFA, four items were retained on the PS scale, six items on the BM scale, and ve items on the RP scale. Cronbachs alphas for the PS, MB, and RP subscales were 0.82, 0.86, and 0.76, respectively; for the GFFM, Cronbachs a was 0.88. The factor loading (l) and R2 (which is equal to l2 in the current analysis) of each item are presented in Table 2. Factor loadings ranged from 0.47 to 0.85. Sample items are given in Table 3. The overall t indices for the three subscales (15 items) in the nal version of the GFFM are presented in Table 4. The PS,

55 (55.0) 45 (45.0) 31 (31.0) 69 (69.0)

203 (52.9) 181 (47.1) 107 (27.9) 277 (72.1)

16 (16.0) 84 (84.0)

48 (12.5) 336 (87.5)

RP and MB subscales demonstrated an excellent overall t (i.e., non-signicant w2 ; w2 =df o3; GFI, AGFI and CFI40.90; and RMSEAo0.05). The three subscales demonstrated a good overall t to a two-scale model (Table 5). Items were loaded only on their presumed latent factors; no double loading was allowed. All error terms were uncorrelated. The PS vs. MB subscales demonstrated an excellent overall t to the model (nonsignicant w2 ; w2 /dfo3; GFI, AGFI and CFI40.90; and RMSEAo0.05). The RP vs. MB scales demonstrated good model ts, despite AGFI being 0.88 (w2 was nonsignicant, w2 =df was less than 3, GFI and CFI exceeded 0.90, and RMSEA was less than 0.05). The PS vs. RP

ARTICLE IN PRESS
T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 Table 2 Item analysis, internal consistency and overall t indices of GFFM Scale/Item # Study 1 (N 100) Mean (SD) GFFM (Cronbachs a) PS (Cronbachs a) 34 37 39 41 MB (Cronbachs a) 15 16 17 20 21 24 RP (Cronbachs a) 8 9 10 11 12 (0.89) (0.82) 4.83 (0.60) 4.46 (1.23) 4.69 (0.79) 4.70 (0.75) (0.86) 3.76 (1.75) 2.39 (1.64) 2.92 (1.86) 3.16 (1.87) 2.44 (1.83) 2.43 (1.83) (0.76) 3.49 (1.77) 4.14 (1.60) 3.11 (1.89) 3.53 (1.79) 3.34 (1.77) 0.79 0.85 0.47 0.82 0.74 0.64 0.77 0.71 0.68 0.75 0.48 0.77 0.53 0.74 0.65 0.62 0.72 0.22 0.67 0.55 0.41 0.59 0.50 0.46 0.56 0.23 0.59 0.28 0.55 0.42 l R2 Study 2 (N 384) Mean (SD) (0.86) (0.79) 4.82 (0.68) 4.54 (1.16) 4.77 (0.70) 4.75 (0.71) (0.83) 2.31 (1.66) 2.21 (1.59) 2.52 (1.77) 3.03 (1.91) 2.36 (1.72) 2.40 (1.81) (0.75) 3.35 (1.83) 4.00 (1.64) 3.32 (1.88) 3.51 (1.72) 3.06 (1.81) 0.65 0.71 0.43 0.76 0.77 0.67 0.68 0.70 0.60 0.61 0.42 0.80 0.63 0.67 0.59 0.42 0.50 0.18 0.58 0.59 0.45 0.46 0.49 0.36 0.37 0.18 0.64 0.40 0.45 0.35 l R2 361

Table 3 Sample items RP 8. 9. 10. 11. 12. MB 15. 16. 17. 20. 21. 24. PS 34. 37. 39. 41.

To avoid climbing to reach up high, I will take advantage of new tools or techniques, such as using a long-handled mop to wipe tiles When walking on steep terrain or going outdoors, I will use an umbrella or cane for support to prevent myself from falling I will sit on a chair when taking a bath or hold some support I need assistance when going out (e.g., I used to take buses, but now I either take a taxi or ask others for a ride) Nowadays, I do less housework that requires more walking, such as sweeping and mopping When there is an obstacle on the ground or oor, I prefer to detour than go over it I go out less during rainy days I will ask others for help when I need something thats too high to reach I will take care to avoid passing close to places where objects are piled up Nowadays, I do less outdoor activities (e.g., trips, community activities, or visiting friends) I have changed my exercise style (e.g., from active to passive, from outdoor to indoor, or less frequent) I dont sleep well because I worry about falling My heart races when I think about falling after climbing to reach something high I frequently recall terrible experiences Ive had falling I have become more sensitive, agitated, irritable, and critical of others

subscales demonstrated acceptable overall t (although RMSEA 0.06 and AGFI 0.89, w2 =df o3; non-signicant w2 ; GFI, and CFI40.90). The estimated correlations for the latent construct of each subscale

pair were positively correlated (F 0:41, 0.83 and 0.68, respectively) (Table 5). Stevens (1996) suggested for factor analysis, the ratios of N (sample size): p (number of variables to be analyzed) from 5 subjects per variable

ARTICLE IN PRESS
362 T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365

to 20 subjects per variable, and proposed that at least 300 samples will lead to good interpretation. Given the relatively small sample size (N 100), a model for the whole inventory was not examined. Concurrent validity of the GFFM was analyzed by comparing scores for risk of falling from the GFFM with those from the FES (Tinetti et al., 1990) using Pearsons r correlations between total and subscale scores. Correlations between the GFFM and FES were highly signicant (r 0:29, p 0:002). Stability of the GFFM over time was determined by the testretest method. The rst 10 subjects in Study 1 were asked to retake the GFFM 2 weeks later. Pearson correlation coefcients were 0.88 (po0:0001). Paired ttests were performed to determine any signicant difference in mean scores; scores on all three subscales were not signicantly different. To examine inter-rater reliability in administering the GFFM, two trained nurses visited the rst 10 elders in the Study 1 sample, who each completed the GFFM. Since the GFFM is a norm-referenced instrument, Spearman rank correlations were used to compare average scores obtained by the two raters. The Spearman rank correlations for these scores were 0.91 (po0:001), 0.94 (po0:001), and 0.89 (po0:001) for the RP, PS and MB subscales, respectively.

4.2. Study 2: further psychometric testing A second descriptive correlational study focused on testing internal consistency and construct validity using CFA. The sample of 384 elders for Study 2 was drawn at random from eligible community-dwelling elders (465 years) in Taipei County and Tao-Yuan County. CFA was used to test the extent to which the data from

administering the GFFM to this sample supported the relationships specied in the 15-item three-factor model. The sample in Study 2 had a mean age of 84.776.0 years; 189 (49.2%) were men and 195 (50.8%) women. The majority were married (62.8%), illiterate (65.6%), Taoist (71.4%), living with family (64.8%), had at least one chronic condition (52.9%), and had no fall during the previous year (87.5%) (Table 1). Internal consistency of the GFFM was analyzed. Cronbachs a for the GFFM as a whole was 0.86, with coefcients of 0.79, 0.83, and 0.75 for the PS, MB and RP subscales, respectively (Table 2). Overall t indices for the three single-scale models of the GFFM administered to community-dwelling elders are listed in Table 6. The PS scale again demonstrated an excellent overall t (i.e., non-signicant w2 ; w2 =df o3; GFI, AGFI and CFI40.90; RMSEAo0.08). The RP and MB subscales demonstrated acceptable model ts, although the RMSEA values were 0.09 and 0.08, respectively; w2 was signicant; w2 =df o4; and GFI, AGFI and CFI40.90. The factor loading (l) and R2 for each item are presented in Table 2; factor loadings ranged from 0.43 to 0.85. The three two-scale models demonstrated a good overall t (Table 7). The PS vs. RP and PS vs. MB models demonstrated good overall t (RMSEA approached 0.08; w2 =df o3; GFI, AGFI and CFI40.90). The RP vs. MB scale demonstrated acceptable model ts; although the RMSEA 0.09 and AGFI 0.89, w2 =df o4, and GFI and CFI40.90. The estimated correlations for the latent construct of each subscale pair were positively correlated (F 0:62, 0.39 and 0.79, respectively) (Table 7). The global GFFM (PS, MB and RP) demonstrated an acceptable overall t (Table 7). The model had a w2 =df 3:06, GFI 0.92, AGFI 0.89, CFI 0.90, and RMSEA 0.07. The latent constructs for each

Table 4 Overall t indices to one-factor model: Study 1 (N 100) Scale w2 PS RP MB df w2 =df GFI AGFI CFI RMSEA 0.55 0.51 0.85 0.99 0.97 0.99 0.97 0.98 0.94 1.0 0.00 1.0 0.00 0.98 0.00

Table 6 Overall t indices to one-factor model: Study 2 (N 384) Scale w2 PS RP MB df w2 =df GFI AGFI CFI RMSEA 0.99 0.96 0.98 0.94 0.98 0.94 0.99 0.07 0.97 0.09 0.97 0.08

1.1 (p 0:58) 2 2.55 (p 0:77) 5 7.65 (p 0:57) 9

5.5 (p 0:06) 2 2.76 19.59 (p 0:001) 5 3.92 28.42 (p 0:001) 9 3.16

Table 5 Overall t indices and estimated correlation coefcients (F) for two-factor model (N 100) Scale PS and RP PS and MB RP and MB w2 35.51 (p 0:10) 26.92 (p 0:80) 50.65 (p 0:20) df 26 34 43 w2 =df 1.37 0.79 1.18 GFI 0.94 0.95 0.92 AGFI 0.89 0.92 0.88 CFI 0.93 0.95 0.91 RMSEA 0.06 0.00 0.04 F 0.68 0.41 0.83

ARTICLE IN PRESS
T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 Table 7 Overall t indices and estimated correlation coefcients (F) of two- and three-factor models (N 384) Scale PS and RP PS and MB RP and MB Global GFMM w2 76.59 (po0:001) 75.05 (po0:001) 167.13 (po0:001) 266.14 (po0:001) df 26 34 43 87 w2 =df 2.95 2.21 3.89 3.06 GFI 0.96 0.96 0.93 0.92 AGFI 0.93 0.94 0.89 0.89 CFI 0.94 0.96 0.91 0.90 RMSEA 0.07 0.06 0.09 0.07 F 0.62 0.39 0.79 MBRP 0.72 MBPS 0.21 RPPS 0.20 363

subscale pair were positively correlated (F 0:20, 0.21 and 0.72, respectively) (Table 7).

5. Discussion The GFFM was developed to meet the need for an instrument that health care providers can use to assess elders fear of falling through PS, adopting an attitude of RP and MB. The nal version of the GFFM contains 15 items: four items in the PS subscale, ve items in the RP subscale, and six items in the MB subscale. Health care providers can complete the whole scale in 5 min. Development of the GFFM was guided by a model derived from qualitative ndings (Guyatt, 1999), rather than relying on mathematical criteria (Bollen, 1989). Hendrick and Hendrick (1986) suggested that the rst step in developing a valid measure is to examine its theoretical construct. Construct validity of the GFFM was examined by CFA instead of exploratory factor analysis (EFA). EFA is often used to test the construct validity of instruments, but that approach is purely exploratory (Pendhauzur and Schmelkin, 1991) since it does not indicate to which factor any item belongs; all items have a factor loading associated with each factor. In CFA, both the number of latent constructs and their relationship to indicators are specied before the analysis. CFA is designed to conrm the pre-specied measurement model. The estimated correlation, F, for the latent construct between Adopting an Attitude of RP and MB were highly positively correlated (0.83 in Study 1 and 0.79 in Study 2). One possible explanation is that RP reects attitudes and MB refers to actions of elderly people to eliminate potential risks for falling. Therefore, the two subscales are highly correlated. Because there is no culturally sensitive measure of fear of falling for Taiwanese elders living in the community, this study aimed to develop such an instrument grounded in qualitative study ndings and to test it in a moderate size sample. Another study aim was to examine the generalizability of the instruments factor structure in a larger sample of community-dwelling elders. The GFFM was originally developed for com-

munity-dwelling elders in an urban area. To examine generalizability, Study 2 tested the GFFM in community-dwelling elders living in suburban and rural areas. The ndings indicate that the GFFM is appropriate for measuring fear of falling among elders living in urban, suburban and rural areas of Taiwan. Several existing instruments for measuring fear of falling among elders are limited to issues related to activity restriction. Moreover, none of them explore and describe fear of falling from the perspective of elders in naturalistic settings. In our newly developed instrument, the GFFM has three subscales: PS, RP and MB, which go beyond activity restriction. The GFFM contains not only negative aspects, such as PS and activity restraints, but also has positive aspects, such as RP, MB, and developing resources by seeking help.

6. Conclusion The GFFM was explicitly developed from a conceptual model based on qualitative study ndings. The GFFM, which can be administered in 5 min, is the rst documented measure of fear of falling administered directly to Taiwanese elders living in urban and rural areas. Although these data are preliminary, they suggest that the GFFM could be used as a quick screening instrument to evaluate fear of falling in communitydwelling elders and as an outcome indicator of nursing interventions. 6.1. Implications for nursing practice Information on the perceived meaning of fear of falling for community-dwelling elders may help nurses understand elders responses to their fear. This instrument may be used to identify elders at high risk for emotional distress. Nursing interventions focusing on cognitive appraisal of the fear of falling experience should be developed, particularly interventions that can reinforce the positive aspects of meaning (e.g., increased vigilance and adjusting behavior). Consequently, the GFFM can function as both a screening tool and an

ARTICLE IN PRESS
364 T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 Guyatt, G.H., 1999. Measuring health-related quality of life in childhood cancer: lessons from the workshop. International Journal of Cancer 12, 143146. Hendrick, G., Hendrick, S., 1986. A theory and method of love. Journal of Perspective Social Psychology 50 (3), 392402. Howland, J., Peterson, E.W., Levin, W.C., Fried, L., Porrdon, D., Bak, S., 1993. Fear of falling among community-dwelling elderly. Journal of Aging and Health 5 (2), 229243. Howland, J., Lachman, M.E., Peterson, E.W., Cote, J., Kasten, L., Jette, A., 1998. Covariates of fear of falling and associated activity curtailment. The Gerontologist 38 (5), 549555. Huang, T., 2005. Managing fear of falling: Taiwanese elders perspective. International Journal of Nursing Studies, in press. Huang, T., Acton, G., 2004. Effectiveness of home visit falls prevention strategy for Taiwanese community-dwelling elders: randomized trial. Public Health Nursing 21 (3), 248257. Jo reskong, K.G., So rbom, D., 1996. Lisrel 8 Users Reference Guide. Science Software International, Chicago. Lachman, M.E., Howland, J., Tennstedt, S., Jette, A., Assmann, S., Peterson, E.W., 1998. Fear of falling and activity restriction: the survey of activities and fear of falling in the elderly (SAFE). Journal of Gerontology: Psychological Sciences 52 (1), 4350. Lawrence, R.H., Tennstedt, S.L., Kasten, L.E., Shih, J., Howland, J., Jette, A.M., 1998. Intensity and correlates of fear of falling and hurting oneself in the next year. Journal of Aging and Health 10 (3), 267286. Maki, B.E., 1997. Gait changes in older adults: predictors of falls or indicators of fear. Journal of the American Geriatrics Society 45 (3), 313320. Nunnally, J.C., 1978. Psychometric Theory. McGraw-Hill, New York. Pendhauzur, E.J., Schmelkin, L.P., 1991. In: Measurement Design and Analysis: Integrated Approach. Lawrence Erlbaum Associate, Hillsdale, NJ, pp. 5280. Powell, L.E., Myers, A.M., 1995. The activities-specic balance condence (ABC) scale. Journals of Gerontology: Medical Sciences 50 (1), M2834. Stevens, J., 1996. Applied Multivariate Statistics for the Social Sciences. Lawrence Erlbaum Association, Hillsdale, NJ, pp.362427. Strickland, O.L., 2000. Deleting items during instrument developmentsome caveats. Journal of Nursing Measurement 8 (2), 103104. Suzuki, M., Ohyama, N., Yamada, K., Kanamori, M., 2002. The relationship between fear of falling, activities of daily living and quality of life among elderly individuals. Nursing & Health Sciences 4 (4), 155161. Tennstedt, S., Howland, J., Lachman, M.E., Kasten, L., Jette, A., 1998. A randomized, controlled trail of a group intervention to reduce fear of falling and associated activity restriction in older adults. Journal of Gerontology: Psychological Sciences 53 (1), 384392. Tinetti, M.E., Powell, L., 1993. Fear of falling and low selfefcacy: a cause of dependence in elderly persons. Journals of Gerontology 48 (Special Issue), 3538. Tinetti, M.E., Richman, D., Powell, L., 1990. Falls efcacy as a measure of fear of falling. Journals of Gerontology 45 (6), P239P243.

outcome measure to evaluate the effect of specic nursing interventions.

6.2. Implications for nursing research The two-stage validation process used in our study tested content and factorial validity of the GFFM and provided support for its 15 items and three subscales. However, as validation is an ongoing process (Anastasi and Urbina, 1997), and all of the subjects in these two studies are community-dwelling elders living in northern Taiwan, research needs to continue investigating the reliability and validity of the GFFM.

Acknowledgements I would like to thank Dr. Yeh, CH and Dr. Chen, ML for their valuable suggestions on this study. My thanks also go to all the study participants for sharing their fear of falling experience. This research was supported by a grant to Dr. Huang from National Science Council, Taiwan (Grant number: NSC90-2314-B-132-064).

References
Anastasi, A., Urbina, S., 1997. Psychological Testin, seventh ed. Prentice-Hall, New York. Arbuckle, J.L., 1995. Amos Version 3.6. SPSS Incorporation, Chicago, IL. Arfken, C.L., Lach, H.W., Birge, S.J., Miller, P., 1994. The prevalence and correlates of fear of falling in elderly persons living in the community. American Journal of Public Health 84 (4), 565569. Bandura, A., 1977. Self-efcacy: toward a unifying theory of behavioral change. Psychological Review 84 (2), 191215. Bentler, P.M., 1988. Comparative t indexes in structural models. Psychological Bulletin 107 (2), 238246. Bollen, K.A., 1989. Structural Equations with Latent Variables. Wiley, New York. Browne, M.W., Cudeck, R., 1993. Alternative ways of assessing model t. In: Bollen, K.A., Long, J.S. (Eds.), Testing Structural Equation Models. Sage, Newbury Park, pp. 136162. Clague, J.E., Petric, P.J., Horan, M.A., 2000. Hypocapnia and its relation to FOF. Archives of Physical Medicine & Rehabilitation 81 (11), 14851488. Franzoni, S., Rozzini, R., Boffelli, S., Frisoni, G.B., Trabucci, M., 1994. Fear of falling in nursing home patients. Journal of Greontology 40 (1), 3844. Friedman, S.M., Munoz, B., West, S.K., Rubin, G.S., Fried, L.P., 2002. Falls and fear of falling: which comes rst? A longitudinal prediction model suggests strategies for primary and secondary prevention. Journal of the American Geriatrics Society 50 (8), 13291335.

ARTICLE IN PRESS
T.-T. Huang / International Journal of Nursing Studies 43 (2006) 357365 Tinetti, M.E., de Leon, C.F.M., Doucette, T., Baker, D., 1994. Fear of falling and fall-related efcacy in relationship to functioning among community-living elders. Journal of Gerontology: Medical Sciences 49 (3), M140147. Velozo, C.A., Peterson, E.W., 2001. Developing meaningful fear of falling measures for community dwelling elderly. 365 American Journal of Physical Medical Rehabilitation 80 (9), 662673. Waltz, C.F., Strickland, O.L., Lenz, E.R., 1997. Measurement in Nursing Research, third ed. F.A. Davis, Philadelphia. Yuan, K.H., Bentler, P.M., 1998. Normal theory based test statistics in structural equation modelling. British Journal of Mathematics Statistical Psychology 51 (2), 289309.