Psychometric Evaluation of the Icelandic Version of the Modified Dynamic Gait Index for Community‐Dwelling Older Adults With Balance Impairments

ABSTRACT

Rationale

Optimal mobility is crucial for healthy aging, particularly among older adults with balance impairments. This research examines the psychometric properties of the modified Dynamic Gait Index (mDGI) translated into Icelandic, highlighting its suitability for evaluating mobility in this demographic group and within the context of healthy aging. Addressing the scarcity of international psychometric research on the mDGI, this study contributes to the translation of geriatric outcome measures into different languages, enhancing clinical applications and international research.

Aim

To assess the reliability and validity of the mDGI among Icelandic older adults experiencing balance impairments.

Methods

This methodological study included 30 participants, aged 67–91 years, receiving outpatient physical therapy for balance impairments. The participants completed two mDGI assessments 4–7 days apart, and additional assessments using the 10‐meter walking test (10MWT), Timed Up and Go (TUG), Activities‐specific Balance Confidence (ABC) scale, and Short Form Health Survey (SF‐36) subscales. Analysis included evaluating the mDGI's total scale and subscales' reliability and validity using Intraclass Correlation Coefficient (ICC _3,1), Standard Error of Measurement (SEM), Cronbach's alpha, and Spearman's rho.

Results

The mDGI demonstrated high relative reliability (ICC _3,1 = 0.95 for total mDGI; 0.73–0.92 for all subscales) and strong absolute reliability (SEM for total mDGI = 1.32; two subscales = 1.17–1.43). Internal consistency was robust (alpha for total mDGI = 0.9; two subscales = 0.86–0.89). Construct validity was confirmed by mDGI's correlations with 10MWT, TUG, and SF‐36 social and physical functioning subscales. No floor or ceiling effects were observed in mDGI total scores.

Conclusion

The Icelandic version of the mDGI provides reliable and valid measures for evaluating balance and gait in older adults with balance impairments. Its sound psychometric properties support its use in similar demographic settings globally, providing a reliable tool for geriatric care practitioners and researchers worldwide.

1. Introduction

Optimal mobility is crucial for healthy aging, facilitating daily activities, access to community resources, and social engagement [1]. Walking ability, a fundamental aspect of mobility, reflects an individual's capacity to move safely and independently [2]. Good walking ability entails navigating diverse contexts and managing challenges in the physical and social environment while maintaining balance and avoiding falls. In advanced age, the widening gap between physical limitations and environmental challenges increases the risk of falls, impacting daily functioning and the quality of life [3].

Health professionals, including physical therapists, play a crucial role in enhancing walking ability for older adults, utilizing standardized outcome measures for guidance and progress monitoring [4]. One such measure, the Dynamic Gait Index (DGI), developed by Anne Shumway‐Cook in 1995 assesses individuals' adaptability in gait performance across diverse daily life tasks and environmental challenges [5]. The DGI has demonstrated strong psychometric properties in older adults (≥ 65 years of age) [5, 6, 7, 8, 9] and in different patient populations, including those with balance impairments related to vestibular disorders [10, 11, 12, 13], However, the DGI has inherent limitations, notably its limited responsiveness and reliance on a single score (ranging from zero to 24) encompassing three facets of walking: gait pattern, level of assistance, and gait speed.

To overcome these limitations, the Modified Dynamic Gait Index (mDGI) was introduced in 2013 by Shumway‐Cook et al. [14]. This revised version retains the eight walking tasks and three facets for each task from the original DGI and expands the scoring system to provide a more comprehensive assessment of walking ability, as detailed in Table 1. The total mDGI score ranges from zero to 64 points, with higher scores reflecting better functioning. The mDGI has been recognized as a valuable tool for assessing walking ability in older adults, providing important insights into their functional mobility [15]. However, to date, most studies evaluating its psychometric properties have relied on datasets from the USA [14] and Italy [16], where the focus on older community‐dwelling adults was limited.

Table 1.

Modified Dynamic Gait Index (mDGI), its four dimensions, eight tasks, three performance facets, and scoring range.

Eight MDGI tasks and four dimensions	Performance facets			Total task score range
	Time	Gait pattern	Level of assistance
Temporala
1. Gait on level surface (usual pace)	0–3	0–3	0–2	0–8
2. Change in gait speed (change pace)	0–3	0–3	0–2	0–8
Posturala
3. Gait with horizontal head turns	0–3	0–3	0–2	0–8
4. Gait with vertical head turns	0–3	0–3	0–2	0–8
5. Gait and pivot turn	0–3	0–3	0–2	0–8
Densitya
6. Steps over obstacles	0–3	0–3	0–2	0–8
7. Steps around obstacles	0–3	0–3	0–2	0–8
Terraina
8. Walking up stairs	0–3	0–3	0–2	0–8
Total score for each performance facet and total mDGI score	0–24	0–24	0–16	0–64

Note: This table is adapted from Anne Shumway‐Cook et al. [14] with permission of authors.

^a Dimensions of environmental demands for mobility [2].

The first data set from the USA, which included 995 participants aged 15–99 years [14], aimed to create and validate the mDGI. This data set was subsequently used to examine the validity of the mDGI across six diagnostic groups [17, 18] and to validate its theoretical foundation [19]. Test‐retest reliability, assessed within a subgroup comprising 257 participants, demonstrated a robust Pearson correlation coefficient of 0.92 for the mDGI total score [14]. Additionally, the values were 0.91 for time and gait pattern, 0.87 for level of assistance, and between 0.86 and 0.90 for the eight mDGI task scores. Interrater reliability, determined through the assessment of videos featuring nine participants by a panel of eight physical therapists, showed kappa coefficients ranging from 0.90 to 0.98 for time, 0.59 to 0.88 for gait pattern, and 0.84 to 1.0 for level of assistance. Internal consistency was robust, with a Cronbach's alpha of 0.97 for the mDGI total score, 0.97 for time and level of assistance facet, and 0.92 for gait pattern facet. Alpha values for the eight mDGI tasks ranged from 0.75 to 0.85 among 995 participants, and in a subsequent study with a subset of the same data set (794 participants), alpha values ranged from 0.66 to 0.97 [18]. No floor or ceiling effects were observed for the mDGI total score, in a subgroup of 91 participants with gait abnormalities. The minimum detectable change at the 95% confidence level (MDC ₉₅) was 7.0 for the mDGI total score, 3.1 for the time facet, 4.0 for the gait pattern facet, and 2.4 for the level of assistance facet among 654 participants with impairment. The second data set from Italy involved 58 participants with neurological disorders aged 20 to 85 years [16, 20] and was instrumental in establishing cut‐off scores for identifying fall risk [20] and determining the minimal clinically important difference (MCID) for the mDGI [16]. Findings from these datasets support the mDGI's psychometric properties and its suitability as an assessment tool for measuring dynamic gait abilities across different populations [14, 16, 17, 18, 19, 20].

Although the mDGI has become a valuable tool in clinical practice and research, its psychometric properties in specific contexts require further validation. These properties, including reliability and validity, warrant investigation when an outcome measure is translated into different linguistic and cultural settings [21, 22]. Thus, this study aimed to evaluate the Icelandic version of the mDGI among older community‐dwelling adults with balance impairments. The assessed psychometric properties include test‐retest reliability (relative reliability), standard error of measurement and minimal detectable change (SEM and MDC ₉₅; absolute reliability), internal consistency, construct validity, and potential floor and ceiling effects.

Grounded in theoretical constructs, we established hypotheses regarding the psychometric properties of the Icelandic mDGI. For reliability, we aimed for mDGI Intraclass Correlation coefficient (ICC) values of at least 0.75 [23], no significant differences between repeated measurements, and internal consistency values of 0.70 to 0.90 [23]. To evaluate construct validity, we identified some key outcome measures associated with dynamic balance and gait, alongside those with potentially weaker or no detectable associations [24]. Specifically, we hypothesized that the mDGI would demonstrate moderate to good negative correlation with basic mobility [25], moderate to good positive correlation with gait‐speed [26] and balance confidence [27], low to fair positive correlation with physical dimensions of health‐related quality of life [28], while displaying little or no correlation with mental health dimensions of health‐related quality of life [28]. Lastly, we hypothesized that total mDGI scores among community‐dwelling older adults would not manifest floor or ceiling effects.

2. Methods

2.1. Research Design and Ethics

This methodological study followed the international Consensus‐based Standards for the Selection of Health Measurement Instruments (COSMIN) [21], with approval from the Icelandic National Bioethics Committee (VSNb2018050025/0301) and adherence to the Helsinki Declaration. All participants provided written informed consent after receiving verbal and written explanation of the study protocol.

2.2. Participants, Sample Size and Recruitment Process

The participants were all clients at an outpatient physical therapy clinic in the capital area of Iceland, referred by their general practitioners, geriatricians, and otolaryngologists. The referral reasons were based on balance impairments of various causes, which had resulted in falls, feelings of unsteadiness when standing or walking, and/or worries about falling. Since the study focused on older adults, we set the minimum age limit for inclusion at 67 years, corresponding to the official retirement age in Iceland [29]. Inclusion criteria: age at least 67 years, living at home, capable of walking six meters without physical assistance from another person, and able to follow verbal instructions. Exclusion criteria: acute illness, lower limb diseases or fractures, and hip or knee joint replacement within the past 3 months.

A minimum sample size of 30 participants was pre‐determined [30, 31], and confirmed with the COSMIN group's “ICC & SEM power” online tool [32]. The estimation, based on an ICC model of agreement, expected a 0.8 correlation between repeated measurements, variance of 10, no systematic differences between repeated measurements, 0.3 target width for the 95% Confidence interval (CI) for ICC, and 0.15 for SEM [33].

In 2018, 53 clients referred to physical therapy balance programs were recruited based on the severity of their impairments. The researchers introduced the study, ensured adherence to the inclusion criteria, and distributed information leaflets during the group sessions. Out of the 53 clients, 45 met the criteria, and eight were excluded due to age, inability to walk without physical assistance from another person, or a serious illness. Thirty‐seven clients initially volunteered, but seven withdrew due to overseas travel (n = 3), surgery (n = 1), illness (n = 2), or work commitment (n = 1).

2.3. The Primary Outcome Measurement ‐ Modified Dynamic Gait Index (mDGI)

In 2015, the mDGI was translated into Icelandic using the FACIT (Functional Assessment of Chronic Illness Therapy) translation methodology model [34]. The translators were instructed to achieve conceptual equivalence between the original and translated versions while considering the need for cultural adaptations [35]. However, no such adaptations were necessary, as the mDGI tasks and administration instructions were found to be culturally relevant and familiar to the Icelandic population. The translation is available in the Icelandic Journal of Physical Therapy [35]. The mDGI administration includes standardized instructions and an obstacle course setup [14]. Participants performed eight tasks, and the administrator assessed task duration, gait pattern, and required level of assistance (Table 1). Scores were recorded for each task (range 0–8), time (range 0–3), gait pattern (range 0–24), level of assistance (range 0–2), and the mDGI total score (range 0–64). Higher scores reflect better balance and gait functioning.

2.4. Other Measurements

2.4.1. The Ten‐Meter Walk Test (10MWT)

The 10MWT assesses comfortable walking speed, reflecting mobility, gait, and balance function [26], where participants cover a distance of 20 m. Time was recorded for the middle 10 m of walking, with 5 m allocated for acceleration and 5 m allocated for deceleration. A higher walking speed, measured in meters per second (m/s), indicates better physical ability.

2.4.2. The Timed Up & Go (TUG)

The TUG evaluates basic mobility, providing insights into balance and fall risk among older individuals [25]. Participants are timed during the sequence of standing up, walking three meters, turning, walking back, and sitting down. A shorter time (in seconds) indicates better mobility.

2.4.3. The Activities‐Specific Balance Confidence (ABC) Scale

The ABC scale evaluates balance confidence using a 16‐item self‐report questionnaire [27]. Participants rated their confidence in daily activities without losing balance or experiencing unsteadiness on a scale of zero (no confidence) to 100 (complete confidence). The total score was obtained by summing the individual item scores and dividing them by the total number of items. The total score ranges from zero to 100 with a higher score indicating greater balance confidence.

2.4.4. The Questionnaire on Health‐Related Quality of Life (Short Form Health Survey, SF‐36)

The SF‐36 evaluates health‐related quality of life through 36 questions [28], including physical and mental health components. Physical health component scales include physical functioning, role‐physical, body pain, and general health, whereas mental health component scales include vitality, social functioning, role‐emotional, and mental health. Scores on all SF‐36 scales range from zero to 100, with higher scores indicating better well‐being.

2.4.5. The Clock Drawing Test

This test screens for cognitive impairments related to spatial and executive functions [36], potentially impacting participants' comprehension of performance test instructions and questionnaire responses. The participants were tasked with drawing a clock and placing their hands to indicate a specific time. Scoring ranges from zero to six points, with higher scores reflecting better cognitive ability.

2.5. Procedures

Two final‐year MSc physical therapy students (RFG and NDO) underwent comprehensive training on the testing battery, which included administration of the mDGI and other standardized measurements. The training was provided by experienced physical therapists (HHS and SAA). The measurements were conducted at an inpatient physical therapy clinic in a controlled and specially prepared environment. The administration of outcome measures followed standardized protocols, and the questionnaires were conducted in an interview format. Assessments were scheduled based on the participants' convenience, aligning with their twice‐weekly group therapy sessions (Tuesdays and Thursdays) with weekend options available. To ensure consistency, participants were encouraged to schedule the second session at the same time of the day as the first.

During the initial visit (Session 1), participants provided informed consent, and RFG administered the mDGI, focusing on usual gait speed, and encouraged participants who used walking aids in their daily life to attempt the test without them. To minimize bias, RFG recorded scores without calculating the outcomes for each mDGI scale. NDO collected background information on age, sex, medical history, medications, falls in the past 12 months, use of walking aids, height, weight, and calculated Body Mass Index (BMI, kg/m²). Subsequently, standardized measures were administered in a fixed order: SF‐36 questionnaire, ABC scale, Clock Test, TUG test, and 10MWT. This session lasted 45–50 min.

The second mDGI assessment (Session 2) took place 4–7 days later (M = 5.8 days, SD = 1.41), at the same time of the day as in Session 1 and was administered by the same tester (RFG). The participants received a reminder phone call the previous day. All participants attended and completed this 15‐min session, which focused solely on mDGI administration.

2.6. Statistical Analyzes

Statistical analyzes were performed using IBM SPSS Statistics for Windows (version 29.0, Armonk, NY: IBM Corp; 2020), with the significance level set at p < 0.05. Continuous background variables and outcome measures are summarized using means (M), standard deviations (SD), medians (where appropriate) and ranges, whereas categorical background variables are expressed as frequencies and percentages. The identification of floor and ceiling effects in the mDGI involved recognizing whether ≥ 15% of the scores clustered at the scale extremes [37]. Test‐retest reliability for the mDGI was evaluated using ICC _3,1, SEM, MDC ₉₅, Bland‐Altman analysis, and paired t‐tests. Data distribution normality was assessed using the Shapiro‐Wilk test.

For relative reliability of all mDGI scales, ICC _3,1 (with 95% CI) was calculated using a two‐way mixed‐effects model for absolute agreement and a single rater, accounting for systematic differences between sessions [38]. Consistency ICC _3,1 was also calculated for comparative purposes, excluding any systematic differences between sessions [38]. Interpretations of ICC _3,1 followed established benchmarks: ≥ 0.90 excellent, 0.75–0.90 good, 0.5–0.74 moderate, and < 0.5 poor reliability [23]. Significant differences between Session 1 and 2 measurements were assessed using paired t‐tests.

Absolute reliability for mDGI‐total and facets was assessed through SEM and MDC ₉₅, with better reliability indicated by lower SEM and smaller MDC ₉₅ values (SEM = SD_d /√2 and MDC ₉₅ = SEM ∗ 1.96 ∗ √2) [23]. Bland and Altman analysis was employed to visually represent the results, illustrating the limits of agreement between two measurements, along with random and systematic errors [23].

Internal consistency, based on mDGI measurements in Session 1, was evaluated using Cronbach's alpha coefficient (α) with a 95% CI. An α ≥ 0.90 indicates strong internal consistency, while α = 0.70–0.90 indicates preferred moderate correlations [23]. Construct validity of the mDGI was evaluated using Spearman's rho correlation (r _s). The interpretation of correlation strength included ≥ 0.75 strong, 0.50–0.75 moderate to good, 0.25–0.50 low to fair, and ≤ 0.25 as little or no correlation [24].

Five participants scored ≤ 4 on the Clock Drawing test [36], indicative of Mini‐Mental State Examination scores of < 24, and were initially excluded from the analyzes. Upon reexamination, it was determined that their exclusion did not impact the study findings. Therefore, all participants were included in the final analysis.

3. Results

3.1. Participants

The age range of the 30 participants varied from to 67–91 years, with men (M = 80 years, SD = 5.9) notably older than women (M = 75 years, SD = 4.7) (p = 0.02). Additional characteristics are presented in Table 2. Among the 13 participants who experienced falls, the identified causes included obstacles (n = 3), slipperiness (n = 1), dizziness (n = 1), and weakness (n = 1), and seven reported other reasons. The majority of participants did not use walking aids. Among the six who did so, one reserved the aid for outdoor use only. Participants reported a variety of health conditions, and five participants scored ≤ 4 on the Clock test. Session 1 scores on standardized outcome measures, excluding the mDGI, are summarized in Table 3.

Table 2.

Background characteristics of participants.

Characteristics (N = 30)	Mean (SD) [range] or n (%)	Characteristics (N = 30)	Mean (SD) [range] or n (%)
Age		Sex
Years	76.7 (5.5) [67–91]	Woman	20 (66.7%)
Younger group	12 (40%)	Man	10 (33.3%)
Older group	18 (60%)
Education		Body composition
Elementary school	10 (33.3%)	Body Mass Index	25.5 (5.4) [17–38]
High‐ or vocational school	8 (26.7%)	Height, m	1.68 (0.09) [1.54–1.93]
College or university	7 (23.3%)	Weight, kg	72.7 (18.3) [43–125]
Other	5 (16.7%)
Living arrangement		Fall history in the past 12 months
Lives alone	15 (50%)	None	17 (56.7%)
Lives with a spouse	15 (50%)	1 fall	13 (43.3%)
		≥ 2 falls	8 (26.7%)
Use of walking aids		Clock Drawing Test
Walks without aids	24 (80%)	6 points	22 (73.3%)
Cane or crutch	3 (10%)	5 points	3 (10%)
Walker (4‐wheeled)	3 (10%)	4 points	2 (6.7%)
		3 points	3 (10%)
Number of prescribed medicationsa
None	1 (3.4%)
1–4	15 (51.7%)
≥ 5	13 (44.8%)
Health conditionsb
Coronary artery disease	5 (16.7%)	Chronic heart failure	2 (6.7%)
Chronic obstructive pulmonary disease	2 (6.7%)	Asthma	2 (6.7%)
Stroke	1 (3.3%)	Parkinson's disease	3 (10%)
Osteo‐ or Rheumatoid arthritis	10 (33%)	Diabetes	2 (6.7%)
Depression	2 (6.7%)	Anxiety	7 (23.3%)
Hearing impairment	8 (26.7%)	Vision impairment	7 (23.3%)

Note: This table illustrates the varied demographic, health, and lifestyle characteristics, highlighting aspects such as age, sex, education, body composition, mobility, medication use, and health conditions. Body Mass Index (BMI) was measured in kg/m². The Clock Drawing Test scores range from zero to 6 points, with higher scores indicating better cognitive function.

^a Information on medications were based on each participants' medication list. One participant did not bring information regarding medication use, so numbers and percentages are based on N = 29.

^b Information on health conditions was based on participants self‐report on their medical diagnoses.

Table 3.

Participants functioning based on scores on standardized outcome measures, other than modified Dynamic Gait Index.

Standardized outcome measures, scalesa	Mean	Standard deviation	Median	Range
Ten‐Meter Walk Test (10MWT), m/sec	1.23	0.2	1.21	1.0–1.9
Timed Up & Go (TUG), sec	9.7	1.7	9.35	7.1–13.9
Activities‐specific Balance Confidence (ABC) scale, 0–100	70.4	15.5	65.6	41.9–98.1
Short Form Health Survey (SF‐36), 0–100
Physical component summary score	43.7	8.4	44.5	25.7–57.7
Physical functioning scale	74.0	20.6	81.3	6.3–100
Role physical scale	41.7	39.6	25.0	0.0–100
General health scale	59.6	18.7	59.4	18.8–87.5
Bodily pain scale	64.96	22.0	62.0	22.0–100
Mental component summary score	49.7	11.5	51.3	22.3–69.9
Vitality scale	55.8	16.9	60.0	10.0–95.0
Social functioning scale	78.3	20.7	81.3	25.0–100
Role emotional scale	66.7	42.9	100	0.0–100
Mental health scale	76.3	17.4	76.0	28.0–100

^a For the 10MWT, ABC scale and all SF‐36 scales, higher score indicates better functioning. For TUG, higher score (time in sec) indicates worse functioning.

3.2. MDGI Descriptives, Floor and Ceiling Effects

Table 4 presents results for the mDGI scales, including the total score, three facets, and eight tasks. It provides key details such as available scoring range, M and SD for Session 1 and Session 2. For Session 1, additional information is given on median, actual scoring range, and the proportion achieving the maximum score, denoted as the ‘ceiling effect.’ A ceiling effect of varying degrees was observed in all mDGI scales, except for the mDGI total, gait pattern facet, and walking with horizontal and vertical head movements (Tasks 3 and 4), where no ceiling effects were observed. The ceiling effect in the level of assistance facet was noted because all participants successfully completed tasks without support, resulting in a maximum score on this facet. Floor effects were not observed.

Table 4.

Modified Dynamic Gait Index (mDGI) test‐retest relative reliability and comparisons of scores in repeated measurements (Session 1 and Session 2), with median and maximum Score in Session 1.

Measure (Available score range) N = 30	ICC 3,1 (95% CI)		Mean (SD)			Ceiling effectb
	Absolute agreement	Consistency	Session 1	Session 2	p valuea	Median (range)	Maximum Score n (%)
mDGI total score (0–64)	0.95 (0.90–0.98)	0.95 (0.90–0.98)	53.8 (6.1)	54.0 (5.6)	0.500	54 (37–63)	0 (0%)
Performance facets scores
Time score (0–24)	0.89 (0.78–0.95)	0.89 (0.77–0.94)	19.7 (3.5)	19,8 (3,4)	0.827	20 (11–24)	6 (20%)
Gait pattern score (0–24)	0.79 (0.61–0.90)	0.79 (0.60–0.89)	18.1 (3.2)	18,3 (3.0)	0.654	19 (9–23)	0 (0%)
Level of assistance score (0–16)	1	1	16.0 (0.0)	16.0 (0.0)	1.000	16 (16)	30 (100%)
mDGI task scores (0–8)
1. Gait on level surface (usual pace)	0.82 (0.66–0.91)	0.83 (0.67–0.91)	7.3 (1.1)	7.5 (0.9)	0.211	8 (4–8)	19 (63.3%)
2. Change in gait speed (change pace)	0.73 (0.51–0.86)	0.73 (0.51–0.86)	6.8 (1.0)	7.0 (1.0)	0.326	7 (5–8)	9 (30.0%)
3. Gait with horizontal head turns	0.78 (0.59–0.89)	0.78 (0.58–0.89)	6.1 (1.1)	6.2 (1.0)	0.293	6 (4–7)	0 (0.0%)
4. Gait with vertical head turns	0.75 (0.54–0.87)	0.75 (0.53–0.87)	6.0 (1.0)	6.0 (0.9)	0.601	6 (4–8)	4 (13.3%)
5. Gait and pivot turn	0.80 (0.60–0.91)	0.83 (0.67–0.91)	6.7 (1.1)	6.5 (0.9)	0.018	7 (4–8)	8 (26.7%)
6. Steps over obstacles	0.79 (0.61–0.90)	0.80 (0.63–0.90)	6.7 (0.9)	6.6 (0.8)	0.096	7 (5–8)	7 (23.3%)
7. Steps around obstacles	0.82 (0.65‐0.91)	0.82 (0.66‐0.91)	6.7 (1.0)	6.8 (0.9)	0.211	7 (5–8)	7 (23.3%)
8. Walking up stairs	0.92 (0.83–0.96)	0.92 (0.83–0.96)	7.0 (0.9)	7.0 (0.8)	0.326	7 (5–8)	12 (40.0%)

Note: This table demonstrates the test‐retest reliability of the mDGI across two sessions, showing high consistency in scores with detailed analysis on median and maximum values, and no significant differences between sessions for most measures.

Abbreviations: CI, Confidence Interval; ICC, Intraclass Correlation coefficient.

^a P‐values for comparisons of all mDGI scores from session 1 and 2, are based on paired t‐tests. Values of ≤ 0.05 reflect a significant difference.

^b Scores from Session 1.

3.3. Test‐Retest Reliability

3.3.1. Relative Reliability

Table 4 includes the ICC _3,1 values for all mDGI scales and comparisons of performances in both sessions. The test‐retest reliability varied, with the highest observed for the mDGI total score (ICC _1,3 = 0.95), time facet (ICC _1,3 = 0.89), and gait pattern facet (ICC _1,3 = 0.79) and the lowest for mDGI Task 2, change in gait speed (ICC _1,3 = 0.73). While most ICC _1,3 values for absolute agreement and consistency were nearly identical, a notable exception in the mDGI Task 5 revealed a distinct coefficient due to significant session differences (ICC _1,3 = 0.80 absolute agreement; ICC _1,3 = 0.83 consistency).

3.3.2. Absolute Reliability

SEM for the 64‐point mDGI total was 1.32 points, for the 24‐point time facet it was 1.17 points, and for the 24‐point gait pattern facet, it was 1.43 points. The MDC ₉₅ values for the total mDGI, the time facet, and the gait pattern facet were 3.66, 3.25, and 3.96, respectively. Bland‐Altman plots, presented in Figure 1a–c, illustrate the mean of two mDGI measurements (on the x‐axis) against the differences between these measurements (on the y‐axis) for the mDGI total, time, and gait pattern scores. The data points are approximately symmetrically distributed around the zero line in each plot, indicating no significant bias between the measurements from the two sessions, with the mean difference (bias) close to zero, suggesting an absence of systematic errors. Each plot shows two outliers that exceed the limits of agreement, marked by dashed lines. A reanalysis of the data, excluding these outliers, had negligible effects on the calculated reliability or validity of the measurements.

Figure 1.

(a–c) Bland‐Altman Plots Depicting the Differences Between Session 1 and Session 2 of the Modified Dynamic Gait Index (mDGI) Scales (Total, Time Facet, and Gait Pattern Facet Scores) Against the Mean of the Two Sessions for Each mDGI Scale.

3.4. Internal Consistency

Internal consistency was most pronounced for the mDGI total score (α = 0.90, 95% CI = 0.83–0.94), followed by time facet (α = 0.89, 95% CI = 0.81–0.94) and gait pattern scores (α = 0.86, 95% CI = 0.77–0.92). For mDGI task scores, α ranged from 0.35 to 0.62.

3.5. Construct Validity

Table 5 presents the correlations between the mDGI scores and other standardized measurements. The 10MWT showed a strong correlation with several mDGI scales, including the total score (r _s = 0.79), time facet (r _s = 0.85), and Task 6 (steps over obstacles) (r _s = 0.77), with moderate to good correlation for the other six mDGI tasks. The TUG showed a moderate to good correlation with the mDGI total score (r _s = −0.73), time facet (r _s = −0.73), gait pattern facet (r _s = −0.58), and five tasks. The ABC scale had a low‐to‐fair correlation with the mDGI time facet (r _s = 0.22) and Tasks 3 (r _s = 0.32) and 4 (r _s = 0.34). SF‐36 scales exhibited varied correlations with mDGI, with only the social functioning scale and Task 5 reaching a moderate to good correlation (r _s = 0.50).

Table 5.

Spearman's rho correlations ^a between the Modified Dynamic Gait Index (mDGI) scales and other outcome measures.

Measuresb (Available score range) N = 30	mDGI Total (0–64)	mDGI performance facets			mDGI tasks (0–8)
		Time (0–24)	GP (0–24)	LA (0–16)	Task 1	Task 2	Task 3	Task 4	Task 5	Task 6	Task 7	Task 8
10MWT, m/sec	0.79***	0.85***	0.49*	N/A	0.38*	0.62**	0.64**	0.64**	0.53**	0.77***	0.55**	0.52**
TUG, sec	−0.73**	−0.73**	−0.58**	N/A	−0.43*	−0.53**	−0.64**	−0.42*	−0.58**	−0.68**	−0.44*	−0.65**
ABC scale, 0–100	0.22	0.26*	0.15	N/A	0.10	0.06	0.32*	0.34*	0.13	0.09	0.14	0.19
PCS score (SF‐36), 0–100	0.22	0.24*	0.19	N/A	0.19	0.14	0.10	0.10	0.35*	0.24*	0.20	0.34*
PF scale (SF‐36), 0–100	0.35*	0.35*	0.35*	N/A	0.35*	0.35*	0.08	0.03	0.13	0.34*	0.25*	0.35*
RP scale (SF‐36), 0–100	0.13	0.16	011	N/A	0.11	− 0.01	0.19	0.1	0.18	0.05	0.27*	0.19
GH scale (SF‐36), 0–100	0.14	0.11	0.10	N/A	0.16	−0.03	−0.11	0.03	0.37*	0.21	0.26*	−0.03
BP scale (SF‐36), 0–100	0.01	0.10	−0.05	N/A	−0.17	−0.06	0.06	−0.05	0.31*	0.09	−0.16	0.31*
MCS score (SF‐36), 0–100	0.17	0.18	0.08	N/A	0.10	−0.10	0.19	0.12	0.36*	0.04	0.17	0.08
VT scale (SF‐36), 0–100	0.11	0.14	0.04	N/A	0.07	−0.02	0.12	0.13	0.20	0.11	0.09	0.02
SF scale (SF‐36), 0–100	0.40*	0.46*	0.26*	N/A	0.06	0.11	0.46*	0.34*	0.50**	0.28*	0.28*	0.34*
RE scale (SF‐36), 0–100	0.10	0.07	0.10	N/A	0.09	−0.12	−0.01	0.04	0.31*	−0.01	0.17	0.25*
MH scale (SF‐36), 0–100	0.11	0.17	−0.11	N/A	−0.11	−0.18	0.17	0.14	0.30*	−0.03	0.16	0.04

Note: This table demonstrates the Spearman's Rho correlations between the mDGI scores and various outcome measures. 10MWT = Ten‐Meter Walk Test in m/sec; TUG = Timed Up & Go test in sec; ABC scale = Activities‐specific Balance Confidence scale, 0–100 points; SF‐36 = Short Form Health Survey scales, do all range from 0–100 points; PCS = Physical Component Summary; PF = Physical Functioning scale; RP = Role Physical; GH = General Health; BP = Bodily Pain scale; MCS = Mental Component Summary; VT = Vitality; SF = Social Functioning; RE = Role Emotional; MH = Mental Health; Task 1 = gait on level surface (usual pace), Task 2 = change in gait speed (change pace); Task 3 = gait with horizontal head turns; Task 4 = gait with vertical head turns; Task 5 = gait and pivot turn; Task 6 = steps over obstacles; Task 7 = steps around obstacles; Task 8 = walking up stairs; GP = gait pattern; LA = level of assistance; N/A = not applicable as the level of assistance scores are a constant.

^a Interpretations of Spearman's rho correlation coefficients > 0.75 as strong***, 0.50–0.75 as good**, 0.25–0.50 as fair*, and < 0.25 as poor [24].

^b For all outcome measures, except TUG, higher score means better functioning.

4. Discussion

This study evaluated the psychometric properties of an Icelandic version of the mDGI in community‐dwelling older adults with balance impairments. The assessment covered test‐retest reliability (relative reliability), SEM and MDC ₉₅ (absolute reliability), internal consistency, construct validity, and potential floor and ceiling effects. In summary, for the total mDGI scale, mDGI time, and gait pattern facets, the research hypotheses were supported, with a few exceptions associated with construct validity.

Participants in this study were aged 67 to 91 years, living at home, and engaged in physical therapy group balance training twice per week due to their balance impairments. While the majority did not use walking aids, slightly over 43% had fallen in the past year, with nearly 27% experiencing recurrent falls. This fall history exceeds the expected 30% annual fall rate among older adults [3, 39] and aligns with findings that approximately 50% of those who fall do so repeatedly [40]. These statistics indicate that the participants were at a higher risk of falling compared to the general community‐dwelling older adult population. This elevated fall risk may be associated with the various health conditions of the participants, including cardiovascular diseases, stroke, depression, diabetes, arthritis, Parkinson's disease, vision impairments, and hearing impairments [3]. Additionally, nearly 45% of the participants took five or more medications (polypharmacy), increasing their fall risk due to potential side effects and interactions [41]. The characteristics of these participants are likely representative of many older adults receiving treatment for balance impairments in outpatient clinical practice, suggesting that our findings on the psychometric properties of the mDGI may be generalizable to similar populations in other countries.

The Icelandic version of the mDGI exhibited a high relative reliability for the total scale (ICC _1,3 = 0.95) and good relative reliability for the mDGI facet subscales (ICC _1,3 = 0.88–1.00), exceeding our hypothesized ICC threshold of 0.75. These high ICC_1,3 values suggest stable measurements over time, with score variations primarily due to individual differences rather than measurement errors [23, 38], indicating these mDGI scales produce reliable results under consistent conditions. The ICC _3,1 results for absolute agreement and consistency were almost identical, indicating that our findings are comparable to reliability studies where ICCs or Pearson correlation coefficients have been calculated [38]. Accordingly, our ICC _1,3 results align with Shumway et al.'s (2013) original test‐retest study, which reported a Pearson correlation coefficient of 0.92 for the mDGI's total score and 0.87 to 0.91 for the facets. Despite differences in participant characteristics, sample sizes, and the language between the current study and Shumway et al.'s study, the comparable ICC _3,1 results further support the reliability and applicability of the mDGI total scale and facet subscales for dynamic gait evaluation across diverse populations and contexts.

The relative reliability of the mDGI task was good for seven out of the eight tasks (ICC _1,3 = 0.75–0.92) and moderate for one task (0.73), yet these values are generally slightly lower than the reported Pearson correlation coefficients in the original study of the mDGI (r = 0.86–0.90) [14]. Our findings indicate the lowest reliability in walking with a change in gait speed (0.73), followed by walking with vertical head turns (0.75), and the highest reliability for walking up stairs (0.92). The high reliability of the walking up stairs task, may reflect the task's relatively simple instructions and scoring, signifying consistent behavior in stair walking, such as using a handrail or not [6, 7, 10, 42]. Discrepancies in the least reliable tasks across studies may stem from variation in tested individuals and learning effects [42]. Consequently, while the reliability of repeated measurements for the total score on mDGI, similar to DGI, may be good to excellent, individual tasks may exhibit lower reliability [10]. The least reliable tasks may represent activities challenging for individuals in daily life, particular for those with inner ear balance system impairments, leading to variable performance [10]. The varied reliability in individual mDGI tasks underscores the importance of including all eight tasks in calculations for the total mDGI and facet scores, for use as outcome measures [17].

Absolute reliability was confirmed by a low SEM for both the total mDGI (1.32) and its performance facets (1.17–1.43), supported by their respective MDC ₉₅ values (3.25–3.96). Notably, the MDC ₉₅ value of 3.66 for the mDGI total was considerably lower than the value of 7.0, as reported by Matsuda et al. [18], whereas the MDC ₉₅ values for the facets were similar. Smaller SEM and MDC ₉₅ values highlight the mDGI's ability to detect subtle performance changes and reflect the variability in repeated measurements within an individual [23]. Our relatively low SEM values are noteworthy, given SEM dependence on sample size, which decreases with more measurements [23]. The Bland‐Altman plots (Figure 1a–c) demonstrated a predominantly unbiased pattern in the mean difference between the two measurements across mDGI total scores, time, and gait pattern facets. The estimated agreement interval, within which 95% of the differences fall, is considered acceptable [43], especially for the mDGI total score. Due to the lack of variability between sessions in the level of assistance facet for the mDGI, it was not suitable for such graphical representation.

The internal consistency of the total mDGI (α = 0.90), time facet (α = 0.89), and gait pattern (α = 0.86) scores was robust, indicating strong coherence within these scales and confirming our hypothesized values within the moderate range of 0.70 to 0.90. The most relevant comparison is with the ‘Gait Abnormality’ subgroup (N = 91) from Matsuda et al. [18], with an age range of 54 to 95 years (M = 80.4 years). Their α values for mDGI total, time facet, and gait pattern scores were 0.96, 0.96, and 0.87, respectively. The higher internal consistency in the original sample may be due to a larger, more homogeneous group, providing more stable estimates [23, 44]. Similar to Matsuda et al. [18], our study found lower α values for individual mDGI tasks (α = 0.35–0.62) compared to the total and facet scores. These lower values are expected, as each task score comprises only three relatively heterogeneous facets. Both the number of items and their homogeneity can potentially raise internal consistency [44, 45]. Thus, individual tasks may not be psychometrically strong enough to stand alone as outcome measures, unlike the mDGI total and facet scales.

The Icelandic version of the mDGI demonstrated construct validity through its correlations with both gait performance measures (10MWT and TUG) and patient‐reported measures (ABC and SF‐36). Although none of these measures captures the most dynamic aspects of gait such as mDGI does, the observed correlation patterns align with the original aim of the mDGI [14].

As hypothesized, the mDGI total and time facet had moderate to good correlations with basic mobility (TUG). While both instruments include timed outcomes used to assess balance and predict the risk of falls, their theoretical constructs differ. The mDGI offers a detailed assessment of how individuals adapt their gait to various daily life tasks and environmental challenges [14], whereas the TUG provides a quick measure of basic mobility, including balance [25], explaining the correlation between their outcomes. The correlation between the mDGI gait pattern facet and the TUG was slightly lower, as the TUG does not assess the quality of movements.

The correlation between the mDGI scales and comfortable gait speed (10MWT) was strongest and higher than hypothesized for the mDGI total and the time facet. While both the mDGI time facet and the 10MWT include timed comfortable walking, explaining their correlation, the 10MWT provides no direct or specific information on balance. However, walking speed is generally considered a measure of physical ability and functional gait [26], which inherently requires balance. Additionally, faster walking speed has been associated with better health, functional status, and longer life expectancy in older adults [46]. In this context, walking speed is regarded as a simple and accessible summary indicator of vitality, as it requires energy, movement control, and places demands on multiple organ systems [46], which supports the strong correlation between the 10MWT and the mDGI total measuring dynamic balance. Similar to the TUG, the 10MWT does not assess the quality of movement, explaining its slightly lower correlation with the mDGI gait pattern facet.

The correlation between the mDGI scales and the ABC scale ranged from low to fair, which was weaker than hypothesized. This result can be expected because, although the constructs behind these scales have similarities, they do differ. The ABC scale focuses on individuals' self‐perceived confidence in situation‐specific real‐life activities that challenge balance and stability, which are not limited to gait alone [27]. In contrast, the mDGI scales measure actual gait performance in challenging movements and environments in the clinic [14]. Self‐perceived ability to maintain balance in real life and actual performance in a safer clinical setting can be quite different [47]. Interestingly, the strongest correlation was observed in mDGI gait tasks involving horizontal and vertical head turns (Tasks 3 and 4), which also were the only tasks without ceiling effects. This may be because these tasks more closely reflect the balance challenges individuals perceive in real‐life situations, compared to the other tasks within this sample.

The mDGI's correlation patterns with the SF‐36 scales unexpectedly revealed the strongest correlation with the social functioning scale, which belongs to the mental health dimension of health‐related quality of life [28]. This finding suggests that individuals with higher mDGI scores might have a greater tendency for social engagement and participation due to better mobility and confidence, potentially enhancing their quality of life and promoting more active social interactions. In contrast, the low to fair correlation of many mDGI scales with the physical functioning scale, and the generally little to low correlation with other SF‐36 scales, were mostly as expected based on potential similarities and differences in the underlying constructs.

Further supporting the mDGI's construct validity, and consistent with previous studies [14], no floor effects were observed in any of the mDGI scales. Ceiling effects were absent in the mDGI total, gait pattern facet, and gait with head turns (Tasks 3 and 4), indicating that these mDGI scales effectively measure a wide range of dynamic gait abilities without the limitations of extreme scores clustering. Notably, the level of assistance facet exhibited clear ceiling effects, as all participants completed the mDGI without a walking aid. Additionally, the tasks involving walking on a level surface and climbing stairs had the highest proportion of participants scoring in the top 15% of the scale, indicating potential limitations in these mDGI scales' ability to differentiate higher levels of performance among older community‐dwelling adults.

Some limitations of this study should be considered when interpreting these results. First, the relatively small sample size of 30 participants, all of whom completed the mDGI without using walking aids despite their balance impairments, may limit the generalizability of our findings to similar populations. It is crucial to acknowledge that, despite the constrained sample size and limited variation in walking ability without aids, we adhered to guidelines for reliability research [30]. This involved the inclusion of older community‐dwelling adults with diverse underlying medical diagnoses and various causes of balance and mobility impairments. Second, the testers in this study were final‐year MSc students in a physical therapy program. Consequently, they lacked extensive experience with the mDGI tool, which relies partially on subjective judgment and requires familiarity with specific test guidelines [14, 21]. This inexperience could potentially introduce variability in the test results due to differences in interpretation and application of the guidelines. To mitigate this, the testers underwent comprehensive training and received supervision throughout the testing process. The training included detailed instruction on the mDGI protocol, practice sessions to develop competency, and ongoing feedback to ensure consistent and accurate application of the measurement tool. This rigorous training and supervision aimed to ensure both the quality of assessment and adherence to the established protocol. Therefore, we believe that novice physical therapists and other health professionals who receive similar training can obtain psychometrically sound outcomes using the mDGI, despite their initial lack of experience. Finally, scheduling flexibility, which allowed participants to choose the timing of their assessments before or after a group session, may have introduced some bias in repeated measurements. The timing could have influenced participants' fatigue levels and, consequently, their performance on the mDGI and other outcome assessments [4]. This scheduling flexibility was adopted to encourage participation and reflects reality in an outpatient clinical practice.

Longitudinal studies are needed on the Icelandic version of the mDGI to assess its responsiveness to changes in balance and gait, including participants using assistive devices during the test, and to investigate the relationship between the mDGI and fall risk. Additionally, international studies of mDGI's psychometric properties of the mDGI should include cross‐cultural validations in participants with various health conditions or age groups to assess its applicability and effectiveness across a wider range of individuals.

5. Conclusions

The findings of this study strongly support the reliability and validity of the Icelandic version of the mDGI for evaluating balance and gait in community‐dwelling older adults with balance impairments. These findings align with international research, affirming the mDGI's credibility in geriatric clinical practice. The robust evidence presented here enhances the potential for widespread implementation of the mDGI in health care settings, and it may foster global collaboration among physical therapists and other health professionals focusing on clinical practice and research on balance and gait.

Ethics Statement

This study was approved by the Icelandic National Bioethics Committee (VSNb2018050025/0301).

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study are available from the Chief Investigator Professor Solveig A. Arnadottir (saa@hi.is) upon reasonable requests for data sharing.