Rasch analysis of items from two self-report measures of motor function: determination of item difficulty and relationships with children’s ability levels

Abstract

Aim  The aim of this article was to determine item measurement properties of a set of items selected from the Gillette Functional Assessment Questionnaire (FAQ) and the Pediatric Outcome Data Collection Instrument (PODCI) using Rasch analysis, and to explore relationships between the FAQ/PODCI combined set of items, FAQ walking scale level, Gross Motor Function Classification System (GMFCS) levels, and the Gait Deviation Index on a common measurement scale.

Method  Rasch analysis was performed on data from a retrospective chart review of parent-reported FAQ and PODCI data from 485 individuals (273 males; 212 females; mean age 9y 10mo, SD 3y 10mo) who underwent first-time three-dimensional gait analysis. Of the 485 individuals, 289 had a diagnosis of cerebral palsy (104 GMFCS level I, 97 GMFCS level II, 69 GMFCS level III, and 19 GMFCS level IV). Rasch-based person abilities and item difficulties based on subgroups defined by the FAQ walking scale level, Gait Deviation Index, and the GMFCS level were compared.

Results  The FAQ/PODCI item set demonstrated necessary Rasch characteristics to support its use as a combined measurement scale. Item groupings at similar difficulty levels were consistent with the mean person abilities of subgroups based on FAQ walking scale level, Gait Deviation Index, and GMFCS level.

Interpretation  Rasch-derived person ability scores from the FAQ/PODCI combined item set are consistent with clinical measures. Rasch analysis provides insights that may improve interpretation of the difficulty of motor functions for children with disabilities.

Abbreviations

FAQ

Gillette Functional Assessment Questionnaire

FAQ-WL

Gillette Functional Assessment Questionnaire walking scale level

GDI

Gait Deviation Index

PODCI

Pediatric Outcome Data Collection Instrument

SPF

Sports and physical function

TBM

Transfers and basic mobility

What this paper adds

• •
   This article provides a Rasch analysis of item-level measurement characteristics of a combined FAQ/PODCI item set which exhibits better content coverage and greater precision than either item set alone.
• •
   Item groupings at similar difficulty levels were consistent with the mean person abilities of subgroups based on FAQ-WL, GDI, and GMFCS level.
• •
   Clinically meaningful integration of self-report measures (FAQ/PODCI), clinical scales (GMFCS), and objective (GDI) assessments are described.

Improvements in mobility, ambulation, and enhanced participation in activities with family and peers are often identified as primary goals of clinical and surgical interventions for children and adolescents with physical impairments. Clinical measurement of ambulation and functional mobility can be accomplished using a variety of measures including instrumented gait analysis, standardized clinical assessments of gross motor skills, and/or self-reported outcomes. Despite concern about the accuracy of self-/proxy report,^1–3 the perspective provided by the affected individual and/or family adds important insight into both goal setting and the assessment of outcome.

The reporting of multiple assessments, however, is complex. The International Classification of Functioning, Disability; and Health (ICF) framework⁴ has become an important organizing framework for health and disability in rehabilitation research. Within the ICF framework, an activity is defined as the execution of a task or action by an individual. Participation is defined as involvement in a life situation. The ICF provides two qualifiers for activity and participation: (1) ‘capacity’– what a child is able to do in an ideal environment, the highest functioning level; and (2) ‘performance’– what a child actually does in the environment in which they live. Since the ICF’s emergence in 2001, the use of these ICF qualifiers, along with ‘capability’– a child’s capacity influenced by environmental factors and choice – have become increasingly important in clinical research, particularly in assessing treatment efficacy and describing the range of function for the child with a disability.^5–7 Existing validated instruments often predate the emergence of the ICF, and may include within a single instrument items that measure capability, capacity, or performance, and generically use words (capable, skill, ability, etc.) in their instrument description that now have additional, specific ICF-related meanings. The fact that existing instruments often blend measurement of different ICF components and qualifiers of functioning and disability complicates the issues faced by clinicians and researchers in their choice of appropriate instruments.

Ideally, a comprehensive battery of instruments chosen to characterize function would be structured to ‘fit’ together to provide the greatest breadth of assessment, with each measure adding complementary information about the person’s status without adding redundancy. Often, however, each instrument in a given battery is reported in isolation from the others – lacking a common context of how each relates to the others. Approaches that provide common measurement across instruments can help to reduce the complexity introduced by the use of multiple instruments.

Rasch analysis can be used to create a common, continuous interval-level measurement scale for estimates of both ‘person ability’ (how much of the underlying construct the person demonstrates) and ‘item difficulty’.⁸ The Rasch terms ‘person ability’ and ‘item difficulty’ have a specific meaning in this context. This person item mapping indicates the probability that an individual of a certain ‘person ability’ can perform specific items based on each ‘item’s difficulty’. Recent outcome assessments of physical function have used Rasch or item response theory analyses to more fully evaluate the measurement properties of the instruments on an item-level basis^9–14 and in designing functional staging systems which have been found to be useful in enhancing clinical decision making.^15,16

The Gillette Functional Assessment Questionnaire (FAQ)^10,17 and the Pediatric Outcome Data Collection Instrument (PODCI)^9,18–21 are self-/parent-report measures of physical abilities commonly used in pediatric clinical practice. Both are often used in conjunction with the Gross Motor Function Classification System (GMFCS)^22,23 to describe function and walking ability in children with cerebral palsy (CP).^24–26 The FAQ consists of a 10-level classification of walking ability (FAQ walking scale level or FAQ-WL) and 22 functional activities rated on a five-point Likert difficulty scale (FAQ-22). The FAQ-WL portion of the FAQ has been validated as a measure of functional walking status.¹⁷ The PODCI consists of 86 items divided into eight subscales. The PODCI is ‘designed to assess overall health, pain, and participation in normal daily activities as well as in more vigorous activities associated with young people.’¹⁸ The transfers and basic mobility (TBM) subscale consists of 11 items, and the sports and physical function (SPF) subscale consists of 21 items (12 tasks and nine conditional responses). PODCI test scores and subscales have been shown to be reliable and valid.¹⁹ Factor- and item-level properties of both the FAQ and PODCI instruments have been previously reported.^9,10 Items contained in both measures (FAQ and PODCI) reflect a blend of ICF activity and participation concepts because the ICF emerged after the development of these instruments. Therefore, the underlying construct being assessed by these instruments is reflective of more general physical functioning.

Rasch analysis can be used to establish the relationship between items from the FAQ-22 and PODCI TBM/SPF items on a common difficulty scale. Subsequent stratification of individuals based on classification systems such as the FAQ-WL and GMFCS and measures of gait impairment such as the Gait Deviation Index (GDI)²⁷ may reveal additional relationships, which will allow clinicians to have a higher level of confidence that certain physical skills tend to be associated with a specific classification of functional mobility or gait impairment.

The purposes of this study were to assess the factor- and item-level properties of a combined FAQ-22 and PODCI TBM/SPF item set, and to explore the associations between the Rasch-ordered combined FAQ-22/PODCI TBM/SPF item set, with participant groupings based on the FAQ-WL, GMFCS, and GDI.

Method

A retrospective medical record review of FAQ (FAQ-WL and the FAQ-22), PODCI (TBM/SPF scale), and GDI data was conducted on a group of children and young adults (<19y) who underwent first-time gait analysis in a tertiary hospital setting between January 2006 and June 2008. GMFCS level was included for children with a diagnosis of CP. FAQ and PODCI data were obtained by proxy report of the parents or legal guardian as part of the routine clinical gait analysis. All FAQ-22 items, nine of 11 PODCI-TBM subscale items, and five of 12 PODCI-SPF subscale items scored on a similar five-point Likert difficulty scale were included for analysis. The two excluded TBM items measure frequency of assistance needed, a different construct, rather than a specific physical skill. The excluded SPF items either measure frequency of assistance needed or are structured as multiple responses to a single item with conditional responses that are not on a five-point Likert difficulty scale which did not allow these items to be included in the analysis. Waiver of informed consent and Health Insurance Portability and Accountability Act authorization were obtained for this study from the local institutional review board. Individuals whose families had opted out of use of medical records for research were excluded.

Statistical analysis

Exploratory and confirmatory factor analysis was conducted using MPlus 5.1 software²⁸ to validate the structure of the combined FAQ-22/PODCI TBM/SPF item set. Model fit was examined via multiple indices including the Confirmatory Fit Index, the Tucker–Lewis Index, and standardized root mean square residual. Confirmatory Fit Index and Tucker–Lewis Index values greater than 0.95 and a standardized root mean square residual less than 0.08 indicate good fit of the model to the data. Data were tested to ensure that the statistical assumptions for Rasch analyses were met. The Rasch rating scale model was implemented using Winsteps software²⁹ to simultaneously determine item difficulty or location (i.e. the difficulty level of each item relative to other items in the scale), person ability (i.e. the ability level of each person relative to other persons in the sample), and item-level fit statistics (degree of variation in the responses relative to the predicted responses) for the combined FAQ-22/PODCI TBM/SPF item set. The Rasch model uses a log odds units (logit) scale, which is linear and additive. Logit distances describe the relative performance on adjacent categories of the response scale. For our Likert scale, there is a point on the latent trait of physical functioning at which ‘some difficulty’ and ‘much difficulty’ are equally likely to be observed. So, on a logit scale at a point 1.4 logits higher, ‘much difficulty’ is likely to be observed eight times for every two times that ‘some difficulty’ is observed. Precision (standard error) of the estimated person ability using the PODCI TBM/SPF items alone, the FAQ-22 items alone, and the combined set of FAQ-22/PODCI TBM/SPF item set was calculated based on the assumption that all items were completed.

All children were subsequently grouped by their FAQ-WL. Children with CP were also separately grouped by their GMFCS level. In each of the subgroups for the FAQ-WL and GMFCS level, the mean Rasch-based person ability score from the combined FAQ-22/PODCI TBM/SPF item set was determined. The differences between mean person ability scores of the FAQ-WL and GMFCS subgroups were tested using analysis of variance (ANOVA; p < 0.05) with least significant difference post hoc testing, correcting for multiple comparisons if the main effect of the factor (FAQ-WL or GMFCS level) was significant. These resultant mean and related 95% confidence intervals were then compared with the item difficulty statistics from Rasch analysis to determine the corresponding sets of items that individuals at each GMFCS level and FAQ-WL could be expected to perform with a 50% probability.

Pearson’s correlations were calculated between the GDI and person ability measures. The GDI was also used to group children. The continuous GDI scores, ranging from 42 to 116, were transformed to a discrete measure by creating five 1-SD bins for GDI scores from 50 to 100 (50.1–60, 60.1–70, etc.). Those with GDI scores below 50 were grouped together and those with GDI scores above 100 were grouped together. The differences in mean person ability scores among GDI bins was tested using ANOVA (p < 0.05) with least significant difference post hoc testing if the main effect of the factor (GDI bin) was significant.

Results

Nine hundred thirty-five children and adolescents had gait analysis during the specified time frame. Of those 935, 485 who had first-time gait analysis met the criteria for inclusion. Two hundred and eighty-nine children had a diagnosis of CP (104 GMFCS level I, 97 level II, 69 level III, 19 level IV) and 196 children had other neurological or musculoskeletal diagnoses (73 orthopedic, 39 neuromuscular, 30 joint disorder, 17 acquired brain injury, 11 genetic, 26 miscellaneous). Mean age was 9 years 10 months (SD 3y 10mo). Demographic data are summarized in Table I.

Table I.   Demographics and breakdown of available data by diagnosis category

	Overall (n=485)	Cerebral palsy (n=289)	Other (n=196)
Age range for all children is 3 to 19 years. Differences in total numbers represent missing data values for some variables. n, number totals; GMFCS, Gross Motor Function Classification System; FAQ, Gillette Functional Assessment Questionnaire; PODCI, Pediatric Outcome Data Collection Instrument; GDI, Gait Deviation Index.
Sex, M/F	273/212	172/117	101/95
Mean age (SD), y	9.9 (3.8)	9.1 (3.8)	11.1 (3.7)
GMFCS (CP only) level
I	–	104	–
II	–	97	–
III	–	69	–
IV	–	19	–
FAQ, n	479	285	194
PODCI, n	465	277	188
GDI, n	480	284	196

Exploratory and confirmatory factor analysis of the combined FAQ-22/PODCI TBM/SPF item set indicated a sufficiently unidimensional underlying construct as only two factors explained more than 5% of the variance. The ratio of these two factors was greater than 12:1, indicating dominance of the single factor. Consistently high communalities (range 0.666–0.904) indicated that approximately 80% of the common variance is represented by a single latent factor. The model goodness of fit statistics demonstrated adequate one-factor model fit (standardized root mean square residual=0.071; Tucker–Lewis Index=0.976; Confirmatory Fit Index=0.979). Data met the statistical assumption of monotonicity necessary for subsequent Rasch analysis. Local dependence in 26 of 535 item pairings was noted, but the items retained. The item fit statistics for the combined item set are shown in Table SI (supporting information published online).

The resultant item person map showing the relationship of the combined set of item difficulties to the distribution of person abilities in the sample is provided in Figure 1. Item difficulties ranged from easiest (‘sit in a regular chair without holding on’ at −3.1 logits) to hardest (‘ice skating or roller skating’ at 2.8 logits; Fig. 1 and Table II). Rasch analyses demonstrated adequate fit of all items except one (‘ride a three-wheel bike [or a two-wheel bike with training wheels]’) which was retained in the analyses. No floor or ceiling effects were noted. The range of item difficulties closely matched the range of person abilities, with 35 of 36 skills (97%) falling within 2 SDs of the mean person ability (Fig. 1).

Figure 1.  Rasch Item Person Map. The frequency distribution of the sample by person ability (left) and item difficulty (right) is demonstrated on a single difficulty continuum. The vertical scale (logits) is an interval scale which represents the relative difficulty of lower extremity function, with lower numbers representing easier skills (less person ability) and higher numbers representing more difficult skills (more person ability). Each ‘#’ represents three children and each ‘.’ represents one child. Items in the right hand column include the Gillette Functional Assessment Questionnaire (FAQ) items (FAQ1:22) and Pediatric Outcome Data Collection Instrument (PODCI) transfers and basic mobility/sports and physical function (TBM/SPF) items (PODCI*). Please refer to Table II for item descriptions. Ideally, these two columns (item responses and person ability) should appear similar, with the distribution of possible item responses and person skill ability spread similarly across the difficulty scale. Distribution shows that most items and abilities surround the mean and 1 SD. M, mean; S, one SD; T, 2 SD of person ability (left) or item difficulty (right).

Table II.   The 22 FAQ skill items and 14 PODCI TBM/SPF Items are shown in order of difficulty on an interval scale

Item no.	Description	Measure (logits)
F, FAQ-22 item; P, PODCI item. Measure ‘logits’ represents a non-dimensional level of difficulty. Higher ‘measure’ scores represent more difficult tasks. Mean and 95% confidence interval of the average ability levels by GMFCS (bold), GDI bin (italic), and FAQ Walking Level (shadow pink) are shown. Average ability represents a 50% probability of successfully achieving a skill of the same level. Probability of success for a given level of person ability or classification grouping is higher for items of lesser difficulty and lower for items of higher difficulty. FAQ Walking Level ≤5 N=30 (14 FAQ-WL 5; 12 FAQ-WL 4; 2 FAQ-WL 3; 2 FAQ-WL 2).
	GDI ≥ 100.1; CI: 2.46, 3.86	3.16
	FAQ Walking Level 10 Mean Ability; 95% CI: 2.45, 3.06	2.76
F21	Ice skate or roller skate	2.71
F11	Jump rope	2.30
	GDI 90.1–100.0; CI: 1.83, 2.67	2.25
F19	Ride 2 wheel bike (without training wheels)	2.02
	GDI 80.1–90.0; CI:1.54,2.08	1.81
F13	Hop on right foot	1.66
F14	Hop on left foot	1.65
	GMFCS Level I Mean Ability; 95% CI:1.36,1.82	1.59
=>P22	Walk more than a mile	1.55
	FAQ Walking Level 9 Mean Ability; 95% CI:1.01, 1.36	1.18
F04	Walk up and down stairs without using the railing	0.92
F07	Runs well including around a corner with good control	0.84
	GDI 70.1–80.0; CI:0.52,0.99	0.76
F22	Ride an escalator, can step on/off without help	0.72
P20	Climb three flights of stairs	0.61
F12	Jumps off a single step independently	0.43
P19	Bicycle or tricycle	0.42
	FAQ Walking Level 8 Mean Ability; 95% CI:0.13, 0.52	0.32
F10	Can get on and off a bus by him/herself	0.25
F02	Walk carrying a fragile object or glass of liquid	0.24
P18	Run short distances	0.24
	GMFCS Level II Mean Ability; 95% CI:0.04, 0.44	0.24
P23	Walk three blocks	0.21
F06	Runs	0.19
P25	Get on and off the bus	−0.13
F18	Kick a ball with left foot	−0.19
F20	Ride 3 wheel bike (or 2 wheel bike with training wheels)	−0.19
F17	Kick a ball with right foot	−0.25
F16	Step over an object, left foot first	−0.29
F15	Step over an object, right foot first	−0.30
	FAQ Walking Level 7 Mean Ability; 95% CI: −0.73, 0.00	−0.37
F08	Can take steps backwards	−0.39
	GDI 60.1–70.0; CI: −0.63, −0.18	−0.40
	GDI 50.1–60.0; CI: −0.64, −0.03	−0.41
F05	Steps up and down curb independently	−0.41
F09	Can maneuver in tight areas	−0.42
P33	Bend over from standing; pick something off floor	−0.72
P21	Climb one flight of stairs	−0.77
F01	Walk carrying an object	−0.82
F03	Walk up and down stairs using the railing	−0.90
	GMFCS Level III Mean Ability; 95% CI: −1.16, −0.72	−0.94
	FAQ Walking Level 6 Mean Ability; 95% CI: −1.30, −0.63	−0.97
P24	Walk one block	−0.98
P7	Put on coat	−1.18
	GDI <=50; 95% CI: −2.06, −0.81	−1.43
	FAQ Walking Level ≤5 Mean Ability; 95% CI: −1.93, −1.01	−1.47
P30	Get on and off a toilet or chair	−1.73
P28	Stand while washing hands and face at a sink	−1.88
	GMFCS Level IV Mean Ability; 95% CI: −2.77, −1.09	−1.93
P31	Get in and out of bed	−2.23
P29	Sit in a regular chair without holding on	−3.13

The Rasch-derived person item map shown in parallel with mean (standard error) participant groupings based on FAQ-WL and GDI for all children and GMFCS level for those with CP is found in Figure 2. This figure illustrates the relationship between the ability level of the person and the relevant tasks that this person is likely and unlikely to be able to accomplish. It also depicts the relationship between the FAQ-WL, GMFCS, and GDI with item difficulties.

Figure 2.  Rasch-derived person item map and functional classification subgroups. Item level difficulty scores for the combined Gillette Functional Assessment Questionnaire (FAQ) FAQ-22/Pediatric Outcome Data Collection Instrument (PODCI) transfers and basic mobility/sports and physical function (TBM/SPF) item set are shown in parallel with mean (standard error) person ability participant groupings based on FAQ-WL 5 to 10 and 1 SD Gait Deviation Index (GDI) bins for all individuals, and Gross Motor Function Classification System (GMFCS) levels I to IV for children with cerebral palsy. The likelihood of an individual being able to successfully perform specific skills is demonstrated. An individual at a given ability level will have a 50% chance of successfully performing a skill of that same difficulty level. For example, a child with a GMFCS level II classification has a 50% probability of being able to run short distances and a 50% probability of being able to walk carrying a fragile object. The same child has approximately a 12.5% probability of being able to walk up/down stairs without a railing because the skill is almost four times as difficult based on the logit scale. The FAQ-22 and PODCI TBM/SPF items are separated only to more clearly identify each. The vertical dashed lines from the PODCI TBM/SPF items to the FAQ-22 items depict where they are located if presented as a single item set. The person ability or item difficulty is an interval scale.

Estimated person abilities based on the FAQ-22/PODCI TBM/SPF item set exhibited strong correlation with classification by FAQ-WL (Spearman’s rho = 0.739; p < 0.001) and GMFCS level (Spearman’s rho = −0.777; p < 0.001). There was a significant difference in mean estimated person ability when grouped by FAQ-WL classifications 6 to 10. Post hoc testing indicated that each estimated mean person ability for each FAQ-WL 6 to 10 was statistically different from all other FAQ-WLs (ANOVA; p < 0.05). There was a significant difference in estimated person ability when grouped by GMFCS level (ANOVA; p < 0.05). The mean person ability for each GMFCS level was statistically different from all other GMFCS levels.

Similarly, the GDI exhibited a good correlation with estimated person abilities based on the FAQ-22/PODCI TBM/SPF item set (Pearson’s rank correlation coefficient = 0.643; p < 0.001). As a continuous variable, each GDI score has a broad range of abilities possible, and for each estimated person ability score there is a broad range of GDI scores possible (r²=0.42). When the person ability scores were grouped into GDI SD bins, there were significant differences in average ability score between the bins (ANOVA; p < 0.05). Post hoc testing revealed that Rasch-derived person scores for each group by GDI 10-point bin was statistically different (p < 0.05) from all others except for GDI 50 to 60 with GDI 60 to 70.

Precision of the combined FAQ-22/PODCI TBM/SPF item set was assessed by calculating the standard error for each ability score. The standard error of the model for each score along the ability spectrum is shown in Figure S1 (supporting information published online). At the high and low end of scoring where there are few items that measure or provide information at these ability levels and few people with those levels of ability, the standard errors are, as expected, higher. Precision is best (standard errors are lowest) in the middle of the ability scale, where many tasks target that ability level and many respondents have their ability. The combined item set measures a greater range of person abilities (content coverage) with better precision than either item set alone.

Discussion

The item difficulties derived from the Rasch analysis in this study resulted in an interval-level measurement scale which enables direct comparison of item difficulties between items from two different instruments of physical functioning. Both measure a blend of ICF activity and participation qualifiers, so the resultant represents not a single ICF aspect but a more general construct of physical functioning. The resultant calibration of items from both instruments collected on the same sample (common people) provides a means to relate scores from the individual measures (FAQ-22 and PODCI TBM/SPF) to each other as well as to scores derived from the combined set. In addition, interpretation is now enhanced as the magnitude of difficulty between skills and person abilities is established on one common continuous interval scale. Individuals with Rasch person ability scores at a given level have a 50% probability of successfully completing skills at that same level of item difficulty. The improved precision and content coverage of the combined FAQ-22/PODCI TBM/SPF item set suggests that the use of both tools together may be an improved measure of physical functioning compared with using either instrument alone. This reflects both the increased number of items in the combined set and the distribution of items along the scale. Additional work to refine the set of items would be useful. Items with local dependence should be divided between different, but still directly comparable, instrument versions.

When the individuals in the sample are grouped by FAQ-WL, GMFCS level, or GDI SD bin, their classification level can be related to groups of items that fall within the confidence band of the group. For example, the relationship between FAQ-WL 8 and FAQ-22 skills ‘runs’ and ‘jump off a single step’ is now established. Individuals’ estimated abilities can be associated with the likelihood of their successful performance of a given skill or set of skills. It broadens the scope of understanding of function beyond the interpretation of each tool or system individually, and establishes the relationship between item difficulties and person abilities. The information is specific enough to be of practical use to clinicians, children, and families. For example, when a child who walks at a FAQ-WL 8 and is able to run describes the difficulty encountered when attempting to play hop-scotch or to skip with friends, the clinician now has a tool to demonstrate that the ability to hop is more difficult than the ability to run. The same child, however, should be expected to have the ability to ‘jump off a single step’ as it is within the range of item difficulty of the person ability estimate.

This study produced a model that matches clinical impression of differences in skill level based on a GMFCS level. For example, the level of difficulty assigned to the ability to ‘walk up/down stairs without a railing’, a common skill used to differentiate between GMFCS levels I and II, clearly falls between the two GMFCS levels on the difficulty scale. Another example, the ability to ‘hop on the right or left foot’, is associated with GMFCS I. The FAQ-22 skill set (designed to be a set of skills more advanced than walking) is associated with GMFCS levels I to III. Specifically, all FAQ-22 items had an estimated difficulty higher than the PODCI item ‘walk one block’, which has an estimated difficulty of −0.98. PODCI skills ‘get in/out of bed’ and ‘get on/off a toilet or chair’ are within a level of difficulty assigned to GMFCS IV. Across all levels, the level of difficulty assigned to skills appears to be consistent with clinical impression.

Understanding each perspective of function (proxy-report instruments, classification systems, and GDI) in the context of the others provides the clinician with a more complete integrated context for interpretation. Although previous reports have demonstrated that measures of gait pathology such as the GDI or the Gait Profile Score distinguish between each pair of community walking levels on the FAQ,^27,30 how they relate in difficulty has not been previously established. This can provide direction for therapies, as well as counseling for families regarding expectations of their child’s abilities. It also can identify mismatches in reported function.

A limitation of this study is that item misfit and response dependency were found within our item set, but were not accounted for within the study. The sample size impact on infit and outfit statistics may also be a potential limitation. A recommendation for further work would be to investigate the impact of this misfit and response dependency upon the scaling characteristics of the grouped item set.

The retrospective nature of our study limited the items and instruments to those that were available on a large common participant sample. Therefore, we focused our analyses on a version of the FAQ-22 and a subset of the PODCI (TBM/SPF). The subset of PODCI items selected for analyses relied on the same five-point Likert difficulty scale as the FAQ-22 in order to maintain consistency of item structure for the Rasch analysis. The excluded items were frequency of personal assistance or use of assistive device questions (sitting/standing [TBM] and walking/climbing [SPF]) not associated with any particular skill. In addition, the PODCI SPF items concerning participation in recreational outdoor activities, pick-up games or sports, competitive level sports, and gym/recess also were excluded as the structure of the questions included a conditional response that is not consistent with other questions. We would expect that if these items were included, the upper range of the PODCI item coverage would be extended toward the more difficult end, as found for the FAQ-22 items.

Another potential limitation is the inclusion of only children who were ambulatory and referred for gait analysis. Hence, our results should be interpreted very cautiously in non-ambulant children with lower skill levels than our study sample and are not an exhaustive query of functional skills, which are important supplements to locomotion. Given that the scope of our work was on application of Rasch analyses to improve our knowledge of the measurement properties of the combined instruments, we did not attempt to directly map the included items to specific ICF components and qualifiers. Future work using a prospective longitudinal design will help to define reliability and responsiveness, including the minimum clinically important difference. Studies of changes in skill ability as a result of intervention are also needed.

Conclusions

The combined FAQ-22/PODCI TBM/SPF item sets ordered using Rasch analysis lays the foundation for a framework to link outcome tools. This study represents an initial attempt to relate person ability and skill difficulty to functional classification and gait impairment. The association between the classification scales and GDI and a specific cluster of items from the ordered item set may help clinicians to better understand the relationship between each type of measure and potentially to guide treatment with a level of confidence that specific skills are within the child’s capability. With continued efforts, the potential for a paradigm shift to an integrated view of how lower extremity function is conceptualized, measured, and reported is possible.

Acknowledgments

This study was funded in part by an American Academy for Cerebral Palsy and Developmental Medicine Planning Grant 2008. The funding agency did not participate in any way in the conceptualization, design, data collection, data analysis, manuscript preparation, and/or publication decisions.