Introduction

Spinal cord injury (SCI) is a major health problem worldwide, with a global incidence of approximately 236–1009/100,000 [1]. SCI is a sudden event that causes paraplegia or tetraplegia, and has lifelong consequences in terms of activities of daily living (ADL), especially walking. Along with motor and sensory deficits, instability of the cardiovascular, thermoregulatory and bronchopulmonary systems are common after SCI [2].

Quality of life (QOL) in patients with SCI is a particularly important goal of treatment [3]. Therefore, rehabilitation has an important role in care for individuals with SCI [4]. Various kinds of orthoses are used for rehabilitation of SCI. Positive effects on the kinematics and temporal-spatial parameters of gait have been demonstrated in those with SCI following a period of gait training with orthoses [5, 6]. Leg orthoses are regarded as an important rehabilitation tool, but their contribution to improving ADL after SCI remains unclear [7, 8].

The aim of this retrospective observational cohort study was to determine the impact of leg orthotic therapy for improving ADL after SCI, using propensity score analysis of data in the Japan Rehabilitation Database.

Methods

The study was approved by the Institutional Review Board of the Japanese Association of Rehabilitation Medicine. The requirement for informed patient consent was waived, as retracing is not possible because of the anonymous nature of the data.

Data source and patient selection

The study included data obtained from the Japan Rehabilitation Database on individuals with a diagnosis of SCI who were discharged between April 2015 and December 2016 to a ward in any of eight participating hospitals for assistance with ADL after the acute treatment phase. The SCI patients included were those were in the rehabilitation phase and had transferred to a hospital after acute treatment and for whom complete data were available for cervical injury, American Spinal Injury Association (ASIA) impairment scale score [9], and manual muscle testing (MMT) scores [10].

The Japan Rehabilitation Database was developed with financial support from the Ministry of Health, Labour and Welfare of Japan [11]. Detailed clinical data were collected for inpatients undergoing rehabilitation at participating hospitals. The database comprises continuous samples from only voluntary hospitals, not random hospitals, and includes unique identifiers for the following data. Individual data were collected on admission by rehabilitation staff. Data for variables not collected on admission were collected at discharge. All data submitted to the Japan Association of Rehabilitation Database were extracted and sent to the study researchers. All personal data entered into the database are coded and any personally identifiable information is deleted.

Rehabilitation programs

Rehabilitation programs for SCI focused on gait and exercise related to ADL [12, 13]. The gym exercise program typically comprised 40–80 min of physical therapy daily for 5–7 days per week. The program included range of motion, muscle strengthening, and basic motion exercises (e.g., rolling over, standing up, and walking), typically with use of orthoses.

Leg orthoses

In Japan, leg orthoses can be prescribed by a rehabilitation physician to facilitate physical rehabilitation in individuals with SCI and are covered by health insurance. It was assumed that leg orthoses were prescribed according to the preference of the attending physicians and hospital policy. Ankle-foot orthoses are generally used in SCI rehabilitation, while knee-ankle-foot orthoses tend to be used in patients with more severe SCI. In Japan, orthotics are generally prescribed as training devices during hospitalization but as assistive devices for daily mobility after discharge.

Study variables and outcomes

Admission data were evaluated on admission. Time-dependent variables (total amount of rehabilitation, presence of orthotics prescription, etc.) and outcome data were evaluated on discharge. The extracted data were age and sex; injury type; surgery; comorbidities (cerebrovascular disease, hypertension, diabetes, chronic kidney disease); performing the surgery; ASIA impairment scale score; MMT scores (range 0 [no palpable muscle contraction] to 5 [normal]); days from SCI to admission for rehabilitation; days from admission to starting rehabilitation; duration of hospital stay; total amount of rehabilitation provided by physical therapists (PTs) or occupational therapists (OTs) per day; Functional Independence Measure® (FIM) score (range 18 [totally dependent] to 126 [totally independent]) [14]; type of hospital; and whether leg orthoses were prescribed during hospitalization. FIM was measured on hospital admission and on discharge. MMT scores were acquired according to standardized ASIA/ISNCSCI testing procedures, and were summarized as upper extremity motor scores (UEMS) [15] and lower extremity motor scores (LEMS) [15, 16]. UEMS and LEMS range from 0 to 50 (maximum score of 5 for each of 5 key muscles of each leg).

The primary outcome was motor FIM effectiveness, calculated as (discharge motor FIM score— admission motor FIM score)/(total motor FIM score [maximum 91]—admission motor FIM score) [14]. Motor FIM effectiveness corrects the ceiling effect that is present in motor FIM gain (change in FIM during hospitalization). The secondary outcome was discharge motor FIM.

Statistical analysis

Outcomes were compared between the leg orthotic group (those prescribed leg orthoses during hospitalization) and the non-leg orthotic group, using inverse inverse probability weighting (IPW) to construct a marginal structural model. The IPW method uses weights based on propensity scores to create a synthetic sample in which measured covariates follow a distribution that is independent of treatment assignment [17]. The propensity score is the probability of treatment assignment and is conditional on observed characteristics. When this score is applied, the distribution of observed covariates should be similar between treated and untreated individuals. IPW attempts to simulate a randomized experimental situation in which both groups are comparable with regard to observed prognostic factors.

Quasi-experimental design using a marginal structural model

In a quasi-experimental design, the probability (propensity score) that an individual would have been treated is used to adjust the estimated treatment effect [18]. Using the propensity score to construct a marginal structural model by IPW requires (1) propensity score estimation, (2) weighting, (3) balance checking, and (4) impact estimation.

The propensity score for allocation to the leg orthotic group was estimated using a multivariable logistic regression model [19] including as indicators the admission data listed in the “Study variables and outcomes” section. These variables included confounders and variables related to exposure and outcome. The c-statistic for goodness of fit was 0.76 in the propensity score model. Within-hospital correction was done using generalized estimation equations, which are commonly used to analyze clustered data, while correcting for the effects of confounders by clustering [20].

The leg orthotic group was weighted by 1/(propensity score) and the non-leg orthotic group was weighted by 1/(1−propensity score). We used descriptive statistics for all individuals, with or without IPW adjustment, to check covariate balance and to estimate effect. The leg orthotic group and non-leg orthotic group were compared using Student’s t-test and the chi-squared test.

Continuous variables that were normally distributed are expressed as the mean (standard deviation). Ordinal and continuous variables that were not normally distributed are expressed as the median [interquartile range]. We conducted all statistical analyses using IBM SPSS Statistics for Windows version 22.0 software (IBM Corp., Armonk, NY). A p-value of less than 0.05 was considered statistically significant.

Results

Of 505 individuals with SCI, 212 were excluded because of incomplete data on leg orthotic prescription, leaving 293 for analysis (Fig. 1). Table 1 shows the demographics for all individuals in the leg orthotic group and in the non-leg orthotic group, before and after IPW adjustment. Seventy-six of the 293 individuals (26%) were prescribed leg orthoses and 217 (74%) were not. Mean age was 53.9 years, and 84% were male. SCI was traumatic in 82% of individuals, and 58% had undergone orthopedic surgery. In terms of comorbidities, 5% had cerebrovascular disease, 19% had hypertension, 11% had diabetes, and 5% had chronic kidney disease. On the ASIA impairment scale, 26% were grade A, 12% were grade B, 23% were grade C, 33% were grade D, 2% were grade E, and the score was unknown in 4% of cases. Admission UEMS was 19.4 and admission LEMS was 18.8. The median interval between SCI and admission for rehabilitation was 58 days, and the median interval between admission for and starting rehabilitation was 1 day. Mean total daily amount of rehabilitation provided by PTs and OTs was 92 min. Mean hospital stay was 146 days. On admission, mean motor FIM score was 32.4. In the unadjusted results, individuals who were prescribed leg orthoses were younger than those who were not. After adjustment by IPW, there was no statistical difference between the two groups in these characteristics.

Fig. 1
figure 1

Flow chart of patient’s selection

Table 1 Demographic and clinical characteristics of individuals with spinal cord injury

Table 2 shows the outcomes in the leg orthotic and non-leg orthotic groups. Individuals who were prescribed leg orthoses had significantly higher motor FIM effectiveness scores than those who were not (0.54 vs. 0.35, p < 0.01). The same result is obtained before and after adjustment by IPW (0.45 vs. 0.36, p = 0.02). Discharge motor FIM was significantly higher in individuals who were prescribed leg orthoses than in those who were not (discharge motor FIM: 64.5 vs. 52.2, p < 0.01). The same result is obtained after adjustment by IPW (discharge motor FIM: 58.9 vs. 53.5, p = 0.02).

Table 2 Outcomes in the leg orthotic and non-leg orthotic groups

Discussion

In this study, a large database of rehabilitation inpatients was used to examine the impact of leg orthotic therapy for improving motor ability in activities of dialing living and rehabilitation outcomes after SCI. After IPW adjustment, prescription of leg orthoses was significantly associated with increased motor FIM effectiveness and discharge motor FIM.

Many studies have investigated the effectiveness of leg orthoses in individuals with stroke, cerebral palsy, or multiple sclerosis [21,22,23,24], although none have evaluated improvement in motor ability. One study involving persons with thoracolumbar SCI reported that an individualized reciprocating gait orthosis significantly improved their ability for ADL and locomotion [25]. In the present study, we investigated the impact of conventional leg orthoses (i.e., not specialized orthoses such as those with electrical stimulation) used in daily clinical practice in individuals with SCI.

Previous studies have found leg orthotic therapy to be an important therapeutic intervention that often leads to better improvement of motor ability after SCI [12, 13]. However, these studies may have been confounded by clustering. We assumed that leg orthoses were prescribed according to the preference of attending physicians. These preferences would cause within-center correlations with the response variables. Such correlation or clustering occurs when outcomes within a center tend to be more similar to each other than to outcomes at other centers. Here, we used generalized estimation equations to adjust for data clustering.

Our study had a large sample size, which is required to achieve statistically reliable results in propensity score analysis. This was possible because we used the Japan Rehabilitation Database, which contains a large proportion of the individuals with SCI who undergo rehabilitation in Japan. This large-scale observational study using propensity score analysis was a practical alternative to a randomized controlled trial, which ethically cannot be conducted because conventional orthoses are commonly used in SCI rehabilitation. However, we believe that a large representative prospective cohort study is feasible and necessary in the future.

There are several mechanisms by which leg orthoses might be associated with favorable rehabilitation outcomes. Leg orthoses improve the biomechanics of standing and walking [26]. This might have allowed individuals to perform rehabilitation exercises more safely (at least for individuals whose do not use a wheelchair as their primary mode of mobility), confidently, and effectively, contributing to the greater capability improvement seen in our study. Another possibility is that they were motivated to perform rehabilitation exercises by prescription of leg orthoses or that leg orthoses made the exercises easier, allowing them to be done without supervision by therapists. Leg orthoses likely made other types of therapy possible.

This study has several limitations. First, selection bias was the major limitation of this study. Prescription of leg orthoses is affected by prior information on patient characteristics and hospital records, so there was a systematic difference between the characteristics of the leg orthotic group and those of the non-leg orthotic group. The non-orthotic group was older and had more comorbidities. This may have created bias on the part of the prescribing physicians that resulted in fewer orthotic prescriptions for that group. Accordingly, these systematic differences must be considered when estimating the influence of leg orthoses on outcomes. Regression adjustment can be used for this purpose, but selection bias generally makes it difficult to analyze causal effects in an observational study. To adjust for selection bias and ensure comparability, we used IPW to adjust for differences in patient characteristics [27]. Second, the database used in this study does not contain detailed information on the type of leg orthoses used, the timing of the prescription, and MMT scores for the hamstrings. Third, we did not have detailed data on within-center practices, so specific qualities and practices that achieved good recovery outcomes could not be identified. Fourth, the Japan Rehabilitation Database comprises samples from only voluntary hospitals, not random hospitals and therefore generalizability of our findings to all individuals with SCI undergoing rehabilitation may be limited. Fifth, we did not have data regarding the details of the rehabilitation patients received.

In conclusion, our data suggest that leg orthotic therapy is associated with improved motor ability for ADL in individuals with SCI undergoing rehabilitation on an inpatient basis. Leg orthoses might be a useful strategy for improving the quality of rehabilitation in these individuals.