Impact of secondary health conditions on the quality of life and wellbeing of Canadians living with spinal cord injury: A comparison of preference-weighted index scores derived from generic standardized instruments

Abstract

Context/objectives

To compare the assessment of the impact of secondary health conditions (SHCs) on the quality of life and wellbeing of Canadians living with spinal cord injury (SCI) using four preference-based outcome measures.

Design

Secondary analysis of data from a cross-sectional, online survey.

Setting

Community.

Participants

Community-dwelling adults (n = 364) living with traumatic or non-traumatic spinal cord injury at least one year post-injury (70% at least 10 years post-injury).

Outcome measures

A modified version of the Spinal Cord Injury Secondary Conditions Scale (SCI-SCS); three health-related instruments (EQ-5D-5L, Health Utilities Index Mark 3 (HUI3), and the Assessment of Quality of Life 8-dimension questionnaire (AQoL-8D)) and a capability wellbeing instrument (ICEpop CAPability measure for Adults (ICECAP-A)).

Results

Across unadjusted and controlled analyses (i.e. controlling for associations between index scores and sociodemographic and impairment characteristics), trends were observed that identified lower levels of quality of life/wellbeing with higher problem ratings for each of the SHCs. Despite the trends, there was considerable variation in mean index scores across instruments, with HUI3 scores the lowest of the health-related instruments and ICECAP-A scores the highest overall. Respiratory problems, depression/mood problems, pressure sores, and autonomic dysreflexia were associated with the lowest levels of quality of life and wellbeing.

Conclusions

Higher problem ratings for SHCs are negatively associated with scores derived from preference-based quality of life and wellbeing instruments. Variation in index scores across instruments – including across the health-related instruments alone – highlights the critical importance of assessing the relative merits of preference-based instruments when using (or considering using) these instruments/estimates in comparative effectiveness research and economic evaluation.

Introduction

Spinal cord injury (SCI) impacts various aspects of one’s overall health, including physical and mental wellbeing, social life, and ability to participate in society (1). Furthermore, individuals living with SCI are prone to experience secondary health conditions (SHCs) across their lifespan, which have been defined as “physical or psychological health conditions that are influenced directly or indirectly by the presence of a disability or underlying physical impairment” (2). People living with SCI are at risk of developing SHCs that are caused either by the impairment (e.g. neurogenic bladder and/or spasticity) or are indirectly associated with living with a long-standing physical condition (e.g. an elevated risk of developing conditions such as obesity, diabetes, and/or pressure sores) (2). A 2013 scoping review, comprising 92 studies, identified the most common secondary conditions or symptoms to include pain, bowel and bladder dysfunction, muscle spasms, fatigue, oesophageal symptoms, and osteoporosis, often with a reported prevalence greater than 50% (3). Rates of contact with the health care system are greater for people living with SCI when compared with the general population (4), and recent evidence has shown the additional burden to the health care system associated with SHCs (5–7). For example, using a multimorbidity index (MMI) comprising 30 SHCs, Hong and colleagues found that the MMI score was associated with increased health care use in persons with non-traumatic SCI (6). Mitton and colleagues have attempted to estimate the health care costs associated with one additional SHC, reporting estimates of $428 for those reporting receiving needed care and $1240 for those reporting not receiving needed care (7).

A further burden associated with SHCs is a reduction in quality of life (1, 8–10). To better understand these impacts, there is a need for methods that facilitate the identification and quantification of different dimensions of health-related quality of life (HRQoL), as well as for measuring changes in these dimensions over time (11). The quantification of these impacts is important for a number of reasons, particularly in the context of conducting comparative effectiveness research and economic evaluation. Broadly, comparative effectiveness research and economic evaluation aim to assess the benefits and harms associated with alternative approaches to providing treatment, as well as assisting patients, clinicians, decision makers, and other stakeholders in making informed decisions across competing demands (12, 13).

Preference-based HRQoL instruments are widely used to measure and value health states in economic evaluation (11, 13, 14). Such instruments comprise two distinct components: a descriptive classification system and a valuation system. The descriptive classification system enables respondents to describe their HRQoL as one of a finite number of health states through a set of standardized questions and response options. For example, an instrument with five questions, each with three levels of response, would have 243 (3⁵) distinct health state descriptions. The valuation system refers to the procedure for assigning scores to each instrument-defined health state. Typically, these scores represent societal preferences in terms of health utility values (also known as ‘index scores’ or ‘preference weights’), reflecting the relative value that society places on living in each health state. The standardized questionnaires allow for obtaining health utility values interpreted on a scale with a maximum upper anchor of 1, representing ‘full health’, the 0 (zero) point representing a health state equivalent to dead, and negative scores interpreted as reflecting health states worse than dead. The generic nature of these questionnaires means the questions and response options should be suitable for use in all clinical contexts as well as general population samples. Although such measures are designed to allow for comparisons within and across health conditions, evidence from multiple clinical contexts, including SCI, has shown that different instruments provide different scores (15–21). The variability in health utility values obtained from these measures when administered to the same individuals raises concerns about comparability and, consequently, the decision about which measure is most suitable to a particular context (14, 22). Despite the variations, preference-based HRQoL instruments are often a key feature of evaluative research alongside intervention-based studies, primarily as the means to estimate the ‘quality-adjustment’ component of quality-adjusted life years (QALYs) in economic evaluation (14). The QALY – a measure of health outcome that reflects both quantity and quality of life – is the recommended expression of health effects for multiple national health technology assessment organizations (23–25).

Outcome measurement in economic evaluation can go beyond the health-related focus of many preference-based instruments, such as the widely-used health status instruments developed by the EuroQol Group (EQ-5D-3L and EQ-5D-5L) (26). An alternative approach stems from Amartya Sen’s work on capabilities, which distinguishes between functionings and capabilities (27). Functionings represent a range of what an individual achieves or manages to be. Capabilities, however, consist of what an individual ‘is able to do’ or ‘can do’ (27). Efforts to quantify self-reported capabilities include the Investigating Choice Experiments for the Preferences of Older People (ICEPOP) program, from which the ICEpop CAPability (ICECAP) measures emerged (28, 29). A range of ICECAP measures have been developed, such as the ICECAP-O (for use with older people) (30), ICECAP-A (for the general adult population) (31), and ICECAP-SCM a supportive care measure for use at the end of life (32). ICECAP measures are also preference-based, with scores on a 0–1 scale, although the interpretation of such scores is different to the health-related measures: a score of 0 (zero) represents the absence of capability, whereas 1 represents full capability (28). Scores from ICECAP instruments cannot be used to estimate QALYs, which means that different decision rules (i.e. non-QALY-based decision rules) are required when using ICECAP instruments to inform decision making (33, 34).

The objective of this study is to describe the impact of SHCs on the quality of life and wellbeing of Canadians living with SCI using four different approaches for estimating preference-weighted index scores (three health-related instruments and one capability wellbeing instrument). The opportunity to explore this research question came from a cross-sectional survey, administered in 2013, that was conducted as part of a mixed-methods evaluation of different preference-based instruments in the context of SCI (17, 35).

Materials and methods

Data source

This study is a secondary analysis of data obtained from a self-completed, forced choice, online cross-sectional survey. A full description of the primary study design and the online survey can be found elsewhere (17). Briefly, study participants were recruited from a list of individuals who participated in a previous Canada-wide SCI research project, the SCI Community Survey (SCICS) (36). Individuals received an email invitation for the online survey if they were 19 years of age or older, were able to communicate in English, had an SCI (regardless of the cause or severity of injury), and consented to be contacted for further participation in SCI research. The survey study was approved by the University of British Columbia Behavioural Research Ethics Board (H12-01138) and Vancouver Coastal Health Authority (#V12-01138); this secondary analysis received approval from the Office of Research Ethics at Simon Fraser University (#30001398).

Instruments

The survey comprised five sections: (i) demographics, (ii) SCI classifications and characteristics, (iii) secondary health complications and conditions, (iv) quality of life and wellbeing, and (v) SCI-specific functioning in activities of daily living. To assess SHCs, a modified version of the Spinal Cord Injury Secondary Conditions Scale (SCI-SCS) was used (37). The modified version (herein, mSCI-SCS) included nine of 16 items from the SCI-SCS and an additional seven items recommended by collaborators from the Praxis Spinal Cord Institute. The 16 SHCs were: muscle spasms (spasticity); depression/mood problems; pressure sore(s); bladder dysfunction; trouble sleeping; joint and muscle pain; neuropathic pain; sexual dysfunction; autonomic dysreflexia; bowel dysfunction; fatigue; fractures; urinary tract infection; respiratory problems; osteoarthritis/degenerative arthritis; cerebrovascular disease, stroke, transient ischemic attack. As in the original SCI-SCS, all SHCs were assessed using a four-level ordinal scale (0 = not experienced in the last three months/insignificant problem, 1 = mild/infrequent problem, 2 = moderate/occasional problem, 3 = significant/chronic problem). Total scores for the mSCI-SCS ranged from zero (‘level 0’ for all SHCs) to 48 (‘level 3’ for all SHCs).

Survey instruments of relevance for the purposes of this secondary analysis include three preference-based HRQoL instruments: the EQ-5D-5L (38), the Health Utilities Index (HUI) (39), and the Assessment of Quality of Life 8-dimension questionnaire (AQoL-8D) (40). Table 1 describes the key properties of these instruments, highlighting some of the differences between them (e.g. the dimensions, the number of questions, and the methods used to value the respective health states), and provides details of the scoring algorithms. The survey also included the ICECAP-A (31, 43). The ICECAP-A includes five dimensions (stability [being able to feel settled and secure], attachment [being able to have love, friendship and support], autonomy [being able to be independent], achievement [being able to achieve and progress], and enjoyment [being able to have enjoyment and pleasure]), each with four levels of response. The United Kingdom (UK) scoring algorithm was used to derive ICECAP-A index scores, which has a scoring ranging from 0.149 to 1.000 (44). Section 4 of the survey also included the International SCI Quality of Life Basic Data Set (SCI-QoL-BDS), which comprises ratings of satisfaction with general quality of life, satisfaction with physical health, and satisfaction with psychological health, all measured on a scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied) (45).

Table 1.

Key properties of the three preference-based health-related quality of life instruments included in the analysis.

Instrument	Dimensions	Number of items used to derive an index score	Number of unique health states	Valuation study participantsa	Elicitation method	Scoring range for the value set
AQoL-8D	8: Independent Living, Mental Health, Coping, Relationships, Pain, Senses, Happiness, Self-worth	35 questions, with between 4 and 6 levels of response	2.4 × ${10}^{23}$	Mental health patients (n = 323) and a random sample of the general public (n = 347) in Australia (40)	TTO and VAS	0.090 to 1.000
EQ-5D-5L	5: Mobility, Self-care, Usual Activities, Pain/Discomfort, Anxiety/Depression	5 questions, each with 5 levels of response	3125 (55)	A random sample of the adult population in four Canadian cities (n = 1073) (41)	TTO	−0.148 to 0.949
HUI3	8: Vision, Hearing, Speech, Ambulation, Dexterity, Emotion, Cognition, Pain	12 questions, with between 4 and 6 levels of responseb	972,000 (65 × 53)	A random sample of the adult population in Hamilton, Canada (n = 256) (42)c	SG and VAS	−0.359 to 1.000

SG, standard gamble; TTO, time-trade-off; VAS, visual analogue scale

^aParticipants included in the derivation of the respective scoring algorithms used in the analysis

^bThe HUI self-administered questionnaire consists of 15 questions (12 of which are used derive HUI3 index scores)

^cThe HUI3 valuation study comprised 504 completed interviews; 256 in a modeling survey and 248 in a validation survey. Feeny and colleagues stated that the estimated HUI3 scoring function, “is based on preference scores obtained from a representative random sample of a general population” (42)

Statistical analysis

Descriptive statistics were computed for the four preference-weighted instruments, the mSCI-SCS, and the SCI-QoL-BDS. Mean index scores (EQ-5D-5L, HUI3, AQoL-8D, and ICECAP-A) were generated, by mSCI-SCS response level, for the 16 SHCs, providing unadjusted (‘raw’) estimates. The ‘primary analysis’ comprised an examination of potential confounders, with the aim of estimating mean index scores after controlling for statistically significant associations between index scores and a range of sociodemographic and impairment characteristics (sex, neurological level, American Spinal Injury Association [ASIA] classification, age, and time since injury). In the examination of potential confounders, the response variable (the respective index score) was continuous. Linear regression was used when the explanatory variable was continuous. Student’s t-tests (for two-level variables) and analysis of variance (for variables with more than two levels) were used for categorical explanatory variables. For each instrument-specific analysis, any explanatory variable that had a statistically significant (P < 0.05) association with the index scores was controlled for using multiple regression. The results from the multiple regression models were then used to derive estimated marginal means, by mSCI-SCS response level. As a sensitivity analysis, two additional variables (living arrangement and household annual income) were explored as potential confounders. The final component of the analysis comprised an instrument-specific ranking of SHCs, based on the impact on the respective index scores. For each SHC, mSCI-SCS response levels were dichotomized to ‘absent’ (response level 0) and ‘present’ (response levels 1, 2, or 3), with the rankings based on the respective index scores in the ‘present’ category. Rankings were explored using the unadjusted and controlled index score estimates. To determine the controlled rankings, analysis of variance models were used, controlling for the same potential confounders as the primary analysis. All analyses were performed using R Statistical Software (v4.2.3) (46).

Results

Of 609 eligible individuals who received the email invitation, 364 (60%) completed the online survey. Demographics, clinical characteristics, and summary scores for the study participants are provided in Table 2. The majority (62.9%) of the sample was male, with 90.1% self-identifying as Caucasian. Mean time since injury was 19.5 years, with 70.1% of participants being at least 10 years post-injury. Approximately three quarters (74.7%) of the sample had experienced a traumatic spinal cord injury. The mean total score for the mSCI-SCS was 18.4, with scores ranging from 2 to 40. On average, satisfaction with psychological health (6.38) was higher than satisfaction with physical health (5.48), as measured by the SCI-QoL-BDS. Across the preference-based instruments, HUI3 (0.248) and ICECAP-A (0.761) had the lowest and highest mean index scores, respectively. While the highest ICECAP-A score was 1.000, there were no ‘full health’ responses for the EQ-5D-5L, HUI3 or AQoL-8D. Table 3 reports the frequency of the 16 SHCs across the response levels of the mSCI-SCS, with joint and muscle pain (90.1%), muscle spasms (spasticity) (85.2%), and fatigue (83.0%) the most common, based on the proportions of participants reporting at least a mild or infrequent problem (i.e. response level 1, 2, or 3).

Table 2.

Demographics, clinical characteristics and summary scores for the study participants. Values are numbers (percentages) unless stated otherwise.^a

Characteristic	n = 364
Age (mean (SD))	50.4 (13.2)
Sex
Female	135 (37.1)
Male	229 (62.9)
Ethnicity
Caucasian	328 (90.1)
Other	33 (9.1)
Relationship status
Married or Common Law	186 (51.1)
Single, never married	105 (28.8)
Divorced, separated, or widowed	69 (19.0)
Education level
High school diploma/GED and below	102 (28.0)
Diploma/Certificate	140 (38.5)
Bachelor’s degree and above	119 (32.7)
Household annual income
<$20,000	56 (15.4)
$20,000–39,999	49 (13.5)
$40,000–59,999	57 (15.7)
$60,000–79,999	47 (12.9)
$80,000–99,999	47 (12.9)
≥ $100,000	53 (14.6)
Neurological level
Tetraplegia	175 (48.1)
Paraplegia	189 (51.9)
ASIA neurological classification of SCIb
A	115 (31.6)
B	31 (8.5)
C	119 (32.7)
D	69 (19.0)
E	8 (2.2)
Uncertain (A or C) or unclassifiable	22 (6.0)
Time (years) since injury (mean (SD))	19.5 (13.8)
Time since injury
1–10 years	109 (29.9)
≥10 years	255 (70.1)
Cause of injury
Traumatic	272 (74.7)
Non-traumatic	92 (25.3)
Preference-based instruments (mean (SD); median (IQR))
AQoL-8Dc	0.573 (0.20); 0.561 (0.31)
EQ-5D-5L	0.492 (0.20); 0.496 (0.31)
HUI3	0.248 (0.22); 0.258 (0.29)
ICECAP-A	0.761 (0.18); 0.828 (0.26)
mSCI-SCS (mean (SD); median (IQR))	18.36 (7.73); 19.00 (11.00)
SCI QoL Basic Data Set (mean (SD); median (IQR))
Satisfaction with general quality of life	6.23 (2.3); 7.00 (3.00)
Satisfaction with physical health	5.48 (2.6); 6.00 (4.00)
Satisfaction with psychological health	6.38 (2.6); 7.00 (3.00)

ASIA, American Spinal Injury Association; GED, General Educational Development; IQR, interquartile range; mSCI-SCS, modified Spinal Cord Injury Secondary Conditions Scale; QoL, quality of life; SCI, spinal cord injury; SD, standard deviation

^aNumbers for some characteristics do not always sum to 364 due to ‘prefer not to answer’ or ‘do not know’ responses

^bASIA neurological classifications were determined using an algorithm developed by the SCI Community Survey research team (36)

^cOne participant was able to complete the survey without completing all items for the AQoL-8D

Table 3.

Frequency of the 16 secondary health conditions across the response levels of the mSCI-SCS (with a timeframe of “the last 3 months”).

Secondary health condition	0 = Not experienced / insignificant problem		1 = Mild or infrequent problem		2 = Moderate or occasional problem		3 = Significant or chronic problem
	n	%	n	%	n	%	n	%
Muscle spasms (spasticity)	54	14.8	111	30.5	125	34.3	74	20.3
Depression/mood problems	147	40.4	110	30.2	78	21.4	29	8.0
Pressure sore(s)	235	64.6	61	16.8	36	9.9	32	8.8
Bladder dysfunction	88	24.2	86	23.6	76	20.9	114	31.3
Trouble sleeping	70	19.2	107	29.4	113	31.0	74	20.3
Joint and muscle pain	36	9.9	85	23.4	125	34.3	118	32.4
Neuropathic pain	97	26.7	79	21.7	58	15.9	130	35.7
Sexual dysfunction	78	21.4	46	12.6	69	19.0	171	47.0
Autonomic dysreflexia	172	47.3	90	24.7	71	19.5	31	8.5
Bowel dysfunction	73	20.1	120	33.0	97	26.7	74	20.3
Fatigue	62	17.0	114	31.3	104	28.6	84	23.1
Fractures	333	91.5	15	4.1	9	2.5	7	1.9
Urinary tract infection	152	41.8	69	19.0	77	21.2	66	18.1
Respiratory problems	277	76.1	51	14.0	22	6.0	14	3.9
Osteoarthritis/degenerative arthritis	172	47.3	89	24.5	48	13.2	55	15.1
Cerebrovascular disease, stroke, transient ischemic attack	355	97.5	4	1.1	3	0.8	2	0.6

mSCI-SCS, modified Spinal Cord Injury Secondary Conditions Scale; n, number.

Unadjusted mean index scores for the 16 SHCs, by mSCI-SCS response level, are reported in Appendix 1 (EQ-5D-5L and HUI3) and Appendix 2 (AQoL-8D and ICECAP-A). With some exceptions, a consistent pattern was observed within and across instruments when looking at the direction of the change in index scores after moving to the adjacent problem rating level. More specifically, of the 192 one-level increases in problem ratings (i.e. moving from 0 to 1, from 1 to 2, or 2 to 3, for each SHC, across all four instruments; 3*16*4 = 192), a lower mean index score was observed on 164 (85.4%) occasions.

Results from the examination of potential confounders are provided in Appendix 3. ASIA classification had a statistically significant impact on EQ-5D-5L, HUI3, and AQoL-8D scores; neurological level had a significant impact on EQ-5D-5L and HUI3 scores; while time since injury impacted AQoL-8D scores only. No variable was identified as a potential confounder for the ICECAP-A, meaning the ICECAP-A index scores in Appendix 2 are appropriate for the purposes of drawing comparisons with ‘controlled’ EQ-5D-5L, HUI3, and AQoL-8D scores. Table 4 reports the estimated marginal means for the EQ-5D-5L, HUI3, and AQoL-8D across 14 SHCs after controlling for the identified (instrument-specific) confounders. Two SHCs (fractures and cerebrovascular disease, stroke, transient ischemic attack) were excluded from the primary analysis because at least 90% of participants reported a level 0 response (‘not experienced/insignificant problem’) (Table 3). Of the one-level increases in problem ratings for the 14 SHCs included in the primary analysis, a lower mean index score was observed on 152 of 168 (90.5%) occasions, with six directional inconsistencies for the ICECAP-A, five for the HUI3, three for the AQoL-8D, and two for the EQ-5D-5L. Eight of the 16 directional inconsistencies were in three SHCs: pressure sores (three times), bladder dysfunction (three times), and automatic dysreflexia (twice). There were no discernable patterns to suggest that directional inconsistencies were more common for a particular one-level increases (e.g. from 0 to 1).

Table 4.

Results from the primary analysis – estimated marginal means, by secondary health condition and response level on the mSCI-SCS, for the EQ-5D-5L, HUI3 and AQoL-8D after consideration of potential confounders.^a

Secondary health condition	EQ-5D-5L				HUI3				AQoL-8D
	0	1	2	3	0	1	2	3	0	1	2	3
Muscle spasms (spasticity)	0.542	0.542	0.484	0.402	0.310	0.290	0.245	0.166	0.595	0.625	0.552	0.473
Depression/mood problems	0.578	0.509	0.401	0.376	0.335	0.270	0.165	0.125	0.699	0.565	0.428	0.392
Pressure sore(s)	0.525	0.465	0.469	0.414	0.288	0.204	0.225	0.181	0.597	0.540	0.516	0.502
Bladder dysfunction	0.564	0.540	0.473	0.439	0.321	0.307	0.206	0.208	0.635	0.586	0.532	0.538
Trouble sleeping	0.627	0.535	0.488	0.373	0.363	0.301	0.252	0.132	0.732	0.634	0.529	0.434
Joint and muscle pain	0.609	0.548	0.529	0.406	0.319	0.325	0.286	0.168	0.714	0.635	0.602	0.455
Neuropathic pain	0.599	0.546	0.482	0.404	0.324	0.308	0.272	0.174	0.686	0.597	0.563	0.470
Sexual dysfunction	0.572	0.549	0.516	0.439	0.344	0.326	0.270	0.188	0.668	0.657	0.570	0.498
Autonomic dysreflexia	0.545	0.497	0.434	0.417	0.309	0.241	0.181	0.208	0.634	0.560	0.470	0.485
Bowel dysfunction	0.583	0.523	0.460	0.413	0.332	0.291	0.221	0.163	0.681	0.596	0.506	0.482
Fatigue	0.629	0.548	0.486	0.364	0.390	0.328	0.230	0.111	0.757	0.649	0.532	0.391
Urinary tract infection	0.533	0.504	0.498	0.426	0.296	0.241	0.251	0.197	0.602	0.570	0.569	0.508
Respiratory problems	0.526	0.462	0.464	0.285	0.283	0.231	0.218	0.029	0.598	0.552	0.439	0.426
Osteoarthritis/degenerative arthritis	0.531	0.522	0.468	0.382	0.292	0.263	0.239	0.149	0.620	0.581	0.503	0.449

mSCI-SCS, modified Spinal Cord Injury Secondary Conditions Scale

^aResponse levels for the mSCI-SCS, all with a timeframe of “the last 3 months”, are as follows: 0 = Not experienced / insignificant problem, 1 = Mild or infrequent problem, 2 = Moderate or occasional problem, and 3 = Significant or chronic problem. Two secondary health conditions (fractures and cerebrovascular disease, stroke, transient ischemic attack) were excluded from the examination of confounders because at least 90% of participants reported a level 0 response (‘not experienced/insignificant problem’)

Based on level 3 estimates for the mSCI-SCS (‘significant or chronic problem’), respiratory problems, fatigue, depression/mood problems, and trouble sleeping were associated with the largest quality of life decrements across the three health-related instruments. Three of these four SHCs (respiratory problems, fatigue, and depression/mood problems) were also among the four largest decrements in ICECAP-A index scores when considering the same 14 SHCs that comprised the primary analysis (see Appendix 2); compared with the results of the health-related instruments, autonomic dysreflexia (0.673) replaced trouble sleeping (0.689) among the four lowest scores when using the ICECAP-A.

Across all instruments, higher ratings of problems for SHCs were associated with lower mean index scores, but the range of values differed considerably. For example, within response level 3 on the mSCI-SCS (significant/chronic problem), the highest HUI3 score (0.208) was lower than the lowest EQ-5D-5L (0.285), AQoL-8D (0.391), and ICECAP-A (0.559) scores. Also, the lowest ICECAP-A score (0.557), was higher than the highest EQ-5D-5L (0.439), HUI3 (0.208), and AQoL-8D (0.538) scores. Fig. 1 further illustrates the differences across instruments, with decreasing linear relationships observed in all four panels, but with significant variation in the location of the mean index scores across the respective 0–1 scoring ranges.

Figure 1.

Jittered scatterplots illustrating the impact of secondary health conditions and associated mSCI-SCS response levels on EQ-5D-5L (top left), HUI3 (top right), AQoL-8D (bottom-left) and ICECAP-A (bottom-right) index scores. For the health-related measures (EQ-5D-5L, HUI3 and AQoL-8D), each marker corresponds to the estimated marginal means reported in Table 4. For the ICECAP-A, the markers correspond to the unadjusted analysis reported in Appendix 2.

mSCI-SCS, modified Spinal Cord Injury Secondary Conditions Scale Response levels for the mSCI-SCS, all with a timeframe of “the last 3 months”, are as follows: 0 = Not experienced / insignificant problem, 1 = Mild or infrequent problem, 2 = Moderate or occasional problem, and 3 = Significant or chronic problem.

In the sensitivity analysis, household annual income had a significant association with the AQoL-8D and ICECAP-A only (see Appendix 3). The re-estimated means for the AQoL-8D and ICECAP-A are reported in Appendix 4. Based on estimates for mSCI-SCS response level 3, the four lowest ICECAP-A index scores were attributed to the same four SHCs identified in the primary analysis. For the AQoL-8D, osteoarthritis/degenerative arthritis (0.451) replaced respiratory problems (0.474) among the four lowest index scores when comparing results from the primary and sensitivity analyses.

Table 5 presents the rankings of 14 SHCs according to their impact on index scores. Based on the controlled rankings, the presence of respiratory problems, depression/mood problems, pressure sores, and autonomic dysreflexia were consistently associated with the lowest index scores (these four SHCs were in the ‘top 5’ rankings for all four instruments). Muscle spasms (spasticity) and joint and muscle pain were the two lowest ranked SHCs across all four instruments. The most notable difference between the unadjusted and controlled rankings was observed for osteoarthritis/degenerative arthritis (11th/5th) and neuropathic pain (13th/8th) when using the HUI3.

Table 5.

Ranking of secondary health conditions with regard to the impact on index scores, by instrument. Rank #1 indicates the largest impact, i.e. the lowest of the 14 mean index scores. For illustration, secondary health conditions are ordered below according to the rankings observed in the controlled EQ-5D-5L analysis.^a

Secondary health condition	EQ-5D-5L		HUI3		AQoL-8D		ICECAP-A
	Controlled	Unadjusted	Controlled	Unadjusted	Controlled	Unadjusted	Unadjusted
Respiratory problems	1	1	1	1	2	2	1
Depression/mood problems	2	4	3	4*	1	1	2
Pressure sore(s)	3	2	2	2	5	8	3
Autonomic dysreflexia	4	3	4	3	3	3	4
Neuropathic pain	5	5	8	13	6	4	5*
Osteoarthritis/degenerative arthritis	6	12	5	11	4	5*	9*
Trouble sleeping	7	7	10*	10	7*	5*	5*
Bowel dysfunction	8*	8*	10*	7	9	9	9*
Fatigue	8*	8*	9	8	7*	5*	7
Urinary tract infection	8*	6	6	4*	11	13	12
Sexual dysfunction	11	10*	7	6	10	10	8
Bladder dysfunction	12	10*	12	9	12	11	11
Joint and muscle pain	13	14	14	14	13	12	14
Muscle spasms (spasticity)	14	13	13	12	14	14	13

^aTwo secondary health conditions (fractures and cerebrovascular disease, stroke, transient ischemic attack) were excluded from the analysis because of the small number of participants reporting problems. Ties are indicated by an asterisk (*)

Discussion

This study explored the impact of SHCs on the index scores of four preference-weighted generic instruments in a sample of individuals living with SCI. The results showed a consistent pattern, where higher problem ratings for SHCs, as measured by the mSCI-SCS, were associated with lower index scores. However, differences were apparent in the extent to which each instrument captured these impacts. Overall, the health-related measures resulted in lower index scores, across all mSCI-SCS response levels, when compared with the ICECAP-A (a capability wellbeing instrument). Among the three health-related measures, the AQoL-8D index scores were most similar to those of the ICECAP-A, with the HUI3 scores being the least similar. These observations are consistent with previous research from a multinational comparison of the ICECAP-A with five health-related measures across seven disease areas (in addition to a ‘disease free’ population) (47). The observation of notable differences between EQ-5D-5L and ICECAP-A scores also confirms previous findings from pairwise comparisons between the ICECAP-A and instruments developed by the EuroQol Group (EQ-5D-5L and EQ-5D-3L) (47–52). Indeed, a consistent conclusion from comparisons of ICECAP-A and EQ-5D instruments is that these are complementary instruments rather than substitutes.

The observed relationship between the self-reported problem ratings for SHCs and HUI3 index scores aligns with the findings from a 2012 study by Craven and colleagues (9). Direct comparisons between our results and the previous findings are hampered to some extent because Craven and colleagues dichotomized their SHC variable to create ‘low impact’ (absent/mild) and ‘high impact’ (moderate/significant) groups. The overall mean HUI3 scores were similar (0.27 in the 2012 study, 0.25 in this study) and only one of 357 people in the 2012 study reported ‘full health’ on the HUI3. In both studies, respiratory problems and depression were among the SHCs with the greatest impact on HUI3 scores for those in the respective ‘most problems’ group (level 3 on the mSCI-SCS in this study, ‘high impact’ in the Craven study). Craven and colleagues concluded by saying, “Although further work is required, the HUI-Mark III data may be a useful tool for calculating quality-adjusted life years, and advocating for additional resources where secondary health conditions have substantial adverse impact on health” (9). Our comparative work adds to the body of evidence highlighting the low index scores (relative to other health-related instruments) that result from using the HUI3 in the context of SCI. As noted in previous research, these low scores are inevitable given the framing of the Ambulation dimension within the HUI descriptive classification system (17, 53). The framing of mobility-related items has also been shown to be the source of poor agreement between AQoL-8D and EQ-5D-5L index scores in a population of individuals receiving assisted ventilation (20). In non-SCI settings, where significant physical impairment is less likely to be reported, the HUI3 has not been shown to be an outlier when comparing differences in the magnitude of index scores generated from health-related preference-based instruments (15, 19).

Craven’s conclusion about the HUI3 (stated above), coupled with our findings, highlights the importance of comparative assessments of preference-based instruments. Viewed independently, it is feasible that any of the health-related instruments could seem acceptable for use in SCI research – Craven and colleagues explicitly state that the HUI3 “may be a useful tool” (9) and Fig. 1 shows the expected pattern, for all instruments, of lower index scores when reporting higher problem levels for SHCs. But what is also clear, and reported across multiple studies, is the systematic differences in index scores across instruments. There is no unequivocally ‘best’ instrument in a particular clinical context, which means that analysts’ judgement is crucial when determining which instrument(s) to use. For example, in the case of SCI and SHCs, examination of content overlap between the dimensions (and response options) of an instrument and the domains of quality of life impacted by a SHC would be an important step.

In the context of the broader quality of life literature (i.e. not limited to preference-based instruments), our findings confirm negative associations observed between quality of life and SHCs (8, 10, 54, 55). For example, Adriaansen and colleagues, in a study exploring SHCs and quality of life in persons living with SCI for at least ten years – a circumstance experienced by more than 70% of the sample in our study – found that four SHCs (musculoskeletal pain, pressure ulcers, problematic spasticity, and constipation) were independently associated with lower quality of life, albeit with relatively weak associations (54).

The frequencies of SHCs in our study are similar to those reported by Strøm and colleagues in a study that explored the prevalence of self-reported physical health conditions in individuals with SCI across 21 countries (56). The most prevalent conditions in the multinational study were pain, muscle spasms (spasticity), and sexual dysfunction. While our study did not explicitly ask about a ‘general pain’ condition, the most prevalent conditions were joint and muscle pain (90.9%) and muscle spasms (spasticity) (85.2%) (Table 3). Our findings and those of Strøm and colleagues showed that respiratory problems were among the least prevalent conditions. As reported in Table 5, respiratory problems were associated with the lowest mean index scores, while joint and muscle pain and muscle spasms (spasticity) were associated with the highest scores. These observations (i.e. that SHCs with relatively low prevalence can have among the largest impacts on quality of life, and vice versa) emphasize the importance of comprehensive assessments of prevalence, quality of life, and wellbeing across a broad range of SHCs. The effects of health conditions post-SCI are complex and enduring (10). The availability of more evidence, quantifying the self-reported impacts of SHCs, provides valuable information for clinicians, decision makers, research funding organizations, and other stakeholders involved in making resource allocation decisions and seeking to positively impact the lives of individuals living with SCI.

Strengths and limitations

Exploring the impact of SHCs on four preference-weighted outcome measures within the same group of respondents is a major strength of this work. This approach provides insight regarding the relative merits of the different instruments, enabling more informed decisions to be made when selecting an instrument for a study, or when deciding which estimates from our paper might be used at a decision point (e.g. as an input in an economic model). The selection of the three health-related instruments was evidence based, with the survey design building on findings from a qualitative study that explored the views of individuals living with SCI towards six established generic, health-related instruments (35). The inclusion of the ICECAP-A allowed us to further highlight the variation in index scores between capability-focused and health-focused instruments. Our findings contribute to the evidence (in SCI and non-SCI contexts) that the ICECAP-A provides information above that garnered from commonly used health-related HRQoL instruments (47, 57).

As with all studies, there are limitations. First, two of the 16 SHCs included in the survey were infrequently reported, resulting in their exclusion from the controlled analysis. This meant the impact of these two SHCs on individuals’ quality of life and wellbeing could not be explored in a comparable way to the other 14 SHCs, neither were the two SHCs included in the ranking exercise. Second, it is important to note that the preference weights used in this study do not all reflect societal values of the same country, with Canadian values used for the EQ-5D-5L and HUI3, UK values for the ICECAP-A, and Australian values for the AQoL-8D; there is no Canadian scoring algorithm for the ICECAP-A or AQoL-8D. Third, caution is needed when generalizing to other SCI samples (e.g. 70.1% of the study participants were at least 10 years post-injury, and 90.1% self-identified as Caucasian). Finally, while not a limitation of this study per se, it is important to acknowledge that evaluative research, such as comparative effectiveness research and economic evaluation (where these instruments are widely-used), is driven by changes in individuals’ health outcomes over time. Data for this secondary analysis was obtained from a cross-sectional study and, therefore, does not provide information about the change in health utility when moving from one state to another (e.g. for someone moving from ‘mild or infrequent’ fatigue to ‘significant or chronic’ fatigue).

Conclusions

While higher problem ratings for SHCs are associated with lower levels of quality of life and wellbeing, variations in index scores are apparent when deriving scores from different preference-weighted outcome measures. The fact that different instruments provide systematically different scores – a ubiquitous finding in comparative research comprising preference-weighted instruments – emphasizes the importance of careful consideration (and the need for further research) when selecting outcome measures that could inform priority setting and resource allocation decisions. The comprehensive and transparent reporting of quality of life estimates, across multiple instruments and self-reported SHC problem levels, provides a wealth of information about the relative importance of SHCs (quantified in terms of quality of life decrements) and a valuable source of data for prioritization activities.

Disclaimer statements

Contributions DGTW was the project lead for the primary survey study. NE and DGTW designed the secondary analysis plan. NE performed data analysis and drafted the manuscript. All authors were involved in the interpretation of results, review of draft manuscripts, and read and approved the final version prior to submission.

Data deposition No open access availability. Individuals interested in gaining access to the survey data should contact the corresponding author.

Funding The secondary analysis reported in this paper was not a funded project. Financial support for the primary survey study was through a grant from the Praxis Spinal Cord Institute (formerly the Rick Hansen Institute) – Rick Hansen Institute Translational Research Program, grant #2012-29: Spinal Cord Injury & Secondary Complications: A Mixed-Methods Evaluation of Preference-Based Instruments. The funding agreement ensured the research team’s independence in designing the study, interpreting the data, and writing and publishing manuscripts. We would also like to thank Health Utilities Inc. for the New Investigator Grant awarded to DGTW during the inception of the primary study.

Conflicts of interest DGTW is a member of the EuroQol Group Association. The opinions expressed in this paper do not necessarily reflect the opinions of the EuroQol Group, or those of the EuroQol Research Foundation. NE, HS, and DGTW report no further conflicts of interest that are directly or indirectly relevant to the content of this article.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/10790268.2024.2391597.