A comparison of the Child Health Utility 9D and the Health Utilities Index for estimating health utilities in pediatric inflammatory bowel disease

Inflammatory Bowel Disease (IBD) is an ongoing childhood condition that can be physically painful and greatly reduces the quality-of-life of affected children. It can be difficult to measure quality-of-life in children, particularly in very young children. Being able to measure quality-of-life for different levels of disease is important to understand the effectiveness of new treatments. Using a generic common measure of quality-of-life enables a comparison of health improvement across different patient populations. This is very useful for decision-makers considering value-for-money when allocating healthcare budgets. The aim was to compare different questionnaires to see which was best able to reflect changes in quality-of-life when IBD worsens in a group of 271 children with IBD. While all questionnaires were able to pick up quality-of-life differences when disease activity changed, the Child Health Utility-9 Dimensions (CHU9D) performed the best. The findings can be used by researchers searching for the right questionnaire to use in studies that look at value-for-money of new treatments for children with IBD.

Ulcerative colitis (UC) and Crohn’s disease (CD), collectively referred to as inflammatory bowel disease (IBD), are a class of chronic gastrointestinal diseases characterized by periods of unpredictable flares of inflammation of the gastrointestinal tract, abdominal pain, diarrhea, fatigue and weight loss, and periods of symptomatic remission [1]. Pediatric IBD is of particular concern because the incidence is increasing [2‐4], growth may be affected [1, 5] and it can have significant quality-of-life impacts [6‐9]. With an increase in costly IBD treatments such as biologics, there is a need for economic evaluations to inform funding. However, measuring preference-based health-related quality-of-life (HRQOL) to generate health state utilities to calculate quality-adjusted life years (QALYs) for use in economic evaluation is challenging in children. In addition to challenges children experience in comprehending abstract concepts, HRQOL attributes featured in adult instruments may not be applicable to children. The validity of applying existing preference-based HRQOL instruments in children has been questioned [10‐13]. Head-to head comparisons in pediatric patient populations are lacking [14, 15]. Further, whether the set of utility weights that underly HRQOL classification systems should be derived from adults valuing pediatric states or directly from children continues to be debated [10, 16].

The Health Utilities Index (HUI) has been used in children and adults [17] and generates utilities using the HUI Mark 2 (HUI2) or Mark 3 (HUI3). The HUI2 has 7 dimensions: Sensation, Mobility, Emotion, Cognition, Self-Care, Pain and Fertility, and the HUI3 has 8 dimensions: Vision, Hearing, Speech, Ambulation, Dexterity, Emotion, Cognition and Pain [17, 18]. The underlying utility weights for HUI2 and HUI3 were developed with adults. While the HUI has been used in numerous patient populations and across several age groups [18], it had not been used in pediatric IBD [7]. The Child Health Utility 9D (CHU9D) was developed in 2009 specifically for children and with children [19, 20]. It features a classification system of nine dimensions relevant to child health: Worried, Sad, Pain, Tired, Annoyed, Schoolwork, Sleep, Daily routine, and Activities [21]. Sets of utility weights obtained from Australian adults or adolescents are available [22, 23].

An earlier study by our team found the CHU9D to be valid and reliable in pediatric CD and UC with moderate correlations observed between CHU9D, HUI2, and HUI3 utilities [24]. A Moderate to strong correlations between the CHU9D, HUI2, HUI3 and the disease-specific IMPACT-III or the generic PedsQL HRQOL measures were observed [25]. Multiple clinical measures may be used to determine disease activity in pediatric IBD. It’s critical to examine and compare utilities associated with different measures to inform health state modeling used in cost-effectiveness analysis. The objective was to directly compare CHU9D using adult and youth tariffs to the HUI2 and HUI3 for determining health state utilities for distinct disease activity levels defined by different clinical scales in children with UC and CD. Such information is vital to assess the discriminant validity of preference-based HRQOL instruments and inform the choice of measures for use in children. This work also aims to derive utilities for health states for cost-effectiveness analysis in pediatric IBD.

The study was approved by the Research Ethics Boards of the Hospital for Sick Children in Toronto, ON (#1000039604), the IWK Health Centre in Halifax, NS (#1015558), and the Janeway Children’s Health and Rehabilitation Centre in St. John’s NL (#13.229). The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. All parent participants provided informed consent and children provided informed assent.

Despite being generic, all HRQOL instruments and tariff sets discriminated moderately well between quiescent, mild, moderate and severe disease activity in CD and UC. Compared to the adult tariff set, CHU9D youth tariff utilities were consistently lower and displayed a greater range across disease activity levels for CD and UC. Similarly, the HUI3 displayed lower utilities and a wider range across disease activity levels in CD and UC compared to the HUI2.

While specific utilities for pediatric CD and UC are lacking, a meta-analysis reported mean utilities of 0.860 with the HUI2 and 0.882 with the HUI3 for digestive system disorders other than IBD, such as liver diseases, gastric ulcer, and other disorders [42]. Combined pediatric chronic diseases had mean utilities of 0.924 using a standard gamble (SG), 0.884 using the HUI2 and 0.834 with the HUI3 [42]. A meta-analysis of eleven adult CD studies determined a mean utility of 0.8403, 95% CI (0.8012, 0.8794) for remission, 0.7533, 95% CI (0.6887, 0.8178) for active disease, 0.8619, 95% CI (0.8016, 0.9223) for mild disease, 0.7318, 95% CI (0.6271 0.8364) for moderate disease and 0.5102, 95% CI (0.3554, 0.6650) for severe disease [43], values comparable to utilities observed in the present analysis. The results suggest that similar to adults, children with IBD can experience a range of utilities across clinically meaningful disease activity levels. A meta-analysis of 15 adult UC studies demonstrated a mean utility of 0.8726, 95% CI (0.8457, 0.8995) for remission, 0.6992, 95% CI (0.5847, 0.8136) for active disease, 0.7834 95% CI (0.7265 0.8403) for mild disease, 0.6969 95% CI (0.3959 0.9978) for moderate disease, and 0.7059 95% CI (0.5065 0.9054) for severe disease [43]. The present study demonstrated comparable results to adults for remission, but lower utilities for more active disease, however active disease utilities fell within the 95% confidence interval reported for adults.

Assigning utilities to health states can be complicated by the existence of multiple clinical measures used to describe disease activity. The wPCDAI has been found to perform better in measuring CD disease activity than other versions of the PCDAI and the Harvey-Bradshaw Index (HBI), developed for use in adult patients [30, 44]. In the present study, correlations between the PGA and clinical measures of disease activity for CD and UC exceeded 0.85, and the ability of the preference-based instruments to distinguish between PGA-defined disease activity levels suggest that when laboratory and other objective clinical data are unavailable, PGA utilities can be used in economic modeling. Dichotomizing disease activity as active or inactive may further facilitate modeling when a variety of scales are used or when data needed for finer disease activity stratification are missing. It’s important to note that multi-item measures of current disease activity may not incorporate how a patient’s prior disease experience or duration of disease influences their health state preferences. For example, long-term healing and achieving stable disease may reduce anxiety, fatigue, and pain which may be observed as improvement in HRQOL over time [45, 46].

A common challenge is that while all instruments aim to capture the construct of HRQOL, they may return different utilities for the same health state. These discrepancies may be due to the different domains, classification systems and underlying weights of each instrument. This raises the issue of comparability and interchangeability of pediatric health utility instruments with each other and with adult instruments. It must also be acknowledged that HRQOL in children differs from adults, and also differs within pediatric age groups from neonate to adolescent [47‐50]. While the HUI2 and HUI3 have long been used in pediatric as well as adult populations, the CHU9D was designed exclusively as a tool for children aged 7 to 17 years [28, 51, 52]. Our previous research reported correlations of 0.62–0.69 between the CHU9D and HUI2 and HUI3, with slightly higher correlations for the youth compared to the adult tariff set [24]. The agreement between CHU9D and HUI2 or HUI3 was greater at higher utilities [24]. That research indicated that the CHU9D Sleep domain, a domain not present in the HUI2 or HUI3, had the lowest domain score when youth tariffs were used but Pain was scored lowest with the adult tariff set. The Pain domain score also ranked lowest for the HUI2 and HUI3. The difference between CHU9D youth and adult tariffs with regard to which domain ranked lowest suggests that youth and adults may place more weight on different attributes [38, 53]. Not surprisingly, the highest correlations between the CHU9D and HUI2 or HUI3 were observed for Pain [24]. Pediatric IBD patients can experience pain and this common domain may have contributed to the ability of CHU9D as well HUI2 and HUI3 to distinguish between levels of disease activity in the present study. This may not be the case for other pediatric conditions. The present study demonstrated that compared to the HUI2, the HUI3 generally demonstrated lower and a wider range of mean and median utilities across activity levels, possibly due to different domains, which may account for its superior ability to distinguish between quiescent and mild disease activity levels. Although differences in utilities between instruments were mostly small, they could impact QALY calculations. A probabilistic analysis that integrates ranges of utilities for a given heath states corresponding to the values observed in this study is recommended.

Although the CHU9D was initially developed with utility weights obtained from adults, an adolescent tariff set was created from a sample of Australian adolescents using a best–worst scaling method [23]. Ratcliffe et al. found that adults placed less weight on mental health impairments (worried, sad, annoyed) and more weight on moderate to severe levels of pain compared to adolescents. Whereas youth tariffs may be more reflective of the adolescent experience and childhood in general compared to adult tariffs, IBD and other chronic pediatric conditions have ages of onset less than ten years of age. Choosing a valuation method that can be used in children to generate underlying weights remains a challenge [14]. At present, guidelines for economic evaluation prefer utilities elicited using a SG or TTO approach [54, 55]. As pressure increases to conduct cost-utility analysis to inform budget allocation for drugs and technologies for children, guidelines will need to reflect newer methods and approaches for eliciting and deriving utilities in children.

The present results can be used to populate health state transition models for cost-utility analysis comparing pediatric IBD treatments, including biologics and less costly biosimilars. Health state transition (Markov) models incorporate the probability of transitioning between health states characterized by different levels of disease activity or severity [56, 57]. The effectiveness of different treatments can be compared by capturing how much time is spent in a given health state (e.g., inactive versus active) over the designated study time horizon. For example, from the present analysis, a CHU9D youth utility of 0.81 (SD 0.17) can be assigned to a disease remission (inactive) health state and 0.69 (SD 0.21) would be assigned to a relapse (active) state in CD [58]. More effective treatments will result in children spending more time in the higher utility inactive state. The better that active disease states can be distinguished from inactive states in terms of utility, the more sensitive an analysis will be to true differences in treatment effectiveness. Thus preference-based HRQOL measures that are better at discriminating between active and inactive disease will be stronger options for conducting cost-effectiveness analysis compared to instruments that are less discriminant.

Including multiple clinical disease activity scales and categorizations, preference-based instruments, and tariff sets in the present study provide researchers with the opportunity to choose utilities that align best with their target study population, preferred health economic protocols, and clinical practice. Further, the present study indicates that in the absence of detailed objective clinical data for the complete scoring of the wPCDAI or PUCAI, a PGA may provide reliable utility estimates across disease activity levels. The similarities in utilities within HRQOL instruments when disease activity is defined by the PGA or wPCDAI in CD and by the PGA or PUCAI in UC suggest that QALY calculations will not be affected and economic evaluations conducted using utilities associated with disease activity described by any of these clinical scales would be comparable. Despite uncertainty arising from variation across instruments and tariff sets, establishing genuine pediatric health utilities in CD and UC addresses an important gap.

While a strength was comparing different pediatric utility instruments within the same groups of patients, the study did not include alternative approaches such as TTO, visual analog scales, or discrete choice options to generate utilities. These approaches can be challenging to administer in children. Another strength was the comparison of utilities between CD and UC populations at the same clinical sites, thereby controlling for variations in practice patterns or standards of care. However, enrollment was limited to three Canadian centres which may limit generalizability. The need to examine sub-groups according to ethnic diversity and gender is growing in importance. The present analysis found that utilities were significantly lower in females with UC as measured by the CHU9D. Such a finding has not been reported previously in IBD except in adult CD patients with utility assessed using the SF-6D [59]. An important limitation was the small samples in some of the disease activity levels. A large proportion of participants in this observational cohort were well managed and in a quiescent state, and few had severe disease activity. This hampered the comparison of utilities between quiescent, mild, moderate and severe activity states. However, when dichotomized into active and inactive disease, there was sufficient statistical power to detect a significant difference in utilities in CD and in UC with all instruments. Another limitation was the skewness of the data. It’s important to consider median utilities and IQRs alongside means and standard deviations to interpret the findings.

In conclusion, the CHU9D calculated with adult and youth tariffs, the HUI2 and the HUI3 discriminated between levels of disease activity experienced by children with CD and UC from quiescent to severe disease activity. These utilities may be valuable for health state transition models assessing the cost-effectiveness of emerging treatments in pediatric IBD.