In this analysis we compared the costs and Quality-Adjusted Life Years (QALYs) of children with cerebral palsy with and without a surveillance program to prevent hip dislocation. The perspective was of the Spanish National Health System,15 with a time horizon of 18 years, coinciding approximately with the follow-up duration of available data. We applied a 3% discount rate to future costs and QALYs.15,16 The paper follows the CHEERS Statement for economic evaluations.17

This is, to our knowledge, the first cost-effectiveness evaluation of a surveillance program to prevent hip dislocation in children with cerebral palsy. The analysis was thus based on a de novo decision analytical model. Similar to previous cost-effectiveness analyses of related interventions, such as screening of developmental dysplasia of the hip,18–20 the model took the form of a decision tree (Fig. 1). A decision tree is appropriate in this case because the data did not indicate a complex pattern of recurring-remitting health conditions. However, to account for the timing of preventive surgery, dislocation or salvage surgery, costs and QALYs associated with each pathway were computed on a yearly basis and summed up for the duration of the time horizon with appropriate discounting.

Figure 1.

Decision tree. Decision node represented by squares and chance node by circles.

(0.21MB).

Based on this model, mean cost and mean QALYs and the incremental cost-utility ratio (ICUR) under both strategies were computed. The ICUR represents the additional cost required to achieve one additional QALY,16 which is then compared with the decision makers’ willingness to pay threshold in order to draw conclusions about the cost-effectiveness of the intervention. In Spain a cost-effectiveness threshold of 20,000-25,000€ per QALY has been recommended.21

• 1)

  Effectiveness of the surveillance program

A systematic review was performed to identify studies that estimated the effectiveness of a screening program to prevent hip dislocation in children/adolescents (0 to 18 years old) with cerebral palsy. Methodological quality of the included studies was assessed with the Scottish Intercollegiate Guidelines Network (SIGN) criteria22 (see online supplementary Appendix I). The methods and results of the systematic review are described in detail in online supplementary Appendix II.

Three articles (reporting on two studies) were included in the final analysis. These two studies were retrospective analyses of the effectiveness of the same program: the Cerebral Palsy Follow-Up Program (CPUP), initiated in southern Sweden in 1994. Hägglund et al.,3,9 in 2005 and 2014, reported results at 10 and 20 years, respectively, comparing a cohort of children born between 1992-1997 (study group 1) and between 1998-2007 (study group 2) included in the CPUP, to a historical control group of children born in 1990-1991 not included in the program. Elkamil et al.,10 in 2011, compared a subsample of the CPUP to a sample of children recruited over the same period and with the same GMFCS (Gross Motor Function Classification System) levels in Norwegian regions who were not in a surveillance program. None of the studies reported data on pain or health-related quality of life.

The overlap of the intervention samples in the included studies precluded a meta-analysis of their results. The cost-utility model was populated based on data from Hägglund et al.,9 which provided the longest follow-up and compared two groups of children from the same region. The methodological quality of this study was evaluated as acceptable, the highest possible quality for retrospective studies according to the SIGN criteria (see online supplementary Appendix I).

In this study, children included in the control and intervention groups were not born in the same period. As a result, at the end of follow-up children in the control group (aged 22-23 years by then) have been at risk of developing dislocation (defined in the study as MP=100%) for far longer than children included in the intervention group 2 (aged 6-15 years at follow-up). Therefore, we compared information only from children included in intervention group 1 (aged 16 to 21 at follow-up) with those included in the control group, since they had a similar follow-up duration that was long enough to detect most cases of hip dislocation. Nine of 103 children developed dislocation of the hip in the control group (8 of them between 3 and 6 years of age, and one at age 16), while two out of 210 children from study group 1 included in the CPUP program suffered from a dislocated hip (relative risk=0.1090; 95% confidence interval [CI]: 0.2-0.49). Information on the probabilities of undergoing preventive primary, revision and salvage surgery were also computed by comparing children from study group 1 with the historical control cohort. These are presented in Table 1.

• 2)

  Resource use and unit costs

The intervention under analysis is the CPUP surveillance program,3,9 which included a standardized physiotherapist and occupational therapist visit twice a year until the age of six years, and once a year thereafter. Inclusion in the program was from identification of a possible cerebral palsy diagnosis, i.e., from birth on most patients, until they reached skeletal maturity. Radiological examinations in the CPUP program are based, since 2007, on the GMFCS, with children in level I not examined radiologically (if they have normal pain-free range of movement), children in level II examined at two and six years of age, and children in level III-V examined annually.9 GMFCS is currently the most widely applied scale for motor function classification in patients with cerebral palsy.23 We assume that each radiological examination involves a visit with an orthopeadic specialist doctor. In order to compute the mean cost of the program we considered the percentage of children in each GMFCS category as reported in Hägglund et al.,9 and shown in Table 1. This distribution by GMFCS was very similar to that reported in a previous study conducted in a Spanish region.24

In the CPUP, decisions regarding preventive surgery were made locally, and the most common types of preventive surgeries performed consisted of adductor–psoas tenotomy25 and varus femoral osteotomy.26 The proportions of surgery types are presented in Table 1, alongside with the mean age of children when undergoing preventive, revision and salvage surgery.

Information on the use of non-surgical treatments to prevent hip dislocation, such as appropriate lying, sitting and standing positions and the use of orthoses, is not provided in Hägglund et al.9 We assume there are not differences across groups in the provision of this usual care, and therefore these costs are not included in the analysis. Furthermore, no information was provided in this study on the follow-up required for children after surgery or who developed dislocation but could not undergo a surgical procedure. In our analysis we assumed, based on clinical expertise, an additional annual visit to the physiotherapist and an additional radiological examination involving an orthopaedic surgeon visit in these children. The impact of this assumption was tested in sensitivity analyses.

Unit costs data (Table 1) were taken from the mean values of the most up-to-date (2013 to 2018) Spanish regional tariffs (see online supplementary Appendix III for references).

• 3)

  Life expectancy and health-related quality of life (HRQoL)

In order to calculate the QALYs associated to each strategy, we combined information on life-expectancy as well as on HRQoL, the latter expressed in terms of QALY weights.

We estimated mortality rates for patients with cerebral palsy until 18 years of age based on data from Hägglund et al.9 There is no evidence of differences in mortality for children under and not under a surveillance program, and neither there is evidence that hip dislocation has an impact on life expectancy in patients with cerebral palsy. Therefore, we applied the same mortality rates (represented as survival curves in online supplementary Appendix IV) for all children in our analysis.

Several studies have shown that hip displacement/dislocation is significantly associated with a lower HRQoL in children with cerebral palsy.27–30 However, these studies have used a measure of HRQoL not suitable for the computation of QALYs weights (e.g. the Child Health Index of Life with Disabilities). QALY measurement in paediatric populations is very challenging.31 One study by Carroll and Downs32 calculated QALY weights for a wide range of health problems in the paediatric population. They considered mild, moderate and severe symptoms for each health problem, including cerebral palsy. In our base case analysis, we applied the reported utilities for children with mild cerebral palsy to children with cerebral palsy not suffering from hip dislocation, and that estimated for children with moderate cerebral palsy to children with cerebral palsy with hip dislocation (Table 1). We explored the impact of this assumption in sensitivity analyses, and estimated the change in utility required for the program to be cost-effective. We assumed that the mean age of children developing dislocation was 5 years of age.3,9 Furthermore, to allow for the fact that undergoing a surgical procedure might have a short-term detrimental impact on HRQoL, we applied a disutility associated to any surgical procedure equivalent to 0.1 QALYs for one-year after surgery. The impact of this assumption is also analyzed in sensitivity analyses.

Data parameters were increased to double and reduced by half the base-case value in one-way deterministic sensitivity analyses. Wider ranges were applied to the assumptions included in the analyses: the number of follow-up visits after surgery was varied from no follow-up to monthly (base-case: annual visits); the disutility associated to any surgical procedure for a one-year after surgery was varied from 0.05 to 0.5 (base-case: 0.1); the time horizon of the study was varied from 10 to 100 years (base-case: 18 years) and discount rate varied from 1% to 5% (base-case: 3%). In addition, we undertook a threshold analysis that computed the value required on the change in QALY weights after dislocation for the intervention to be considered cost-effective.

We also conducted a probabilistic sensitivity analysis to characterize the joint uncertainty in the model using 1,000 simulations in a Monte Carlo simulation. The results of the probabilistic sensitivity analysis are presented in terms of a cost-effectiveness plane and cost-effectiveness acceptability curves, which indicate the probability that an intervention is cost-effective for different values of the willingness to pay for a QALY. Probability distributions for each parameter are shown in Table 1.