Economic evaluation of operative versus nonoperative treatment of a humeral shaft fracture: economic analyses alongside a multicenter prospective cohort study (HUMMER)

Purpose Operative treatment of a humeral shaft fracture results in faster recovery than nonoperative treatment. The cost-effectiveness, in terms of costs per Quality-Adjusted Life Year (QALY) gained (Dutch threshold €20,000-€80,000) or minimal important change (MIC) in disability reduced (DASH 6.7), is unknown. The aim of this study was to determine cost-utility and cost-effectiveness of operative versus nonoperative treatment in adults with a humeral shaft fracture type 12A or 12B. Methods This study was performed alongside a multicenter prospective cohort study. Costs for health care and lost productivity until one year after trauma were calculated. The incremental cost-utility ratio (ICUR) was reported in costs per QALY (based on the EuroQoL-5D-3L (EQ-5D)) gained. The incremental cost-effectiveness ratio (ICER) was reported in costs per MIC (based on the DASH score at three months) reduced. Results Overall, 245 patients were treated operatively and 145 nonoperatively. In the operative group, the mean total costs per patient (€11,925 versus €8793; p < 0.001) and QALYs (0.806 versus 0.778; p < 0.001) were higher. The ICUR of operative treatment was €111,860 per QALY gained (i.e., €3132/0.028). The DASH was 7.3 points (p < 0.001) lower in the operative group. The ICER of operative treatment was €2880 per MIC in disability reduced (i.e., €3132/7.3*6.7). Conclusion Due to the limited effect of treatment on quality of life measured with the EQ-5D, the ICUR of operative treatment (€111,860 per QALY gained) exceeds the threshold. However, the incremental costs of €2880 per clinically meaningful difference in DASH are much lower and suggest that operative treatment for a humeral shaft fracture is cost-effective. Supplementary Information The online version contains supplementary material available at 10.1007/s00068-022-02160-1.


Background
In an era of budget restraints on health care costs, efficient resource use is crucial and data on cost-effectiveness of treatment are gaining importance in health care budget allocation [1][2][3]. In the Netherlands, costs of injuries account for 5% of the total health care budget and 8% of the indirect costs resulting from all diseases [3]. However, there seems to be a paucity of evidence in the area of cost-utility and costeffectiveness of treatment of orthopedic trauma injuries [2]. Multiple studies have shown that long bone fractures are costly in terms of direct medical costs and lost productivity [4,5]. The burden on society of long bone fractures can be attributed to the costs of surgery, possible reinterventions, and the physical rehabilitation of patients [5]. When comparing upper extremity injuries, upper arm fractures resulted in the highest costs per case (€4440) in the Netherlands [6]. Cumulative medical costs in the Netherlands of patients, List of authors in the HUMMER Investigators are listed in acknowledgement.

3
admitted due to a humeral shaft fracture only, added up to €10.6 million in 2012 [7].
Humeral shaft fractures pose a burden on society as they make up 3% of all orthopedic injuries [8]. In the Netherlands, the overall incidence rate of patients admitted for a humeral fracture per year has risen by 132% to 7.2 per 100,000 person years from 1986 to 2012, partly attributable to an aging population [7]. Incidence rate is characterized by a bimodal age distribution, affecting both young and elderly patients, which influences the pattern of health care costs [7,9]. Fractures in young employed persons can induce high costs due to the absence of work and lost productivity [4]. Furthermore, it is established that especially medical costs of humeral shaft fractures in elderly women are substantial due to extended nursing home admission or homecare [7,9].
Humeral shaft fractures can be managed operatively or nonoperatively, with both treatments resulting in high union rates and excellent results [8]. Nonoperative treatment is mostly performed using a functional brace [10]. Operative treatment mostly includes plate osteosynthesis, intramedullary nailing (IMN), or external fixation for limited indications [8]. The primary results of the HUMMER study indicate, based on functional and clinical outcomes, that operative treatment should be the preferred treatment option for these fractures, as it is associated with faster functional recovery and fewer complications such as nonunion than nonoperative treatment [11].
These findings are not yet supported by data on health care consumption and costs [12]. Policy-makers need the detailed information provided by cost-utility (CUA) and cost-effectiveness analyses (CEA) to adequately balance costs and effects with suitable thresholds of efficiency in order to provide well-informed advice on health care budget allocation [9,[13][14][15]. Therefore, the aim of this study was to determine cost-utility and cost-effectiveness of operative versus nonoperative treatment of adult patients with a humeral shaft fracture. The hypothesized was that operative treatment would be cost-effective, due to earlier functional recovery and lower costs for follow-up and lost productivity outweighing higher costs for initial treatment.

Setting and participants
These economic analyses were performed alongside the observational HUMMER study [16]. The study was exempted by the local Medical Research Ethics Committee (no. MEC-2012-296) and registered at the Netherlands Trial Register (NTR3617). Patients were eligible if they (1) were aged 18 years and older (with no upper limit), (2) had a closed fracture of the humeral shaft (AO type 12A or 12B; confirmed on X-ray), (3) had provided written informed consent, and if operatively treated, (4) had an operation within 14 days after presentation to the Emergency Department [17]. Patients were excluded if they had sustained other traumatic injuries or were known to have pre-existing disorders that were expected to affect bone healing, treatment, or rehabilitation of the affected arm (e.g., polytrauma, open fractures, pathological fractures, bone disorders (excluding osteoporosis), rheumatoid arthritis, or pre-existing impaired upper extremity function). Furthermore, patients with expected problems with follow-up (e.g., no fixed address or cognitive impairment) or insufficient comprehension of the Dutch language were excluded. Full details on inclusion and exclusion criteria are available in the published study protocol [16].
Treatment was left to the treating physician and consisted either of operative treatment with plate osteosynthesis or IMN, or nonoperative treatment with a splint, plaster, collar and cuff, or hanging cast, followed by a Sarmiento brace.

Outcomes measures
The effect measure for the CUA was the Quality-Adjusted Life Years (QALYs). The mean increase in QALYs during one year was calculated using the EuroQol-5D-3L (EQ-5D), a validated questionnaire recommended for assessing quality of life in trauma patients, especially for economic assessments [18][19][20]. Participants completed the EQ-5D at two and six weeks and three, six, and 12 months after initiation of treatment. The EQ-5D descriptive system consists of five health domains (mobility, self-care, usual activities, pain/ discomfort, and anxiety/depression) with three answer levels (no problem, moderate problem, or severe problem). Utility scores were calculated to express the health status descriptions ranging from zero to one, in which zero is death and one is full health.
The effect measure for the CEA was the Disabilities of the Arm, Shoulder, and Hand (DASH) score at three months, as at that time, a clinical difference was expected [21]. The DASH is a validated, 30-item (scored 1-5), self-report questionnaire with an overall score ranging from 0 (no disability) to 100 (severe disability), reflecting functional outcome and pain of the upper extremity [22,23]. The minimal important change (MIC) of the DASH is 6.7 points [21].

Health care consumption and productivity loss measurement
These economic analyses were performed from a societal perspective, following Dutch guidelines [24,25]. Data on health care consumption and work absenteeism were collected at each scheduled follow-up contact using a custommade questionnaire based upon the Medical Consumption Questionnaire (iMCQ) and the iMTA Productivity Cost Questionnaire (iPCQ) [26,27]. Data were gathered until one year after trauma. Health care consumption included intramural and extramural medical care directly associated with diagnosis, treatment, and rehabilitation of the patient with a humeral shaft fracture. Missing data of hospital care consumption were collected during the close-out visits at each hospital.

Cost calculation
Reference prices of health care resources were derived from the Dutch manual for costing in economic evaluations where possible (Supplemental Table S1-2) [28]. Other reference prices for cost categories were calculated based on data derived from the participating academic and non-academic hospitals, surgical equipment and implant firms, the NZa (Nederlandse Zorgautoriteit; Dutch Healthcare Authority), the CVZ (College voor Zorgverzekeringen; Health Care Insurance Board), or obtained from home care firms [24,29]. Reference unit costs for 2020 (€) were used or adjusted to 2020 (€) costs with the national consumer price index [30]. Inflation was taken into account. Costs were calculated by multiplying the frequency of resource use by the unit prices per cost category. Comparison with US costs was done after applying the exchange rate (€1 = US$1.21) [31].
Indirect societal costs due to work absence were calculated using the friction cost method [18]. Costs for lost productivity were defined as the costs associated with production loss and replacement due to illness, disability, and premature death [32]. Costs for lost productivity were calculated by multiplying the cumulative duration of work absence in hours within the first 85 days after injury with the costs related to work absenteeism for different five-year age groups for employed persons aged 18-68 years (Supplemental Table S1) [33].

Statistical analysis
Data were analyzed using the Statistical Package for the Social Sciences version 25 (SPSS, Chicago, Ill., USA). Missing data were not imputed. Data were averaged for patients for whom data were available. Analysis was performed according to intention to treat and all statistical tests were two-sided. Chi-squared analysis was used for statistical testing of categorical data. Functional outcomes that were repeatedly measured over time were compared between treatment groups using linear mixed-effects regression models, as described before [11]. The models included fixed effects for treatment group, age, gender, and the individual fracture types. Continuous data were analyzed using a Mann-Whitney U test. For the pairwise comparison of the mean costs, the bootstrap 95% confidence interval (95% CI) was computed based on 1000 replications. Since the time horizon was one year, no discounting was required for costs and health utilities. Results were reported following the CHEERS Checklist for reporting economic health evaluations [34]. A p value < 0.05 was taken as a threshold for statistical significance in all statistical tests.
The incremental cost-utility ratio (ICUR), comparing operative versus nonoperative treatment, was expressed in terms of incremental mean total costs per mean QALY gained and calculated by dividing the difference of the mean total costs by the difference of the mean increase in QALYs over 12 months. The Dutch threshold of maximum costs per QALY was used (ranging from €20,000 up to €80,000 per QALY) [9,[13][14][15]35].
The incremental cost-effectiveness ratio (ICER), comparing operative treatment versus nonoperative treatment, was reported in terms of incremental costs for a clinically meaningful difference (6.7 DASH points reduced at the three months' time point). The ICER was calculated by dividing the difference of the mean total costs of the two interventions by the difference of the mean DASH score at three months and multiplied by 6.7. This ratio, with a different time interval used in the numerator and denominator, was chosen in order to compare clinical expected differences to the total costs of treatment, as a difference in DASH score was expected at three months and treatment of a humeral shaft fracture usually does not exceed one year [16].

Patient characteristics and employment details
Between October 23, 2012 and October 3, 2018, 390 patients were included of whom 245 (62.8%) were treated operatively and 145 (37.2%) nonoperatively. Compared with the nonoperative group, patients in the operative group were younger (median age of 53 (P 25 -P 75 35-66) versus 62 (P 25 -P 75 49-71) years; p < 0.001) and more often male (45.6% versus 35.2%; p = 0.044) ( Table 1). Furthermore, patients in the operative group were significantly more often employed (55.5% versus 42.8% in the nonoperative group; p = 0.016) and worked more hours per week (38 versus 32 h in the nonoperative group; p = 0.016). Twenty patients were lost to follow-up due to mortality (N = 4) or withdrawal of consent (N = 16). The total number of patients available for follow-up varied per follow-up moment, as 55 patients did not show up at least one follow-up visit.

QALY and DASH
The mean increase in QALYs during one year was 0.028 higher after operative treatment (mean of 0.806 (95% CI 0.801-0.811) versus 0.778 (95% CI 0.771-0.784) in the nonoperative group; p < 0.001), which was mostly attributable to a faster increase in health-related quality of life in the first six months.
There was a significant and clinically meaningful difference in DASH score of 7.3 points between the operative and nonoperative group at three months follow-up, in favor of the operative group (mean of 22.3 (95% CI 19.9-24.6) versus 29.6 (95% CI 26.6-32.6) in the nonoperative group; p = 0.001).

Health care consumption and work absence.
An overview of the mean health care consumption and work absenteeism per patient is shown in Table 2. Patients in the operative group were all admitted to the hospital (N = 145, 100%) for a median stay of 2 (P 25 -P 75 2-4) days. In the nonoperative group, 26 (17.9%) patients were admitted for a median stay of 2 (P 25 -P 75 2-3) days. Patients in the operative group had significantly more medical imaging units during their primary stay compared to the nonoperative group (median 4 (P 25 -P 75 2-4) versus 2 (P 25 -P 75 2-2) units; p < 0.001). During follow-up, patients in the nonoperative group had significantly more medical imaging, used more devices for immobilization, and had more outpatient clinic visits. Besides that, compared with the operative group, a doubling of surgical reinterventions was found in the nonoperative group (12. Although the operative group resumed work seven days earlier (26 versus 33 days in the nonoperative group), there was no significant difference in work absence in days (p = 0.253).

Health care costs and costs for lost productivity
An overview of the mean health care costs per patient is shown in Table 3. The mean total costs were significantly higher in the operative group (€11,925 versus €8793 in the nonoperative group; p < 0.001) ( Table 3). In addition, the mean total hospital costs per patient of primary stay were significantly higher in the operative group (€5159 versus €1093; p < 0.001). The mean costs of surgery attributed to almost half of the costs of primary stay (€2434). The mean follow-up costs per person were significantly lower in the operative group (€1377 versus €2306; p < 0.001). The mean costs for ambulance transport, medical imaging (primary stay), initial treatment, and hospital admission days (primary stay) were significantly higher in the operative group. The mean costs of devices for immobilization (initial treatment), medical imaging (follow-up), and mean costs related to revision surgery and consequent hospital admission days were significantly lower in the operative group.

Cost-utility analysis
Operative treatment resulted in higher mean total costs per person until 12 months of €3132 (95% CI €1325-€4940; p < 0.001). The mean change in QALYs until 12 months was 0.028 (p < 0.001) higher in the operative group.

Cost-effectiveness analysis
The mean difference in DASH score was 7.3 points (p < 0.001) in favor of operative treatment, resulting in Table 3 The mean costs (2020) (€) by treatment group The exchange rate was €1 = US$1.21 [31] P values < 0.05 are shown in boldface Data are presented as a mean with a bootstrap 95% CI N* represents the number of patients for whom data were available per follow-up moment

Discussion
This study showed that operative treatment of a humeral shaft fracture results in higher mean costs per person over one year of €3132 (95% CI €1325-€4940; p < 0.001) than nonoperative treatment. The mean difference in QALYs (0.028; p < 0.001) during one year in favor of the operative group demonstrates that operative treatment results in a higher health-related quality of life during the first year after trauma. This difference is statistically significant but small, therefore incremental costs per QALY gained are high (€111,857; i.e., €3132/0.028). The clinical and statistically significant difference of 7.3 DASH points (p < 0.001) in favor of the operative group exceeds the MIC and results in incremental costs for operative treatment of €2880 for a measurable change in disability. The different measures of efficacy used in these economic analyses should be carefully weighted in the decision-making process. Economic evaluations with QALYs may be preferred in order to allow for comparison across populations with different medical conditions. However, a humeral shaft fracture does not necessarily affect a patient's self-reported health-related quality of life as the injury may have little effect on some of the measured domains of the EQ-5D (i.e., anxiety and depression), resulting in marginal differences in QALYs gained [21]. Due to the limited effect of a humeral shaft fracture on quality of life, the costs per QALY (€111,857) exceed the threshold set by society. The difference in functional outcome measured by the DASH score was shown to be more specific than the health-related quality of life measured in QALYs [21]. An ICER calculated with the DASH score cannot be compared to other injuries, but it does show the relatively low incremental costs of operative treatment for a clinically meaningful difference and suggests that operative treatment for a humeral shaft fracture is cost-effective.
The results of the cost calculations are comparable with results from previous research. Polinder et al. (2013) described comparable direct health care costs of upper arm fractures of €4440 per case (versus €5116 in this study), taking into account inflation and the more detailed health care resource use described in this study [6]. Bonafede et al. (2013) determined higher direct health care costs (US$10,842 (≈ €8960) versus €7589) and higher costs for lost productivity (US$4868 (≈ €4023 versus €2894 in this study) per humeral fracture [4]. However, costs were calculated by multiplying the total number of hours reported absent multiplied by an average rate per hour (human capital approach) instead of assuming that productivity costs are only incurred during the period until the moment the employee is replaced, the so-called friction period [4,32]. Meerding et al. (2006) described similar total costs of Fig. 1 The relative contribution of various cost categories. The exchange rate was €1.00 = US$1.21 [31]. Only cost categories representing more than 1% of the total costs are shown. ED Emergency department, FU Follow-up, HLOS Hospital length of stay, PS Primary stay humeral shaft fractures in the Netherlands, namely €9430 per patient, with also hospital care costs and costs for lost productivity as main cost drivers [36].
Patients' preferences shape clinical decision-making which therefore could be influenced by employment status. It is desirable that employed patients return to work as soon as possible, especially knowing that costs for lost productivity account for more than a quarter of the total costs of treatment of a humeral shaft fracture and added up to €5.4 million in the Netherlands for admitted patients alone in 2012 [7]. Hendy et al. (2020) identified no advantage for faster return to work after operative or nonoperative treatment of humeral shaft fractures [37]. This study showed that employed patients were treated operatively more often, but there was no significant difference in work absence in days or costs for lost productivity between treatment groups. However, the underlying differences between the treatment groups, specifically the male predominance, younger median age, and overrepresentation of employed patients, who also worked significantly more hours per week, in the operative treatment group, result in an underestimation of the advantage of their earlier return to work in terms of costs for lost productivity.

Strengths and limitations
The strengths of this study include a large multicenter prospective cohort methodology measuring health utility, a formal economic costing approach, and a societal perspective for costs. Furthermore, this study design ensures great external validity by allowing for variation between hospitals (e.g., differing policies on follow-up procedures and allocation of resources).
A limitation of these cost analyses is that both groups included multiple treatment strategies with different costs of material (Supplemental Table S1). Moreover, costs were based on Dutch prices and practices and therefore may vary depending on the health care system used. Furthermore, the follow-up duration of 12 months did not take into account the late complications of nonunion or the need for revision surgery after more than one year. Lastly, the lack of an upper age limit for age inclusion may have (slightly) skewed the results, based on life expectancy and working situation.

Conclusion
This study showed that operative treatment of a humeral shaft fracture is more expensive than nonoperative treatment, but results in a higher health-related quality of life and significantly less disability. Due to the limited effect of a humeral shaft fracture on quality of life measured with the EQ-5D, the cost-effectiveness of operative treatment in terms of costs per QALY (€111,857) exceeds the acceptability limit. However, the incremental costs of €2880 per clinically meaningful difference in DASH are much lower and suggest that operative treatment for a humeral shaft fracture is cost-effective.