Chronic back pain (CBP) is responsible for the largest amount of years lived with disability world-wide with tremendous direct and indirect costs.1,2 Intervention options are numerous but vary in effectiveness. Although therapy effects on pain and quality of life in CBP are grossly known, the effects are usually studied during highly controlled cost-effectiveness trials. For example, multidisciplinary rehabilitation leads to moderate effect sizes in nonspecific back pain,3,4 and surgery has better short-term outcome in lumbar disc herniation compared to conservative care.5 Pain interventions based on medication or injections may be effective in patients with severe radicular pain.6
Although these trials provide insight in the effects of the intervention on group level with selected patients, they may not be sufficiently capable to detect changes in health care costs in patients with complex CBP, who frequently suffer from comorbidities and CBP-related subdiagnoses.7 For (cost)effectiveness trials, this undesirable heterogeneity between studies leads to impossibilities of pooling results in reviews9 and may also lead to moderate effects of trials in general.8 Hence, generalization of randomized trials is often limited because of strict use of “narrow” inclusion criteria as well as to sampling bias. To properly control for heterogeneity, trials should control for deconditioning,10 altered central pain processing,11 comorbidity12 and psychosocial barriers for recovery including depression,12 anxiety,13 or somatization.14 To get a realistic impression of the magnitude and effects in healthcare costs of patients with complex CBP in real-world data, large cohorts with registry data may provide additional evidence.15 These real-world data, however, frequently suffer from biases related to regression to the mean, selection, and confounding by indication.
Although it is known that health care costs of patients with CBP are much higher compared to controls,12 a subset of patients with complex CBP may utilize a disproportionate amount of health care costs. It is unknown whether health care costs of patients with complex CBP can be reduced post-intervention and which factors are predictive for a reduction in costs. Therefore, the aims of this study were to analyze the health care costs of patients with complex CBP 2 years before until 2 years after tertiary multidisciplinary spine care. Study questions were: Do patients with complex CBP reduce healthcare costs after a multidisciplinary intervention?, Which patient characteristics are associated with a reduction in healthcare costs after multidisciplinary intervention? , What were the effects of possible biases related to real-world data?
MATERIALS AND METHODS
Patients referred to The Groningen Spine Center, the Netherlands were invited to participate. Patients came from urban and rural areas. In The Netherlards, all inhabitants are insured. Patients were included between November 2011 and November 2014. To be eligible, patients were older than 18 years, insured at health care insurer Menzis, and report CBP with or without radiation to leg(s) or arm(s) as primary pain site. As little exclusion criteria as possible were applied. Patients were excluded if they did not sign informed consent, when complaints were considered unrelated to their back or when patients were unable to fill out the baseline questionnaire.
A before and after study design comparing health care costs from 2 years before to 2 years post-intervention.
The Groningen Spine Center is a multidisciplinary center including neurosurgery, neurology, orthopedics, rehabilitation and anesthesiology. Interventions were chosen based on an evidence-based medicine approach, including clinical expertise, research evidence, and patient preferences.16 Physician assistants performed initial triaging to one or a combination of medical specialists for an outpatient consultation. During the first visit to the spine center, if necessary, two to four specialists with different focus areas can assess the complaints to find the best treatment modality. If necessary, the case could be discussed afterwards in a multidisciplinary meeting, where other medical specialists (i.e., radiology, rheumatology, psychiatry) could be consulted. Next, all patients received information and advice to cope optimally with their complaints. In general, patients with multifactorial CBP were offered treatment in primary care (i.e., physiotherapy or exercise or posture therapy) or outpatient interdisciplinary rehabilitation including psychology, physiotherapy, and occupational therapy in secondary or tertiary care. Surgery was offered to patients with specific CBP based on herniated disks or stenosis. Anesthesiology was provided to those patients with segmental root problems. Furthermore, patients could also be referred to primary care. When deemed indicated, combination of interventions could be deployed, following a stepped care approach. Information and advice were also provided in case of diagnostic purposes (second opinions).
After referral, patients were sent an online or paper questionnaire. Before participation, patients received information about the purpose of this study and signed informed consent. The Medical Ethical Committee of the University Medical Center Groningen, The Netherlands provided a waiver for this study. Research was in accordance to the declaration of Helsinki and good clinical practice.17 For reporting of the study, the consolidated health economic evaluation reporting standards (CHEERS) were applied and adapted for the design of the study.18
Baseline descriptive and health-related characteristics of patients were gathered. The dependent variable in this study is health care costs.To investigate correlations of healthcare costs with health characteristics, those characteristics were included which were previously correlated with CBP. These include demographics, work-related factors, disease characteristics and psychological functioning. 19
Health care declarations were obtained from the healthcare insurer and were matched with the patients’ clinical profile based on anonymous numbers. Based on the costs retrieved, the authors identified possible related health care costs to CBP and decisions on whether costs were concerned CBP-related were blinded to whether these had occurred prior or post-intervention. Whenever there was doubt on the relatedness of the declaration to the patients’ CBP, it was decided to include the declaration (e.g., a generic medicament prescription such as tramadol could be both CBP and not-CBP-related). CBP-related costs were divided into medical specialist care, allied care including physiotherapy, pharmaceutical care as regular medication by doctor's prescription, alternative care, and supportive and assistive help: ergonomic house or mobility-related adaptations including wound healing patches.
All secondary measures are presented in Table 1.
To avoid any biases in costs due to missing data, Multiple Imputation by Chained Equations (MICE) were used to impute missing data. Ten imputed datasets were made. Only full-year declaration blocks per patient were imputed because imputation of more detailed declarations (e.g., on missing data of alternative medicine declarations) was not possible, because of uncertainty if respective costs were actually made.
To answer the first study question, health care costs of patients were calculated according to five artificial time blocks ranging from 2 years before admission to 2 years after spine care (−2 and −1; respectively 1 and 1 year before- and +1 and +2; 1 and 2-year post-intervention) and 1 block representing the time between baseline and discharge. The time blocks were analyzed and the health care declarations made to the health care insurer were obtained and stratified in four blocks of 1 year each. All costs were corrected for inflation with 2017 as a reference year.26 Additionally, differences over time per intervention were reported. Repeated measures analysis of variance (ANOVA)s were used to analyze differences over time.
To answer study question 2, first, a linear regression was performed to identify relevant predictors that might be associated with the effect on health care costs. To study the intervention effect on costs, an intervention indicator was created, indicating whether a specific declaration was made pre (0) or post-intervention (1). The intervention indicator was inserted as independent variable in the regression analysis. The cost reduction related to treatment might differ by individual characteristics such as, age, sex, or disease or work-related characteristics because the context of patients may enable or disable a change in behavior. To test this, these characteristics were interacted with the treatment indicator in separate fixed-effects models.
To study regression to the mean, a conservative adjusted sample was constructed. This adjusted sample included only the health care costs of 2 years before compared to costs 2 years post-intervention assuming that patients were not admitted to the intervention based on high health care costs 2 years before admission. A natural logarithmic transformation was applied to control for skewness. To study relevant interaction effects of the intervention effect and predictors, a fixed-effect model was constructed. The interaction of every variable was tested in a separate fixed-effect model. P < 0.05 was considered statistically significant. Analyses were performed with SPSS-23 and STATA-15.
A total of 1830 patients were admitted to the Groningen Spine Center (GSC) and invited to participate in the study. Of these 997 patients were included. See Figure 1 for the flow diagram and reasons for exclusion. Included patients were on average 5.1 years older compared to nonparticipants (P < 0.01) and women were included more compared to men (P < 0.01). Descriptive statistics of the patient sample are presented in Table 2. The mean intervention time (GSC-0) was 201 days. The interventions in the total sample consisted of 546 patients receiving information and explanation, 203 patients followed interdisciplinary rehabilitation, 67 underwent surgery, 141 medication or injections, 37 a combination intervention of the above-mentioned, and for 3 patients the intervention could not be retrieved due to missing information in the patient record.
Study Question 1: Health Care Costs
Included patients filed a total of 428,158 declarations of which 128,666 were deemed CBP-related, during the 4.6 years they were followed, resulting in a total amount of €26.4 million of which €7.5 mln CBP-related. Some patients changed health insurer during the inclusion period; therefore, 287 missing blocks of a total of 4703 (5.8%) declaration blocks were imputed. There appeared no significant differences in results between analyses in complete cases and imputed data.
Figure 2 graphically presents the mean health care costs per patient and mean CBP-related health care costs over the study period. Non–CBP-related health care costs rose during the study period (P < 0.01; Table 3 and Figure 2). CBP-related health care costs rose pre-intervention, are highest during GSC, decrease 2 years following GSC, and were lower at that time compared to 1-year pre-intervention. The majority of the reduction in CBP-related health care costs appeared to be due to a reduction in medical specialist care and allied care consumption (P < 0.01) (Table 3). Results for the RM-ANOVAs show that there were no relevant increases in supportive devices, although results are significant. GP costs 2 years before GSC are significantly lower compared to other years (P < 0.01). Per intervention, similar patterns were observed, with highest CBP-related health care costs during the intervention and significant reductions in the post-intervention period. Post hoc tests revealed that differences between before- and post-intervention blocks were nonsignificant (Table 3).
Study Question 2: Patient and Disease-related Factors Predicting Reduction of Health Care Costs
Treatment is associated with a statistically significant reduction in CBP-related health care costs of 34% (P < 0.01) (Table 4). The following values at baseline predicted higher health care costs per year: higher age (1% increase; P < 0.01), female sex (42% increase; P < 0.01), higher body mass index (BMI) (2% increase; P < 0.01), lower EQ-5D (59% higher costs per point QoL; P < 0.01), longer pain duration (20% increase with pain >1 year; P < 0.01), and comorbidity (25% increase; P < 0.01).
Study Question 3: Effects of Bias
In the adjusted sample, costs of 2 years before and to 2 years post-GSC are compared. Results show a statistically significant cost reduction of 21% (P < 0.01), with similar predictors being significant (see Table 4).
Sex and BMI significantly interact with the treatment effect, leading to cost reduction. For example, a higher baseline BMI results in a smaller reduction in CBP-related health care costs (2% point smaller reduction per 1 kg/m2 increase in BMI). In Table 5, the fixed-effect model is presented.
CBP-related health care costs of patients with complex CBP included in the GSC indicate that rising CBP-related health care costs pre-intervention decreased post-intervention to values lower than pre-intervention. This contrasts to the total health care costs, that kept rising over the years, before, during, and post-intervention. This implies a cost-saving effect of multidisciplinary spine care on CBP-related costs, which was confirmed using several analytical methods. This study provides complimentary evidence that multidisciplinary spine care could be cost-saving from an insurers’ perspective, whereas we found no differences between intervention types. The results complement and point in the same direction as controlled studies, adding to the robustness of the observation that this type of care is beneficial from a health economic perspective. Additionally, this study is the first to fill an identified research gap.9
Although all estimates indicate a reduction in CBP-related health care costs, the different methods produce different estimates of the percentage decrease in health care costs. The most conservative estimate, comparing 2-year pre-intervention to 2 years post-intervention was calculated to control for regression to the mean. This led to a significant decrease of 21% for all costs; however, the model would not only control for regression to the mean, but probably also partially bias the estimated intervention effect toward zero, leading to a too conservative estimate. The largest difference in health care costs was observed when comparing patients 1 and 2 years pre-intervention to patients 1 and 2 years post-intervention, which estimates a significant decrease of 34% for CBP-related costs. A previous study showed that the (self-reported) burden of patients with discogenic low back pain were high and reflect our data.27
The major limitation of cohort studies compared to RCTs is the lack of a control group, which allows controlling for regression to the mean (in this case regression to the mean may indicate that patients get referred to the Groningen Spine Center on the basis of high health care costs 1-year pre-intervention). Two theories are further explored. First, in the pre-interventional 11.5 years that patients report recurrent pain, patients encounter exacerbations of their chronic condition, leading to increase of costs before the intervention, which would lead to a cost reduction post-intervention regardless of any intervention. Second, patients ran out of primary and secondary care options and are referred to tertiary care and have a stable declaration pattern over the years before intervention. Lower costs post-intervention would then imply true effect of the treatment. Although the long-term utilization patterns should be subject for further study, our conservative estimates in this study imply that effects are unlikely to be fully driven by regression to the mean and are indicative that back pain-related health care costs decline post-intervention. The mean estimates would lead to a decrease of 34%, which equals a decrease of €484 annually post-intervention per patient. For the entire sample (N = 997), this would imply a cost saving of €482,548; however, this should be returned after the intervention because intervention costs were high.
It is remarkable to see the differences in cost patterns between total health care costs and CBP-related costs. There were no system or reimbursement changes during the study period that could explain the decrease in CBP-related costs. The reasons of the rising total health care costs (about 80% increase) can partially be explained by the following biases. First, inflation; estimated at 9%.26 Second, in the Netherlands, health care costs keep rising in general; estimated at 26%.28 Third, patients were 4.5 years older in the last block compared to the first; estimated at 19%.29 Fourth, post hoc data analyses showed that there were some cases in which a dramatic increase in health care costs was observed because of severe and very expensive treatments such as Pompe disease (increase costs >€300k), total hip replacement (€105k), heart surgery, and cancer (multiple medicaments such as lenalidomide). We assume that patients referred to spinal treatment are currently not in active phases of other life-threatening disease that requires immediate attention. The first 3 biases explain 54% of the 80% rise in total health care costs. Against a strong trend of rising total health care costs, the CBP-related costs decreased, which would suggest that the effect of CBP-related costs of the intervention could even be substantially larger.
A particular strength of this study was the use of real-world data and the long follow-up period. This is among the first studies to report on real-world data. Objective data instead of self-report data were used which excludes recall bias. The data from this real-world cohort reflect costs savings derived from RCTs,9 however, are much more applicable to the broad population of patients with CBP. Another strength was that declarations of a large group of almost 1000 persons were retro- and prospectively retrieved from the health care insurer with little missing data (5.5%). As previously stated, the main limitation of this study is the study design because no control group was available that received usual or no care. Furthermore, unknown confounders may have not been controlled for.30 Further research should aim at comparing real-world data with a relevant control group to provide further evidence on the economic and clinical effect of multidisciplinary spine care.31 A second limitation is that this study focused solely on direct medical health care costs from the insurance perspective. In a previous study, it was concluded that approximately 87% to 89% of total societal burden was related to losses in productivity.32 This means that the results from this study may underestimate the total societal burden.
The results of this study indicate that multidisciplinary spine intervention decreases back pain-related health care costs for 34%. BMI and sex are associated with a decrease in healthcare costs post-intervention. Controlling for biases related to real world data led to a significant decrease of 21%. These results are valuable from the perspective of the patients, government, health insurers, and clinicians.
- Multidisciplinary spine intervention is associated with 34% (P < 0.01) reduction in back-related costs 2 years after intervention.
- Reduction in costs after intervention is related to being male and lower BMI.
- Reduction in CBP-related health care utilization in patients with complex CBP can be achieved after a multidisciplinary spine intervention.
- The results are robust to controlling for background characteristics and are unlikely to be fully driven by regression to the mean.
1. Dagenais S, Caro J, Haldeman S. A systematic review of low back pain
cost of illness studies in the United States and internationally. Spine J
2. de Vroome EM, Uegaki K, van der Ploeg CP, et al. Burden of sickness absence due to chronic disease in the dutch workforce from 2007 to 2011. J Occup Rehabil
3. Kamper SJ, Apeldoorn AT, Chiarotto A, et al. Multidisciplinary biopsychosocial rehabilitation for chronic low back pain
. Cochrane Database Syst Rev
4. Hartvigsen J, Hancock MJ, Kongsted A, et al. What low back pain
is and why we need to pay attention. Lancet
5. Peul WC, van Houwelingen HC, van den Hout WB, et al. Surgery versus prolonged conservative treatment for sciatica. N Engl J Med
6. Foster NE, Anema JR, Cherkin D, et al. Prevention and treatment of low back pain
: evidence, challenges, and promising directions. Lancet
7. Schneider S, Mohnen SM, Schiltenwolf M, et al. Comorbidity of low back pain
: representative outcomes of a national health study in the federal republic of germany. Eur J Pain
8. Maixner W, Fillingim RB, Williams DA, et al. Overlapping chronic pain conditions: implications for diagnosis and classification. J Pain
2016; 17: (9 suppl): T93–T107.
9. van Dongen JM, Ketheswaran J, Tordrup D, et al. Health economic
evidence gaps and methodological constraints in low back pain
and neck pain: Results of the research agenda for health economic
evaluation (RAHEE) project. Best Pract Res Clin Rheumatol
10. Bousema EJ, Verbunt JA, Seelen HA, et al. Disuse and physical deconditioning in the first year after the onset of back pain
11. Roussel NA, Nijs J, Meeus M, et al. Central sensitization and altered central pain processing in chronic low back pain
: fact or myth? Clin J Pain
12. Gore M, Sadosky A, Stacey BR, et al. The burden of chronic low back pain
: clinical comorbidities, treatment patterns, and health care costs in usual care settings. Spine (Phila Pa 1976)
13. Vlaeyen JW, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. Pain
14. Bener A, Verjee M, Dafeeah EE, et al. Psychological factors: anxiety, depression, and somatization symptoms in low back pain
patients. J Pain Res
15. Garrison LP Jr, Neumann PJ, Erickson P, et al. Using real-world data for coverage and payment decisions: the ISPOR real-world data task force report. Value Health
16. Sackett DL, Rosenberg WM, Gray JA, et al. Evidence based medicine: what it is and what it isn’t. 1996. Clin Orthop Relat Res
17. Internationaal richtsnoer voor ‘good clinical practice’ voor het onderzoek met geneesmiddelen; vertaling naar de nederlandse praktijk. rev version ed. Den Haag: GCP Begeleidingscommissie; 2003.
18. Husereau D, Drummond M, Petrou S, et al. Consolidated health economic
evaluation reporting standards (CHEERS) statement. Value Health
19. van der Hulst M, Vollenbroek-Hutten MM, Groothuis-Oudshoorn KG, et al. Multidisciplinary rehabilitation treatment of patients with chronic low back pain
: A prognostic model for its outcome. Clin J Pain
20. EuroQol Group. EuroQol—a new facility for the measurement of health-related quality of life. the EuroQol group. Health Policy
21. Lamers LM, McDonnell J, Stalmeier PF, et al. The Dutch tariff: results and arguments for an effective design for national EQ-5D valuation studies. Health Econ
22. Jensen MP, Karoly P, Braver S. The measurement of clinical pain intensity: a comparison of six methods. Pain
23. Linton SJ, Hallden K. Can we screen for problematic back pain
? A screening questionnaire for predicting outcome in acute and subacute back pain
. Clin J Pain
24. Linton SJ, Boersma K. Early identification of patients at risk of developing a persistent back problem: the predictive validity of the orebro musculoskeletal pain questionnaire. Clin J Pain
25. Hays RD, Sherbourne CD, Mazel RM. The RAND 36-item health survey 1.0. Health Econ
26. Statistics Netherlands. Consumer prizes; price index. Available at: http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=83131ned
. Updated November 6, 2018. Accessed November 6, 2018, 2018.
27. Geurts JW, Willems PC, Kallewaard JW, et al. The impact of chronic discogenic low back pain
: costs and patients’ burden. Pain Res Manag
28. Vektis. Health care thermometer; 10 years of healthcare legislation in the netherlands. Available at: https://www.vektis.nl/intelligence/publicaties/zorgthermometer-10-jaar-zorgverzekeringswet
. Accessed June 5, 2019.
30. Gill J, Prasad V. Improving observational studies in the era of big data. Lancet
31. Briss PA, Zaza S, Pappaioanou M, et al. Developing an evidence-based guide to community preventive services--methods. The task force on community preventive services. Am J Prev Med
2000; 18: (1 suppl): 35–43.
32. Lambeek LC, van Tulder MW, Swinkels IC, et al. The trend in total cost of back pain
in the netherlands in the period 2002 to 2007. Spine (Phila Pa 1976)