As many as 80% to 85% suffer from low back pain (LBP) at some point in life.1 For most LBP patients, with or without concomitant leg pain, the symptoms are resolved within 3 months, but for some pain remains.2 Surgery may be considered as a last option for a limited patient group that does not respond to any other treatments.3,4 Persistent LBP is associated with substantial quality of life reductions and an increased burden for the society in terms of health care costs and productivity losses.5–8
Most patients experience improvement following surgery but as many as 40% report suboptimal results, depending on the diagnosis.9 Jonsson et al10 showed that the societal cost is, on average, reduced over 2 years after lumbar spine surgery. This analysis included all patients having surgery irrespective of outcome. There is likely a correlation between surgical outcome and long-term costs. That is, patients with good outcome probably consume less health care resources and return to work faster and more frequently than those with worse outcome. Currently, there is little knowledge on cost patterns before and after lumbar spine surgery depending on the actual outcome of surgery.
This study's objective was to describe the association between societal costs and the outcome of lumbar spine surgery based on patient reported outcome measures (PROMs) collected in Swedish spine register Swespine combined with resource utilization data from Swedish national registers.11
MATERIALS AND METHODS
Data Sources and Research Database
This study utilized a research database consisting of data from six Swedish regional and national registers; LISA from Statistics Sweden, the National Board of Health and Welfare's National Patient Register (PAR), spine surgery register Swespine, the Social Insurance Agency register, the Prescribed Drug Register, and administrative database VEGA from the Västra Götaland region. All patients enrolled in Swespine between 2000 and 2012 and all patients with a recorded healthcare visit due to LBP in VEGA between 2008 and 2012 were included in the database. Swespine includes data on PROMs at surgery (baseline) and at follow-up 1, 2, 5, and 10 years after surgery. All data utilization was consistent with relevant Swedish legislation on personal data, Personuppgiftslagen 1998:204. The Stockholm ethics vetting committee has approved the study (2013/2225-31/5).
All lumbar spine surgery patients (vertebrae L1–S1) registered in Swespine 2000 to 2012 were identified in the research database. Patients were stratified by the following lumbar diagnosis groups; Degenerative Disc Disease (DDD) (ICD-10 M99.0, M99.1, M99.1K, M99.2–9, Q76.3, M40∗, M47.2–9 M47.9K, M47.9X, and M53.2–9), Disc herniation (LDH) (M51.0, M51.0K, M51.1, M51.1A, M51.1K, M51.2–9), spondylolisthesis (LSL) (M43∗), and lumbar spinal stenosis (LSS) (M48.0, M48.0W, M48.0K, M48.1–4, M48.4K, M48.5, M48.8, M48.8K, M48.8W, M48.9). Asterisks indicate that all subcodes are included.
DDD is defined as the clinical entity composed of the combination of:
- (a) Localized low back pain of
- (b) mechanical type and
- (c) X-ray/Magnetic resonance imaging findings in segments where
- (d) you can provoke pain on clinical examination.
Patients were excluded if global assessment (GA) was not reported or reported as “no pain prior to surgery” at the 2-year post-surgery follow-up.
Patient Reported Outcome Measurements
Four different PROMs prospectively collected in Swespine were used in the analysis:
Visual analog scale (VAS) for leg and back pain12 is measured from 0 indicating “No pain” to 100 indicating “Worst pain imaginable,” using a vertical line on which the patients indicate their pain levels.
Oswestry disability index (ODI) for functional ability13–15 is measured on a scale from 0 to 100 going from minimal disability (0–20) to bed bound (81–100).
EuroQol five dimensions questionnaire (EQ-5D-3L) measures generic health-related quality of life (HRQoL). An EQ-5D index score was generated using UK social preference weights.16 The index ranges from –0.594 to 1, where 0 implies HRQoL of being dead and 1 is perfect health.
Global assessment (GA) of back and leg pain. Patients assess their back/leg pain at surgery and at follow up 1, 2, and 5 years after surgery. GA ranges from 0 to 5, as follows; 0: “no pain prior to surgery”; 1: “pain has disappeared”; 2: “pain is much improved”; 3: “pain somewhat improved”; 4: “no change in pain”; and 5: “pain has worsened.”17
Definition of Surgical Outcome Groups
Since the study's objective was to analyze the impact of different patient reported outcomes of surgery on societal costs, decision rules to categorize the outcome of surgery were needed. There is no standardized definition of successful outcome of spine surgery based on PROMs,18 but a consensus that PROM score changes should reflect changes both statistically significant and clinically important for the patients.19 The minimal clinically important difference (MCID) has been designed to identify the minimal impact of a medical intervention considered beneficial for the patient.17 The approach was therefore to categorize patients in outcome groups using MCID values for the available PROMs in the database (ODI, VAS, EQ-5D, and GA). Categorizing patients into outcome groups based on MCIDs for each PROM separately would yield three different category sets. To have a unison algorithm we instead defined patients into surgical outcome groups using the thoroughly validated GA20 as an external criterion. The approach followed these steps:
- 1. Patients were categorized into five groups by their GA response at 2-year follow up, and the mean changes in the other PROMs (ODI, VAS for leg and back pain, and EQ-5D) from baseline to 2-year follow up were then calculated for each GA group.
- 2. These mean changes were then compared with MCID thresholds identified in the literature (the defined level of change in a PROM denoting that the pain situation “worsened” or “improved”) for each respective PROM.
- 3. The GA based categorization of surgical outcome was then defined in accordance with how the changes in the PROMs by each GA level corresponded to the MCID thresholds.
New MCID thresholds by these outcome groups were then estimated from the PROM changes, as defined in Equation 1,17 and compared with the thresholds identified in the literature, to further investigate the validity of the categorization.
Equation 1: MCID thresholds
To define surgical outcomes on the most relevant criterion for the respective diagnoses, patients with spinal stenosis, and disc herniation were analyzed using GA and VAS of leg pain, whereas spondylolisthesis and DDD were analyzed using GA and VAS of back pain.
Costs were considered from a societal perspective. Monthly costs for the outcome groups were estimated from 3 years before to 3 years after surgery. Direct costs included specialist visits, inpatient care, surgery, and drug costs related to the back condition. Unit costs were sourced from regional price lists.21–23 Indirect costs were measured as days with sickness insurance benefit in patients in working age (19–65 yrs) and valued with the human capital approach, that is, by assuming that the cost of 1 day's work absence equals the average gross daily wage (€147).24 Costs were reported in EUR (€) 2016.
Data management was conducted using a MySQL server (Oracle Corporation, Redwood Shores, California) and statistical analyses using the software package Stata 14 (StataCorp LLC, College Station, Texas).
The study identified 12,350 lumbar spine surgery LBP-patients who reported GA back or leg pain 2 years after surgery. A total of 2282 patients were evaluated using GA back pain and 10,068 using GA leg pain, in accordance with their diagnoses.
PROMs and MCID Threshold Values
The MCID improvement and deterioration thresholds identified in the literature are presented in Table 1. Estimated mean changes in PROMs from baseline to 2 years after surgery for all GA responses 1 to 5 to determine the use of GA in defining the outcome groups are presented in Table 2.
The mean changes estimated at GA level 1 and 2 fulfilled all the lower MCID thresholds for improvement identified in the literature. Considering the upper thresholds, GA level 1 and 2 fulfilled all thresholds, except for EQ-5D. GA level 3 fulfilled the lower ODI threshold, but not the VAS and EQ5D ones, and none of the upper thresholds. Patients reporting GA level 4 were on average below all improvement thresholds. They did not fulfil any deterioration thresholds. GA level 5 was below all improvement thresholds and fulfilled the VAS deterioration threshold, however not the ODI deterioration threshold.
Based on this, patients who reported GA level 1 or 2 were categorized as having a “successful” outcome, patients reporting GA level 3 were categorized as “undetermined,” and those who reported GA level 4 or level 5 as having an “unsuccessful” outcome. Using this categorization, patients defined as having a successful outcome on average fulfilled all improvement thresholds, except for the upper EQ-5D threshold whereas patients with undetermined outcomes fulfilled the lower ODI and EQ-5D improvement thresholds, none of the VAS improvement thresholds and none of the upper improvement thresholds. The patients defined as having unsuccessful outcomes did not fulfil any improvement thresholds (Table 1).
As presented in Table 1, the MCID thresholds for improvement and deterioration estimated from our sample by the outcome groups corresponded well with the literature identified thresholds.
Study Sample Stratified by Outcome Groups
Of the identified patients, 8247 (67%) had successful outcomes, 2001 (16%) undetermined outcomes, and 2102 (17%) unsuccessful outcomes. Age was on average lower in the successful and undetermined outcome groups. The sex ratio was similar in all outcome groups. There were no outstanding differences between the groups in terms of education. In the successful group, about 10% more patients were born in Sweden, and income was in general slightly higher. Successful outcome patients had on average less comorbidities, and fewer spine surgical reinterventions. Patient characteristics are summarized in Table 3.
Mean EQ-5D values at surgery and 1, 2, and 5-year follow up for the outcome groups are presented in Figure 1. As seen there, all groups were at fairly similar levels at surgery (unsuccessful: 0.31; undetermined: 0.34; successful: 0.37), but differences increased at follow-up. The successful group experienced a substantial increase which kept constant over the whole period, resulting in a follow up score of 0.80, whereas the undetermined and unsuccessful values initially increased to year 1, followed by a marked decrease to year 2, ending up at 0.56 (undetermined) and 0.42 (unsuccessful) at year 5.
The largest direct cost component in all outcome groups was inpatient costs. The monthly direct cost trajectory from 3 years before and 3 years after surgery is shown in Figure 2. The successful group was associated with lowest monthly costs over the whole time period, followed by the undetermined and the unsuccessful groups. The costs for all groups increased gradually up to the point of surgery. After surgery, there is a marked difference in the trajectories between successful and unsuccessful. For those unsuccessful, the cost increase after surgery seems to increase with the same pace as before surgery. Whereas for those successful, the increasing cost pattern is broken at the point of surgery and after 6 months monthly cost remained at a fairly stable level.
None of the groups reduced their costs enough to reach the average monthly cost level of 3 years before surgery. However, costs in the successful group were on average lower the second and third year after surgery than in the year leading up to surgery (Table 4).
Indirect costs were the major component of total costs, accounting for 82% in the successful, 86% in the undetermined, and 83% in the unsuccessful group. All groups ended up with lower indirect costs as compared with the year leading up to surgery, however, with a very small cost difference in the unsuccessful group. The successful group returned to the cost level of 3 years before surgery, but as shown in Figure 3 the costs increased markedly in the undetermined and unsuccessful groups. These results are presented in Table 4.
Three years after surgery, the cost level had not returned to the level of 3 years before surgery in neither of the groups, although the successful group was close. The differences between the groups increased markedly after surgery. The successful and undetermined group ended up at levels below the cost level of the year leading up to surgery, whereas the unsuccessful group had higher costs 3 years after surgery than in the year leading up to surgery. This is summarized in Table 4 and Figure 4.
Mean monthly total costs over the whole study period by outcome and diagnosis group are presented in Figure 5. The differences between the successful outcome group and the undetermined and unsuccessful groups are more pronounced than the difference between the two latter.
Societal costs differ depending on the outcome of lumbar spine surgery and diagnostic group. Based on this study's criteria for surgical outcome, the majority (67%) were defined to have successful outcomes, 16% were defined as undetermined, and 17% as unsuccessful. Patients with successful outcomes on average incurred significantly lower costs both before and after surgery compared with those with undetermined or unsuccessful outcomes. Total monthly costs on average increased 33% and 78% more from the third-year pre-surgery to the third post-surgery for the unsuccessful group compared with the undetermined and successful. For the successful, costs after surgery almost returned to 3-years pre-surgery levels. For the other groups, no similar downward cost trend was observed.
For patients with successful outcomes HRQoL was markedly increased and costs reduced compared to pre-surgery. For non-successful surgeries, the patients’ HRQoL was not improved or even worsened. Improved methods to identify patients with potentially higher probability of responding to surgery could lead to improved health outcomes and cost savings for society. How such patient selection could be achieved has no clear answer, but potential suggestions include improved clinical diagnostic methods, patient profile based algorithms predicting outcomes, or health care provider level organizational changes. Regardless, this study shows that payers and health care providers should perceive a value of investing in developing such methods.
Several PROMs are collected in Swespine (ODI, VAS-pain, EQ-5D, and GA). When assessing the health outcomes of an intervention it could be useful to consider several PROMs as they capture different dimensions of health with different granularity. However, for the purposes of this study it was required to categorize outcome based on one PROM. We used GA, which is a carefully validated outcome measure for spine surgery.34 One limitation of GA is that no baseline estimate pre-surgery is collected making it impossible to establish MCID values. A thorough exercise was conducted to validate the GA categorization with MCIDs for the other PROMs. GA corresponded well with the outcomes of ODI, VAS, and EQ-5D when defining the PROM outcomes as improved or failed using MCID. This suggests that GA is a valid instrument to classify patients into surgical outcome groups.
The fact that only two MCID deterioration thresholds were identified is a limitation, however not surprising, since MCIDs are usually estimated as improvement rather than deterioration thresholds. It has been suggested that this is because relatively few patients experience pain deterioration, which reduces the statistical power of such analysis.12
To our knowledge this is the first study assessing differences in societal costs related to outcome of spinal surgery. The cost estimates are mainly applicable for Sweden as only Swedish data sources were used. Though the costs in absolute terms are not directly transferrable to other geographies it is likely that observed cost patterns in relation to surgical outcome are relatively similar between developed countries with comparable health care and sick insurance systems.
Retrospective observational studies have some inherent limitations, for example, missing or incorrect data is difficult to control. However, high coverage rates of the included registers strengthen the study's validity. The 2015 coverage of PAR was 99% and that of Swespine was 77%.35
Although this study aimed to include all LBP-incurred costs, the registers did not include data on diagnostic imaging or indirect costs in terms of informal care, and the sickness insurance register includes only work absences longer than the period covered by the employer and employee (14 days).
To control for attrition bias, that is differences between patients that have and have not responded to the 2-year GA follow-up, independent t tests and chi-squared were run. We compared patient characteristics between our sample and the patients who did not respond to GA at the 2-year follow up. Small differences were found in some patient characteristics, however minor enough not to impact the conclusions (data not shown).
Patients with unsuccessful outcome consumed significantly more resources, both 3 years before and 3 years after lumbar spine surgery. This information may be helpful when creating more advanced algorithms for identifying patients that will benefit from surgery.
1. Hoy D, March L, Brooks P, et al. Measuring the global burden of low back pain
. Best Pract Res Clin Rheumatol
2. Andersson GB. Epidemiological features of chronic low-back pain. Lancet
4. Bhangle SD, Sapru S, Panush RS. Back pain made simple: an approach based on principles and evidence. Cleve Clin J Med
5. 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
6. Maetzel A, Li L. The economic burden of low back pain
: a review of studies published between 1996 and 2001. Best Pract Res Clin Rheumatol
7. Montazeri A, Mousavi S. Quality of life
and low back pain
. Handbook of Disease Burdens and Quality of Life
Measures. 2010; New York, NY: Springer, 3979–3994.
8. Buchbinder R, Blyth FM, March LM, et al. Placing the global burden of low back pain
in context. Best Pract Res Clin Rheumatol
9. Chou R, Baisden J, Carragee EJ, et al. Surgery for low back pain
: a review of the evidence for an American Pain Society Clinical Practice Guideline. Spine (Phila Pa 1976)
10. Jonsson E, Olafsson G, Fritzell P, et al. A Profile of Low Back Pain
11. Svensk Ryggkirurgisk förening. Swespine Årsrapport 2016 [Annual report 2016]; 2016.
12. Mannion AF, Balague F, Pellise F, et al. Pain measurement in patients with low back pain
. Nat Clin Pract Rheumatol
13. Fairbank JC, Couper J, Davies JB, et al. The Oswestry low back pain
disability questionnaire. Physiotherapy
14. Fairbank JC. The use of revised Oswestry Disability Questionnaire. Spine (Phila Pa 1976)
15. Fairbank JC. Oswestry disability index. J Neurosurg Spine
16. Dolan P. Modeling valuations for EuroQol health states. Med Care
17. Hagg O, Fritzell P, Nordwall A, et al. The clinical importance of changes in outcome scores after treatment for chronic low back pain
. Eur Spine J
18. Chapman JR, Norvell DC, Hermsmeyer JT, et al. Evaluating common outcomes for measuring treatment success for chronic low back pain
. Spine (Phila Pa 1976)
2011; 36 (21 suppl):S54–S68.
19. Solberg T, Johnsen LG, Nygaard OP, et al. Can we define success criteria for lumbar disc surgery?: estimates for a substantial amount of improvement in core outcome measures. Acta Orthop
20. Parai C, Hagg O, Lind B, et al. The value of patient global assessment in lumbar spine surgery
: an evaluation based on more than 90,000 patients. Eur Spine J
21. Region Västra Götaland, Regional price list 2016.
22. Karolinska University Hospital, Regional price list 2016.
23. Region South, Regional price list 2016.
25. Parker SL, Mendenhall SK, Shau DN, et al. Minimum clinically important difference in pain, disability, and quality of life
after neural decompression and fusion for same-level recurrent lumbar stenosis: understanding clinical versus statistical significance. J Neurosurg Spine
26. Haugen AJ, Grovle L, Brox JI, et al. Estimates of success in patients with sciatica due to lumbar disc herniation depend upon outcome measure. Eur Spine J
27. Ostelo RW, de Vet HC. Clinically important outcomes in low back pain
. Best Pract Res Clin Rheumatol
28. Iversen T, Solberg TK, Wilsgaard T, et al. Outcome prediction in chronic unilateral lumbar radiculopathy: prospective cohort study. BMC Musculoskelet Disord
29. Glassman SD, Copay AG, Berven SH, et al. Defining substantial clinical benefit following lumbar spine arthrodesis. J Bone Joint Surg Am
30. Maughan EF, Lewis JS. Outcome measures in chronic low back pain
. Eur Spine J
31. Cleland JA, Whitman JM, Houser JL, et al. Psychometric properties of selected tests in patients with lumbar spinal stenosis. Spine J
32. Lauridsen HH, Manniche C, Korsholm L, et al. What is an acceptable outcome of treatment before it begins? Methodological considerations and implications for patients with chronic low back pain
. Eur Spine J
33. Fritz JM, Irrgang JJ. A comparison of a modified Oswestry Low Back Pain
Disability Questionnaire and the Quebec Back Pain Disability Scale. Phys Ther
34. Hagg O, Fritzell P, Oden A, et al. Simplifying outcome measurement: evaluation of instruments for measuring outcome after fusion surgery for chronic low back pain
. Spine (Phila Pa 1976)
35. Täckningsgrader 2015 Jämförelser mellan nationella kvalitetsregister och hälsodataregistren (Coverage 2015 Comparison between national quality registers and the health data registers). Available at: http://www.socialstyrelsen.se/publikationer2015/2015-12-8
. The National Board of Health and Welfare; 2015. Accessed February 14, 2019.
Keywords:Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.
cost patterns; low back pain; lumbar spine surgery; minimal clinically important difference; quality of life; register study; surgery outcomes; Sweden