Geographical variation in common orthopedic procedures in Norway: a cross-sectional population based study of 175,973 procedures

Background and purpose Standardized surgery rates for common orthopedic procedures vary across geographical areas in Norway. We explored whether area-level factors related to demand and supply in publicly funded healthcare are associated with geographical variation in surgery rates for six common orthopedic procedures. and methods Cross-sectional population based study of hospital referral areas in Norway. We included all adult admissions for arthroscopy for degenerative knee disease, arthroplasty for osteoarthritis of the knee and hip, surgical treatment for hip fracture, and decompression with or without fusion for lumbar disc herniation and lumbar spinal stenosis in 2012-2016, a total of 175,973 admissions. Variation in age and sex standardized rates was estimated using extremal quotients, coecients of variation and systematic components of variance. Association between surgery rates and urbanity, unemployment, low-income, high level of education, and mortality, was explored with linear regression analyses.


Introduction
The number of orthopedic surgical procedures per population, i.e. the surgery rate, vary for all common orthopedic procedures in Norway (1-3), as well as in most other countries (4)(5)(6).For some procedures, like hip fracture repair where the e cacy of the treatment and positive outcomes are not disputed, the variation is small and likely to describe differences in incidence.However, surgery rates for several other orthopedic conditions show marked geographic variation.
Norway is a country with a universal health system that strives to provide equitable access to healthcare for all inhabitants.The prevalence of diseases in the relatively homogeneous population is assumed to vary minimally with area of residence.What causes variation in surgery rates under these conditions?Socioeconomic factors are known to impact health and the demand for healthcare (7).Supply of health services can have an in uence on utilization as well (7).We examine these factors in the setting of orthopedic healthcare in Norway.
Geographic variation has been a eld of interest in international research for a long time (8).It can be caused by differences in morbidity or preferences in the patient population, but it can also be unwarranted as a result of differences in medical practice and supply of procedures (9).The risk of unwarranted variation is overuse, underuse or wrong utilization of the health services (10).Since geographic variation can be system dependent, it is necessary to examine factors associated with such variation in Norwegian healthcare.
The aim in this study is to explore whether area-level factors related to demand and supply in publicly funded healthcare are associated with geographical variation in surgery rates for six common orthopedic procedures.

Setting
The public health care system in Norway has an equal distribution of monetary resources, and uniform training and licensing for healthcare staff.Most patients are treated at the public hospital serving their residential address.Orthopedic surgery is provided free of cost to patients in the public health care system.Some private hospitals have government funding through contracts with the public regional health authorities, and are thus part of the public health care system.A minority of orthopedic patients receive treatment in private hospitals without public funding, where the patients or their insurance providers cover expenses of surgery.These hospitals do not report activity data to the national registry.The quality of treatment is considered equivalent in public and private hospitals, and the presumption is backed by a recent study showing equivalent effectiveness of spine surgery in public and private hospitals (11).
Non-operative treatment for orthopedic conditions is available in the public health care system, but referral pathways vary and display great inconsistency.Data on the structure of and access to nonoperative treatment was not available in this study.

Study design and Data sources
This is a cross-sectional population based study based on data from the Norwegian Patient Registry (NPR).NPR covers all publicly funded specialist healthcare services in Norway, including treatment in private institutions and specialists contracted to regional health authorities (12).The data supplied by the NPR included all admissions in the period of 2012 -2016 for de ned orthopedic diagnoses and procedures.We retrieved online available data on nine socioeconomic variables possibly associated with geographic variation from the Statistics Norway (SSB) national register.We retrieved online data on public and private hospitals from three national medical quality registers (the Norwegian Arthroplasty Registry, the Norwegian Hip Fracture Registry and the Norwegian registry for spine surgery).The Norwegian Medical Association provided data on the numbers of surgeons according to area of residence.

Patient selection
We included all cases who underwent one of the following surgical procedures; arthroscopy for degenerative knee disease, arthroplasty for osteoarthritis of the knee, arthroplasty for osteoarthritis of the hip, surgical treatment for hip fracture, decompression with or without fusion for lumbar disc herniation and decompression with or without fusion for lumbar spinal stenosis.Additional le 1 shows the inclusion criteria in terms of the International Classi cation of Diseases (ICD-10) and NOMESKO Classi cation of Surgical Procedures (NCSP) codes (see Additional le 1).Cases 18 years and older were included for all the above mentioned procedures, with the exception of arthroscopy for degenerative knee disease, for which cases aged 50 years and older were included.We chose this age group to maximize inclusion of those with complaints related to osteoarthritis, who are not expected to bene t from arthroscopic procedures (3).Patients where municipality of residence was not available or registered as 'abroad' were excluded (less than 1 % of admissions).The 19 geographical areas correspond to hospital referral areas (HRAs) of health trusts in Norway.Treatment was recorded according to patients' area of residence (postal code), and not based on which hospital or health trust delivered the treatment.

Statistics
For each of the six procedures we calculated surgery rates per 100,000 adult population (relevant age group), per year and hospital referral area.Rates were directly standardized for age and gender using the relevant age group of the population of Norway on 1 January 2016 as the standard population.We extracted necessary population data from SSB tables 07459 and 10826.
The extent of variation in standardized surgery rates for each procedure across areas was estimated using the extremal quotient (EQ; maximum rate divided by minimum rate), the coe cient of variation (CV; standard deviation of rates divided by mean of rates), and the systematic component of variance (SCV) (13,14).
To quantify supply related factors we analyzed the impact of the numbers of hospitals and surgeons.We calculated the number of hospitals per 100,000 population for each hospital referral area and procedure.
Private hospitals included both hospitals with partial public funding and hospitals with no public funding.We performed the main analyses using data for the total number of hospitals per area, including both public and private hospitals.In addition, we performed sensitivity analyses to detect the possible effect of the number of public and private hospitals separately.We calculated the number of surgeons per 100,000 population for each hospital referral area.For knee and hip procedures, we used data on orthopedic surgeons listed according to their residential address in 2016.For spine procedures, the number of surgeons was the sum of orthopedic surgeons and neurosurgeons listed according to their residential address in 2016.We also performed sensitivity analyses to detect possible effects of the numbers of orthopedic surgeons and neurosurgeons separately.
To quantify demand related factors we calculated proportions and rates per hospital referral area for nine socioeconomic variables.Based on a correlation analysis and an assessment of relevance, we chose to include the following factors to quantify demand; the proportion of population in urban areas (SSB table 05212), the unemployment rate (SSB table 10594), the proportion of persons living in low-income households (SSB table 06947), the proportion of persons with a high level of education (SSB tables 09429 and 09434), and the mortality rate (SSB table 12983).
We performed a linear multivariate regression analysis for each procedure.We evaluated model performance using adjusted R-squared, and considered a p-value <0.05 statistically signi cant.
We used R version 4.03, a free software program for statistical computing, for all statistical analysis.
Ethics, funding, and potential con icts of interest Patients were not involved in the conception, design, or conduct of this research.Helse Førde health trust holds a license from the Norwegian Data Protection Authority to process health data for the national healthcare atlas service from 2016.Since 20 July 2018, the basis for the processing of data has been the General Data Protection Regulation.One of the authors (MH) received support from Helse Førde health trust in Norway through grant number 38595/2019.No con icts of interest were declared.

Results
Table 1 shows the numbers of procedures, the surgical rates and their extremal quotients, and the measures of variance.The number of knee arthroscopies decreased during the period from 8,857 arthroscopies in 2013 to 4,172 in 2016.The numbers were stabile for the other procedures.
Arthroscopy for degenerative knee disease was the procedure with the highest level of variation (SCV 10.3, 4.5-fold) and the highest standardized rate among the eligible population.There was also considerable variation of the rates for decompression for lumbar spinal stenosis (SCV 4.9, 2.1-fold) and lumbar disc herniation (SCV 3.8, 2.6-fold).Arthroplasty for osteoarthritis of the knee showed moderate variation (SCV 2.6, 1.9-fold), while the variation was relatively low for arthroplasty for osteoarthritis of the hip (SCV 0.8, 1.4-fold).Surgical treatment for hip fracture showed least variation in surgery rates (SCV 0.2, 1.2-fold).Figure 1 illustrates the geographical variation in standardized surgery rates for the included procedures.
Table 2 shows the national mean, lowest and highest values per hospital referral area for factors associated with variation.The association between surgery rates and socioeconomic and supply factors were weak for arthroscopy for degenerative knee disease and decompression for lumbar disc herniation and spinal stenosis.Higher rates of arthroscopy for degenerative knee disease were associated with more orthopedic surgeons (adjusted coe cient 24.8, 95% CI 2.7 to 47.0), and a smaller proportion of the population living in urban areas (adjusted coe cient -13.3, 95% CI -25.4 to -1.2).For decompression for lumbar spinal stenosis, higher rates were associated with more hospitals (adjusted coe cient 22.4, 95% CI 4.6 to 40.2), and a larger proportion of the population living in urban areas (adjusted coe cient 2.1, 95% CI 0.4 to 3.8).For decompression for lumbar disc herniation, higher rates were associated with lower mortality (adjusted coe cient -192.6,95% CI -384.2 to -1.1).For lumbar spine procedures, a sensitivity analysis including the number of neurosurgeons and orthopedic surgeons as separate variables did not strengthen the association.
Surgery rates for knee and hip arthroplasty, and surgical treatment for hip fracture were not associated with the supply and demand factors included in this study.
A sensitivity analysis of public and private hospitals included as separate variables did not render notable associations for any of the procedures included in the study.
Additional le 2 includes full univariable and multivariable models of the linear regression analysis with standardized surgery rates as outcome.

Discussion
The main nding in this study was that variation in standardized surgery rates ranged from almost none to tenfold across the six orthopedic procedures in publicly funded Norwegian hospitals between 2012 and 2016.The variation was very high (SCV >10) for arthroscopy for degenerative knee disease for patients 50 years and older, moderate (SCV 3-5) for the decompression procedures for lumbar disc herniation and lumbar spinal stenosis, and low (SCV <3) for arthroplasty for osteoarthritis of the knee or hip, and surgical treatment for hip fracture.
The rates for arthroscopy for degenerative knee disease decreased markedly in all hospital referral areas.This indicates that practice is being adapted to recommendations in new guidelines (15).Persisting variation in arthroscopy rates might suggests, however, that different areas are adapting asynchronously.Other European countries, as well as in the US, report similar patterns of decrease (4,16,17).
We observed that higher rates of arthroscopy for degenerative knee disease had a weak association to more orthopedic surgeons and a larger proportion of the population living in rural areas.The reason for this could not be further explored in this study, but may relate to local practice differences, availability of services or speci c lifestyle demands.
Studies from Ireland, Korea and the UK (4,5,18) report an increase in knee arthroplasty with deprivation.
Rates for knee and hip arthroplasty have also been associated with numbers of surgeons, hospital capacity (5,18,19), and socioeconomic factors (19,20).Our ndings align with those reported by de Pina et al. (6) and Mäkelä et al. (21), who found no association with supply or socioeconomic factors.The relatively low variation in arthroplasty for osteoarthritis of the hip is similar to that seen in Finland, where variation decreased from threefold in 1998 to 1.9-fold in 2005 (21).
As surgical treatment of a hip fracture is considered both effective and necessary, we did not expect to nd any association with supply or demand related factors.The small amount of variation in hip fracture surgery is in line with other reports (4,22) and variation is generally considered to re ect the relatively small differences in incidence.
Variation in spine surgery procedures, similar to that found in the current study, has been observed both internationally and in Norway (2,4,23), and is related to more uncertainty about the effectiveness of the procedures (24).Bederman et al. ( 23) found that higher rates were associated with lower income, but found no association with supply of physicians.The number of hospitals was negatively associated with spine surgery rates in Ireland (4).We found no notable association between spine surgery rates and the area-level supply or demand related factors analyzed in our study.

Strengths and limitations
The main strength of this study is the availability of nationwide individual-level data on orthopedic procedures performed at publicly funded hospitals, which gives a comprehensive picture of variation in public healthcare.
The majority of orthopedic surgery, and all hip fracture surgery, in Norway is performed in hospitals with public funding.The proportion of private hospitals ranged, in our study period, from 12 to 30 percent (arthroscopy for degenerative knee disease 19 out of 64 hospitals, decompression for lumbar disc herniation and lumbar spinal stenosis 10 out of 43, arthroplasty for osteoarthritis of the knee 9 out of 60, and arthroplasty for osteoarthritis of the hip 7 out of 60).Most private hospitals perform both publicly funded and privately funded treatment, but report only publicly funded activity to the NPR.Accordingly, absence of data on privately funded activity is a limitation in our analysis, as the data re ects most but not all activity.For some of the studied procedures, private activity has been reported to increase variation (3,11).Further, the number of hospitals is a crude measure, as it does not account for differences in hospital capacity.
Non-operative treatment is an important option for many orthopedic conditions.While this is generally acknowledged, the lack of uniformity in conservative care in the Norwegian public health care system might be a driver of variation in surgery rates.We consider it a limitation that we could not account for this, due to the availability of data.
The number of surgeons includes practicing surgeons nationwide, but remains a rough estimate as the surgeons area of residence might not always coincide with the area of work, especially in the larger capitol area.Furthermore, we have not included the varying number of surgeons in training, who independently perform some of the included procedures in many hospitals.The proportion of surgeons in training varies between geographical areas, but we did not have access to these data.Finally, some surgeons are not permanently employed, but contracted through sta ng agencies.And data quantifying this are not available.
It is a limitation of the study design that data on demand-related factors were available on area, and not individual, level.This precludes us from exploring variation in surgery rates in light of demand in the patient groups involved.Hence, the analysis of associations between surgery rates and demand-related factors only gives rough estimates of this relationship.Nevertheless, geographic variation research conducted at a national level, using area-level units of analysis while crude, is important in signaling potential inequality and treatment underuse or overuse (25).

Interpretation
The associations found in our study between area-level demand and supply-related factors and surgery rates were weak.We might assume that the included orthopedic surgery rates are not notably associated with regional numbers of hospitals or surgeons, or with income, education, unemployment, health level (estimated by mortality as a proxy) or with urbanization.This may be a result of longstanding efforts in Scandinavia to facilitate equity by universal health care and tax paid education.
On the other hand, one could argue that area-level factors re ecting demand are not detailed enough to detect the associations that may exist between socioeconomic factors and surgery rates.Qualitative analyses and more detailed multilevel analyses, that include individual-level data of factors known to in uence surgery rates, are needed to further explore this.This is being done in two ongoing Norwegian studies on geographic variation in hip and knee arthroplasty and lumbar spine surgery, using a mixed methods design consisting of multilevel analyses on registry data and qualitative data collection from focus groups (general practitioners) and individual interviews (patients and surgeons).
While sociodemographics and healthcare supply do not seem to explain variation in common orthopedic procedures based on our ndings, we acknowledge that there are more factors known to impact utilization of health care.These include differences in preferences among surgeons, the effect of shared decision-making, as well as capacity and structural aspects of the health care system (7,26).

Conclusions
Variation in surgery rates was particularly high for arthroscopy for degenerative knee disease, such as meniscal tears and osteoarthritis, in Norway between 2012 and 2016, and these rates decreased considerably during the ve-year period, in line with guidelines and international practice.Area-level factors re ecting socioeconomic circumstances, health and supply have a weak association to orthopedic surgery rates.Whether this re ects the equity of universal health care services, or if area-level factors are not detailed enough to detect an existing association is being explored in two ongoing Norwegian studies.

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download.

Figures Figure 1
Figures

Table 1
Numbers of procedures, surgical rates and variation statistics Rate; number of procedures per 100 000 adult population, EQ; extremal quotient, CV; coefficient of variation, SCV; systematic component of variation.* Patients 50 years and older

Table 2 :
National mean, lowest and highest values per hospital referral area (HRA) for factors associated with variation.