Introduction
Knee injury is a potent and potentially modifiable risk factor for knee osteoarthritis, conferring an estimated three- to six-fold increased risk of osteoarthritis later in life [
1,
2]. Data are consistent with a causal interpretation in those circumstances where significant injury has clearly preceded the development of osteoarthritis. Indeed, anterior cruciate ligament transection or destabilisation of the medial meniscus are used in animal models of surgically-induced osteoarthritis [
3]. The strength of association, plausible causal interpretation, and potential to be modified mean that injury prevention and rehabilitation are now key priorities within public health agendas for osteoarthritis prevention [
4,
5].
There remains much about the relationship between injury and osteoarthritis that requires further investigation, including the effects on osteoarthritis incidence and progression of incurring different types of injury at different ages [
6‐
8]. A fundamental requirement is knowledge of the rate of occurrence of knee injuries across the entire lifespan. As few as 20% of soft tissue knee injuries are presented to and managed in emergency departments and hospitals [
9,
10] and 50% of all knee injuries are associated with non-sporting activities [
11,
12], particularly those sustained in middle age onwards [
13].
Yet most studies of the incidence of soft-tissue knee injury have tended to focus their attention either on injuries captured within a single healthcare setting (for example, emergency department [
13,
14] or hospital admissions [
15‐
17]) or in high risk subpopulations (for example, children/adolescents [
17,
18], sports participants [
19‐
21], physically active compulsory military conscripts [
15,
22], active duty armed service personnel [
23]) or to a particular type of injury (for example, cruciate ligament injury [
17,
24,
25]. The need for population-wide estimates across multiple injury types has been identified as a gap by recent reviews [
26,
27]. The incidence rates for lower limb fractures have been the subject of several previous database studies (for example, [
28]). We, therefore, sought to provide occurrence estimates for a range of soft-tissue knee injuries presenting to primary and secondary healthcare for persons of all ages in a general population.
Discussion
This regional, population-wide study based on routinely recorded healthcare data from primary and secondary care over a nine-year period provides a broad description of the pattern of soft-tissue knee injuries presented to healthcare across the life course. Injury registries (for example, the Scandinavian ACL reconstruction registries [
33]) and surveillance networks (for example [
21]) are highly valuable sources of rich epidemiological data often on well-defined, specified injuries, in particular settings, or in special or high-risk populations. However, broader studies, such as the present one, are also justified on the grounds that knee injury represents a wide spectrum of exposure at multiple levels of the ‘injury pyramid’ (for example, hospitalisations, accident and emergency attendances, general practitioner visits, and those not presented to formal healthcare) [
34,
35]. Minor injuries may be more prone than severe injuries to misclassification [
35] and they may be less strongly related to future risk of osteoarthritis [
2]. Nevertheless, we argue that they are still potentially relevant to understanding population exposure to injury and subsequent osteoarthritis development. Previous similarly broad population-wide studies have been either limited in size [
11,
14,
36] or have focused on a particular class of injury [
37].
Although based on routine recording as opposed to standardised, reference standard diagnostic criteria, our mean annual incidence estimates for clinically diagnosed soft tissue knee injury appear consistent with incidence rates from previous studies. Additional file
2 provides a data extraction table with key features in study design and published estimates from the current study and previous comparable studies. Our overall annual incidence (720 per 100,000) is higher than the incidence rate of 278.7 per 100,000 reported in the Olmsted County study. The Olmsted County study used a narrower case definition of new, isolated, acute injuries based on final diagnosis [
11]. Conversely, our estimates are somewhat lower than the 1,080 and 1,160 per 100,000 estimates from studies conducted in Finland and Denmark, respectively, in the mid-1980s [
14,
36]. Those previous studies included fractures as well as soft-tissue injuries.
Most comparable data exists for cruciate ligament injury. The observed peak age appears marginally younger in our study than in a national, population-wide study of ligament injuries from insurance data in New Zealand [
37]. It is also younger than that reported in Scandinavian cruciate ligament reconstruction registries [
38] and from cruciate ligament injury estimates from hospital in- and outpatient data in the Swedish National Patient Register [
25]. It is possible that this may reflect earlier diagnoses recorded in primary care captured in SHR. Yet our annual incidence estimate for cruciate ligament sprain (71 per 100,000) closely approximated the incidence rate from Swedish hospital data [
25]. They were also similar, within the 10- to 64-year age band, to the figure of 81 per 100,000 reported by Frobell
et al. [
24]. That study was based on injuries presenting to an emergency unit in the same geographical region but using magnetic resonance imaging confirmation of all suspected cases of acute anterior cruciate ligament injury. This peak of severe soft-tissue knee injuries in adolescence (10 to 19 years) and emerging adulthood (18 to 29 years) [
39,
40] most likely is associated with sports participation. The substantial seasonal variation in injury rates, especially among men, would be consistent with this. Although the current study gathered no data on the inciting event or mechanism of injury, the peak months coincide with the skiing (March) and football seasons (spring-autumn) with dips in vacation periods and breaks in the sporting seasons.
Our study confirms and extends the distinctive age-related pattern in injury rates observed in women [
9,
37] which show higher rates in adulthood after a decline from peak rates in late adolescence. In the present study, the increase was seen to begin in the age band 35 to 39 years and was evident across several injury codes, notably cruciate and collateral ligament sprains and injuries to multiple structures. In the same age band we observed a temporary interruption in the otherwise smooth decline in incidence rates among men. We speculate that this pattern of findings reflects an increased risk of injury during childrearing years (the mean maternal age at childbirth in Sweden is 31 years). A strong association between parity and subsequent risk of knee osteoarthritis has been reported in Danish registry data [
41] and recently confirmed in US females in the Multicenter Osteoarthritis Study (MOST) cohort [
42]. Injury was not taken into account in the former study and was not implicated as an important mechanism in the latter although this did rely on recall of lifetime injury at age 50 to 79 years. Injury risk in this stage of life may warrant closer investigation as part of a strategy for injury prevention in the general population.
A number of limitations in the present study should be noted. Firstly, we have not undertaken any independent verification of the accuracy of injury diagnoses within SHR although we have shown that our estimates for cruciate ligament injury are quite consistent with studies using more stringent reference standards [
24]. However, this accuracy cannot be assumed to apply across all types of injury. Accurate diagnosis of soft-tissue knee injuries is challenging even in specialist settings. Frobell
et al. [
24] found, for example, that half of all patellar dislocations identified on MRI were missed by clinical examination in patients presenting to an orthopaedic emergency unit with acute torsional injury and rapid effusion. We would expect misclassification of diagnoses to affect estimates for specific injury codes more than the overall estimates for all soft-tissue knee injury codes combined. Nevertheless, healthcare records are likely to under-estimate the true occurrence of minor knee injuries [
43]. This tendency toward conservative estimates could be exacerbated by restricting case ascertainment to those consulting and diagnosed by a physician although we found very few persons in a calendar year who had only consulted a non-physician for their knee injury (data not shown). Secondly, the lack of data on side-specific injury in the register data contributes to being unable to count more than one injury per person within the same calendar year. We, therefore, did not attempt to calculate an event rate based on injuries per person-time. The extent to which person-based incidence rates will be lower than these event rates depends on what proportion of injured individuals experience a recurrent or contralateral injury within the same calendar year. For severe soft-tissue knee injuries this rate of contralateral injury is likely to be low: previous studies suggest between 0.4% and 2.5% of individuals with an anterior cruciate ligament injury would experience the same injury on the contralateral knee [
44]. To our knowledge there are no reliable estimates of recurrence and contralateral injury for all of the other types of knee injury we have investigated. Thirdly, our estimates are averages of rates for the period 2004 to 2012. However, we found no strong evidence of a trend in injury rates between 2004 and 2012 on visual inspection. During this period, the resident population in Skåne region increased by 9% (compared with 6% for Sweden as a whole) driven mainly by net immigration from abroad [
45]. No dramatic changes in age composition were seen and, therefore, we think it is unlikely that important underlying changes in injury rates were masked by our approach to use average rates. Further, during the time period there was, to the best of our knowledge, no shift in policy, diagnostic coding or methods of diagnostics that would substantially have influenced case ascertainment. Fourthly, we did not include fractures in our estimates. These have been the subject of previous reports but are clearly also relevant to future osteoarthritis risk. Within the Skåne Healthcare Register, the annual incidence rates of fractures of the distal femur, patella and proximal tibia/fibula are, respectively, 13, 25 and 36 per 100,000 persons 20 years old and older (Rosengren B
et al., submitted manuscript). Finally, our data are from Southern Sweden and their generalizability to other populations cannot be assumed. Regional data on lifestyle factors associated with the incidence of knee injury is limited but Sweden has higher than average rates of educational attainment [
46] and relatively high levels of regular physical activity and sports participation compared with other European countries [
47]. Although obesity has been increasing over recent decades [
48,
49], the prevalence of obesity in Sweden has historically been in the lowest tertile internationally [
50]. Further, the adjustment for missing data to private care is a crude adjustment as the potential influence of socioeconomic status and profession for knee injury and health care seeking behaviour is not accounted for.
Our findings carry several implications for understanding osteoarthritis risk. Firstly, severe knee injury such as cruciate ligament injury, while known to be a potent risk factor for osteoarthritis, is nevertheless a relatively infrequent exposure (71 per 100,000 persons per year) compared with the incidence rate of diagnosed knee osteoarthritis (650 per 100,000 persons per year) [
51]. Causal mechanisms made up of other components must be responsible for a large proportion of osteoarthritis cases, even allowing for ligament injuries not captured in routine healthcare. The effect of minor injuries on osteoarthritis risk remains a challenge for epidemiologic research in this field. They are more numerous in the population than severe injury but are subject to greater misclassification (for example, due to inaccurate recall or being recorded as ‘flare-ups’ of osteoarthritis in older adults). Injuries are potentially recurrent events. One advantage of reporting annual incidence estimates as opposed to first-ever incidence rates is that one can see clearly the continued presentation of acute meniscal tears, and unspecified and superficial injuries into middle and old age. One question that arises from this is whether minor injury after osteoarthritis disease initiation and diagnosis contributes to the progression of osteoarthritis. Finally, the empirical induction period [
52] between severe injury and osteoarthritis diagnosis may be several decades. The peak age for severe knee injury is 15 to 19 years while the incidence of diagnosed knee osteoarthritis continues to increase to age 75 years [
51]. Such long intervals are beyond the current scope of electronic healthcare databases and injury registries. This fact is an argument in favour both of their long-term investment and in alternative study designs using markers of disease initiation as earlier outcome measurement.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
GP, CB and ME conceived of the study and participated in its design. CB carried out the analyses. AJ, RF contributed to the analysis and interpretation of the data. All authors contributed to the drafting of the manuscript and read and approved the final manuscript.