The burden of hip and knee osteoarthritis in Finnish occupational healthcare

Background Osteoarthritis (OA) is a leading cause of disability and pain especially among older adults, but it is also known to affect working age individuals, often leading to reduced productivity and increased healthcare usage. The aim of this study was to determine the burden of hip and knee OA in Finnish occupational healthcare. Methods This was a retrospective registry study utilizing the electronic medical records of the largest private and occupational healthcare provider in Finland. All consented patients with hip or knee OA were identified. A subcohort of occupational healthcare (OCH) patients was then compared to an age- and gender-matched control group without OA. Patient demographics including comorbidities were determined and healthcare contacts, medication prescriptions, and sick leaves were compared between the two groups. The study period was from January 1st, 2012 to April 30th, 2020. Results 51,068 patients with hip or knee OA were identified (all OA cohort) and 35,109 of these formed the occupational healthcare subcohort. Most of the OA patients were female and belonged to the age group 50–59 years. The point prevalence of hip/knee OA at the end of the study period was 5.6% for the occupational healthcare subcohort. OA patients had 2.2 times more healthcare contacts and 2.8 times more overall sick leave days compared to the age- and gender-matched control cohort. Etoricoxib was the most commonly prescribed medication at OA-related visits (21.8% of patients). Opioids were prescribed to 10.6% of patients at OA-related visits and the most prescribed opioid was a combination of codeine and paracetamol (4.8% of patients). 5054 OA patients (14.4%) had a contraindication for non-steroidal anti-inflammatory drugs (NSAIDs). Conclusions This retrospective registry study utilizing real-world data provides new evidence on the disease burden of hip or knee osteoarthritis from the electronic medical records of Finnish occupational healthcare customers. OA patients had more comorbidities, more healthcare contacts, more sick leave days, and more analgesic prescriptions compared to an age- and gender-matched control cohort without OA. Supplementary Information The online version contains supplementary material available at 10.1186/s12891-021-04372-9.

Health and Wellness Survey in the US found that a substantial proportion of workers suffer from OA pain, and workers with OA have significantly higher healthcare resource utilization and costs compared to those without OA [6,7]. Further studies have shown that working age adults with OA have a lower employment rate compared to the age group without OA [3], and that OA patients are at an increased risk of work loss due to illness or disability [8,9].
Even though previous research has clearly demonstrated that OA is a substantial burden also for the working age population, more information is needed on the clinical characteristics, treatment options and healthcare resource utilization of employed OA patients specifically. The aim of this study was to investigate these parameters in patients with either hip or knee OA compared to age-and gender-matched controls, utilizing the electronic medical records (EMRs) of occupational healthcare customers at Terveystalo, which is Finland's largest private and occupational healthcare provider. Our main hypothesis was that the diagnosis of OA is a significant burden in Finnish occupational healthcare (OCH), demonstrated by the higher use of medical services and more sick leave days recorded for OA patients compared to controls without OA.

Methods
In Finland, all employees are entitled to preventive occupational healthcare financed and arranged by the employer. Additionally, employers can choose to provide employees with access to medical care, and care coverage at least at the general practitioner level is often included. Electronic medical records (EMRs) of Terveystalo, the largest private and occupational healthcare provider in Finland, were utilized in this retrospective registry-based study. Terveystalo has about 300 clinics around Finland, and in 2019 the total number of medical doctor visits was 3.7 million, or approximately 15% of the nationwide total visits [10]. The study period was from January 1st 2012 until April 30th 2020.
Two OA patient cohorts were formed: the all OA cohort consisting of consented patients with hip or knee OA, and the OCH subcohort from the OCH patients with hip or knee OA. The study inclusion criteria were diagnosis of either hip or knee OA (ICD-10 codes M16* or M17*, respectively), active patient consent available, being adult (≥18 years) at the date of first diagnosis, and health registry data available and accessible. If a patient was diagnosed with both hip and knee OA, the first recorded diagnosis was used for analyses. An additional inclusion criterion for the OCH subcohort was active OCH status. Patients were followed from the first M16/ M17 diagnosis (index date) until the end of follow-up, which was defined as the end of the study period (30-Apr-2020), death or end of OCH customership. Separate OCH follow-up periods with a maximum of 3 months in between were combined into a single period. If the break between OCH periods was more than 3 months, only the first continuous follow-up period containing the OA diagnosis was included in the analyses.
In order to determine the burden of hip and knee OA, a one-to-one age (birth year), gender, and length of follow-up matched control cohort of patients without OA was formed from the OCH patients. The controls were followed starting from the index of the corresponding cases. The exclusion criteria for the control cohort were diagnoses of hip, knee, or other OA (ICD-10 codes from M15* to M19*) or arthralgia (M25.5).
The EMRs included in this study were diagnoses, visits, procedures, prescriptions, laboratory results, sick leaves, and demographic characteristics.

Prevalence
The point prevalence at the end of the study was calculated by dividing the number of prevalent (i.e. alive) and consented OCH patients by the number of prevalent and consented patients recorded at the Terveystalo EMRs on the 30th of April 2020.The point prevalence was reported by gender and type of OA in six age groups.

Comorbidities and treatment practices
The frequency of comorbidities and co-diagnoses was assessed from the EMRs based on ICD-10 codes (threecharacter level) overall during the follow-up period. A patient was counted only once even if the same diagnosis code was recorded multiple times. Of special interest were OA-related diagnoses (M15-M19), arthralgia (M25.5), rheumatoid arthritis (M05-M06) and contraindication diagnoses for non-steroidal anti-inflammatory drug (NSAID) use.
OA medication prescriptions for the OCH subcohort were reported separately for OA-related visits and all visits. For controls, prescriptions from all visits were reported. Of special interest were pain medication prescriptions and prescriptions for medications that are contraindications for NSAID use.
An NSAID contraindication was recorded for those patients whose EMRs contained either a diagnosis (ICD-10-code) that is a contraindication for NSAID use, or a prescription for a medication that is a contraindication for NSAID use (or both) according to the national treatment guidelines [11] during the follow-up period (see Additional file 1 for list of diagnoses and medications). Diagnosis codes and prescriptions from all healthcare contacts, not only OA-related contacts, were taken into account here.

Healthcare resource utilization
For OA patients, healthcare resource utilization (HCRU) consisted of both total and disease-specific healthcare contacts and sick leaves. For disease specific HCRU, an ICD-10 diagnosis code M16* or M17* was required as a diagnosis for the visit/sick leave. The controls had no OA-specific HCRU, so only total HCRU was calculated for them. All healthcare contacts with diagnosis codes (including e.g. physician's telephone consultation) were included in the calculations for healthcare visits. It should be taken into account that only one primary diagnosis code is recorded for each sick leave period in the Terveystalo EMRs, but it is possible to record multiple diagnosis codes per visit, and for healthcare contacts all recorded diagnosis codes were taken into account in the analysis.
Overlapping records of sick leaves were combined into one continuous sick leave period and counted only once, whereas subsequent, non-overlapping sick leaves were counted individually. Sick leaves beginning prior to the index date or ending after the end of follow-up were included, however, only the sick leave days occurring during the follow-up period were counted.
HCRU data is presented per patient-year (cumulative number of events/cumulative follow-up). Hence, the data is adjusted for differences in follow-up that would otherwise bias those with longer follow-up to more likely have higher HCRU.

Statistical methods
Data management and statistical analysis were performed using Python 3.7.4 with pandas, spicy stats, and other standard data analysis packages. Only pseudonymized data was used in the analyses to protect patient privacy. The Chi-squared test was used to test for differences in frequencies for categorical variables (comorbidities and prescriptions). The fold-changes between the cases and controls for the comorbidities were calculated by dividing the frequency of cases by the frequency in controls. P-values less than 0.05 were considered statistically significant and no multiple testing correction was applied.

Study cohort formation and baseline characteristics of the patients
The number of consented patients at Terveystalo EMRs at the date of data extraction (30-Apr-2020) was 1,134, 643, of whom 412,086 patients were OCH customers ( Fig. 1). The all OA patient cohort consisted of 51,068 patients diagnosed with either hip or knee OA (n = 9040 hip and n = 42,028 knee) during the study period. For the 3025 patients who had both hip and knee OA, the first recorded diagnosis was used for further analyses. An OCH subcohort of 35,109 patients (n = 6416 hip and n = 28,693 knee) with OCH customership was formed from the all OA cohort. A randomly selected, one-to- Fig. 1 The formation of the study cohorts. The all OA cohort (n = 51,068) including adult patients diagnosed with hip or knee OA was formed from the total number of consented patients at the end of the study period. The occupational healthcare (OCH) subcohort (n = 35,109) included those OA patients with occupational healthcare customership. The control cohort without OA (n = 35,101) was formed from the occupational healthcare customers in the total number of consented patients one age (birth year), gender and OCH follow-up length matched control cohort of patients without OA was formed for the OCH subcohort. No suitable controls were found for eight elderly OA patients with OCH customership, so the control cohort was eight patients smaller.
Just over half of the patients included in the study were female ( Table 1). The age at index was 56.6 years for the all OA cohort and 53.3 years for the OCH subcohort. Most of the patients belonged to the age group 50-59 years. Body mass index (BMI) was available more often for the OCH subcohort than the all OA cohort patients. For those with BMI data available, the mean BMI in the OA cohorts was 30.1 for the hip/knee OA patients, and 27.6 for the patients in the control cohort. Patients in higher BMI groups were seen more often in the OA cohorts, for example, 7.0% of knee OA patients had a BMI higher than 40, compared to only 2.1% of the control cohort patients. Smoking status was only available for between 23.4 and 33.7% of the patients. For those with smoking status available, the majority (68.1 to 74.0%) were non-smokers. The mean follow-up time for the OCH subcohort was 3.4 years (3.4 years for hip and 3.5 years for knee OA patients), and 3.5 years for the control cohort (Table 1).

Comorbidities and other diagnoses
Comorbidities and other co-diagnoses were assessed from the EMRs based on ICD-10 codes (three-character level) overall during the study follow-up period. The most common comorbidity diagnosis was J06 (ICD-10), acute upper respiratory infection, which was detected in the EMRs of 63.2% of the OCH subcohort patients and 43.8% of the controls (Table 2). M54, dorsalgia was the second most common comorbidity, and it was 1.8-times more common in the OCH subcohort compared to controls. The big difference seen between OA patients and controls for the diagnosis code M25 (Table 2), is explained by the fact that the diagnosis code M25.5, arthralgia, was an exclusion criterion for the control cohort. The prevalence of arthralgia diagnosis in the OCH subcohort was studied in more detail. Of the OCH subcohort patients, 28.9% were diagnosed with arthralgia prior to OA (n = 10,155), and the median time from arthralgia Mean follow-up (years) n/a n/a n/a n/a n/a n/a 3.   Table 2, there were several rarer comorbidities with a prevalence less than 10%, such as type 2 diabetes mellitus (ICD-10 code E11) and chronic obstructive pulmonary disease (COPD; ICD-10 code J44). The prevalence of type 2 diabetes mellitus was 6.5% for hip OA patients, 8.2% for knee OA patients, 7.9% for hip/knee OA patients, and 4.5% for controls (fold change 1.8, p-value < 0.001). COPD was also more common in the OA patients than controls, but the overall prevalences were lower (1.4% for hip OA patients, 1.2% for knee OA patients, 1.2% for hip/knee OA patients, and 0.7% for controls; fold change 1.8, p-value < 0.001).

OA prevalence
The point prevalence of hip or knee OA in the OCH subcohort at the end of the study period was 5.6% (5.2% for males and 5.9% for females). Knee OA was 5.5-times more common with a point prevalence of 4.4%, compared to 0.8% for hip OA prevalence. In addition, 1325 patients (0.3%) had both hip and knee OA. The prevalence of OA for both males and females increased with age (Fig. 2). The highest prevalence numbers were seen in the age group 60-64 years, where the prevalences of hip OA were 2.2 and 2.3% for males and females, respectively, and the corresponding numbers for knee OA were 10.9 and 12.8%.

Current treatment practices
OA medication prescriptions for the OCH subcohort were extracted separately from OA-related visits and all visits to Terveystalo. The most prescribed OA medication was etoricoxib, which was prescribed to 21.8% of OA patients at OA-related visits, and to 52.7% when all visits were taken into account ( Table 3). All of the OA medications listed in Table 3 were prescribed to OA patients more often than to controls. The biggest differences between OA patients and controls were seen in the prescriptions for chondroitin sulfate, glucosamine and hyaluronic acid, which were prescribed almost exclusively to OA patients compared to controls (Table 3).
We also looked at the prescription data based on the step-wise approach described in the national treatment guidelines [12], where the first-line pharmaceutical treatment for hip or knee OA is paracetamol, followed by NSAIDs and finally opioids. Nearly half (47.6%) of the patients had no pain medication prescriptions from OA-related visits and 1.9% had a prescription for paracetamol only, and no NSAID or opioid prescriptions. 40.0% had an NSAID prescription but no opioid prescriptions, and the remaining 10.6% had an opioid prescription. The most prescribed opioid was the combination of codeine and paracetamol, which was prescribed to 4.8% of the OA patients at OA-related visits (6.8% of hip OA patients, and 4.4% of knee OA patients; Table 3). When all visits were taken into account, a total of 12,987 OA patients (37.0%) had an opioid prescription in their EMRs, which was 2.5-fold more than was recorded for controls (n = 5155; 14.7%).
Hyaluronic acid was prescribed to nearly 3% of OA patients in the OCH subcohort (all visits, Table 3). However, hyaluronic acid is also available as a medical device and these treatments are not seen in the prescription data. Therefore, we also looked at the joint injection procedure codes from the EMRs. The procedure code TNX10, which is used most often for joint injections in Terveystalo EMRs, was recorded for 1994 OA patients (5.7%) in the OCH subcohort (n = 1827, 6.4% knee, and n = 167, 2.6% hip) compared to only 255 controls (0.7%).
NSAIDs were the most used pain medication for OA patients, but several patients also had NSAID contraindications. In the OCH subcohort, the EMRs of 4026 patients (11.5%) had an ICD-10 code that is a contraindication for NSAID use. Furthermore, 1857 patients (5.3%) had a prescription for a medication that is a contraindication for NSAID use (see Methods for list of diagnosis codes and medications). When these two groups were combined, a total of 5054 OA patients (14.4%) had a contraindication for NSAIDs based on either a diagnosis code or a prescription. An NSAID contraindication was almost twice as common in the OCH subcohort compared to the control cohort, where only 2787 patients (7.9%) had a contraindication for NSAIDs. Fig. 2 Prevalence of hip/knee OA in the OCH subcohort. The prevalence of hip or knee OA in the OCH subcohort was calculated at the end of the study period (April 2020) and reported by gender, type of OA and age group. The overall point prevalence was 5.6%, and knee OA was 5.5-times more common than hip OA We also looked at the number of healthcare contacts per patient year by type of professional visited. The type

Discussion
Even though the prevalence of OA is known to increase with age, hip and knee OA are a significant burden also for the working age population [3,5,7]. In this retrospective registry study conducted in the data lake of Finland's largest private and occupational healthcare provider, the overall prevalence of hip or knee OA in the occupational healthcare subcohort was 5.6% at the end of the study period. As expected, the prevalence increased with age, peaking in the age group 60-64 years.
In a recent study with data from the United States, the prevalence of OA in the age group 50-64 years was 14.8% [3], which is very similar to the 12.5% seen in this study for the same age group. The prevalences reported here are also comparable to the latest Finnish prevalence data from 20 years ago, where the prevalence of hip or knee OA in the age group 55-64 years was 14.4% for males and 11.2% for females [13]. In contrast to the other studies mentioned above, which reported even higher prevalences in the age groups over 65 years, in this study the prevalence of OA was lower in the age group 65 years and older compared to 60-64 years (12.6% vs 15.4%). This can be readily explained by the fact that in Finland the most common retirement age is around 65 years, and very few people are covered by occupational healthcare after they turn 65. Therefore, most of the OA patients in the older age groups receive treatment for their OA from the public healthcare sector and other data sources would be needed to access that data. Since data for age groups older than 65 years is not comprehensive in the data source used for this study, we also did not report the OA prevalence for the all OA cohort.
Importantly, our study shows that hip or knee OA diagnosis increases HCRU, as demonstrated by the increase in both sick leave days and healthcare contacts compared to controls without OA. Out of the 22.8 sick leave days per patient year recorded for OA patients (vs. 8.1 days for controls), 6.3 days were caused by OA specifically. Therefore, comorbid conditions were responsible for 8.4 days of the increase seen for OA patients. A similar pattern was seen for healthcare contacts, with 1.1 of the 6.1 healthcare contacts per patient year due to OA specifically. In general, HCRU was similar for hip and knee OA patients, although patients with hip OA had slightly more sick leave days compared to patients with knee OA (25.6 vs 21.8 sick leave days per patient year). Other studies have also described substantial increases in days missed from work due to OA. For example, Kotlarz et al. reported that OA increased annual absenteeism from 5.5 to 9.2 days for women and from 5.2 to 9.7 days for men [5], and Sharif et al. showed that OA cases had a 90% higher hazard ratio of work loss Fig. 3 Healthcare resource utilization for the OCH subcohort and controls. The number of sick leave days (A), sick leave periods (B), and healthcare contacts (C) in the OCH subcohort and control cohort per patient year are shown. OA patients had 2.6 times more sick leaves, 2.2 times more sick leave periods and 2.2 times more healthcare contacts compared to controls without OA due to illness or disability compared to non-OA controls over a 2 year period [8].
On average, the OA patients had a higher BMI and more comorbidities recorded in their EMRs compared to the age-and gender-matched controls without OA. For example, dorsalgia was 1.8 times more common in the OA patients. Other studies have also reported higher BMIs [3,7] and a higher comorbidity burden [7,14] for patients with OA compared to those without. Furthermore, 14.4% of the OA patients in this study had a contraindication for NSAID use based on either diagnosis codes and/or prescription data. This data supports the findings from previous studies and highlights the need for a multidisciplinary approach in the treatment of OA, and the importance of considering possible comorbid diagnoses and other prescriptions.
OA treatment guidelines, including the Finnish national guidelines, emphasize the role of nonpharmacological interventions, rehabilitation, and exercise programs in the treatment of OA [12, 15,16]. According to the national treatment guidelines [12], pharmaceutical treatment for OA pain should be initiated with paracetamol, followed by NSAIDs, whereas opioids are seen as the last resort. Based on these guidelines, the number of paracetamol prescriptions reported here (16.0% on OA-related visits) is quite low. However, paracetamol is also available over the counter from Finnish pharmacies, and patients may also have a prescription from the public healthcare sector, which is not seen in this data. Furthermore, it should be noted that while 16.0% of OA patients had a paracetamol prescription, it was the only analgesic prescription for only 1.9% of the patients, indicating that most of these patients had an NSAID/opioid prescription as well.
Based on recent literature from other European countries, it seems that around one third of OA patients have NSAID prescriptions or use NSAIDs for OA pain. For example, Spitaels et al. reported that in Belgium 29.4% of knee OA patients had an oral NSAID prescription and 2.3% had a prescription for a cyclooxygenase-2 (COX-2) selective NSAID [17]. In a Dutch study, 30% of hip/knee OA patients used NSAIDs with diclofenac being the most commonly used drug in this group (15%) [18]. In our study, NSAIDs were prescribed to 40% of OA patients in the OCH subcohort, and the COX-2 selective NSAID etoricoxib was the single most prescribed analgesic (21.8% on OA-related visits).
Even though opioid use is seen as the last resort in the national treatment guidelines, opioid prescriptions were relatively common, with 10.6% of patients having an opioid prescription from an OA-related visit. The vast majority of the opioid prescriptions reported here were for weak opioids, including tramadol. The Dutch and Belgian studies reported very similar values, with opioid use recorded for 12% of OA patients in the Netherlands [18], and weak and strong opioid prescriptions to 6.1 and 4.3% of OA patients, respectively, in Belgium [17].
Looking at prescriptions from all visits, not just those with an OA-diagnosis, 37.0% of the OA patients in the OCH subcohort had an opioid prescription, which is 2.5 times higher compared to the controls. This is comparable to a recent study from Sweden, where the 12 month prevalence of opioid use among hip/knee OA patients was 23.7%, and 2.1-fold higher compared to those without hip/knee OA [19]. In a follow-up study, Thorlund et al. reported that more than half of the incident opioid dispensations to hip/knee OA patients within the first year after diagnosis were inappropriate according to current treatment guidelines [20]. Considering the updated treatment guidelines and the results from a recent meta-analysis, which concluded that opioids provide minimal relief for OA symptoms and cause discomfort in a majority of patients [21], it is important to pay close attention to the pattern of opioid prescriptions for patients with hip or knee OA.
To our knowledge, this the first real-world data study focusing specifically on employed, working age individuals diagnosed with either hip or knee OA. The data source used in the study, the Terveystalo data lake, provides a geographically comprehensive data set, and with 1.1 million consented patients covers up to 21% of the Finnish population. As occupational healthcare must be provided by the employer according to Finnish law, there is also no selection of patients due to societal status or position. Additional strengths include the fact that data was collected in a real-world setting without stringent inclusion and exclusion criteria, and the fact that diagnoses, prescriptions, sick leave days and visits can all be accessed from the same data source.
However, like all real-world data studies, this study also has some limitations. Some information may not have been consistently recorded for all patients, potentially affecting the study population and other outcomes. The data source does not contain data on drug purchases, so the use of medication in daily dose equivalents cannot be determined. Furthermore, the data source does not currently contain any patient reported outcome measures (PROMs), such as pain or physical function measures or patient global assessments, which would be relevant for OA patients [22,23]. Such measures would also be required to truly measure the value of healthcare [24]. With the increasing prevalence of OA due largely to an aging population and the increasing prevalence of obesity [2,4], we expect the OA-associated HCRU to increase correspondingly. This will form a challenge for the healthcare system and highlight the need to build value-based care pathways for patients suffering from chronic diseases such as OA, requiring the implementation of PROMs into clinical practice.