Abstract Body

Background and purpose: Current epidemiological research on knee osteoarthritis is largely limited to codified data from electronic health records. This study investigates what the use of narrative data on top of codified data adds on the epidemiology of knee osteoarthritis.

Methods: A retrospective cohort study was conducted using the Integrated Primary Care Information database including 2.5 million patients from Dutch general practices. An algorithm was developed to identify patients (aged ≥30 years) diagnosed with knee osteoarthritis with codified data and/or keywords in narrative data. Annual prevalence and incidence rates were calculated from 2008 to 2019.

Results: The prevalence with codified data increased from 2.12% (95%CI 2.09-2.14) in 2008 to 5.36% (95%CI 5.33-5.39) in 2019. Adding narrative data showed on average 2.08 times higher prevalence; 5.12% (95%CI 5.09-5.16) in 2008 to 10.3% (95%CI 10.2-10.3) in 2019. The incidence rate with codified data increased from 4.72 per 1000 person-years (95%CI 4.43-5.03) in 2008 to 5.77 per 1000 person-years (95%CI 5.60-5.95) in 2019. Adding narrative data showed on average 1.80 times higher incidence; 9.01 per 1000 person-years (95%CI 8.60-9.44) in 2008 to 10.6 per 1000 person-years (95%CI 10.4-10.9) in 2019.

Conclusions: Including narrative data on top of codified data showed on average 2 fold higher prevalence and 1.8 fold higher incidence. This indicates that codified data alone from general practices registries seriously underestimates healthcare demand of knee osteoarthritis. Further research is required to understand when general practitioners use osteoarthritis codes instead of another code with osteoarthritis free text notes in the medical journal.