Burden of disease due to hip, knee, and unspecified osteoarthritis in the Peruvian social health insurance system

Introduction: Since its introduction by the World Health Organization (WHO), the concept of burden of disease has been evolving. The current method uses life expectancy projected to 2050 and does not consider age-weighting and time-discounting. Our aim is to estimate the burden of disease due to hip, knee, and unspecified osteoarthritis using this new method in the Peruvian Social Health Insurance System (EsSalud) during 2016. Methods: We followed the original 1994 WHO study and the current 2015 Global Burden of Disease (GBD) methods to estimate disability adjusted life years (DALY) due to osteoarthritis, categorized by sex, age, osteoarthritis type, and geographical area. We used disability weights employed by the Peruvian Ministry of Health, and the last update issued by WHO. Results: Overall, EsSalud reported 17.9 new cases of osteoarthritis per 1000 patients per year. Annual incidence was 23.7/1000 among women, and 72.6/1000 in people above 60 years old. Incidence was 5.6/1000 for knee osteoarthritis and 1.1/1000 for hip. According to the 1994 WHO method, there were 399,884 DALYs or 36.6 DALYs/1000 patients per year due to osteoarthritis. 12.4 and 2.2 DALYs/1000 patients per-year were estimated for knee and hip osteoarthritis, respectively. Using the 2015 GBD method, there were 1,037,865 DALYs or 94.9 DALYs/1000 patients per year. 31.4 and 5.3 DALYs/1000 patients per year were calculated for knee and hip osteoarthritis, respectively. Conclusions: In the Peruvian social health insurance subsystem, hip, knee, and unspecified osteoarthritis produced a high burden of disease, especially among women and patients over 60. The 2015 GBD methodology yields values almost three times higher than the original recommendations. Open Peer Review


Introduction
Globally, the musculoskeletal disorders are a group of chronic diseases with high disability rates, which have been increasing, especially among people above 60 years old 1,2 . Worldwide, osteoarthritis is the sixth cause of disability-adjusted life years (DALYs), representing 3% of global burden of disease 3 . Knees and hip osteoarthritis produce the greatest disability in patients, with 6%, and 4% of the burden of disease due to knee and hip osteoarthritis, respectively 1,2 . Overall, the prevalence of osteoarthritis is homogeneous among men and women between 30 to 60 years old, but increases among women after that 4,5 .
The Peruvian burden of disease studies estimated that 308,804 DALYs were lost in the Peruvian population due to musculoskeletal disorders, representing 6% of the disease burden from all health conditions during 2004 6,7 . Osteoarthritis was the health condition with the seventh greatest disability rates and caused 165,636 DALYs, which represents 3% of the total disease burden 6,7 . 98% of DALYs from osteoarthritis were attributed to the years lost due to disability (YLD), since osteoarthritis is not a primary cause of death. It mostly affected women, being the fourth biggest cause of DALYs in women 6 . Among people between 45 and 59 years old, osteoarthritis was the second biggest cause of disability, with 109,804 DALYs, and in people above 60 years old represented 7% of total DALYs [7][8][9] . Without knowing the impact of osteoarthritis on people's lives, it is difficult to propose solutions, invest resources for prevention, and mitigate disability, especially among elders, who constitute a large group within the Peruvian social health insurance system (EsSalud), which cares for approximately 37% of people who seek medical attention 10,11 . Hence, there is an important need to estimate the burden of disease produced by osteoarthritis in this healthcare subsystem 11 .
Since its introduction by the World Health Organization (WHO), the concept of burden of disease has been evolving. The original 1994 recommendations used tables based on 1966 data, employed different life expectancy for men and women, gave less weight to the extremes of life, and penalized years when they were far from the current person age 12 . The current method recommended by WHO in its 2015 Global Burden of Disease (GBD) study uses life expectancy projected to 2050 and does not consider ageweighting or time-discounting 13 . Depending on the age distribution of a population, these new recommendations would probably yield higher estimates. Therefore, our aim is to estimate the burden of disease due to hip, knee, and unspecified osteoarthritis in EsSalud during 2016 using two different methods.

Population
Our study was conducted using nationwide data collected by the Peruvian social health insurance system during 2016. We included records of all patients older than 15 years who were attended due to osteoarthritis (International Classification of Diseases, ICD-10: M15-M19) between January and December 2016. We excluded patients whose ICD-10 codes for osteoarthritis were registered during hospitalization and patients who were previously treated for osteoarthritis. Sample size calculation estimated that a minimum of 20,000 records were needed to find differences of at least four DALYs/1000 people between subgroups with 80% power and 95% significance.
All medical attentions at EsSalud between January and December 2016 were reviewed. These attentions had been registered using three different electronic systems. The Hospital Management System (SGH, by its acronym in Spanish) and the Health Services Management System (HSS, by its acronym in Spanish) record inpatient and outpatient attentions delivered at secondary and tertiary care level facilities. Meanwhile, the Health Information System for Primary Care Centers (SISCAP, by its acronym in Spanish) registers health attentions delivered at primary care facilities. All entries that fulfilled our selection criteria were included. In addition, we reviewed all death certificates issued by EsSalud during the study period, and selected those patients who had osteoarthritis registered as primary or contributory cause of death. Identities of patients on databases were kept confidential during data management and analysis.

Variables
Covariates included sex, age, health center, geographical area, osteoarthritis of the hip (ICD-10: M16), osteoarthritis of the knee (ICD-10: M17), osteoarthritis of first carpometacarpal joint (ICD-10: M18), polyosteoarthritis (ICD-10: M15), unspecified osteoarthritis (ICD-10: M19), death due to osteoarthritis, and time between initial attention for osteoarthritis and death. Additionally, the following indicators were estimated for each person: Life expectancy at time of care delivered: Life expectancy was estimated according to two methods: The original WHO burden of disease study from 1994 5 employed the West extended model life tables from level 26 12 to calculate life expectancy at the time the disability started (in our case, time of osteoarthritis diagnosis). It then applied an age-weighted function, defined by: Cxe -βx , where "x" is the age in years, "C" is the constant of age weighting adjustment (value: 0.16458) and "β" is the age weighting parameter (value: 0.04). This function draws a curve assigning different weights to ages, giving greater values to adult ages since they were considered "more productive". Additionally, a 3% discount by year was applied, trying to capture the fact that people appreciate years in the immediate future than those further away more. The second method, employed in the 2015 GBD study, uses the maximum worldwide life expectancy projected to 2050, does not differentiate between men and women, and does not consider weights or discounts 13 .

Years of life lost (YLL):
In case the patient died because of osteoarthritis, the YLL was calculated as the life expectancy at the time of death using the two methods described above 12,13 . Living patients were assigned a zero value for YLL.
Years lost due to disability (YLD): YLDs were estimated as the average duration of the illness at age of onset, times the disability weight (0 = maximum health, 1 = death) 12 . Given that osteoarthritis is a chronic condition that lasts until death, the average duration of the illness was considered as the life expectancy at the time of initial diagnosis. For this estimation, we used the two methods previously described 12,13 . Two disability weights were used: 0.165, which is the value employed by the 2015 GBD study for severe musculoskeletal diseases of lower limbs 14,15 ; and 0.28, which was used in the Peruvian burden of disease studies for osteoarthritis severe enough to seek medical attention [6][7][8][9][10] .

Disease-adjusted life years (DALY):
This was the sum of the years of life lost (YYL) and the years lost due to disability (YLD) for each patient.

Statistical analysis
We described numerical variables using means and standard deviations. Categorical variables were described using frequencies and proportions. Osteoarthritis incidences were calculated by dividing the number of new cases registered by the number of insured patients in EsSalud during 2016. YLL, YLD, and DALYs were estimated, summing these metrics in total and by subgroups using the 1994 WHO and the 2015 GDB methods, and the 2015 GBD and the Peruvian Ministry of Health (MINSA) disability coefficients, which means four iterations were calculated for each metric. In addition, DALYs incidence ratios per thousand patients per year were calculated by dividing total number of DALYs by the total number of insured patients registered in EsSalud during 2016. The statistical analysis used STATA v14.0 (Statacorp, College Station, Tx). Code used for the analysis is available as Extended data 16 .

Ethical statement
The Institutional Review Board (IRB) of the Edgardo Rebagliati Martins National Hospital (HNERM) approved this study (#832-2019-195). The IRB waived the requirement for consent from the patients as the study was conducted with an anonymized dataset.

Results
During 2016, the Peruvian social health insurance system attended 196,003 patients for a first time visit due to osteoarthritis. Among them, 65.5% (n=128,323) were women, the mean age was 60.9±15.1 years, 62.6% (n=122,705) had polyosteoarthritis or unspecified osteoarthritis, 31.0% (n=60,788) had osteoarthritis of the knee, 5.9% (n=11,472) osteoarthritis of the hip and 0.5% osteoarthritis of other joints 17 . We estimated 17.9 new cases of osteoarthritis per 1000 insured patients in 2016. The incidence of osteoarthritis in women was 23.7/1000 insured patients per year, and in patients over 60 years old the incidence was 72.6/1000 insured patients per year. The incidence of polyosteoarthritis or unspecified osteoarthritis was 11.2/1000 patients per year, the incidence of knee osteoarthritis was 5.6/1000 patients per year, and hip 1.1/1000 patients per year. The geographical region with higher incidence was the Northern Coast/Highlands with 20.2/1000 patients per year, and the lowest was found in the Amazon Rainforest with 13.3/1000 patients per year (Table 1).   (Table 3).

Discussion
The incidence for osteoarthritis overall was 17 9 . In 2015, EsSalud carried out a study of the burden of disease among the insured patients and estimated 131,220 DALYs due to osteoarthritis (12.3 DALYs/1000 insured) 10 .
As observed in Table 4, there has been an upward trend in the DALY estimates despite using the same methodology. This could be due to different approaches to measure incidence. Previous studies have used a variety of sources including epidemiological surveillance, number of attentions, population-based surveys, medical chart reviewing, and scientific papers [6][7][8][9][10] . It is possible then, that the upward trend is just echoing a better registry of cases.
In the present study, we only used electronic records of attentions, and employed all national data instead of small samples. This could lead to better capturing of cases, otherwise overlooked in previous studies. On the other hand, it is possible that our data collection strategy overestimates the number of new cases. Since we based our estimations on the ICD-10 diagnosis entered for each attention, it is not certain that all cases were confirmed at the time of the medical visit. In addition, we could have included mild cases not usually considered for surveillance or research purposes.
We also observed discrepancies when stratified by age group. In our study, the estimated DALYs were higher for insured patients between 45-59 years old, with 150.0 DALYs/1000 patients per year, followed by people older than 59 years, with 134.8 DALYs/1000 patients per year, and people between 15-44 years old, with 42.8 DALYs/1000 patients per year. In contrast, the 2012 Peruvian study found that the elderly group produced the most DALYs/ 1000 people 9 . The same study found that the elderly group delivered a burden of disease almost nine times the one registered for 15-44 years old people (26.0 vs 3.0 DALYS/1000 inhabitants); meanwhile, our study found a much more reduced gap of only three times (134.8 vs 42.8 DALYS/1000 inhabitants). One possibility is that our study identified more patients with osteoarthritis in the younger group than previous studies.
Estimations of burden of disease would differ depending on the methods used and the disability coefficients assigned to the disease. The use of different life expectancy values (West 26 vs GBD 2050) and weights/discounts affects not only the absolute values of DALYs but also the estimations within subgroups 13 . In addition, the original 1994 WHO method differentiated life expectancy values for men (80 years) and women (82.5 years), penalized the extremes of life ages, and discounted the value of years away in time, reducing the DALYs contributed by men and younger people [22][23][24] . The 2015 GBD method tried to correct these differences by using the maximum projected life expectancy for 2050 (91.9 years) without differences between sexes, and discarded age-weighting and time-discounts. The intended effect is to increase the sensitivity of the method to estimate DALYs, especially in the extremes of life. In our study, the 2015 GBD methodology yielded values almost three times higher than the original recommendations, and reduced the gaps between sexes and age groups.
Another important component when calculating DALYs is the disability weights. The original 1994 WHO methodology recommends using a disability weight of 0.22 for diseases that limit recreation, occupation, education or procreation activities, and 0.40 if a disease limits two or more of these activities. Instead, the 2015 GBD methodology recommends using different weights depending on the severity and location of the disease, giving a disability factor for hip and knee osteoarthritis of 0.165, corresponding to musculoskeletal problems, lower limbs, severe. On the other hand, the MINSA and EsSalud studies have consistently used a 0.28 weight for osteoarthritis regardless of age or sex, because it considers this disease severe enough to seek medical attention. Using the MINSA disability weights instead of the 2015 GBD recommendations increases the absolute values of DALYs by approximately 70%.

Conclusions
In the Peruvian social health insurance subsystem, which covers almost 40% of the population, polyosteoarthritis, unspecified  osteoarthritis, knee and hip osteoarthritis produced a high burden of DALYs lost, especially among patients over 60 years old and women. The 2015 GBD methodology yields values almost three times higher than the original recommendations, and the disability weights used by MINSA produced estimates 70% higher than using the 2015 GBD weights.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes "…employed different life expectancy for men and women…"; what is meant is that the standard life expectancy against which years of life lost are estimated had a difference between men and women.