Impact of Osteoarthritis Severity in an Employed Population
Self-report of OA severity has been shown to be accurate and relevant in clinical practice. This study provides a practical application of using self-report as an indicator of OA severity and demonstrates that as self-rated OA severity increases, there is a greater burden relative to workers without OA.
Many of the demographic differences that were identified between the OA and non-OA cohorts are consistent with what may be expected regarding the epidemiology and risk factors for OA (older, female, non-Hispanic white, greater comorbidities, tendency toward obesity). These variables were included as covariates in the multivariable analyses of quality of life and productivity, and thus the observed differences in these outcomes were likely related to the presence of OA.
Since epidemiologic data on individuals with OA younger than 45 years is sparse, it is interesting to note that the prevalence of OA was 4.2% for all workers 20–39 years of age, and that among workers with OA, 16.9% were in this age group. Additionally, approximately half of the workers (50.7%) with OA in this age group rated their OA as at least moderate severity. These data indicate that OA is likely to be more prevalent and have a greater impact in a younger population than has previously been thought based on the consideration of OA primarily as an age-related disease.
Workers with OA reported significantly lower health status relative to non-OA workers as measured by SF-6D utility values. The observed differences of -0.04 points and -0.08 points for moderate and severe OA relative to non-OA, respectively, were clinically significant; differences of at least 0.03 points are proposed to be clinically meaningful. Pairwise differences in health status among OA severity levels also exceeded 0.03, suggesting clinical significance. This trend of poorer health with increasing self-rated OA severity is consistent with previous observations using the EuroQol (EQ-5D) health index in US and European OA populations.
OA affects physical functioning, and it is therefore not surprising that effects were greater on physical components (PCS) of HRQoL than on mental components (MCS). Since differences of greater than 3 points between groups are considered clinically significant, the differences in PCS scores were clinically meaningful as well as statistically significant. In contrast, there was little change in the MCS, and although the score among workers with mild OA was statistically higher than among workers without OA, this is likely due to measurement error, possibly because of the large population. Of note, both the PCS and MCS scores are normed to the US population, enhancing generalizability.
The frequency of general pain and arthritis pain was significantly greater among workers with OA relative to those without, and higher at increasing levels of OA severity. Although OA pain and its treatment are associated with reduced productivity and increased costs, it remains to be determined whether these effects are related to pain severity, frequency, or both. Pain is likely to be only one of several factors that contribute to patients' perceptions of OA severity, and while experiencing pain in the past month was included as a covariate, pain severity was not. Nevertheless, the pain-related interference, which increased with greater OA severity and was significantly higher than non-OA workers, is consistent with an association between OA severity and specific activities of daily living, and provides a foundation for the potential impact of OA severity on work impairment.
Workers with OA were characterized by significantly greater work and activity impairment relative to those without OA, and the separation between cohorts increased with greater OA severity. Among workers with moderate and severe OA, approximately one-third (33.2%) and one-half (47.4%) of worker productivity was lost, respectively, compared with 17.3% among non-OA workers. While several studies evaluated OA-related absenteeism, few data exist on presenteeism, despite presenteeism being suggested as the primary source of lost productivity in the general arthritis population. This study confirms presenteeism as the primary source of work impairment in workers with OA, and characterizes the magnitude of this impairment as 3–4 times greater than that due to absenteeism, resulting in loss of more than a day's work/week among workers with moderate and severe OA.
Interestingly, lost productivity decreased with increasing age among workers with OA. The reasons underlying this observation cannot be established based on the data from this study, and few other studies have evaluated OA and its impact across these age groups. Nevertheless, several suggestions may be proposed to account for these data. First, it is possible that at least some of the differences in lost productivity between the younger and older age groups may be explained by the concomitant increase in self-employment and part-time work observed with increasing age. These types of employment, especially the former, may potentially allow workers to benefit from a more flexible work environment, thereby reducing lost productivity. In this regard, it should also be noted that while all individuals were employed at the time of the survey, no information was available on whether workers switched jobs, received accommodations at work for OA disability, or transitioned out and back into the workforce over time. Such workforce transitions have previously been shown to be common among workers with arthritis. Second, coping strategies and self-efficacy, which are prognostic factors for outcomes in individuals with OA, are likely to be different among the age groups. Such strategies may also relate to duration of disease, which was not captured in the current study.
The consequences of lost productivity were profound as manifested by their impact on costs. Although indirect costs were the primary cost driver, including among workers without OA, the magnitude of work impairment was especially apparent among workers with moderate and severe OA, resulting in unadjusted indirect costs that were 87% and 166% higher, respectively, than among workers without OA, and adjusted indirect costs 44% and 72% higher, respectively.
The magnitude of the unadjusted indirect costs ($6,903, $10.968, and $15,596 for mild, moderate, and severe OA, respectively) is consistent with estimates for employees with mild ($6,096), moderate ($13,251), and severe ($17,214) self-rated OA in a clinical practice-derived database. However, they are in contrast to other studies that reported low indirect costs. Since there are no standardized methods for estimating indirect costs in OA, these discrepancies may be attributed to differences in populations or methodologies. It is likely that inadequately accounting for presenteeism in other studies contributed to underestimation of costs. Differences in methodology may also account for the observation that our estimated mean medical costs of workers with OA ($4,403 across all severity categories) were lower than in other recently reported studies ($6,984-$8,201).
The data additionally show that more workers with OA reported use of traditional healthcare, and had associated higher costs across categories than the non-OA workers. The primary driver of direct costs was hospitalizations, likely due to the high cost per event. Although workers with OA were prescribed significantly more medications than those without OA, this was not included in the costs. Furthermore, whether these visits and medications were specifically related to OA could not be ascertained.
It is worth noting that after adjusting for covariates, non-traditional health provider visits (e.g. acupuncturists, herbalists, etc.) were used by a significantly greater proportion of individuals only among workers with mild OA relative to workers without OA. There is scarce information on utilization and outcomes of non-traditional provider care for OA, possibly because these visits are not generally included in insurance plans or claims databases.
It should be recognized that in the clinical setting, a variety of covariates are likely to impact healthcare-seeking behavior by patients, management strategies by providers, and work-loss-related compensation among employers. However, the resource utilization and total cost analyses were not adjusted for covariates since we wanted to provide a more complete perspective of the burden among these workers; in particular, unadjusted costs are often used to characterize the overall cost burden. Additionally, our multivariable models contained covariates that were outcomes in the resource utilization and costs analysis. Consequently, as a result of using unadjusted values, a conservative approach should be used for interpreting the implications of these analyses.
Strengths of this study include our ability to capture a wide variety of outcomes and the focus on an employed population, since disease burden in an active workforce is of economic importance to a variety of stakeholders. The large sample size and use of population-level analyses based on weighted assessments to reflect the demographic composition of the US population are additional strengths that enhance the generalizability of this study. Conversely, the large sample size may also be considered a limitation, since it is likely that some of the statistical significance could be ascribed to the large population.
Additional limitations include the use of self-report and that the OA diagnosis was not clinically confirmed. The latter could potentially introduce selection bias, since it is possible that workers in both cohorts may not have been clear about whether they had the correct diagnosis for inclusion/exclusion in their respective cohorts. However, given the large sample size, it is likely that the number of workers inappropriately placed in the cohorts would not substantially bias the results. Furthermore, this potential for bias does not preclude patient report as an important resource for evaluating outcomes, especially related to productivity.
Since the WPAI was not OA specific, the observed relationships between OA and productivity should be considered associative rather than causal. Similarly, the higher resource utilization and costs cannot be ascribed specifically to OA, since there are no claims linking resource use with the disease and symptoms of interest. Nevertheless, lost productivity in workers with OA was higher with increasing OA severity, and resulted in significantly greater indirect costs among workers with OA that may be of special concern to employers.
Although information was obtained on salary ranges and education, the type of employment or worker occupation was not considered. Type of employment is not only a risk factor for OA resulting from specific occupational activities, but is also likely to affect productivity, since workers with OA may be likely to avoid certain work activities. A similar limitation is that the joint affected was not determined; specific sites of OA (lower back, neck and knee) have been shown to be predictors of greater productivity losses and may also differentially affect healthcare resource utilization and associated costs. It is important to recognize that these factors may have implications for management strategies, and their absence in our analysis may reduce the generalizability of the results.
Although the number of prescribed drugs was captured by patient report, this information could not be used to estimate pharmacotherapy costs, since the source of our data was not a medical claims database. Thus, direct costs are likely to be underestimated. Furthermore, derivation of annual costs was based on extrapolation of 6-month data to 1 year, and may not adequately reflect annual resource utilization.
Discussion
Self-report of OA severity has been shown to be accurate and relevant in clinical practice. This study provides a practical application of using self-report as an indicator of OA severity and demonstrates that as self-rated OA severity increases, there is a greater burden relative to workers without OA.
Many of the demographic differences that were identified between the OA and non-OA cohorts are consistent with what may be expected regarding the epidemiology and risk factors for OA (older, female, non-Hispanic white, greater comorbidities, tendency toward obesity). These variables were included as covariates in the multivariable analyses of quality of life and productivity, and thus the observed differences in these outcomes were likely related to the presence of OA.
Since epidemiologic data on individuals with OA younger than 45 years is sparse, it is interesting to note that the prevalence of OA was 4.2% for all workers 20–39 years of age, and that among workers with OA, 16.9% were in this age group. Additionally, approximately half of the workers (50.7%) with OA in this age group rated their OA as at least moderate severity. These data indicate that OA is likely to be more prevalent and have a greater impact in a younger population than has previously been thought based on the consideration of OA primarily as an age-related disease.
Workers with OA reported significantly lower health status relative to non-OA workers as measured by SF-6D utility values. The observed differences of -0.04 points and -0.08 points for moderate and severe OA relative to non-OA, respectively, were clinically significant; differences of at least 0.03 points are proposed to be clinically meaningful. Pairwise differences in health status among OA severity levels also exceeded 0.03, suggesting clinical significance. This trend of poorer health with increasing self-rated OA severity is consistent with previous observations using the EuroQol (EQ-5D) health index in US and European OA populations.
OA affects physical functioning, and it is therefore not surprising that effects were greater on physical components (PCS) of HRQoL than on mental components (MCS). Since differences of greater than 3 points between groups are considered clinically significant, the differences in PCS scores were clinically meaningful as well as statistically significant. In contrast, there was little change in the MCS, and although the score among workers with mild OA was statistically higher than among workers without OA, this is likely due to measurement error, possibly because of the large population. Of note, both the PCS and MCS scores are normed to the US population, enhancing generalizability.
The frequency of general pain and arthritis pain was significantly greater among workers with OA relative to those without, and higher at increasing levels of OA severity. Although OA pain and its treatment are associated with reduced productivity and increased costs, it remains to be determined whether these effects are related to pain severity, frequency, or both. Pain is likely to be only one of several factors that contribute to patients' perceptions of OA severity, and while experiencing pain in the past month was included as a covariate, pain severity was not. Nevertheless, the pain-related interference, which increased with greater OA severity and was significantly higher than non-OA workers, is consistent with an association between OA severity and specific activities of daily living, and provides a foundation for the potential impact of OA severity on work impairment.
Workers with OA were characterized by significantly greater work and activity impairment relative to those without OA, and the separation between cohorts increased with greater OA severity. Among workers with moderate and severe OA, approximately one-third (33.2%) and one-half (47.4%) of worker productivity was lost, respectively, compared with 17.3% among non-OA workers. While several studies evaluated OA-related absenteeism, few data exist on presenteeism, despite presenteeism being suggested as the primary source of lost productivity in the general arthritis population. This study confirms presenteeism as the primary source of work impairment in workers with OA, and characterizes the magnitude of this impairment as 3–4 times greater than that due to absenteeism, resulting in loss of more than a day's work/week among workers with moderate and severe OA.
Interestingly, lost productivity decreased with increasing age among workers with OA. The reasons underlying this observation cannot be established based on the data from this study, and few other studies have evaluated OA and its impact across these age groups. Nevertheless, several suggestions may be proposed to account for these data. First, it is possible that at least some of the differences in lost productivity between the younger and older age groups may be explained by the concomitant increase in self-employment and part-time work observed with increasing age. These types of employment, especially the former, may potentially allow workers to benefit from a more flexible work environment, thereby reducing lost productivity. In this regard, it should also be noted that while all individuals were employed at the time of the survey, no information was available on whether workers switched jobs, received accommodations at work for OA disability, or transitioned out and back into the workforce over time. Such workforce transitions have previously been shown to be common among workers with arthritis. Second, coping strategies and self-efficacy, which are prognostic factors for outcomes in individuals with OA, are likely to be different among the age groups. Such strategies may also relate to duration of disease, which was not captured in the current study.
The consequences of lost productivity were profound as manifested by their impact on costs. Although indirect costs were the primary cost driver, including among workers without OA, the magnitude of work impairment was especially apparent among workers with moderate and severe OA, resulting in unadjusted indirect costs that were 87% and 166% higher, respectively, than among workers without OA, and adjusted indirect costs 44% and 72% higher, respectively.
The magnitude of the unadjusted indirect costs ($6,903, $10.968, and $15,596 for mild, moderate, and severe OA, respectively) is consistent with estimates for employees with mild ($6,096), moderate ($13,251), and severe ($17,214) self-rated OA in a clinical practice-derived database. However, they are in contrast to other studies that reported low indirect costs. Since there are no standardized methods for estimating indirect costs in OA, these discrepancies may be attributed to differences in populations or methodologies. It is likely that inadequately accounting for presenteeism in other studies contributed to underestimation of costs. Differences in methodology may also account for the observation that our estimated mean medical costs of workers with OA ($4,403 across all severity categories) were lower than in other recently reported studies ($6,984-$8,201).
The data additionally show that more workers with OA reported use of traditional healthcare, and had associated higher costs across categories than the non-OA workers. The primary driver of direct costs was hospitalizations, likely due to the high cost per event. Although workers with OA were prescribed significantly more medications than those without OA, this was not included in the costs. Furthermore, whether these visits and medications were specifically related to OA could not be ascertained.
It is worth noting that after adjusting for covariates, non-traditional health provider visits (e.g. acupuncturists, herbalists, etc.) were used by a significantly greater proportion of individuals only among workers with mild OA relative to workers without OA. There is scarce information on utilization and outcomes of non-traditional provider care for OA, possibly because these visits are not generally included in insurance plans or claims databases.
It should be recognized that in the clinical setting, a variety of covariates are likely to impact healthcare-seeking behavior by patients, management strategies by providers, and work-loss-related compensation among employers. However, the resource utilization and total cost analyses were not adjusted for covariates since we wanted to provide a more complete perspective of the burden among these workers; in particular, unadjusted costs are often used to characterize the overall cost burden. Additionally, our multivariable models contained covariates that were outcomes in the resource utilization and costs analysis. Consequently, as a result of using unadjusted values, a conservative approach should be used for interpreting the implications of these analyses.
Strengths of this study include our ability to capture a wide variety of outcomes and the focus on an employed population, since disease burden in an active workforce is of economic importance to a variety of stakeholders. The large sample size and use of population-level analyses based on weighted assessments to reflect the demographic composition of the US population are additional strengths that enhance the generalizability of this study. Conversely, the large sample size may also be considered a limitation, since it is likely that some of the statistical significance could be ascribed to the large population.
Additional limitations include the use of self-report and that the OA diagnosis was not clinically confirmed. The latter could potentially introduce selection bias, since it is possible that workers in both cohorts may not have been clear about whether they had the correct diagnosis for inclusion/exclusion in their respective cohorts. However, given the large sample size, it is likely that the number of workers inappropriately placed in the cohorts would not substantially bias the results. Furthermore, this potential for bias does not preclude patient report as an important resource for evaluating outcomes, especially related to productivity.
Since the WPAI was not OA specific, the observed relationships between OA and productivity should be considered associative rather than causal. Similarly, the higher resource utilization and costs cannot be ascribed specifically to OA, since there are no claims linking resource use with the disease and symptoms of interest. Nevertheless, lost productivity in workers with OA was higher with increasing OA severity, and resulted in significantly greater indirect costs among workers with OA that may be of special concern to employers.
Although information was obtained on salary ranges and education, the type of employment or worker occupation was not considered. Type of employment is not only a risk factor for OA resulting from specific occupational activities, but is also likely to affect productivity, since workers with OA may be likely to avoid certain work activities. A similar limitation is that the joint affected was not determined; specific sites of OA (lower back, neck and knee) have been shown to be predictors of greater productivity losses and may also differentially affect healthcare resource utilization and associated costs. It is important to recognize that these factors may have implications for management strategies, and their absence in our analysis may reduce the generalizability of the results.
Although the number of prescribed drugs was captured by patient report, this information could not be used to estimate pharmacotherapy costs, since the source of our data was not a medical claims database. Thus, direct costs are likely to be underestimated. Furthermore, derivation of annual costs was based on extrapolation of 6-month data to 1 year, and may not adequately reflect annual resource utilization.
SHARE
