To compare the hospital inpatient costs between nonobese and obese patients and estimate the economic burden of obesity in primary total knee arthroplasty (TKA).
To compare the hospital inpatient costs between nonobese and obese patients and estimate the economic burden of obesity in primary total knee arthroplasty (TKA).
A cost identification study was conducted in a consecutive cohort of 530 patients who underwent TKA between 2006 and 2007 at a university-affiliated tertiary referral center in Melbourne, Australia. Total hospital inpatient costs incurred at the study institution associated with the index surgery and subsequent related emergency presentations and readmissions during the episode of care were captured. Predictor variables of interest were obesity and body mass index (BMI), and the outcomes of interest were total hospital inpatient costs for the index surgery and episode of care, defined as the first 12 months following TKA. Multivariate linear regression techniques were used to examine the association between the predictors of interest and hospital costs, adjusting for clinically relevant variables.
Economic data were analyzed in 521 patients, of which 317 (60.8%) were obese. Obesity was associated with higher inpatient index surgery costs (+$1,226.89 [95% confidence interval (95% CI) $82.25, $2,371.52]; P = 0.036) and episode of care costs (+$1,821.36 [95% CI $244.93, $3,397.79]; P = 0.024). Each unit increase in BMI was also associated with higher inpatient index surgery costs ($128.91 [95% CI $34.53, $223.28]; P = 0.008) and total episode of care costs ($158.79 [95% CI $28.54, $289.05]; P = 0.017).
The estimated significant additional annual obesity-related expenditure reported in this study establishes a rationale to trial and evaluate interventions that target weight loss in obese patients undergoing TKA from both a quality of life and economic perspective.
Arthritis is one of the leading causes of disability in developed nations. In Australia, it affects 16.5% of the population and imposes a significant economic burden (1–4). Furthermore, its prevalence is projected to increase to 25% by 2030 (3) because of the expected rise in the number of Australians with obesity (5), a major risk factor for osteoarthritis (OA) (6). Obesity-related OA in Australia accounts for more than $220 million in direct health care costs per year (7). This represents 15% of the total health care costs of OA (7).
Total knee arthroplasty (TKA) is considered the treatment of choice for end-stage knee arthritis, and as the prevalence of obesity continues to climb (5), so too has the number of TKAs performed each year (8). The number of primary TKAs in Australia has increased by 25% since 2006, and in the US this increased by 53% between 2000 and 2004 (9).
Despite the overrepresentation of obesity in patients presenting for TKA (10), there is a paucity of literature examining the influence of obesity, or indeed other factors, on the costs of primary TKA. To date, the only known predictor of the costs of primary TKA is illness severity (11, 12). In this study, we sought to determine the predictors of the costs of primary TKA. Given the higher complication rate reported in patients with obesity (10), we hypothesized that obesity would be independently associated with higher hospital inpatient costs for both the index procedure and in the following 12 months.
We estimated significant additional annual obesity-related expenditure associated with total knee arthroplasty.
These findings establish a rationale to trial and evaluate interventions that target weight loss in obese patients undergoing total knee arthroplasty from both a quality of life and economic perspective.
This study was undertaken on a consecutive cohort of TKA patients enrolled in a joint replacement registry held at St. Vincent's Hospital Melbourne (SVHM), Australia, a 460-bed university-affiliated tertiary referral center. The registry has been described in detail elsewhere (10). In summary, it collects comprehensive baseline data on patient demographics, diagnoses, details of surgery, comorbidities, perioperative interventions, length of stay, and discharge destination. Followup via active surveillance is for 12 months post–index procedure, with capture of outcomes, including death, wound complications (breakdown, infections, hematomas, and dehiscence), joint infections (according to Centers for Disease Control and Prevention classification), medical complications, and unplanned procedures and/or readmissions. Functional and quality of life assessments are undertaken at baseline and every 12 months. Passive surveillance of readmissions or outpatient encounters at SVHM are also undertaken via review of the hospital's electronic patient administration system up to 12 months post–index procedure. Data from the registry have been used previously to show that preoperative obesity confers a higher risk of adverse outcomes post–primary TKA (10).
The present study targeted all patients admitted to SVHM to undergo elective primary TKA between January 1, 2006, and December 31, 2007. Of the 533 eligible patients, 529 (99.2%) had complete data. A further 8 patients were excluded, as they underwent a second elective hip or knee replacement within 12 months of the index TKA.
During the study timeframe, 572 elective primary TKAs were performed in the 521 study patients. Thirty-nine patients underwent staged bilateral knee replacements and for these patients, only their most recent TKA was included in the analysis. There were no simultaneous bilateral TKAs performed. Surgery and care of patients undergoing primary TKA at SVHM is standardized through the use of clinical pathway protocols, which have been validated in a randomized controlled trial (13).
Cost data were derived from administrative databases maintained by SVHM. Like all Australian public hospitals, SVHM routinely tracks all services provided to individual inpatients and assigns relevant service costs. Services comprise specific individualized items such as medical imaging, pharmacy, pathology, and surgical procedures, including prosthetics used and operating room time, as well as more general care, such as time spent on wards and consultation by medical, nursing, and allied health staff. This method of costing is known as the “bottom-up” approach. For the present study, we collected total inpatient costs associated with the index surgery, as well as associated with any (nonrehabilitation, see below) readmissions in the following 12 months, with the total combined costs defined herein as “episode of care” costs. We defined “episode of care” in accordance with the National Library of Medicine (medical subject headings) definition (14). Index surgery costs were inclusive of all inpatient costs associated with index surgery from admission until discharge to home or rehabilitation. Episode of care costs were inclusive of all index surgery costs as well as all associated readmissions to the hospital or the emergency department within 12 months of the index procedure. Cost data were extracted by an administrative staff member at SVHM, who was independent of the study and blinded to the patients' body mass index (BMI).
The perspective of the above-mentioned costs was that of the government, predominantly the state (Victorian) government and to a lesser extent the federal government. In Australia, public hospitals are fully funded by the 2 levels of government. There are no out-of-pocket expenses for patients.
Costs associated with inpatient rehabilitation are routinely collected in only 1 of the 2 rehabilitation campuses affiliated with SVHM. Because of this, and also because TKA patients at SVHM can undergo inpatient rehabilitation at a number of non-SVHM–affiliated institutions, rehabilitation costs were not included in our analysis.
Readmission data were captured via ongoing surveillance of patients using several methods. Following the index surgery, all individual patient medical records were reviewed after each outpatient appointment and all related readmissions were recorded in the joint replacement registry. These data were cross-referenced with all admissions recorded on the electronic patient administration system at SVHM. All readmissions that occurred within 90 days of the index surgery were considered to be related admissions. All orthopedic readmissions beyond 90 days other than for a planned elective procedure not involving the original knee joint were also considered readmission related to the index procedure. Readmissions to other units within SVHM beyond 90 days were assessed individually, and only those readmissions clearly unrelated to the index surgery were excluded. For example, this occurred when patients with a medical condition that existed prior to undergoing TKA required admission for ongoing medical management of that preexisting condition.
Obesity defined as a BMI ≥30 kg/m2 and BMI as a continuous variable were the predictor variables of interest. Outcomes of interest were inpatient costs for the index TKA and inpatient costs for the entire episode of care (up to 12 months post–index TKA), excluding rehabilitation costs.
Student's t-tests and chi-square tests were applied to comparisons of continuous and categorical variables, respectively. Our data on costs followed a normal distribution. As the cost data were symmetrically distributed, we examined the association between obesity/BMI and costs using 4 multivariate linear regression models: model A = obesity and index surgery costs, model B = BMI and index surgery costs, model C = obesity and episode of care costs, and model D = BMI and episode of care costs. Each multivariate model adjusted for potential confounding by age, sex, primary problem necessitating TKA, diabetes mellitus, cardiovascular disease, and overall health status using the Charlson Comorbidity Index (CCI). The CCI is a widely used and validated measure consisting of a weighted scale of 17 comorbidities expressed as a summative score, which is derived from both the number and seriousness of comorbid diseases (15). The CCI was calculated using comorbidity data recorded during the preoperative medical and anesthetist assessments on the day of surgery and subsequently age adjusted (16). Each multivariate model also adjusted for the use of computer navigation. We also extracted and analyzed cases that had incurred a complication or adverse event only to compare the average cost between obese and nonobese patients. All analyses were performed using SPSS for Windows, version 18.0.
Both the SVHM arthroplasty registry and the present study were approved by the SVHM Human Research Ethics Committee.
The mean ± SD age of the 521 study patients was 70.6 ± 8.6 years and 158 (30.3%) were men. The mean ± SD BMI was 32.0 ± 6.01 kg/m2 and 317 patients (60.8%) were obese. The mean ± SD cost of the index admission was $15,441 ± $6,136 and the mean ± SD cost per episode of care was $16,454 ± $8,474. The distribution of costs is summarized in Table 1. The mean ± SD length of stay for the index procedure was 5.6 ± 2.1 days and 410 (78.7%) were discharged directly home. In total, 123 patients (23.6%) experienced an adverse event or complication any time from the index procedure to 12 months afterward, with 74 events (14.2%) resulting in a readmission to the emergency department or to the wards. There were 4 deaths. One occurred at 6 weeks due to a direct complication of surgery and 3 occurred more than 9 months after the index procedure, all due to cancer. One patient (BMI 32.1 kg/m2) presented with a prosthetic infection at 5 months and underwent removal of prosthesis, but because this took place at another facility, associated costs were not available.
|Cost category||Mean ± SD||% of total cost|
|Medical (surgical)||1,231.79 ± 794.25||7.5|
|Medical (nonsurgical)||149.27 ± 106.99||0.9|
|Nursing||6,706.29 ± 4,974.56||40.8|
|Allied health||664.13 ± 427.17||4.0|
|Imaging||172.19 ± 265.75||1.1|
|Pathology||110.17 ± 154.93||0.7|
|Pharmacy||531.36 ± 663.41||3.2|
|Operating room*||6,821.74 ± 2,598.31||41.5|
All of the patients underwent a fully cemented TKA using a standard medial parapatellar approach. Procedures were performed by a team of 14 surgeons using implants purchased from 4 different manufacturers. The type of knee implant used varied, but individual surgeons did not alter their manufacturer or implant types during the study timeframe. Both standard and computer-assisted techniques were adopted.
Demographic and clinical variables comparing nonobese and obese patients are shown in Table 2. The incidence of obesity among women was higher (65.3%) than among men (50.6%; P = 0.002), and there was a 4-year difference in the mean ± SD age of obese (69.0 ± 8.2 years) and nonobese (73.0 ± 8.6 years) patients (P < 0.001). There were also higher prevalences of cardiovascular disease (77.0 versus 68.6%; P = 0.044) and diabetes mellitus (24.0 versus 13.7%; P = 0.006) in obese patients, but no significant differences in total medical comorbidities or the mean age-adjusted CCI (Table 2). The prevalence of current smoking was low in both groups. The antecedent diagnosis was OA for the majority of obese (95.0%) and nonobese (91.2%) patients. For the remaining patients, the primary problem was rheumatoid arthritis. Operative time from skin incision to wound closure was similar for obese and nonobese patients (mean ± SD 92.4 ± 18.7 versus 94.4 ± 22.5 minutes; P = 0.270). the mean ± SD hospital length of stay for the index surgery was 5.6 ± 2.2 days for obese patients compared to 5.7 ± 2.0 days for nonobese patients (P = 0.601). There was no significant difference in inpatient rehabilitation referrals between obese (20.8%) and nonobese patients (22.0%; P = 0.820), and the mean ± SD length of stay for rehabilitation admissions was 13.5 ± 9.3 days for obese patients and 14.2 ± 10.9 for nonobese patients (P = 0.716). The incidence of complications and adverse events was significantly higher in obese (28.1%) compared to nonobese patients (17.2%; P = 0.006), as were emergency presentations and readmissions in the 12 months following TKA: 18.0% in obese patients compared to 8.8% in nonobese patients (P = 0.005).
|Nonobese (n = 204)||Obese (n = 317)||P|
|Age, mean ± SD years||73.0 ± 8.6||69.0 ± 8.2||< 0.001|
|Total comorbidities, mean ± SD||2.7 ± 1.4||2.8 ± 1.5||0.347|
|Age-adjusted Charlson Comorbidity Index, mean ± SD||1.9 ± 2.2||2.1 ± 2.3||0.459|
|Cardiovascular, no. (%)||140 (68.6)||244 (77.0)||0.044|
|Diabetes mellitus, no. (%)||28 (13.7)||76 (24.0)||0.006|
|Current smoker, no. (%)||13 (6.4)||23 (7.3)||0.833|
|Operative time, mean ± SD minutes||94.4 ± 22.5||92.4 ± 18.7||0.270|
|Computer navigation, no. (%)||38 (18.6)||44 (13.9)||0.184|
|Index surgery length of stay, mean ± SD days||5.7 ± 2.0||5.6 ± 2.2||0.601|
|Discharged rehabilitation, no. (%)||45 (22.0)||66 (20.8)||0.820|
|Rehabilitation length of stay, mean ± SD days||14.2 ± 10.9||13.5 ± 9.3||0.716|
|Complications/events, no. (%)*||35 (17.2)||89 (28.1)||0.006|
|Unplanned readmissions, no. (%)*||18 (8.8)||57 (18.0)||0.005|
The adjusted multivariate linear regression shown in Table 3 demonstrates a statistically significant association between obesity and higher inpatient costs (+$1,227 [95% confidence interval (95% CI) $82, $2,372]; P = 0.036). Using BMI as a continuous variable, the cost of the index procedure increased by $129 (95% CI $35, $223; P = 0.008) for every unit increase in BMI (Table 4). Lower costs were also independently predicted by female sex in both model A ($−1,548 [95% CI $−2,711, $−384]; P = 0.009) (Table 3) and model B ($−1,648 [95% CI $−2,814, $−482]; P = 0.006) (Table 4).
|Model A||β||P||95% confidence interval|
|Age-adjusted Charlson Comorbidity Index||−79.79||0.513||−319.45, 159.89|
|Diabetes mellitus||572.27||0.407||−781.39, 1,925.92|
|Current smoker||199.38||0.860||−2,025.24, 2,424.01|
|Computer navigation||1,044.33||0.154||−393.00, 2,481.66|
|Model B||β||P||95% confidence interval|
|Age-adjusted Charlson Comorbidity Index||−72.16||0.553||−311.10, 166.79|
|Diabetes mellitus||506.87||0.462||−844.73, 1,858.48|
|Current smoker||306.96||0.786||−1,914.82, 2,528.74|
|Computer navigation||1,029.08||0.158||−401.52, 2,459.53|
The adjusted multivariate linear regression shown in Table 5 demonstrated a statistically significant association between obesity and higher episode of care costs (+$1,821 [95% CI $245, $3,398]; P = 0.024). Using BMI as a continuous variable, episode of care costs increased by $159 (95% CI $29, $289; P = 0.017) for every unit increase in BMI (Table 6). Lower total episode of care costs were again independently predicted by female sex in both model C ($−2,203 [95% CI $−3,805, $−601]; P = 0.007) (Table 5) and model D ($−2,284 [95% CI $−3,893, $−675]; P = 0.006) (Table 6).
|Model C||β||P||95% confidence interval|
|Age-adjusted Charlson Comorbidity Index||−230.95||0.170||−561.03, 99.14|
|Diabetes mellitus||210.06||0.825||−1,654.24, 2,074.35|
|Current smoker||1,792.00||0.251||−1,271.82, 4,855.82|
|Computer navigation||1,085.22||0.282||−894.32, 3,064.76|
|Model D||β||P||95% confidence interval|
|Age-adjusted Charlson Comorbidity Index||−219.80||0.191||−549.57, 109.98|
|Diabetes mellitus||174.92||0.854||−1,690.48, 2,040.33|
|Current smoker||1,902.39||0.223||−1,163.98, 4,968.75|
|Computer navigation||1,031.30||0.305||−943.02, 3,005.63|
Obese patients were more likely to experience a complication or adverse event (28.1 versus 17.1%), but among affected patients, there was no difference in the mean ± SD episode of care costs between obese and nonobese patients ($23,681 ± $15,954 versus $20,925 ± $10,147; P = 0.345).
To our knowledge, the present study is the first to analyze the economic impact of obesity in primary TKA. Our study, which involved a consecutive series of patients undergoing primary TKA over a 2-year period, found that hospital inpatient costs were significantly higher in obese patients both during the index admission and in the following 12-month episode of care. This was despite the fact that all TKAs followed a standardized program of care (13) and that length of stay and discharge disposition were similar for obese and nonobese patients (10).
The higher complication and readmission rates reported among obese patients in this series are likely to have contributed to the significant difference in hospital costs. However, among patients who experienced a complication or adverse event, there was no difference in the mean episode of care costs between obese and nonobese patients. Therefore, higher costs among the obese patients arose from higher rates of adverse events, rather than differences in the nature or severity of the events themselves.
The number of primary TKAs performed each year continues to rise both in Australia as well as internationally (8, 9, 17, 18). The Australian National Joint Replacement Registry (NJRR) recorded 57,485 primary TKA procedures during our study timeframe, 27,172 in 2006, and 30,313 in 2007 (19, 20). Over the same time, the proportion of obese patients presenting for TKA also increased (21–23). Although we cannot claim that our study cohort is representative of TKA at the national level, as a designated center for total joint arthroplasty the referral pattern at the study institution is statewide. Based on our results, if we extrapolate the proportion of obese patients in our study (61%) to NJRR numbers, as many as 35,066 patients who underwent primary TKA in 2006 and 2007 were obese. Multiplying this number by the excess hospital costs per episode of care associated with obesity provides an estimate of the financial burden each year for inpatient costs alone, attributable to obesity, of $32 million (95% CI $4–$60 million). Since commencement of this study in 2006, the number of primary TKAs has further risen to 34,126 in 2009, representing a 25% increase in just 4 years (8). We would caution, however, that further multicenter studies that include outpatient and rehabilitation costs are needed in order to accurately determine obesity-related costs in TKA.
Without rehabilitation and outpatient data, it is possible that we have over- or underestimated the overall financial burden of obesity in TKA; however, the latter is a more likely scenario given the higher complication rate in obese compared to nonobese patients. We have previously reported a significantly higher prosthetic infection rate in obese compared to nonobese patients in this same patient cohort (10), and prosthetic infection is associated with multiple outpatient visits. In a review of the hospital costs associated with prosthetic infection in total hip replacement, the mean ± SD number of outpatient visits following revision surgery in a cohort of 29 patients was 54.6 ± 35.1 in the first 12 months following surgery (24). We could find no published data on outpatient costs associated with primary TKA, but in general, higher rates of hospital outpatient visits have been reported in obese compared to nonobese patients (25). Inpatient rehabilitation hospital charges have also noted to be significantly higher for obese compared to nonobese patients undergoing TKA in a US study (26).
We found that every unit increase in BMI was associated with increasing inpatient costs for both the index surgery and episode of care. This suggests that even small amounts of weight loss may reduce the costs of primary TKA. Weight loss programs specifically tailored for patients scheduled for TKA are currently nonexistent. A number of diet and lifestyle trials have reported modest weight loss and improved function in overweight patients with knee OA over 4–18-month timeframes (27–30). However, these studies have targeted community cohorts rather than patients with end-stage OA on waiting lists for TKA, when arguably patients may lack the motivation to delay surgery for participation in a weight loss program. The cost-effectiveness of such programs would also need to be analyzed to determine if program costs are health economically sound.
We demonstrated an economic cost of $129–$159 for each single unit increase in BMI for hospital inpatient costs in TKA patients, and the question remains: would a single unit decrease in BMI offset the cost of a weight loss program? While this remains unknown in TKA patients, in general weight loss programs appear to be cost effective. However, both costs and economic benefits can vary significantly dependent on the type of program used. Short-term weight management programs appear to be inexpensive and cost effective for low-level (2–6 kg) weight loss, with 1 internet-based study reporting a cost of $25.92 US dollars per kilogram of weight loss (31). Pharmacologic weight loss interventions are more costly than lifestyle interventions, and while their use may result in significantly greater short-term weight loss, this does not appear to result in greater savings in total health care resource utilization over conservative weight management programs (32). Bariatric surgery is by far the most expensive weight loss option in terms of initial outlay, estimated to cost approximately $20,000 US dollars, but is also the most cost-effective weight loss option in the longer term (33–35), with 1 study demonstrating that the net cost of coverage for laparoscopic adjustable gastric banding is reduced to 0 by approximately 4 years after band placement in severely obese individuals (BMI ≥35 kg/m2) (33). To our knowledge, the cost-effectiveness of weight loss programs in TKA patients has yet to be tested; however, based on the literature, weight loss interventions are in general cost effective and therefore should be considered for obese patients undergoing TKA. Dependent on the intervention, cost savings may not be realized in the short term in TKA patients; however, they are likely to prove cost effective in the longer term, particularly considering that such programs confer other health benefits beyond those related to surgery (36, 37).
An unexpected finding in our study was the association between female sex and lower inpatient costs. Higher numbers of medical comorbidities and longer operative time in men were also noted, which provides a likely explanation. Men recorded a significantly higher number of comorbidities (mean ± SD 3.1 ± 1.5) compared to women (mean ± SD 2.7 ± 1.4; P = 0.008), and the operative time was significantly longer for men (mean ± SD 97.4 ± 20.5 minutes from skin incision to wound closure) compared to women (mean ± SD 91.4 ± 19.9 minutes from skin incision to wound closure; P = 0.002). Comorbidity has been reported to correlate with increased inhospital costs in both hip and knee replacements (11, 12). Longer operating room times have also been reported in men compared to women undergoing primary TKA and are thought to be due to anatomic and physiologic differences between sexes (38).
The strength of this study lies in the near-complete capture of inhospital economic data in a large consecutive series of patients undergoing primary TKA with 12 months of followup. However, several limitations warrant mention. First, our study was performed at a single institution and the sourcing of patients from a single center was a potential source of selection bias. There are very few studies with which ours can be compared and we can find no local data for comparison of obesity rates in primary TKA. The mean age of our cohort was similar to Australian Orthopaedic Association Joint Registry data, but our ratio of women was higher than that of the registry during the study timeframe (20). Second, we focused only on hospital inpatient costs in the first 12 months following TKA. As mentioned, this is likely to have underestimated the true difference in TKA costs between obese and nonobese patients. Furthermore, the cost perspective was only of the health care system, meaning no consideration of broader (societal) costs.
We recognize that TKA is a cost-effective procedure, particularly when conducted in a high-volume center, and is associated with significant reductions in direct costs attributed to arthritis compared to conservative management (39, 40). In this regard, we are not recommending that TKA be limited among obese patients. Rather, we wish to highlight the significant additional annual expenditure associated with obesity, and the need to develop and test interventions that target weight loss in obese patients undergoing TKA.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Choong had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Dowsey, Liew, Choong.
Acquisition of data. Dowsey, Liew, Choong.
Analysis and interpretation of data. Dowsey, Liew, Choong.