Trends in the epidemiology of total shoulder arthroplasty in the United States from 1990–2000




Longitudinal trends in epidemiology and utilization of total shoulder arthroplasty (TSA) have not been previously reported. We evaluated trends in the distribution of age, race, hospital volume and teaching status, outcomes, and indications for TSA during the last decade.


TSA cases (n = 12,758) were extracted from the 1990–2000 Nationwide Inpatient Sample databases. TSA trends were obtained for 3 time periods: 1990–1993 (period I), 1994–1997 (period II), and 1998–2000 (period III).


Between 1990 and 2000, there were minor increases in the rate of TSA in most age groups. Ninety-three percent of the patients undergoing TSA in all 3 time periods were white. An increased proportion of patients were operated on in high volume hospitals in period III as compared with period I. Patients discharged to inpatient rehabilitation facilities after surgery had longer lengths of in-hospital stays as compared with those discharged home. Osteoarthritis was diagnosed in an increasing proportion of patients undergoing TSA (56.6% in period I versus 70.9% in period III).


There was a minor increase in the rate of TSA, and almost no change in use of TSA by nonwhites from 1990 through 2000. Efforts to understand and narrow this apparent underutilization of TSA among nonwhites are required. Further research should determine whether the observed shift of TSA to high volume centers improves surgical outcomes.


Shoulder pain is one of the most common complaints in primary care and in rheumatologic, orthopedic, rehabilitation, and other practice settings (1). Shoulder pain originates from a range of pathoanatomic entities including involvement of the rotator cuff tendons, bursae, and glenohumeral joint structures. Most shoulder disorders respond to conservative medical and rehabilitative management. Patients with advanced, symptomatic, disabling glenohumeral joint disorders are typically candidates for total shoulder arthroplasty (TSA). It is important for referring clinicians (e.g., rheumatologists, primary care physicians, and rehabilitation specialists) to make timely referrals for TSA to optimize procedure outcomes and minimize disability.

Osteoarthritis, rheumatoid arthritis, and traumatic arthritis are the most common indications for shoulder arthroplasty. Other indications include avascular necrosis or osteonecrosis of the humeral head (2–6). Reliable data on whether these indications have changed over time are not available. Osteoarthritis commonly presents as a gradual increase in pain, stiffness, and decreased motion of the affected joint. Although epidemiology of shoulder (glenohumeral) osteoarthritis is not well studied, osteoarthritis of the knee is more prevalent in older persons and in women (7, 8). Other factors such as genetic susceptibility, nutrition, osteoporosis, joint injury, and body weight may also be associated with the occurrence and progression of knee osteoarthritis (7, 9). Specific risk factors for glenohumeral osteoarthritis may differ from those associated with knee osteoarthritis because the shoulder is not a major weight-bearing joint. It has been reported that approximately 91% of patients with long-standing rheumatoid arthritis ultimately develop shoulder involvement (10).

Studies to date have reported favorable outcomes after shoulder arthroplasty, citing more than 90% success in intermediate to long-term followup (11). TSA has been shown to provide predictable pain relief and functional restoration for many glenohumeral disorders including osteoarthritis (3, 12–16). Improvement in range of motion, low rate of surgical complications, and a high survivorship after TSA has also been reported (11, 17–22).

In spite of its widely accepted utility, the epidemiology of TSA remains poorly characterized. Reliable estimates for the incidence of TSA in the US are not available. This contrasts with other procedures such as hip and knee arthroplasty, for which such estimates are available from various sources. As the treatment of glenohumeral disorders becomes increasingly specialized, it is important to describe recent longitudinal trends in the epidemiology and outcomes of TSA. This may help in understanding changes in the utilization patterns of TSA with respect to age, race, hospital volume, and indications for the procedure, and may also help in planning future health care needs. TSA practice patterns and patient populations that need attention or improvement can also be determined.

Information on 12,758 patients undergoing TSA between 1990 and 2000 in the US is available from the Nationwide Inpatient Sample databases. Using these data, we evaluated trends in the utilization of TSA and in the distributions of age, race, hospital volume and teaching status, and indications for TSA during the last decade.


Data description.

The Nationwide Inpatient Sample databases for the years 1990–2000 were used for our study. The Nationwide Inpatient Sample is a part of the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality. The Nationwide Inpatient Sample is the largest all-payer inpatient care database that is publicly available in the US, and contains 5–8 million records of inpatient stays per year from ∼1,000 hospitals. This represents a 20% stratified sample of community hospitals in the US (23). To ensure maximal representation of US hospitals, the Nationwide Inpatient Sample databases sampled according to 5 important hospital characteristics: geographic region (Northeast, North Central, West, and South); ownership (public, private not-for-profit, and private investor-owned); location (urban, rural); teaching status (teaching hospital, non-teaching hospital); and bed size (small, medium, and large).

The HCUP assigned validation and quality assessment of these datasets to an independent contractor (24). The validation was performed by reviewing univariate statistics for all numeric data elements, frequency distributions for all categorical and some continuous data elements, checking range against standard norms, and performing edit checks that identify inconsistencies between related data elements. The Nationwide Inpatient Sample was also extensively validated against the National Hospital Discharge Survey, and confirmed to perform very well for many estimates (25).

Sample selection.

Records that included an International Classification of Diseases, Ninth Edition, Clinical Modification procedure code of 81.80 for TSA (n = 12,758) were selected from the Nationwide Inpatient Sample datasets. Each record in the datasets represented a single patient admission and had a unique identification number.

Demographic and outcome measures.

The Nationwide Inpatient Sample databases categorized patient race as white, black, Hispanic, or other (which included Asian or Pacific Islander and Native American). We categorized age into strata of <50, 50–59, 60–64, 65–79, and ≥80 years. Patient comorbidity was assessed using the Charlson Index, as modified for administrative data by Deyo et al (26, 27). This index measures comorbidity by assigning scores of 1, 2, 3, or 6 to each of 30 select comorbid conditions. These scores were then added to generate a single comorbidity index score. The comorbidities used to calculate Charlson Index were based on discharge diagnoses from the index admission included in the Nationwide Inpatient Sample databases. Charlson Index was further categorized into 0, 1, or >1.

The indication for TSA was ascertained from the primary diagnosis code assigned to a case. Mortality data were used to indicate whether a patient died in or out of the hospital, or was discharged. Length of stay was calculated in days by subtracting the admission date from the date of discharge. Length of stay longer than 100 days was replaced as missing (n = 6). Discharge status of the patient was categorized into either discharge home or to an inpatient facility (i.e., short-term hospital, skilled nursing facility, intermediate care facility, or another type of inpatient facility). Discharge home was either with or without home health care. Patients who left the hospital against medical advice or died during hospitalization were replaced as missing (n = 125) for this variable.

Hospital teaching status was obtained from the American Hospital Association Annual Survey by the HCUP. A hospital was considered to be a teaching hospital if it had an American Medical Association-approved residency program, was a member of the Council of Teaching Hospitals, or had a ratio of full-time equivalent interns and residents to beds of 0.25 or higher. Hospitals were characterized as urban if they resided in a metropolitan statistical area, and as rural if they resided in a non-metropolitan statistical area. Hospital volume per year for shoulder arthroplasty (TSA and hemiarthroplasty) was calculated using unique hospital identifiers provided in the databases. Annual hospital volume of shoulder arthroplasty was further categorized into <5 procedures, 5–9 procedures, 10–14 procedures, 15–24 procedures, and ≥25 procedures, based on clinically meaningful cut-offs.

Statistical analysis.

The study period was divided into 3 time periods (period I 1990–1993, period II 1994–1997, and period III 1998–2000). Distributions of demographic and other variables across the 3 time periods were examined using means, medians, and proportions in percentages. Multivariate logistic regression analyses were conducted to assess the relative change in each of the demographic and procedure outcome variables for period III using period I as reference, and controlling for period II and other potential confounders. Race had many (27.5%) missing values. Therefore, logistic regression analyses were performed using an indicator variable for missing race values as well as with the missing values, and compared for consistency. Since results with both approaches were similar, only the ones using missing values of race as an indicator variable are presented.

Survey sampling techniques using weights were used to estimate the total number of patients undergoing TSA in a given year and also to obtain stratified estimates by age groups. These estimates were then divided by the total US population for a particular age group in a given year to obtain rates per 100,000 persons (28, 29).

Statistical analyses were conducted using Intercooled STATA for Windows version 8.0 (Stata Corporation, College Station, TX) and SAS for Windows version 8.02 (SAS Institute Inc., Cary, NC).


There was a minor increase in the rates of TSA from 1990–2000 for most age groups (Figure 1 and Table 1). However, as compared with 1999, there was a slight decline in TSA rates in 2000 for most age groups. The distribution of TSA by age groups did not vary in the 3 time periods, with the odds ratios varying between 0.7 and 1.1 for the proportion of TSA performed in period III as compared with period I for a given age group. Most of the procedures were performed in patients 65–79 years of age (54.8% in period I, 55.0% in period II, and 54.0% in period III). The racial composition of patients undergoing TSA also remained essentially unchanged during the past decade, with whites comprising the vast majority of the patients (93% of procedures in all 3 time periods). The proportion of women undergoing TSA reduced by 13.1% from period I to period III.

Figure 1.

Rates of total shoulder arthroplasty per 100,000 people in the US by age group (1990–2000). Rates calculated by dividing total number of estimated patients undergoing total shoulder arthroplasty by the total US population in the respective age group according to data from the Census Bureau (resident populations as of July 1 of each year from 1990–1999, and April 1 for the year 2000). Patients younger than 50 years were excluded.

Table 1. Temporal trends in characteristics of patients undergoing total shoulder arthroplasty in the US (1990–2000)*
CharacteristicsTime periodChange, period I to period IIIAdjusted OR (95% CI), period III versus period I
I (n = 2,976)II (n = 4,985)III (n = 4,797)
  • *

    Unless otherwise indicated, values are the percent of patients. Data missing for age (n = 1, 0.01%), sex (n = 1, 0.01%), location/teaching status (n = 23, 0.2%). Period I = 1990–1993; period II = 1994–1997; period III = 1998–2000; OR = odds ratio; 95% CI = 95% confidence interval.

  • Adjusted for period II, age, sex, race, hospital location/teaching status, hospital volume, and Charlson score, excluding the variable used as outcome. Indicator variable used for missing race values in all regression models, except the ones for race as outcome.

  • Due to high percentage of missing data (43.0% for period I, 21.5% for period II, 24.1% for period III) percentages are expressed as total excluding missing data.

  • §

    Annual hospital volume for shoulder arthroplasty procedures.

Age, years     
 <508.48.57.2−14.30.7 (0.6–0.9)
 50–5910.911.713.725.71.1 (1.0–1.3)
 60–6411.39.910.6−6.20.9 (0.8–1.1)
 65–7954.855.054.0−1.51.0 (0.9–1.1)
 ≥8014.614.914.60.01.1 (1.0–1.3)
Female sex65.762.757.1−13.10.7 (0.6–0.8)
 White93.293.292.5−0.80.8 (0.7–1.1)
 African American3. (1.0–2.0)
 Hispanic2. (0.8–1.9)
 Other1.81.31.5−16.70.8 (0.5–1.3)
Charlson score     
 067.065.668.21.81.0 (0.9–1.1)
 116.816.414.1−16.10.8 (0.7–0.9)
 > (1.0–1.3)
Hospital volume§     
 <525.017.713.4−46.40.5 (0.5–0.6)
 5–929.628.320.3−31.40.7 (0.6–0.7)
 10–1419.919.017.4−12.60.8 (0.7–0.9)
 15–2416.323. (1.3–1.7)
 ≥ (2.6–3.4)
Hospital location/teaching status     
 Rural11.615.512.57.81.6 (1.4–1.9)
 Urban non-teaching56.754.839.6−30.20.5 (0.5–0.6)
 Urban teaching31.529.547.851.71.6 (1.5–1.8)

The distribution of TSA by hospital volume changed substantially from period I to period III. Hospitals with annual volumes of <5 and 5–9 shoulder arthroplasties per year made up 25.0% and 29.6% of the total TSA cases in period I, respectively, whereas this proportion decreased to 13.4% and 20.3%, respectively, in period III. On the other hand, for hospitals in higher volume categories of 15–24 and ≥25 shoulder arthroplasties per year, the proportion of TSA procedures performed increased by 41.1% and 181.5%, respectively. Similarly, the proportion of TSAs performed in urban teaching hospitals increased by 51.7% (from 31.5–47.8%) from period I to period III.

TSA had low inpatient mortality throughout the 3 time periods, with mortality rates in the range of 0.24–0.34% (Table 2). The proportion of patients discharged to other inpatient facilities after TSA increased significantly from period I to period II (11.0–14.4%) but remained relatively steady between periods II and III. Overall, the median length of in-hospital stay decreased from 4 days in period I to 3 days in period III. However, there was a marked difference throughout the 3 time periods in the lengths of stay for cases discharged home versus those discharged to other inpatient facilities, with patients discharged to other inpatient facilities having longer lengths of stay.

Table 2. Temporal trends in mortality and health care resource utilization of patients undergoing total shoulder arthroplasty in the US (1990–2000)*
OutcomeTime periodChange, period I to period IIIAdjusted OR (95% CI) period III versus period I
I (n = 2,976)II (n = 4,985)III (n = 4,797)
  • *

    Values are median (minimum, maximum) unless otherwise indicated. Data missing for mortality (n = 11, 0.1%), patient discharge status (n = 51, 0.4%), length of stay (n = 26, 0.2%). See Table 1 for definitions.

  • Adjusted for period II, age, sex, race, hospital location/teaching status, hospital volume, and Charlson score. Indicator variable used for missing race values in all regression models.

  • Adjusted ORs not calculated because of very few outcomes.

Mortality, no. (%)10 (0.34)12 (0.24)12 (0.25)−26.5
Discharge to inpatient facility, no. (%)326 (11.0)718 (14.4)696 (14.5)31.81.7 (1.5–2.0)
Length of stay, days     
 Discharged home4 (1, 44)3 (0, 28)2 (0, 29)−50.0
 Discharged to inpatient facilities7 (2, 92)4 (1, 66)4 (0, 55)−42.9

Osteoarthritis was the primary diagnosis for the majority of cases undergoing TSA during the 3 time periods (56.6% in period I, 63.4% in period II, and 70.9% in period III), with a significant increase (25.3%) in period III compared with period I (Table 3). Conversely, other major indications for TSA such as fractures of the proximal humerus, rheumatoid arthritis, and aseptic necrosis of the humeral head either declined or remained steady in their share of TSAs from period I to period III. These trends for indications of TSA were similar even after all 15 diagnoses (primary and secondary) were considered instead of only the primary diagnosis (data not shown).

Table 3. Temporal trends in selected diagnosis to total shoulder arthroplasty cases in the US (1990–2000)*
Primary diagnosis (ICD-9-CM)Time periodChange, period I to period IIIAdjusted OR (95% CI) period III versus period I
I (n = 2,976)II (n = 4,985)III (n = 4,797)
  • *

    Values are the percent of patients unless otherwise indicated; total does not equal 100% because all diagnoses are not included in table. ICD-9-CM = International Classification of Diseases, Ninth Edition, Clinical Modification. See table 1 for additional definitions.

  • Adjusted for period II, age, sex, race, hospital location/teaching status, hospital volume, and Charlson score. Indicator variable used for missing race values in all regression models.

  • Includes primary diagnosis codes 715.11, 715.12, 715.21, 715.22, 715.31, 715.32, 715.81, 715.82, 715.91, 715.92.

  • §

    Includes primary diagnosis codes 812.00, 812.01, 812.02, 812.09, 812.10, 812.11, 812.12, 812.19, 812.20, 812.30.

Osteoarthritis of shoulder region56.663.470.925.31.7 (1.5–1.9)
Fractures of upper end of humerus§12.511.18.1−35.20.7 (0.6–0.9)
Rheumatoid arthritis (714.0)−50.0
Aseptic necrosis of humerus head (733.41)−29.8


We analyzed data from 12,758 patients undergoing TSA performed from 1990–2000 to determine trends in the epidemiology and utilization of TSA. To our knowledge, longitudinal trends in TSA have not been previously reported. We found that the rate of TSA slightly increased for most age groups in the last decade. The proportion of cases with a diagnosis of osteoarthritis, the major indication for TSA, increased steadily. Whites accounted for 93% of TSA in all 3 time periods. An increasing proportion of patients were operated on in high volume urban teaching hospitals over the course of the decade. Finally, inpatient mortality rates remained very low (<0.5%) throughout the observation period.

Shoulder arthropathies are among the most commonly encountered conditions in primary care as well as rheumatology, physiatry, and other clinical settings. It is important for referring physicians to make timely referrals for TSA so that procedure outcomes can be optimized. Primary care physicians, internists, and medical specialists are often the first point of contact for patients with shoulder disorders. Therefore, it is also important for them to understand the epidemiology of TSA. This understanding will help clinicians to better educate themselves and their patients on the suitability, usefulness, and population characteristics of the procedure.

Trends in national rates of TSA by age group have not been previously reported from studies using a population- based sampling design. The American Academy of Orthopaedic Surgeons has reported the number of overall TSA procedures per year based on the 1991–2000 National Hospital Discharge Survey (30). However, some trends observed in this report, such as the sudden increase in TSA procedures from 5,000 in 1996 to 11,000 in 1997 are difficult to explain and are not supported by our analyses. It is likely that some estimates provided by the American Academy of Orthopaedic Surgeons have wide confidence intervals as compared with our estimates (where maximum standard error was <10% for estimates of overall number of TSA procedures/year; data not shown). The small increase in TSA rates in our study is also in contrast to rates of total knee arthroplasty in the US population, which have rapidly increased over the past decade (30). A possible explanation for the relatively minor increase in TSA rates is the discretionary nature of the procedure. Since the shoulder is not a weight-bearing joint, patients may delay or avoid TSA with fewer functional consequences as compared with knee arthroplasty. Another possibility is that TSA is a technically demanding procedure, which is not performed routinely by many orthopedic surgeons, as compared with hemiarthroplasty and other joint replacement procedures (31, 32). Therefore, when treatment selection (between TSA and hemiarthroplasty) of glenohumeral osteoarthritis is largely left to the individual surgeon's preference (16). High volume surgeons are significantly more likely to perform TSA as compared with low volume surgeons (33). High volume surgeons have a lower proportionate share in total annual volume of TSA, leading to a lower growth rate of TSA.

The vast majority of TSA procedures in our study were performed on white patients (93%) as compared with other racial groups. The underlying epidemiology of advanced shoulder arthritis in various racial and ethnic groups is unknown. However, unless there is an increased prevalence of arthritis in whites, these rates would suggest that nonwhites are much less likely to receive TSA relative to their representation in the general population (34, 35). Even more striking was the consistency of this underutilization throughout the past decade. In contrast to TSA, the proportion of black and Hispanic patients receiving total knee arthroplasty increased by 54.8% and 111.1%, respectively, in the last decade (36). This may be due partly to widely reported disparities in access to knee arthroplasty procedures in the literature (37–43) as compared with TSA for which such data are scarce. Also, African American patients have higher rates of knee osteoarthritis than white patients (44); we do not know whether such racial differences exist in shoulder osteoarthritis. Other possible factors responsible for the lack of progress in reducing the apparent underutilization of TSA by nonwhites are unknown. Due to missing information on race for many patients in our study, we did not calculate TSA rates by race relative to the racial distribution in the general US population. This emphasizes the need for cautious interpretation of our results and merits further studies in this area. We also found that the proportion of women undergoing TSA significantly diminished in the years studied. We cannot comment on whether this represents underuse of the procedure because our data do not provide information on the prevalence of advanced shoulder arthritis in men and women. Women have been previously reported to underutilize hip and knee arthroplasty (45). The trend of declining proportion of women utilizing TSA in the past decade needs further evaluation.

The extensive volume-outcomes literature in the past decade has reported better surgical outcomes, including those for shoulder arthroplasty, at high volume centers (46–49). In our study, there was a significant and substantial shift in the distribution of TSA by hospital volume such that high volume hospitals increased their share of procedures during the course of the study. This shift was also observed towards urban teaching hospitals. This phenomenon of voluntary regionalization of TSA patients may be partly attributed to the evolution of dedicated shoulder surgeons at high volume teaching centers during the past decade. However, in spite of this trend towards regionalization, most TSA were still performed in relatively low volume hospitals (those performing <9 shoulder arthroplasties/year), and only 26% of TSA were performed in the highest volume centers (≥25 shoulder arthroplasties/year) in period III of our study.

During our study period, discharge of patients to inpatient facilities after TSA increased by 31.8%. In contrast, the overall median length of stay after TSA decreased. It is likely that increased transfers to rehabilitation centers (50) and other inpatient facilities after surgery as well as a trend towards an earlier discharge home resulted in shorter postoperative lengths of stay during the later period of our study. In addition, the median length of in-hospital stay was considerably longer for patients discharged to inpatient facilities than for those discharged home. This is in contrast to knee arthroplasty where postoperative lengths of stay are similar for patients discharged home versus those discharged to inpatient rehabilitation centers (36). A possible explanation may include the presence of well-defined clinical pathways for knee arthroplasty, which allows stable patients to be transferred to rehabilitation facilities after a certain postoperative period. Similar pathways have not been developed for TSA, although our data point to the possible benefits of developing such pathways.

The 4 major diagnoses associated with TSA were osteoarthritis, fractures of the proximal humerus, rheumatoid arthritis, and aseptic necrosis of the humerus head. The proportion of TSA cases with osteoarthritis increased by 25.3% during the study period. This may be due to the evolution of better surgical techniques and dedicated shoulder surgeons, which led to an increased likelihood of patients with osteoarthritis being diagnosed and undergoing surgery in the later parts of our study. Although hemiarthroplasty or internal fixation, and not TSA, are the preferred surgical strategies for most cases of proximal humeral fractures (51, 52), this diagnosis occurred in 12.5% of patients undergoing TSA in period I and 8.1% of patients in period III. This practice needs further evaluation and the reasons for the surgeon's preference for TSA over hemiarthroplasty or internal fixation in such cases need to be examined. The proportion of patients with rheumatoid arthritis undergoing TSA showed a minor decline, which may reflect better disease-modifying therapies in the last 2 decades.

Our study had some limitations. First, we used the entire US population as the denominator when calculating rates of TSA by age groups instead of using the population at risk (those with advanced glenohumeral disorders). This methodology was necessitated by the lack of data on the prevalence of shoulder disorders. The approach has been used previously in studies on knee arthroplasty (36) and is useful in assessing trends. Second, although the majority of our patient population had nearly complete information on most variables, 27.5% of patients had missing race information. To account for this limitation, logistic regression analyses were conducted using an indicator variable for missing race values as well as with the missing values, and both analyses had similar results. Third, the Nationwide Inpatient Sample databases have not been independently validated against clinical data. However, it has been previously reported that surgical procedures are generally coded very accurately (53), and estimates for many procedures obtained from the Nationwide Inpatient Samples are comparable to the National Hospital Discharge Survey (25). Lastly, trends in short-term patient outcomes after discharge from the hospital and long-term patient outcomes such as pain relief, range of motion, and 1–5 year mortality and morbidity rates following surgery cannot be ascertained from our databases. The salient strengths of our approach are the use of data on a wide spectrum of patients and hospitals, the ability to generalize to the entire US population, our ability to adjust analyses for sociodemographic indicators and medical comorbidities, and evaluate longitudinal trends over an extended period of time for a procedure with relatively unknown epidemiology.

In conclusion, the rates of TSA showed minor increases in most age groups during the 10 years of our study. There was a significant and persistent underutilization of TSA by nonwhites. Efforts to enhance utilization by nonwhites should be explored. There was an increasing proportion of TSA performed at high volume teaching centers, which may help to further improve outcomes after surgery. However, this hypothesis needs further examination. The shift of TSA towards high volume centers may also reflect the evolvement of dedicated shoulder surgeons and shoulder practices at high volume centers over the past decade. Safe postoperative clinical pathways for TSA may help in reducing length of in-hospital stay for patients discharged to inpatient facilities, although the effect of increased discharge to inpatient facilities on outcomes and cost effectiveness of TSA needs to be ascertained.