To describe the extent to which patients were offered a choice between 2 or more hospitals for total knee replacement (TKR); to examine the association between having a choice of hospital for TKR and satisfaction with the surgery; and to identify population groups less likely to be offered a choice.
We studied a population-based sample of 932 Medicare beneficiaries who underwent elective TKR in 2000. We surveyed patients about their participation in choosing a hospital and their satisfaction with surgery. We examined whether lack of hospital choice influenced satisfaction with surgery after adjusting for age, sex, preoperative function, and socioeconomic status.
Among 932 TKR recipients (mean age 74 years, 67% women), more than half (53%) reported having a lack of hospital choice. After adjusting for socioeconomic status, patients reporting lack of choice were approximately twice as likely to be dissatisfied with the results of surgery as patients who reported choosing among 2 or more hospitals for TKR (odds ratio [OR] 2.09, 95% confidence interval [95% CI] 1.13–3.87). Results of logistic regression revealed that patients reporting lack of choice were more likely to be women (OR 1.52, 95% CI 1.14–2.04), >80 years of age (as compared with 65–70 years; OR 1.63, 95% CI 1.03–2.57), living in suburban areas (OR 1.68, 95% CI 1.23–2.30), nonwhite (OR 1.57, 95% CI 0.86–2.87), and were less likely to have TKR performed by a high-volume surgeon (OR 0.71, 95% CI 0.53–0.96).
More than half of the patients did not have a choice in selecting the hospital where they had TKR. Patients reporting lack of choice were more likely to be dissatisfied with surgery. Interventions to address preferences for hospital may improve satisfaction with care for patients with advanced knee arthritis.
Substantial gains in life expectancy have increased utilization of elective surgeries (1–4). In a few circumstances, elective surgeries are life saving, but the vast majority are performed to improve quality of life (5–8). The decision to undergo elective surgery largely depends on patients' preferences, cultural traditions, and beliefs (9–12). After agreeing to undergo an elective procedure, patients are faced with the decision of where the procedure will take place and who will perform it.
Hospital and surgeon selection may affect outcome either because the centers and providers vary in quality, or because having (or not having) a choice influences outcome by psychological or other mechanisms. A complete analysis of hospital choice involves several specific questions, including who is more likely to have a choice in selecting a hospital, why patients tend to choose one hospital over another, and whether patients who reported having a choice of hospital ultimately have better outcomes or greater satisfaction (10).
Several studies have examined hospital factors that influence patients' choice (13, 14). Hospital volume, teaching status, proximity to the patient's residence, and the reputations of both the hospital and the surgeon were among the most frequently reported factors (15–17). A few studies suggested that women, older patients, rural patients, and urban patients tend to have different priorities in identifying factors that influence their hospital choice (18–20). The question of whether choosing the hospital has any effect on patient-based outcomes has received little attention. A few studies found that control over choice of hospital or physician led to greater satisfaction with the hospital stay and physician services (21–23). To our knowledge, no studies have examined whether the opportunity to choose a hospital affects satisfaction with elective surgery, and no studies have documented heterogeneity in the choice of the hospital among different population groups.
Choice of hospitals for elective surgery may be studied in several ways, such as studying a homogenous population undergoing a variety of elective procedures, or a heterogeneous population undergoing a single common elective procedure. We believe the latter approach permits a clearer examination of possible racial and sex disparities in the opportunity to choose a hospital. Therefore, we chose to study hospital choice among a national sample of Medicare beneficiaries undergoing elective total knee replacement (TKR).
TKR is one of the most common elective surgical procedures performed in the US, with an annual volume >350,000 procedures (4). The number of TKRs is growing every year (4). Approximately two-thirds of TKR recipients are Medicare beneficiaries (24). TKR is one of the safest medical procedures, with a mortality rate of just 0.6% in the first 90 postoperative days (24).
The objectives of our study were to describe the extent of hospital choice in Medicare beneficiaries undergoing TKR; to describe factors influencing hospital choice and patterns of patient heterogeneity of choice; to examine the association of control over hospital choice with satisfaction with TKR; and to assess whether patients' demographic and clinical characteristics influence the extent to which they have a choice of hospital. Based on prior evidence, we hypothesized that a substantial number of TKR recipients did not have an opportunity to choose a hospital for TKR; that women, older patients, and nonwhite patients were less likely to choose among several hospitals; and that patients who had an opportunity to choose a hospital for TKR were more likely to be satisfied with the surgery.
PATIENTS AND METHODS
We surveyed a retrospectively assembled cohort of TKR recipients to elicit information on factors related to hospital choice and satisfaction with the surgery.
Selection of study participants.
Using claims-based validated algorithms (24), we identified Medicare beneficiaries from Ohio, Illinois, North Carolina, and Tennessee who underwent elective primary TKR in 2000. These 4 states had TKR rates near the national average and provided variability in hospital volume, race, and ethnicity. A stratified random sample of TKR recipients was selected by a 2-stage process. First, hospitals were randomly selected and stratified by volume with probability proportional to size, and then patients were randomly selected from hospitals, with the number of patients selected varying according to hospital volume strata.
Data sources and elements.
The data for these analyses came from Medicare claims and a patient survey. Medicare claims data included age, sex, an adaptation of the Charlson comorbidity index calculated from the data on the TKR admission (25–27), eligibility for Medicaid (an indicator of low income), and hospital and surgeon volume. We defined high-volume surgeons as those performing >25 primary TKRs per year in the Medicare population. Hospital volume was divided into 4 strata: 1–25, 26–100, 101–250, and >250 primary TKR per year in the Medicare population. High-volume hospitals were defined as those with annual volume >250 TKR in the Medicare population. These volume strata were suggested by a panel of 3 orthopedic surgeons.
The data on preoperative functional status, education, residency (urban/suburban/rural), body mass index (BMI), and satisfaction with the surgery were obtained from a survey administered to patients by phone or self-report mailed questionnaires (whichever the patient preferred) 2 years after the index elective primary TKR. Preoperative functional status was measured using a scale summarizing data on limp, walking distance, stair climbing, and use of walking support. We asked about these activities because they are recalled with reasonable accuracy (kappa values in the range of 0.40 to 0.50 for the comparison of recalled and prospectively obtained data) (28). Another section in the survey focused on utilization of medical services related to knee problems prior to the TKR. This portion of the survey included questions regarding the number and subspecialty of medical providers seen regarding the knee problem prior to surgery; extent of hospital choice; and whether patients, to the best of their knowledge, used the hospital closest to their residence. The survey asked patients to indicate the factor or factors that influenced their choice of hospital for TKR. The possible factors included convenient location, hospital reputation, and recommendation of their personal care physician and/or family and friends. These factors were suggested by published literature or expert opinion (13, 15–17).
Definition of outcomes and principal predictor variables.
We created a dichotomous variable, referred to as “lack of choice” throughout this report, indicating whether or not a patient reported having a choice between 2 or more hospitals. Patients were classified as dissatisfied if they answered “very dissatisfied” to the question regarding overall satisfaction with the surgery (the question included 4 responses: very satisfied, somewhat satisfied, somewhat dissatisfied, and very dissatisfied). In the analyses intended to identify population groups that are less likely to have a choice between 2 or more hospitals for TKR, the lack of choice variable was considered the principal outcome. In the analyses intended to determine whether lack of choice was associated with dissatisfaction with the surgery, the lack of choice indicator was the principal predictor.
For the analyses regarding patient factors associated with lack of choice, we first examined the bivariate association between lack of choice and each variable, including age, sex, race, obesity, comorbidity, residency, Medicaid eligibility, and education. Variables exhibiting odds ratios (ORs) >1.5 or <0.67 and/or P values <0.1 were advanced into a multivariate logistic regression model.
Association of lack of choice with dissatisfaction with TKR.
We first examined the crude association between lack of choice and dissatisfaction with TKR using a chi-square test. To examine the sensitivity of the relationships observed in the bivariate analyses to adjustment for potential confounders, we built a set of multivariate models. The first model included lack of choice variable and age (a categorical variable with 5-year increments, i.e., 65–70, 71–75, 76–80, and >80), sex, education (less than high school versus at least high school), race (white versus nonwhite), worse preoperative functional status, obesity (defined as BMI > 30), comorbidity, whether the respondent reported going to the closest hospital, whether the operation was performed in a high-volume center and/or by a high-volume surgeon, and residency (rural, suburban, urban). Factors that exhibited P values >0.1 and that did not change the lack of choice parameter by >10% were removed from the final model, allowing for a truly parsimonious model.
Heterogeneity in valuing specific hospital preference factors.
These analyses were somewhat exploratory and were restricted to bivariate associations between each hospital preference factor and specific patient characteristics such as age, sex, race, Medicaid eligibility, education, comorbidity, obesity, and residency. All analyses were performed using SAS 8.2 statistical package (SAS Institute, Cary, NC).
Using the stratified sampling procedures described above, we selected a cohort of 1,642 patients from the 18,223 Medicare beneficiaries who underwent elective primary knee replacement surgery in Ohio, Illinois, North Carolina, and Tennessee in the year 2000. Among these 1,642 patients, 20 patients (1.2%) died between the time we chose the sample and the time we contacted patients, 22 patients (1.3%) could not be located because of incorrect addresses, and 3 patients were ineligible because they did not have TKR. Of the eligible 1,597 surviving patients with valid addresses, 365 (23%) never responded to the 4 letters of invitation, 230 (14%) refused to participate, and 1,002 (63%) agreed to participate. Of these 1,002 patients, 932 (93%) returned completed questionnaires. Thus, our survey had a 60% overall response rate. Study participants were similar to those who declined to participate with respect to sex and postoperative complications, but were less likely to be poor, more likely to be white, and on average were 1 year younger than patients who declined (data not shown).
Of 932 study participants, the mean age was 74 years, the majority were white, and two-thirds were women. Thirty-four percent of study participants lived in a rural area and 20% in an urban area. Twenty-three percent had not graduated from high school. Seven percent had income levels making them eligible for Medicaid (Table 1). Ninety-six percent had knee osteoarthritis, 12% had at least 2 comorbidities, and 30% were obese (BMI > 30) (Table 1). The patients reported an average duration of symptoms of 10 months (range 3–24 months) prior to TKR. A total of 197 (21%) had both knees replaced.
OA = osteoarthritis; IQR = interquartile range; TKR = total knee replacement.
Age >80 years
< High school
≥ High school
Symptoms duration, median (IQR), months
Had TKR on contralateral knee
OA present on contralateral knee
Health care utilization
Health care professionals seen for knee problem
Primary care provider
Reported lack of hospital choice
Relied on doctors in making decision
Went to the closest hospital
A total of 445 surgeons operated on the 932 patients. Twenty-nine of these surgeons (6.5%) performed operations in more than 1 hospital, and no surgeon performed operations in more than 2 hospitals.
Health care utilization.
We asked study participants which health care professional they had seen for their knee problem prior to their TKR. Only half of the sample indicated seeing their primary care physician (PCP) for their knee problem prior to their TKR. Forty percent reported seeing at least 2 specialists (Table 1). Fewer than 10% reported seeing a rheumatologist or physical therapist. Of the patients who were not seen by a PCP for their knee problem prior to TKR, 403 (87%) of 469 only saw an orthopedic surgeon. Of those who reported seeing more than 1 health care professional, 366 (98%) had seen an orthopedic surgeon.
Extent of choice.
A total of 496 study participants (53%) reported that they did not have a choice between 2 or more hospitals for TKR (Table 1). Among those who had a choice, 238 (53%) reported relying on doctors to make a decision about hospital selection. Patients who did not have a choice were more likely to have TKR performed in the hospital closest to their home (OR 1.80, P < 0.0001). Bivariate analyses indicated that older patients (>80 years of age), women, nonwhites, poorer patients, and those who lived in suburban areas were more likely to report that they did not have a hospital choice (Table 2). Those factors were advanced into multivariate models, in which only older age (OR 1.62), female sex (OR 1.52), and suburban residence (OR 1.68 compared with rural) remained statistically significantly (P < 0.05) associated with lack of hospital choice (Table 2). Patients who reported seeing only an orthopedic surgeon were as likely to report lack of hospital choice as patients who did not limit their health care utilization to orthopedic surgeons only prior to TKR (52.9% versus 53.5%).
Table 2. Association of select patient characteristics with lack of hospital choice between 2 or more hospitals for TKR*
Did not have a choice, %
TKR = total knee replacement; OR = odds ratio; 95% CI = 95% confidence interval; OA = osteoarthritis; BMI = body mass index.
< High school
≥ High school
BMI > 30
BMI ≤ 30
Preoperative functional status (lowest quartile)
Patient had seen multiple health care professionals
Went to the closest hospital
Went to the high-volume hospital
Was operated on by high-volume surgeon
Lack of hospital choice and satisfaction with the surgery.
In examining the association between lack of hospital choice and satisfaction with the surgery, we used multivariate logistic regression and adjusted for age, sex, education level, surgeon volume, poverty status, obesity, and preoperative functional status. These analyses showed that patients who reported lack of hospital choice were twice as likely to be very dissatisfied with the surgery than patients who reported having a choice (adjusted OR 2.09, 95% confidence interval [95% CI] 1.13–3.87).
Factors influencing hospital choice.
These analyses were restricted to study participants choosing between 2 or more hospitals (n = 436). The reputation of the surgeon was the most frequent factor influencing hospital choice, reported by 68% of patients choosing between 2 or more hospitals for TKR (Figure 1). Approximately half of the eligible patients reported reputation of the hospital, convenient location for the patient, convenient location for family and friends, and recommendation of the surgeon among factors influencing their hospital choice (Figure 1). Approximately one-third of the patients eligible for these analyses reported that the recommendation of family and friends and the recommendation of the PCP affected their hospital choice. Only ∼16% mentioned hospital volume among the reasons for hospital choice.
Differences among subgroups of the cohort in the factors that influenced hospital choice.
Men and women reported that a similar set of factors influenced their choice of hospital. Similarly, preoperative functional status and medical comorbidity did not influence the set of factors governing hospital choice. In contrast, older patients (>80 years of age) were more likely to listen to recommendations of family and friends (12.4% versus 6.5% of patients <80 years; OR 2.03, 95% CI 1.02–4.06). Patients with less than a high school education were significantly less likely to report that hospital volume influenced their choice than more educated patients (8% versus 19%; OR 0.35, 95%CI 0.16–0.80). Obese patients were more likely than nonobese patients to report that convenient location for themselves and for family and friends influenced hospital choice (61% versus 52%; OR 1.45, 95% CI 0.96–2.18). Similarly, obese patients were more likely to report that the recommendation of health care professionals had an important influence on their choice (65% versus 53%; OR 1.63, 95% CI 1.07–2.47). White patients were more likely than nonwhites to report that any of the listed factors influenced their hospital choice (Figure 2). Finally, poorer patients were also less likely to base their decision on any of the suggested factors, or on the recommendation of health care professionals in particular.
We studied a population-based sample of 932 Medicare beneficiaries who underwent elective primary TKR in 2000. Our goals were to better understand the extent to which patients had a choice between 2 or more hospitals; whether lack of hospital choice was associated with dissatisfaction with the surgery; why patients chose one hospital over another; and how different population groups varied in the factors that influenced their choice of hospitals. More than half of our study participants reported that they did not have a choice of hospitals for their TKR. Women, older patients, nonwhites, poorer patients, and suburban patients were less likely to report having a choice of hospitals for TKR. Furthermore, TKR recipients in our sample who reported lack of hospital choice for TKR were 2 times more likely to be very dissatisfied with the procedure. Among factors influencing hospital choice, the surgeon's recommendation, the reputation of the hospital and the surgeon, and the convenience of the hospital's location were each cited by more than half of the patients who had a choice. In contrast, only 16% indicated that the volume of TKRs performed by the hospital influenced their hospital choice. Different population groups varied in the reasons that influenced their hospital choice. Obese patients, for example, indicated a greater reliance on both convenience of hospital location and health provider recommendation than nonobese patients.
Our results are consistent with several studies examining the effect of provider choice on satisfaction. Woodside et al (21) found that personal control over choice of hospital correlated with increased satisfaction with patient stay, as well as with satisfaction with physicians and nursing staff. Because satisfaction was measured after discharge, the authors suggested that the rating might be a good proxy for quality outcome. Schmittdiel et al (23) found that patients in a large group-model health maintenance organization with the opportunity to select their own PCP were 16–20% more likely to be satisfied with their care. Forrest et al (29) found that patients who believed they had a greater choice of PCPs tended to rate the patient-PCP relationship more favorably. Although most of these studies focused on the choice of PCP or hospital stay, our study was, to our knowledge, the first to suggest an association between satisfaction with elective surgery and opportunity to choose the hospital for such a procedure.
It has been shown that patients' family and friends influence patients' decisions about whether to undergo the joint replacement surgery (30). The results of our study showed that family and friends also influence patients' choice of the hospital for undergoing TKR. Patients who are satisfied with the results of their TKR are more likely to suggest the procedure to friends and relatives. Therefore, a better understanding of factors affecting satisfaction may help increase satisfaction with TKR, and indirectly help increase promotion of TKR by word of mouth, which may result in the increased acceptance of TKR leading to decreasing burden of knee OA-related disability among patients with end stage knee OA.
Our results summarizing factors that influence hospital choice were also generally consistent with the literature. Heischmidt and Heischmidt (31) found that patients place considerable weight on the overall hospital reputation. Previous research has demonstrated that convenience, defined as a multidimensional concept including convenient location and general familiarity with the structure and personnel of the referring facility, is one of the primary factors affecting the general practitioner's choice of hospital when referring patients for elective surgery (17). Several studies found that the larger the hospital, the more likely that hospital would be chosen. However, Adams and Wright (13) found that older residents saw larger hospital size as a drawback. The results of our study indicated that procedure volume influenced hospital choice for a small proportion of patients. A large body of evidence exists on the association between volume and outcomes of many surgical procedures including TKR (24). Because many patients indicated that their PCPs influenced their choice of hospital for TKR, discussion about hospital volume and outcomes of TKR may enrich the dialogue between the physician and the patient choosing hospitals for elective TKR.
The finding that suburban patients were less likely to have a choice may be explained by the fact that although rural patients have to travel to reach any hospital performing TKR, and urban patients have several such hospitals within a short distance, suburban patients may receive most of their care from a single community hospital in close proximity to their residence, limiting their knowledge and choice. The substantial influence of the recommendation of the surgeon and PCP may reflect a passive approach to hospital choice even among patients who had the opportunity to choose among several hospitals. Many patients may choose the hospital based on where their surgeon operates. The fact that the nonwhite population expressed less preference across all factors influencing hospital choice suggests that the factors driving hospital choice among racial and ethnic minorities may lay beyond the universe of factors already described in the literature and included in our questionnaire. Due to limited power, the 36% increase in lack of hospital choice among nonwhite patients compared with white patients undergoing TKR did not reach statistical significance in multivariate analyses. More research, including qualitative studies, is needed to better understand factors that are important in hospital choice for these vulnerable populations. The additional difference in factors influencing obese and nonobese patients' choice further suggests that we might not understand exactly who constitutes a vulnerable population, as well as what issues are most important to best enable health care utilization for all populations.
The strengths of our study include a population-based sample and access to several data sources. However, our study also had several limitations. The survey was conducted 2 years after TKR, and thus required postoperative recall of some preoperative factors such as functional status. Furthermore, postoperative outcomes might affect patient recall such that those less satisfied with their surgical outcome might be systematically less likely to report having had a choice.The survey did not include questions directly inquiring about choice of surgeon. Furthermore, it is unclear if data from 1 year in 4 states can be generalized to the entire US.
The results of our study indicate that more than half of patients did not have a choice in selecting hospitals for TKR. Patients reporting lack of choice were more likely to be dissatisfied with surgery. Interventions to address preferences for choice of hospital among women, older patients, patients living in suburban areas, and nonwhite patients may enhance care and improve satisfaction with care for patients with advanced knee arthritis.