Living in immigrant communities does not impact total knee arthroplasty outcomes: experience from a high-volume center in the United States

This is a retrospective study which was conducted using prospectively acquired longitudinal registry data from a large volume orthopedic hospital. The hospital mainly serves the tristate area (New York, New Jersey and Connecticut), which has a high density of immigrants, permitting us to study the association of neighborhood immigrant proportions on TKA outcomes [22].

Patients at least 18 years old were prospectively enrolled in the registry between May 1, 2007 and July 1, 2011 after providing informed consent. Patients included in this study completed questionnaires, including patient reported outcomes, both prior to TKA and 2 years after TKA. Registry patients who lived outside the tristate area were excluded from our analysis so as to focus on the main catchment area for the hospital and permit greater generalizability by excluding patients with resources to travel long distances for care. We excluded patients who underwent revision TKA, bilateral TKA and patients who had a contralateral TKA within 2 years of the index surgery so that the study reflects outcomes of the index surgery. We also excluded patients who had TKA secondary to fractures, avascular necrosis or inflammatory diseases such as rheumatoid arthritis, psoriatic arthritis and ankylosing spondylitis for the study to represent OA patients.

Baseline covariates collected in the registry included date of birth, sex, body mass index (BMI), race (black, white, Asian, other), ethnicity (Hispanic, non-Hispanic), highest education level completed (some high school, high school graduate, some college, college graduate, master’s degree or higher) and living status (alone or with other(s)). The patient-reported outcomes collected in the registry were the pre-operative (pre-op) and 2-year postoperative (post-op) Knee injury and Osteoarthritis Outcome Score (KOOS), from which Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and function scores are derived, and the Hospital for Special Surgery (HSS) Expectations Survey before surgery. The WOMAC is a validated lower extremity specific scale and commonly used as a patient-reported instrument after TKA. WOMAC results are scored between 0 and 100 points, with higher scores indicating better outcomes. Lower extremity pain and function are assessed using 2 subscales; a higher score indicates better status and a score less than 40 indicates significant pain and poor function [23]. The primary outcomes of the study were 2-year WOMAC pain and function scores. The Hospital for Special Surgery (HSS) Expectations Survey is a validated instrument which is specific for TKA outcomes and assesses multiple domains relevant to TKA [24]. Administrative data including comorbidities and patients’ addresses were collected from hospital records. The addresses were used to permit geocoding to link individual level data to census tract defined area-based measures (ABMs).

Geocoding is the process of converting addresses to a precise geographic coordinate (defined by latitude and longitude) which one can use to place markers or positions on a map (Fig. 1). This facilitates spatial analysis using geographic information systems (GIS) and Enterprise Location Intelligence systems. The American Community Survey/Census Bureau release multiple complex datasets containing area-based measures (ABMs) every 10 years. These ABMs are objective data collected by the Census Bureau and broken down into smaller subsections to characterize the neighborhood. This data is presented using smaller subunits - zip codes (which average 150,000 residents), census tracts (average 4000 residents), and census blocks (average 1000 residents). For measuring neighborhood social factors, census tracts are more consistent than census blocks or zip codes [25]. Census tracts are designed to be homogenous with respect to population characteristics, economic status, and living conditions [20].

Each patient in the registry with a geocodable address was assigned to a census tract using ArcGIS [26]. ABMs were then extracted for each census tract using data released by the Census Bureau. Our primary ABM of interest was “foreign born”. The term “foreign-born” and “immigrant” are used interchangeably and refer to anyone who is not a US citizen at birth. This includes naturalized US citizens, lawful permanent residents, temporary migrants, humanitarian migrants, and persons illegally present in the United States [15]. Within each census tract, we categorized the immigrant proportion (IP) as low (< 10%), moderate (10–40%) and high (≥40%), since this categorization has been utilized in similar studies and has been associated with health outcomes [21]. For sensitivity analyses, we dichotomized IP values as < 20% IP and > 20% IP and also kept IP as a continuous variable and performed all primary outcomes analyses. The other ABM included in this study was the percent of the population living below poverty as a measure of neighborhood socioeconomics. The tristate area has a total of 7762 census tracts. (Fig. 1).

Continuous variables are summarized as mean ± standard deviation and compared using t-tests for binary comparisons or ANOVA tests when comparing more than two subgroups. Categorical variables are summarized as frequency and percent and compared using chi-squared tests. Spearman correlation coefficients were calculated for neighborhood poverty and IP to determine the strength of the relationship between these two crucial ABMs. We used descriptive statistics to compare respondents and non-respondents of the 2-year post-op WOMAC scores. Univariate and multivariable linear mixed effects models, with census tracts treated as random effects, were used to analyze the impact of individual and census tract-level factors on pre-op and 2-year post-op WOMAC pain and function scores. Multivariable models were adjusted for individual-level age, sex, BMI, comorbidities (none versus at least one), and living status (alone versus with other(s)), as well as and neighborhood poverty. 2-year post-op models were additionally adjusted for baseline WOMAC scores. Separate linear mixed effects models were calculated to assess the interactions between IP and sex, as well as IP and individual education level. There was less than 1% missing data and thus we did not impute missing values.

Due to the reduced sample size in the ≥40% IP cohort, we performed a matched analysis to further assess the impact of IP after controlling for age, sex, BMI and number of comorbidities. Patients from high IP neighborhoods were matched to one patient from a low IP neighborhood and to one patient from a moderate IP neighborhood. Matching factors included age (within 5 years), sex (perfect match required), BMI (within 1 kg/m2) and number of comorbidities (0 or ≥ 1, perfect match required). Univariate and multivariable linear mixed effect models were performed, treating each matched pair as a random effect. Multivariable models were adjusted for individual living status (alone versus with other(s)) and census tract poverty. SAS version 9.4 was used for the analysis.