Study design and data sources

This study was a retrospective cohort analysis of the data from a national de-identified database. Thus, it was exempt from Institutional Review Board approval. The American College of Surgeons National Surgeons Quality Improvement Program (ACS-NSQIP) 2021 database was queried, with THA defined by the Current Procedural Terminology code 27130. The ACS-NSQIP is a nationally validated program that has about 700 voluntarily participating hospitals and whose case selection was performed systematically.

Patient inclusion and exclusion criteria

The following criteria were used to generate a homogenous study sample. Only elective cases were included. Cases with bone fracture stated in the diagnosis were excluded. The exclusion criteria also included 1) patients admitted to the hospital for longer than one day preceding surgery, 2) patients with preoperative documentation of end-stage renal disease, metastatic disease, sepsis, or bleeding diathesis, and 3) patients with American Society of Anesthesiologist Physical Status (ASA) classification 5 (moribund, not expected to survive without procedure) [12]. There were no modifications made to the ACS-NSQIP classifications for these variables. These inclusion and exclusion criteria were used to generate a sample that could be generalized to the general population of patients for elective THA. These criteria were decided by the researchers a priori. All cases received either spinal anesthesia or general anesthesia as the principal anesthetic technique. All cases were required to have a documentation of home support (living alone versus with others at home).

Measurements and data handling

The primary independent variable was home support (living alone versus with others at home). Other independent variables included age, sex, body mass index (BMI), hypertension, insulin-dependent diabetes, current smoker, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), chronic steroid use, functional status, fall within 6 months, dementia, ASA classification, and anesthetic technique (spinal versus general).

The primary endpoint was discharge disposition (home versus non-home). For those discharged home, requiring home services was a major secondary endpoint.

Secondary endpoints also included functional status at discharge, postoperative delirium, hospital length of stay, and 30-day event rates for unplanned resource utilization, wound complications, systemic complications, bleeding requiring transfusion, and mortality. Unplanned resource utilization included unplanned readmission and return to the operating room. Wound complications included superficial surgical site infection (SSI), deep incisional SSI, organ space SSI, and wound dehiscence. Systemic complications included cardiac arrest, myocardial infarction, stroke, reintubation, pneumonia, deep venous thrombosis, pulmonary embolism, bleeding, sepsis, septic shock, and acute kidney injury.

Statistical analysis

R version 4.3.1 (R Core Team, Vienna, Austria) was used to perform statistical analysis [13]. Patients living alone were identified, and those with others at home were matched using propensity scores computed by the following patient demographics and medical comorbidities: age, sex, BMI, hypertension, diabetes, smoking, COPD, CHF, chronic steroid use, functional status, ASA classification, anesthetic technique, fall history, and dementia. A 1:1 logistic regression covariate estimates with no replacement was used. Cases with missing data were excluded from the analysis in order to promote the creation of balanced cohorts, but with two exceptions: those with unknown functional status were assumed to be independent, and those with unknown fall history were assumed to have not fallen within 6 months given the observed distribution of responses for these two variables. Propensity scoring has been utilized widely to reduce selection bias in cohort studies [14, 15]. Standardized mean differences were used to assess balance among matched pairs, with a standardized mean difference < 0.1 denoting adequate balance [16]. A maximum propensity score difference between groups was not specified a priori.

The associations between living alone and various endpoints outcomes were first examined by univariate analyses. Categorical variables were assessed by Pearson’s chi-squared test without continuity correction while hospital length of stay was tested by Student’s t-test. Additionally, chi-square methodology was used to calculate odds ratios (OR), 95% confidence intervals (CI), and fragility indices [17]. The fragility index indicates that changes of how many subject-dependent classifications alter the statistical significance of a hypothesis test. A greater number in the fragility index suggests a more robust difference between groups [18, 19].

A multivariable analysis was performed on the primary outcome. Multiple logistic regression modeling was used and adjustments were made for age, sex, BMI, hypertension, diabetes, smoking, COPD, CHF, chronic steroid use, functional status, fall history, dementia, ASA classification, anesthetic technique, and home support when univariate testing showed significance at α ≤ 0.05. The selection of these variables was decided a priori and based on the availability of the data for analysis. As with matching, only complete cases were analyzed. Missing data were not imputed. Backwards stepwise model adjusted by Akaike information criterion was used to determine the subsequent variable selection. Results of independent predictors were presented as adjusted odds ratios (AOR) and 95% confidence interval (CI), and the c-statistic was used to assess model discrimination [20].

Continuous variables and categorical variables were presented as mean (standard deviation) and frequency (%), respectively. All hypothesis tests were two-sided, with α ≤ 0.05 as having a significant difference.

Statistics on postoperative home support

To date, there is a paucity of evidence to help determine what percentage of patients receiving primary THA are living independently or with home support. Based on 2021 data from the ACS-NSQIP, approximately 30% of patients receiving primary THA with reported home support status were found to be living alone. However, this metric may not be valid as over half of these cases did not report home support status. Studies by Iwata et al. (2023) and Fleischman et al. (2018) showed approximately 20% of their study participants undergoing primary THA were reported to be living alone, however these metrics were collected from single-institution databases and may not be generalizable to larger populations [10, 21].

Social factors affecting patient outcomes

Given that a significant number of patients receiving primary THA are living independently, there has been much research aimed at determining whether living and social conditions, such as caregiver support at home, affect patient outcomes postoperatively. Fleischman et al. (2018) showed no increase in complications or unplanned clinical events for patients living alone compared to those living with others. Additionally, there were no significant differences in functional outcomes or pain relief in patients living alone or living with others [21].

However, prior systematic reviews have examined the impact of social determinants on patient-reported outcomes and adverse events following TJA [22, 23]. Karimi et al. (2023) found that patients with more social deprivation had a higher proportion of non-home discharge and lower improvements from baseline patient-recorded outcome measures following TJA [22]. A systematic review and meta-analysis by Wylde et al. (2019) showed evidence that social support could be a prognostic factor for some patient-reported outcome measures following total joint replacement [23]. Notably, both reviews reported appreciable limitations in their findings due to the methodological quality of available studies and to the inconsistent measurement of social support or deprivation. To combat the complex and multidimensional nature of social support, future studies must specify the social factors they are measuring and utilize metrics that effectively capture social determinants.

Factors affecting discharge location

Numerous retrospective studies have investigated predictors of discharge destination following TJA [8–10, 24, 25]. Additionally, multiple authors are credited with developing and validating pre-operative risk assessment tools to predict discharge destination for patients undergoing TJA [26–29]. Interestingly, two out of four known discharge prediction tools utilize home caregiver status in their scoring [27, 28]. While our study focused solely on living alone as a predictive factor of discharge destination, other studies identified multiple predictors for non-home discharge including older age [8–10, 24, 25], female gender [9, 30], certain comorbidities including obesity [8, 30] and pulmonary disease [9, 30], postoperative functional status [8, 30], and even patient expectation of discharge destination [25]. The studies that included living alone in their analysis had findings consistent with our study [8, 10, 25].

Factors affecting hospital length of stay

Similarly to discharge destination, prior studies also have looked for factors that influence hospital length of stay after TJA [9, 21, 24, 31]. Correspondingly, Fleischman et al. (2018) determined that living alone was associated with longer inpatient stays [21]. Other retrospective studies found strong associations linking length of stay with age [24], diabetes [9], pre-existing cognitive impairment [31], discharge destination [9], and even day of surgery [9]. These results suggest that a combination of patient and organizational factors play as drivers in determining hospital length of stay.

Implications of our findings

Our findings suggest that patients living alone may require additional support postoperatively as well as longer inpatient stays to receive sufficient patient education and coordinate support services at home. Patients living alone may lack the physical, cognitive, and emotional support and interactions needed in the initial recovery phase following THA, necessitating a discharge destination that provides ongoing caregiver support. Fortunately, there is no evidence to date associating lack of home support with negative postoperative outcomes following THA. However, animals studies have shown that familial support reduces POCD, possibly via inhibition of neuroinflammation and activation of the lateral habenula-ventral tegmental area neural circuit after surgery [32–34]. These novel findings from animal studies suggest that the presence of home support may influence postoperative neuropsychological outcomes, however there is a need for clinical studies to better understand these phenomena and their implication in patients living alone where social interactions are very limited. In this context, it is worth noting that there was a trend for living alone patients to have a higher incidence of POD (OR: 1.64 [0.98 to 2.76], P = 0.06). Since the diagnosis of POCD requires neurocognitive assessment before and after surgery [6], there is no data on the incidence of POCD in this retrospective analysis.

Given that patients undergoing THA have unique medical and social backgrounds and require varying levels of support in the postoperative period, it is crucial that healthcare providers risk-stratify patients early in the preoperative phase to ensure proper discharge planning and deliver the optimal level of care throughout recovery and rehabilitation. Our study emphasizes the importance of including home support in these risk assessments to obtain a more comprehensive view of our patients’ postoperative trajectories.

Strengths

The study is strengthened by reporting of fragility indices for all outcomes, complementing the P value with an interpretable measure of patient classification on statistical significance [18, 19]. At a minimum, the classification of 142 discharge dispositions would need to be changed to the opposite class to alter the statistical significance of these findings. This study is further strengthened by propensity score matching those living alone to those with others at home in order to control for selection bias [16]. The cohort living alone before propensity score matching had an overrepresentation of elderly patients, females, non-diabetics, and recent falls. Matching produced balanced cohorts, as evidenced by post-matching univariate analyses (all P > .05) and a marked reduction in absolute standardized mean differences.

Limitations

There are several important limitations that are inherent to the study design. This was an observational study, and as such the defined endpoints and covariates were limited to those captured by the ACS-NSQIP database. Home support was not documented in 22486 of 28163 patients otherwise meeting inclusion criteria, reducing the power of this study and potentially introducing selection bias. This supports the use of propensity score matching as one method to reduce sampling bias, though matching is not without limitations [35]. Additionally, the characters of individuals living with the supported patients are not reported in the database; factors like age, functional status, health literacy, and chronic illness are likely influential but to an unknown degree. The screening test for postoperative delirium is also not documented. Thus, it is a distinct possibility that the screening methodology has not been validated for the 93 patients with dementia (2.9% of the matched sample). The study design also did not allow for exclusion or stratification by intraoperative factors, such as transfusion requirements. This study design was used to allow the application of our findings to general population of patients with elective hip arthroplasty. Instead, some perioperative complications, such as transfusion requirements, are studied as secondary endpoints in the analysis. Lastly, the results of this study cannot be generalized to patients meeting the a priori exclusion criteria.