Predictors of In-Hospital Mortality, Cost, and Length of Stay following Periprosthetic Fracture of the Hip

Abstract

Purpose

The frequency of total hip arthroplasty (THA) has increased significantly. The morbidity and economic burden of periprosthetic hip fractures (PPHFx) following THA is high. The objective of this study is to identify predictors of in-hospital mortality, length of stay (LOS), and hospital costs associated with PPHFx.

Materials and Methods

A retrospective analysis of the National Inpatient Sample (NIS) for all patients diagnosed with periprosthetic fracture around the internal prosthetic hip joint (International Classification of Disease, Tenth Revision, Clinical Modification [ICD-10-CM], M97.01XA, and M97.02XA) between 2016 and 2019 was conducted. Predictors of outcomes following PPHFx were assessed using univariable analyses; those with a P-value below 0.2 were included in the final multivariable models. Logistic regression was used to determine predictors of in-hospital mortality and linear regression for assessment of hospital cost and LOS predictors.

Results

The study included 77,565 patient encounters. Significant predictors of in-hospital mortality included age, female sex, and comorbidities, such as metastatic cancer and liver disease. Predictors of increased LOS included black race, comorbidities, such as paralysis and fluid and electrolyte disorders, primary payment source, household income, hospital size, and type. Hospital costs showed an association with race, comorbidities, hospital type, and procedure.

Conclusion

In the setting of PPHFx, higher comorbidity, particularly fluid and electrolyte, paralysis, and metastatic cancer, showed an association with all three outcome variables. In addition to comorbidity, advanced age, particularly nonagenarians, showed an association with in-hospital mortality, while cost was primarily driven by the procedure performed.

INTRODUCTION

A remarkable surge in the frequency of total hip arthroplasty (THA) has occurred over recent decades and this upward trajectory is expected to continue, with projections indicating an increase to 635,000 procedures by 2030¹⁾. In fact, hip arthroplasty was the fourth most performed orthopaedic surgery in the United States in 2018, accounting for a substantial 599,500 cases²⁾. This increasing trend is not unique to the United States, as global data indicates a noteworthy increase in the incidence of THA^3,4). This growing prevalence of THA can be attributed to its proven efficacy⁵⁾, along with technological advancements for the purpose of enhancing the longevity of implants⁶⁾. As a result, joint reconstruction procedures have evolved to encompass a broader spectrum of patients, including the younger population⁷⁾. While the trend in postoperative complications has declined⁸⁾, a significant increase in the absolute number of complications, particularly periprosthetic hip fracture (PPHFx), is anticipated⁹⁾.

The incidence rate of PPHFx is 1.4 per 1,000 THA procedures per year¹⁰⁾. Management of PPHFx is notably complex, primarily due to the elevated levels of morbidity and mortality associated with this complication¹¹⁾. The reported mortality rates vary significantly, with rates as high as 10% within the first-month post-surgery¹²⁾, although more recent data indicates a reduced rate, nearing 3%¹³⁾. PPHFx carries a substantial economic burden, surpassing that of primary THA. This increased cost is attributed to various factors, encompassing heightened preoperative expenses related to preoperative management and imaging, the cost of the surgical procedure itself, and the cost of implants¹⁴⁾. The postoperative length of stay (LOS) also contributes to the economic burden¹⁵⁾.

Although institutional studies have identified predictors of mortality^16-18), cost¹⁴⁾, and LOS¹⁹⁾ in the context of PPHFx, a comprehensive large-scale studies that investigate predictors of outcomes and the associated economic burden are significantly limited²⁰⁾. The National Inpatient Sample (NIS), a large, multi-institutional, retrospective database encompassing data from all-payer sources to address this knowledge deficit, was used in the current study. The purpose of this study is to identify the predictors of mortality, cost, and LOS after PPHFx.

MATERIALS AND METHODS

This study did not require institutional review board (IRB) approval because it did not meet the definition of human subject research at Cleveland Clinic. Public health surveillance activities such as this study do not require IRB review/approval at our institution.

1. Data Source

This study utilized data from the NIS from 2016 to 2019²¹⁾. The transition from the International Classification of Diseases, 9th edition (ICD-9) to the 10th edition (ICD-10) coding system occurred in 2015, and as such, the study adopted the ICD-10 coding system for data extraction and analysis²¹⁾. The NIS comprises a stratified sample, representing approximately 20% of all discharges nationwide, and provides national estimates²¹⁾.

2. Study Population

The study population included encounters with patients 18 years or older who were diagnosed with periprosthetic fractures around the internal prosthetic hip joint, as indicated by the International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10-CM) codes M97.01XA and M97.02XA, were included in the analysis.

3. Data Collection

Data was extracted from the NIS for the variables of interest, including age, sex, race, comorbidities, primary payment source, median income quartile, admission day, hospital bed size, and hospital type. Table 1 shows all measured values for each variable. Of note, continuous variable age was categorized as <70, 70-79, 80-89, and ≥90 years. The median income quartile represents the median household income quartile associated with a patient’s ZIP Code. Hospital bed size is defined by the Healthcare Cost and Utilization Project (HCUP)²¹⁾ and varies by hospital type and region. For example, in the Northeast, rural hospitals are categorized as small (1-49 beds), medium (50-99 beds), and large (≥100 beds). The R package Comorbidity²²⁾ was utilized for capture of comorbidity profiles of 30 variables for patient encounters, including hypertension, cardiac arrhythmias, electrolyte disorders, and metastatic cancer. Of note, as bone metastases accounted for fewer than 30 encounters (3% of metastatic cancer cases), we did not distinguish between bone and soft tissue metastases for conditions such as metastatic cancer. The Elixhauser comorbidity index using the package’s built-in mapping algorithm based on Quan et al.²³⁾ and scoring weights based on van Walraven et al.²⁴⁾ were used. The Elixhauser scores were stratified according to five categories (0, 1, 2, 3, ≥4). ICD-10 procedure codes were used for identification of treatment modalities. Encounters without procedure codes were categorized as nonoperative, while previously validated ICD-10 codes were used for classification of open reduction and internal fixation (ORIF), Revision, and ORIF plus Revision based on previous methodology^25,26) (Supplementary Table 1).

Table 1.

Baseline Demographics and Encounter Characteristics (n=77,565)

Characteristic	Value
Age (yr)	78.0 (77.5-78.5)
<70	20,515 (26.4)
70-79	20,690 (26.7)
80-89	24,655 (31.8)
≥90	11,705 (15.1)
Sex
Male	25,745 (33.2)
Female	51,820 (66.8)
Race
White	68,745 (88.6)
Black	3,300 (4.3)
Hispanic	3,320 (4.3)
Other	1,280 (1.7)
Asian or Pacific Islander	615 (0.8)
Native American	305 (0.4)
Elixhauser score
0	3,825 (4.9)
1	9,710 (12.5)
2	14,975 (19.3)
3	15,265 (19.7)
≥4	33,790 (43.6)
Comorbidity*
Hypertension (uncomplicated)	38,275 (49.3)
Cardiac arrhythmias	23,135 (29.8)
Fluid and electrolyte disorders	21,800 (28.1)
Hypertension (complicated)	19,280 (24.9)
Hyperthyroidism	17,160 (22.1)
Chronic pulmonary disease	16,560 (21.3)
Depression	15,390 (19.8)
Renal failure	13,995 (18.0)
Congestive heart failure	13,045 (16.8)
Obesity	10,185 (13.1)
Primary payment source
Medicare	64,105 (82.6)
Private insurance	8,865 (11.4)
Medicaid	2,545 (3.3)
Other	1,415 (1.8)
Self-pay	605 (0.8)
No charge	30 (0.0)
Median income quartile
1	17,300 (22.3)
2	20,575 (26.5)
3	20,235 (26.1)
4	19,455 (25.1)
Admission day
Weekday	59,300 (76.5)
Weekend	18,265 (23.5)
Hospital bed size
Large	39,815 (51.3)
Medium	22,435 (28.9)
Small	15,315 (19.7)
Hospital type
Urban teaching	56,440 (72.8)
Urban non-teaching	15,755 (20.3)
Rural	5,370 (6.9)
Outcomes
In-hospital mortality	1,285 (1.7)
Length of stay	5.0 (4.5-5.5)
Cost (US$)	19,936 (19,695-20,177)

Values are presented as median (interquartile range) or number (%).

Data derived from the NIS (National Inpatient Sample).

All adult patients diagnosed with a hip periprosthetic fracture between 2016 and 2019 were included.

*Comorbidity shows the ten most common comorbidities.

4. Statistical Analysis

The dependent variables included in-hospital mortality, LOS, and cost. Costs were standardized to 2019 using the Bureau of Labor Statistics CPI Inflation Calculator²⁷⁾. Logarithmic transformation was applied to LOS and costs and inverse log-transformed for interpretation. The following predictor variables were considered: age, sex, race, comorbidities, primary payment source, median income quartile, admission day, procedure type, hospital bed size, and hospital type. Discharge weights were used to generate national estimates. Of note, the cost was calculated using the HCUP Cost-to-Charge Ratio for Inpatient Files (CCR for Inpatient Files), which uses hospital-specific cost-to-charge ratios²⁸⁾. While charges represent the amount billed, costs “reflect the actual expenses incurred in the production of hospital services” and thus prevent confounding due to hospital-level cost-to-charges differences²⁸⁾.

To address the inherent database clustering of the NIS, our approach employed an LMM (linear mixed-effects model) that incorporated random intercepts for continuous variables alongside a GLMM (generalized mixed-effects model) for categorical variables, utilizing random intercepts to account for clustering effects as recommended by the HCUP²⁹⁾. These were used for assessment of predictors in a univariable model for each outcome. The final multivariable analysis included all predictors with a P-value less than 0.2³⁰⁾. The multivariable analysis used reference variables: <70 years, male sex, white race, absence of comorbidity, Medicare as the primary payment source, fourth median quartile, weekday admission, procedure type, small hospital size, and rural hospital type. Independent predictors were identified based on their significance at an alpha level of 0.05 in the final model. Assumptions of the models were examined and met³¹⁾. All NIS guidelines were adhered to in order to reduce additional bias³²⁾. All computations were executed using R version 4.3.1³³⁾.

RESULTS

1. Baseline Demographics and Encounter Characteristics

A total of 77,565 patient encounters of adults diagnosed with PPHFx between 2016 and 2019 were found. Table 1 shows a median age of 78.0 years (IQR 77.5-78.5), prominently featuring patient encounters in the 70-79 (26.7%) and 80-89 (31.8%) age groups. Sex distribution indicated a higher prevalence of females (66.8%). The racial composition was predominantly white (88.6%), with smaller representation of black (4.3%), Hispanic (4.3%), and other racial backgrounds. The Elixhauser score showed substantial variability; 43.6% of patient encounters had of 4 or higher. Common comorbidities included hypertension (49.3%), cardiac arrhythmias (29.8%), and electrolyte disorders (28.1%). Medicare served as the primary payment source for the majority (82.6%) of patient encounters, followed by private insurance (11.4%) and Medicaid (3.3%). Medicare is an American federal program covering individuals 65 or older, while Medicaid is a state and federal program providing healthcare for low-income populations. Income quartile distribution was well-balanced across quartiles. Most admissions occurred on weekdays (76.5%). Hospital characteristics showed a majority of large-sized facilities (51.3%) and urban teaching hospitals (72.8%). In-hospital mortality was 1.7%, while the median LOS was 5.0 days (IQR 4.5-5.5). The median cost of care was US$19,936 (IQR US$19,695-20,177). Revision arthroplasty was the most commonly performed procedure (32.3%), followed by ORIF (31.8%), nonoperative (20.8%), and ORIF plus Revision (15.1%).

2. Predictors of In-Hospital Mortality following PPHFx

Table 2 shows the significant predictors of in-hospital mortality following PPHFx. Encounters of patients aged 70-79 years showed an odds ratio (OR) of 2.5 (95% confidence interval [CI] 1.1-5.7, P=0.022) compared to those under 70, while the 80-89 age group had an even higher OR of 5.0 (95% CI 2.3-11.1, P<0.001). The most striking increase in risk was observed in ages 90 years and above, with an OR of 11.0 (95% CI 4.8-25.5, P<0.001). Female sex had an OR of 0.6 (95% CI 0.4-0.9, P=0.010) when compared to male counterparts. Comorbidities contributed significantly to mortality risk, with metastatic cancer (OR 8.4, 95% CI 2.2-32.7, P=0.002), liver disease (OR 8.2, 95% CI 3.6-18.9, P<0.001), paralysis (OR 6.9, 95% CI 1.8-27.1, P=0.006), fluid and electrolyte disorders (OR 4.1, CI 2.6-6.4, P<0.001), and pulmonary circulation disorders (OR 3.6, 95% CI 1.8-7.2, P<0.001) showing the most substantial associations.

Table 2.

Predictors of In-Hospital Mortality following Periprosthetic Hip Fracture

Predictor	OR	95% CI	P-value
Age (yr)
70-90	2.5	1.1-5.7	0.022
80-89	5.0	2.3-11.1	<0.001
≥90	11.0	4.8-25.5	<0.001
Sex
Female	0.6	0.4-0.9	0.010
Comorbidity
Metastatic cancer	8.4	2.2-32.7	0.002
Liver disease	8.2	3.6-18.9	<0.001
Paralysis	6.9	1.8-27.1	0.006
Fluid and electrolyte disorders	4.1	2.6-6.4	<0.001
Pulmonary circulation disorders	3.6	1.8-7.2	<0.001
Weight loss	2.9	1.6-5.2	<0.001
Neurological disorders	2.4	1.4-4.1	0.001
Coagulopathy	2.2	1.2-4.1	0.015
Diabetes, complicated	2.1	1.2-3.8	0.014
Cardiac arrhythmias	1.8	1.1-2.9	0.011

Data derived from the NIS (National Inpatient Sample).

All adult patients diagnosed with a hip periprosthetic fracture between 2016 and 2019 were included.

Logistic regression was used with in-hospital mortality as the outcome variable. Only statistically significant variables (P<0.05) are shown in the table. In addition to significant predictors, the final model included hospital bed size, day of admission, and comorbidities, including congestive heart failure, chronic pulmonary disease, depression, drug abuse, lymphoma, obesity, peripheral vascular disease, renal failure, solid tumor, and valvular disease.

Reference variables were age <70 years, male sex, absence of comorbidity, and Medicare as a primary payment source.

OR: odds ratio, CI: confidence interval.

3. Predictors of Hospital LOS following PPHFx

Significant predictors of LOS were race, comorbidities, primary payment source, median income quartile, and hospital type (Table 3). Of note, black race showed an association with a substantial increase in LOS by 9.1% (95% CI 4.3%-13.9%, P<0.001). Several comorbidities, including paralysis (35.4% increase, 95% CI 24.1%-46.6%, P<0.001), fluid and electrolyte disorders (30.6% increase, 95% CI 28.4%-32.7%, P<0.001), and weight loss (25.9% increase, 95% CI 22.1%-29.7%, P<0.001), were linked to notably longer LOS. Socioeconomic factors also played a crucial role; patient encounters covered by ‘Other’ payment sources showed a 12.9% increase in LOS (95% CI 5.7%-20.0%, P<0.001) and patient encounters covered by Medicaid showed a 9.3% increase (95% CI 3.8%-14.8%, P=0.001). In addition, patient encounters listed with lower median income quartiles showed higher percentage increases in LOS. Hospital characteristics, including medium (3.6% increase, 95% CI 0.8%-6.5%, P=0.013) and large bed sizes (12.8% increase, 95% CI 10.1%-15.4%, P<0.001), and urban teaching hospital types (6.3% increase, 95% CI 2.3%-10.3%, P=0.002), showed an association with higher percentage increases in LOS.

Table 3.

Predictors of Hospital LOS following Periprosthetic Hip Fracture

Predictor	Change in LOS (%)	95% CI	P-value
Race
Black	9.1	4.3-13.9	<0.001
Comorbidity
Paralysis	35.4	24.1-46.6	<0.001
Fluid and electrolyte disorders	30.6	28.4-32.7	<0.001
Weight loss	25.9	22.1-29.7	<0.001
Blood loss anemia	21.5	14.2-28.7	<0.001
Psychoses	21.2	11.7-30.7	<0.001
Metastatic cancer	20.9	11.0-30.7	<0.001
Neurological disorders	18.1	15.1-21.0	<0.001
Obesity	13.8	11.0-16.7	<0.001
Pulmonary circulation disorders	13.6	8.8-18.5	<0.001
Coagulopathy	13.5	9.7-17.2	<0.001
Lymphoma	13.2	1.9-24.5	0.027
Liver disease	12.8	7.0-18.6	<0.001
Congestive heart failure	12.2	8.7-15.6	<0.001
Cardiac arrhythmias	10.3	8.1-12.5	<0.001
Deficiency anemia	9.8	5.6-14.0	<0.001
Drug abuse	9.5	3.4-15.6	0.003
Renal failure	5.5	1.8-9.2	0.005
Primary payment source
Other	12.9	5.7-20.0	<0.001
Medicaid	9.3	3.8-14.8	0.001
Median income quartile
1	4.4	1.5-7.3	0.003
2	3.6	0.9-6.3	0.010
Hospital bed size
Medium	3.6	0.8-6.5	0.013
Large	12.8	10.1-15.4	<0.001
Hospital type
Urban teaching	6.3	2.3-10.3	0.002

Data derived from the NIS (National Inpatient Sample).

All adult patients diagnosed with a hip periprosthetic fracture between 2016 and 2019 were included.

Linear regression was used with hospital length of stay as the outcome variable. The length of stay was log-transformed prior to analysis and reversed for interpretation. Only statistically significant variables (P<0.05) are shown in the table. In addition to significant predictors, the final model included admission day, sex, and comorbidities, including AIDS, alcohol abuse, chronic pulmonary disease, depression, diabetes, hypertension, peptic ulcer disease, peripheral vascular disease, solid tumor, and valvular disease.

Reference variables were white race, absence of comorbidity, Medicare as a primary payment source, fourth median income quartile, small hospital bed size, and rural hospital type.

LOS: length of stay, CI: confidence interval.

4. Predictors of Cost following PPHFx

Significant predictors of hospital cost included age, race, primary payment source, comorbidities, income quartile, and hospital bed size and type (Table 4). Race is a significant factor, with a change in cost of 7.9%, 9.9%, and 23.3% for Black, Hispanic, and Asian or Pacific Islanders, respectively, compared to their white counterparts. Comorbidities such as metastatic cancer (24.6%), paralysis (24.5% increase), fluid and electrolyte disorders (20.0% increase), and other medical conditions showed an association with substantial cost escalations. In addition, urban teaching hospitals showed an association with a decrease in costs of 12.7% when compared to rural non-teaching hospitals. The treatment procedure showed the most substantial change in cost; nonoperative treatment showed an association with a decrease in cost (–74.6%). In comparison, ORIF and combined ORIF plus revision treatment pathways showed an association with an increase in cost of 35.0% and 40.6%, respectively.

Table 4.

Predictors of Cost following Periprosthetic Hip Fracture

Predictor	Change in cost (%)	95% CI	P-value
Race
Asian or Pacific Islander	23.3	12.6-34.1	<0.001
Hispanic	9.9	4.8-15.0	<0.001
Black	7.9	2.8-13.0	<0.001
Comorbidities
Metastatic cancer	24.6	15.4-33.9	<0.001
Paralysis	24.5	11.1-37.8	<0.001
Fluid and electrolyte disorders	20.0	17.7-22.2	<0.001
Coagulopathy	17.0	13.0-20.9	<0.001
Weight loss	14.2	10.2-18.2	<0.001
Liver disease	13.8	7.9-19.7	<0.001
Deficiency anemia	11.4	7.2-15.7	<0.001
Other neurological disorders	10.9	7.9-13.9	<0.001
Pulmonary circulation disorders	10.4	5.2-15.6	<0.001
Cardiac arrhythmias	9.8	7.6-12.0	<0.001
Drug abuse	6.7	0.4-13.0	0.004
Obesity	6.5	3.5-9.4	<0.001
Congestive heart failure	6.2	2.5-9.8	<0.001
Peripheral vascular disorders	5.7	1.9-9.5	<0.001
Diabetes, complicated	5.5	2.1-8.9	<0.001
Diabetes, uncomplicated	5.4	2.1-8.7	<0.001
Valvular disease	5.0	1.4-8.6	0.001
Chronic pulmonary disease	5.0	2.6-7.3	<0.001
Renal failure	4.2	0.3-8.1	0.004
Hospital type
Urban teaching	–12.7	–16.8 to –8.5	<0.001
Procedure
ORIF and revision	40.6	37.8-43.4	<0.001
Revision	35.0	32.2-37.7	<0.001
Nonoperative	–74.6	–77.2 to –72.0	<0.001

Data derived from the NIS (National Inpatient Sample).

Change in cost is represented in dollars.

All adult patients diagnosed with a hip periprosthetic fracture between 2016 and 2019 were included.

Linear regression was used with hospital cost as the outcome variable. The cost was log-transformed prior to analysis and reversed for interpretation. Only statistically significant variables (P<0.05) are shown in the table. In addition to significant predictors, the final model included admission day, hospital size and comorbidities, including AIDS, alcohol use disorder, blood loss anemia, depression, hypertension, lymphoma, peptic ulcer disease, and psychoses.

Reference variables were White race, Medicare as a primary payment source, absence of comorbidity, ORIF and rural hospital.

ORIF: open reduction and internal fixation, CI: confidence interval.

DISCUSSION

1. Predictors of In-Hospital Mortality following PPHFx

Our findings showed an in-hospital mortality rate of 1.7%, which aligns with some previous research. For example, Lamb et al.¹³⁾ estimated a pooled 30-day in-hospital mortality of 2.4% (95% CI 1.6%-3.4%). The literature shows a wide range of mortality rates, with some studies reporting as high as 10%¹²⁾. This variation may be attributed to differences in patient populations³⁴⁾, fracture types and management protocols¹⁶⁾, and study design, including sample size¹³⁾. We found 14 significant predictors of in-hospital mortality; among these age and sex were unmodifiable. As in our study, previous literature has found an association of older age with 30-day mortality after primary hip fracture³⁵⁾. Of particular interest, our analysis found that female sex was a protective factor. While this is not a unique finding³⁶⁾, no mechanism has been determined.

Ten of the significant predictors of in-hospital mortality were comorbidities (i.e., metastatic cancer, liver disease, paralysis fluid and electrolyte disorders, pulmonary circulation disorders, weight loss, neurological disorders, coagulopathy, complicated diabetes, and cardiac arrhythmias). While previous studies have used established systems such as the American Society of Anaesthesiologists score³⁶⁾ and the Nottingham Hip Fracture score¹⁸⁾, few studies have examined specific comorbidities. Finlayson et al.¹⁶⁾ found significance of neoplasia in agreement with our finding of metastatic cancer. However, there was some discordance as no significant association for respiratory disorders or diabetes was found¹⁶⁾. The differences may be due to the time period used (12 months vs. one month for the current study) and differences in study design as we segmented diabetes according to uncomplicated and complicated (end organ damage)¹⁶⁾. Similar to our study, Nasser et al.³⁶⁾ reported that metastatic malignancy and dementia (a subset of our finding of neurological disorders) were significant predictors of mortality after PPHFx. Varady et al.³⁷⁾ found correlation between cardiac arrhythmia, pulmonary circulation disorders, and diabetes and mortality in primary hip fractures. Likewise, our study found these variables, which may suggest some overlap between predictors of primary hip fracture and PPHFx. Little research evaluating liver disease, paralysis, rapid weight loss, and coagulopathy as predictors of mortality following PPHFx is available, indicating a need for further conduct of additional research³⁸⁾.

2. Predictors of Hospital LOS following PPHFx

Association of longer LOS with lower quality of life and delayed recovery has been reported in other patient populations³⁹⁾. Our findings showed a median LOS of 5.0 days for PPHFx patient encounters. This is comparable to Haughom et al.⁴⁰⁾, who reported a mean LOS of 5.2 days, while Ragland et al.⁴¹⁾ reported a longer mean LOS of 7.74 days. Direct comparison is challenging due to the difference in data presentation (median vs. mean) and the skewness of our LOS data. Among 17 different comorbidities associated with increases in LOS, paralysis, fluid and electrolyte disorders, weight loss, blood loss anemia, psychoses, and metastatic cancer showed association with >20% increases in LOS.

Our analysis found an association of black race with a 9.1% increase in the LOS compared to white race. Although specific research on this topic is scarce, underlying causes may be multifactorial. Nayar et al.⁴²⁾ observed that black patients faced significant delays in both the radiographic assessment and surgical intervention for hip fractures compared to white patients, leading to longer hospital stays. Likewise, Johnson et al.⁴³⁾ suggested that health disparities in receiving surgical treatment for acute orthopaedic trauma could play a role in prolonged LOS. In addition, other underlying causes, such as access to care preoperative optimization, may affect LOS in this population⁴⁴⁾.

Our findings also showed a correlation between lower median income quartile and extended LOS. Previous research suggests that patients of lower socioeconomic status backgrounds often lack adequate social support, which may increase the complexity of discharge from the hospital as compared to patients of higher means⁴⁵⁾. In addition, we noted that certain hospital characteristics, including medium and large bed size and urban teaching hospital, showed an association with longer LOS. This observation aligns with findings from Doctoroff et al.⁴⁶⁾, which noted that academic medical centers, which frequently manage complex medical cases, show a trend toward higher LOS compared to non-academic or rural hospitals.

3. Predictors of Cost following PPHFx

In our study the median cost for treating PPHFx was US$19,936. This contrasts with Ragland et al.⁴¹⁾, who reported an average cost of US$30,979±843.28. The discrepancy can be attributed to Ragland et al.⁴¹⁾ focusing on patients who underwent specific treatments like ORIF or revision surgeries, whereas our study included a broader patient cohort based on diagnosis codes, which included patients who received nonoperative management. Our findings also indicated that costs increased by >20% with metastatic cancer, paralysis, and fluid and electrolyte disorders. These conditions likely contribute to higher costs due to increased risk of complications, the need for more complex surgical treatments, and extended LOS. Previous research has established an association between a higher comorbidity burden as these patients may have a higher risk of complications necessitating more interventions⁴⁷⁾, more intricate surgical treatment⁴⁸⁾, and longer LOS⁴¹⁾. In keeping with previous literature, our study found an association between revision surgery and combined revision with ORIF and significant cost increases⁴⁹⁾. The higher cost of revision and revision with ORIF surgery compared to ORIF in hip periprosthetic fractures has been previously demonstrated, primarily due to its higher complexity, increased operating time, greater reliance on intraoperative imaging, higher implant costs, and increased risk of complications⁴¹⁾ .

Despite no concordant increase in LOS, Hispanic and Asian Pacific Islanders, races showed association with higher costs, underscoring factors beyond LOS that affect cost. Our results align with previous literature, which reported disparities in the type and cost of care for non-White racial minorities⁵⁰⁾. In adjacent medical fields, these racial differences have been attributed to differences in healthcare access and utilization⁵¹⁾, and other systemic barriers⁵²⁾. Non-white racial groups may have less access to high-quality surgeries or follow-up care after THA, which may result in greater acuity, necessitating more resources^53,54). In addition, structural and systemic variables, such as limited insurance coverage, transportation, and prior experiences of discrimination, may discourage timely care seeking, leading to more severe presentations and delays in care⁵⁵⁾. Contrary to expectations, urban teaching hospitals, although LOS was longer, showed association with lower costs. This may be due to proficiency bias in managing complex cases, as suggested by Lindahl et al.⁵⁶⁾, indicating that higher volume and specialized expertise can reduce costs. Of note, our analysis is limited to costs associated with inpatient stays. Additional research examining post-discharge care, including transfers to other healthcare facilities, could provide a more comprehensive understanding of overall cost dynamics.

4. Limitations

This study has several limitations. First, ICD-10 coding is subject to misclassification due to coding errors or omissions, which can impact the accuracy of ICD-10 diagnosis codes for periprosthetic fractures and comorbidities. Likewise, ICD-10 procedure codes may be subject to misclassification of the procedure performed. The introduction of ICD-10 procedure codes increased the number of codes to improve specificity; however, the change inadvertently added complexity compared to its predecessor or other coding systems, which may lead to misclassification²⁶⁾. This is particularly relevant for complex procedures such as revision THA, where Lygrisse et al.²⁶⁾ reported significant inaccuracies in coding. Second, the set of predictors used in the study was limited by what is available in NIS. Factors affecting outcomes, such as Vancouver fracture classification, time to surgery, and implant type, are not included in the NIS and, thus, were not part of the analysis. Third, the NIS captures only inpatient hospital stays; thus, we are unable to report on outcomes occurring outside of the inpatient hospital stay. Fourth, in this study, cases were identified by the diagnosis of periprosthetic fracture around the hip. By design the study did not differentiate whether the diagnosis was related to a total hip or hemi arthroplasty. Fifth, while the Elixhauser comorbidity index was used to group ICD-10 codes into 30 comorbidity categories, this approach may reduce the specificity of individual conditions. However, given the large number of ICD-10 codes, further segmentation into specific conditions would diminish statistical power. Despite these limitations, the current study is a large database study that includes a comprehensive and unique set of factors: comorbidities, patient demographics, hospital characteristics, and outcomes. All NIS guidelines were adhered to in order to reduce additional bias³²⁾.

CONCLUSION

This study identified major drivers of mortality, cost, and LOS in PPHFx patients. In general, higher comorbidity burden showed association with worse outcomes, with fluid and electrolyte disorders, coagulopathy, paralysis, pulmonary circulation disorders, liver disease, metastatic cancer, weight loss, and neurological disorders showing association with all three outcome variables. In addition to these factors, female sex showed association with lower mortality. Race, primary payment source, income quartile, hospital size and type, and additional comorbidities showed association with LOS and cost. Considering the high morbidity and mortality of PPHFx, our findings can be used to identify and target predictors in preoperative management. Conduct of additional prospective studies can build on our findings.

Supplementary Materials

Supplementary data is available at https://hipandpelvis.or.kr/.