Robotic versus conventional single-level lumbar fusion in frail older adults: analysis of the National Readmission Database, 2016–2020 Authors: Po-Wen Chen, Tsung-Hsi Yang, Tao-Chieh Yang, Se-Yi Chen Published: 2025 ABSTRACT BackgroundLumbar fusion is a common intervention for degenerative spinal conditions, with robotic-assisted techniques offering improved precision. However, evidence comparing robotic and conventional fusion in frail older adults is limited. This study evaluated short-term postoperative outcomes in this high-risk population.Materials and methodsThis retrospective study analyzed data from frail adults aged ≥ 60 years who underwent single-level lumbar fusion between 2016 and 2020 using the National Readmission Database. Frailty was assessed with the Hospital Frailty Risk Score (HFRS). Outcomes—including in-hospital mortality, complications, hospital charges, and length of stay (LOS)—were compared between robotic and conventional fusion groups using propensity score matching (PSM). Key covariates used in the matching process included age and severe liver disease. Logistic regression provided adjusted odds ratios (aOR) with 95% confidence intervals (CI).ResultsAmong 29,938 patients identified, 20,227 met inclusion criteria, and 3135 patients remained after PSM. Robotic surgery was associated with significantly higher hospital charges compared with the conventional approach (mean 203,700 USD versus 151,200 USD; β = 52.51, 95% CI 26.41, 78.60; p < 0.001). No significant differences were observed in in-hospital mortality (0.4% for both; OR 1.03, 95% CI 0.14, 7.79, p = 0.978), LOS (5.3 versus 5.6 days; β = −0.36, p = 0.263), or complication rates between groups.ConclusionsRobotic-assisted lumbar fusion in frail older adults leads to higher total hospital charges without short-term clinical benefit compared with conventional techniques. Further research is needed to assess long-term outcomes and justify the use of robotic surgery in this population.Level of evidence: 3Supplementary InformationThe online version contains supplementary material available at 10.1186/s10195-025-00890-7. Introduction Lumbar fusion surgery has become the standard treatment for spinal instability and degenerative spinal conditions, particularly in older adults [1]. The increasing prevalence of spinal disorders among the aging population has correspondingly increased the performance of lumbar fusion procedures. In the USA, the number of elective spinal fusion procedures grew by 62.3% between 2004 and 2015, particularly among older adults; however, the procedure remains controversial because 80% of patients with degenerative spinal disorders are reported to improve without surgery [2]. The upward trend parallels the growing reliance on lumbar fusion as the main intervention for degenerative spinal conditions, emphasizing the need for optimized surgical approaches, especially among high-risk patient populations. Older adults undergoing lumbar fusion often present with a complex array of medical challenges, including frailty [3], diabetes [4], chronic kidney disease [5], and multiple comorbidities [6]. Frailty has been identified as a significant predictor of adverse outcomes, including postoperative complications, prolonged hospital stays, and increased healthcare costs [7]. Given the multifaceted nature of the medical conditions of these patients, surgical decision-making becomes more critical, and the need for safer, more effective treatment options is paramount. Robotic-assisted surgical techniques have been introduced to enhance precision, reduce intraoperative variability, and improve clinical outcomes by minimizing soft tissue trauma and ensuring more accurate hardware placement [8]. Robotic-assisted lumbar fusion has been shown to reduce complications and improve recovery times [9]. However, these potential benefits come at a considerably higher cost, with hospital charges for lumbar fusion often exceeding $50,000, and robotic-assisted procedures frequently incurring even greater medical costs [10]. This raises concerns about the cost-effectiveness of robotic surgery, particularly for frail older adults with multiple comorbidities. However, there have been recent reports of improved accuracy in screw placement and enhanced surgical outcomes through greater precision of the robotic procedures, as well as greater safety by incorporating minimally invasive techniques while simultaneously minimizing radiation exposure to both patients and surgeons [11]. Another recent study described differences in robotic systems, reporting that various spine surgical robots have flourished in the last decade. On the basis of spatial location of a spine surgical robot and an operator during surgery, surgical robots are classified as on-site-controlled robots and remote-controlled robots, although the authors found no significant differences in precision between them [12]. Despite the increasing adoption of robotic spine surgery, particularly in the USA and parts of Europe, its routine use remains limited due to concerns about cost-effectiveness, training requirements, and the lack of definitive outcome data. Limited evidence is available comparing its clinical and economic outcomes directly with those of conventional surgery, particularly in high-risk populations. As such, more comprehensive analysis is needed to inform clinical decision-making. Therefore, this study aimed to compare the clinical and economic outcomes of robotic versus conventional single-level lumbar fusion surgery in frail older adult patients. Methods Study design and data source This retrospective, population-based, observational study included data of older adult inpatients extracted from the United States (US) Nationwide Readmissions Database (NRD), a publicly available, all-payer database developed by the Agency for Healthcare Research and Quality (AHRQ) as part of the Healthcare Cost and Utilization Project (HCUP) of the US Department of Health and Human Services (DHHS). The NRD is sourced from the HCUP State Inpatient Databases and offers an accurate representation of total US hospitalizations and readmissions, regardless of insurance provider. It includes verified patient linkage numbers, enabling the tracking of individuals across hospitals within a given year, while maintaining strict adherence to privacy guidelines. The NRD covers a full calendar year of data, with diagnoses and procedures recorded using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM), and procedure codes (ICD-10-PCS) starting from data year 2016. An overview of NRD structure is available at: https://hcup-us.ahrq.gov/nrdoverview.jsp. Study population The present study included the data of older adults (aged 60 years and older), diagnosed with frailty, who we admitted for single-level lumbar fusion surgery between January 1 and September 30 from 2016 to 2020. All data were extracted from the NRD and analyzed retrospectively. Frailty was assessed using the Hospital Frailty Risk Score (HFRS). Patients who underwent emergent surgery; had missing data on sex, dataset sample weight, or outcomes; or underwent multilevel fusion, were excluded. The remaining patients were then categorized into two groups by type of single-level lumbar fusion procedure received: those who underwent robotic surgery and those who received conventional surgery. Follow-up Following the index admission, patients were considered “at-risk” for hospitalization and were included in the follow-up period until 31 December of each admission year or until death. Ethics statement This study used secondary, anonymized administrative data, with no direct patient involvement. The use of deidentified data complies with ethical research standards and aligns with data protection regulations, as it was originally collected for purposes other than this specific study. Therefore, the signed informed consent of included patients was also waived. Defining frailty Frailty was assessed using the HFRS, which was developed and validated in a cohort of 1.04 million hospitalized older adults [13]. The HFRS was calculated on the basis of the ICD-10 Claims codes recorded during each patient’s index admission. Patients in the two fusion surgery groups were also categorized into three frailty risk groups: low frailty risk (HFRS score < 5, referred to as “non-frail”), medium frailty risk (HFRS score 5–15), and high frailty risk (HFRS score > 15). This low-cost assessment, based on ICD-10 codes, can be easily integrated into hospital information systems. It performs comparably to existing frailty and risk stratification tools and is especially well suited for claims-based research [14]. Although the original validation study established the above cutoffs to most effectively distinguish between individuals with different outcomes, patients in the present study with an HFRS score ≥ 5 were classified as frail on the basis of the data distribution, as previously reported [15]. While originally derived from UK acute care data, the HFRS has been successfully applied in elective surgical populations using large US administrative databases. Prior studies have demonstrated its utility in predicting short-term outcomes after nonemergent procedures in older adults [16, 17]. Primary and secondary outcomes assessed The primary postoperative outcomes measured were short term, including in-hospital mortality, total hospital charges (referring exclusively to the billed amount for the index hospitalization, rather than actual resource utilization or hospital-incurred costs), and postoperative complications, including acute myocardial infarction (AMI), cerebrovascular accidents (CVA), venous thromboembolism (VTE), sepsis, infection, respiratory failure, mechanical ventilation, acute kidney injury, shock, and bleeding. Secondary outcomes included length of stay (LOS), prolonged LOS was defined as LOS ≥ 75th percentile, as well as 30-day and 90-day readmission rates. The most common principal diagnoses of 90-day readmissions were also analyzed. Covariates Covariates included: age, sex, insurance status (primary payer), Charlson Comorbidity Index (CCI), and major comorbidities, including hypertension, diabetes, obesity, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD), severe liver disease, and rheumatic disease. Additionally, weekend admissions, hospital location/teaching status, and hospital bed size—classified as small, medium, or large on the basis of HCUP criteria considering region, urban/rural setting, and teaching status—were also included because these variables are available in the NRD. The ICD codes used to identify patients’ diagnoses, comparison groups by surgeries received, comorbidities, and complications are provided in Supplementary Table 1. Statistical analysis The NRD uses a complex survey design. In the present study, to generate national estimates using the NRD, the discharge-level weight provided by the NRD was applied to all analyses. Descriptive statistics were employed to summarize patients’ demographic and clinical characteristics, and hospital data presenting categorical variables as counts and weighted percentages, while continuous variables were presented as means with standard errors (SE). Comparisons between groups for categorical variables were performed using the Rao–Scott chi-squared test, and weighted mean differences for continuous variables were assessed using survey methods that account for stratification, clustering, and sampling weights, ensuring robust statistical inferences within the context of complex survey design. To reduce potential confounding, propensity score matching (PSM) 1:10 was used to match patients who underwent robotic surgery to those who underwent conventional single-level lumbar fusion. The propensity score was estimated using logistic regression that included key baseline variables showing significant group differences, specifically age (in years) and whether or not they had severe liver disease. The matching procedure followed a one-to-many approach [18], starting with the “best” matches and proceeding to the “next-best” matches until no further matches were possible. Logistic regression analysis was utilized to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for outcomes, adjusting for covariates with significant differences between the matched groups. A sensitivity analysis by study year (2016–2020) was performed to evaluate whether the outcomes between robotic and conventional procedures changed over time, potentially suggesting a learning curve effect. A two-sided p-value of < 0.05 was considered statistically significant. All statistical analyses were carried out using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Results Patient selection The study population selection process is shown in Fig. 1. A total of 29,938 older adults (≥ 60 years) with frailty were identified. After applying the inclusion and exclusion criteria, 20,227 patients were included in the analysis. Following PSM, 3135 patients remained for subsequent analyses. Using NRD weighting system, this weighted sample represents approximately 5813 hospitalizations across the entire USA (Fig. 1).Fig. 1Flow diagram of patient selection Flow diagram of patient selection Characteristics of the study population Table 1 and Supplementary Table 2 present patients’ baseline demographic and clinical characteristics before and after PSM. Before matching, the mean age of the study population was 71 years, and 61% of the patients were female. Significant differences were observed in various patient characteristics before matching, including age (p < 0.001) and severe liver disease (p = 0.015). (Supplementary Table 2). Table 1Demographic and clinical characteristics of frail older adult patients undergoing single-level lumbar fusion, after PSMCharacteristicsAll patients (n = 3135)Robotic surgeryp-ValueYes (n = 285)No (n = 2850)DemographyAge, years70.2 ± 0.170.0 ± 0.470.2 ± 0.10.685 60–64693 (22.3)63 (23.4)630 (22.2)0.999 65–69825 (26.4)75 (25.5)750 (26.5) 70–74825 (26.3)75 (26.8)750 (26.3) 75–79561 (18.0)51 (17.5)510 (18.0) 80–84165 (4.9)15 (5.0)150 (4.9) 85–8955 (1.6)5 (1.6)50 (1.7) 90 + 11 (0.4)1 (0.3)10 (0.4)Sex0.963 Male1231 (39.0)110 (38.8)1121 (39.0) Female1904 (61.0)175 (61.2)1729 (61.0)Insurance status/primary payer0.174 Medicare/Medicaid2514 (80.3)225 (80.1)2289 (80.3) Private including HMO507 (16.4)52 (18.2)455 (16.2) Self-pay/no-charge/other111 (3.3)6 (1.7)105 (3.5) Missing321Weekend admission0.230 No3109 (99.2)284 (99.7)2825 (99.1) Yes26 (0.8)1 (0.3)25 (0.9)Major comorbidities Hypertension2440 (78.4)227 (80.4)2213 (78.2)0.359 Diabetes980 (31.6)93 (32.4)887 (31.5)0.756 Obesity894 (28.0)80 (28.2)814 (28.0)0.943 Chronic pulmonary disease788 (25.1)65 (22.5)723 (25.3)0.261 Chronic kidney disease676 (22.1)69 (25.3)607 (21.8)0.254 Severe liver disease0 (0.0)0 (0.0)0 (0.0)– Rheumatic disease218 (7.1)17 (6.2)201 (7.2)0.507Hospital bed size0.424 Small607 (19.1)69 (22.4)538 (18.8) Medium801 (24.4)78 (26.6)723 (24.2) Large1727 (56.5)138 (51.0)1589 (57.0)Hospital location/teaching status0.088 Metropolitan non-teaching661 (19.5)57 (17.6)604 (19.6) Metropolitan teaching2370 (76.1)213 (73.3)2157 (76.4) Non-metropolitan hospital104 (4.4)15 (9.1)89 (4.0)CCI0.298 0985 (31.7)87 (29.5)898 (31.9) 1822 (25.9)66 (23.1)756 (26.1) 2532 (16.7)54 (19.3)478 (16.5) 3325 (10.2)37 (13.1)288 (9.9) 4+ 471 (15.5)41 (15.0)430 (15.6)Continuous variables are presented as mean ± SE; categorical variables are presented as unweighted counts (weighted percentage)CCI Charlson Comorbidity IndexaExcluding patients with in-hospital mortalitybThousand US dollarsp-Values < 0.05 are shown in boldTable 2Associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSMOutcomesAll patients (n = 3135)Robotic surgeryOR/betac (95% CI)p-ValueYes (n = 285)No (n = 2850)In-hospital mortality13 (0.4)1 (0.4)12 (0.4)1.03 (0.14, 7.79)0.978LOS (days)a5.6 ± 0.15.3 ± 0.35.6 ± 0.1−0.36 (−1.00, 0.27)0.263Total hospital chargesa,b155.6 ± 3.0203.7 ± 13.3151.2 ± 2.852.51 (26.41, 78.60) < 0.001Any complications820 (25.6)78 (26.4)742 (25.5)1.05 (0.79, 1.39)0.747 AMI16 (0.5)2 (0.6)14 (0.5)1.11 (0.25, 5.00)0.892 CVA131 (3.9)12 (4.3)119 (3.9)1.13 (0.53, 2.42)0.747 VTE51 (1.5)1 (0.4)50 (1.6)0.25 (0.03, 1.85)0.175 Sepsis52 (1.6)2 (0.8)50 (1.7)0.46 (0.11, 1.91)0.286 Infection273 (8.2)21 (7.3)252 (8.3)0.88 (0.56, 1.37)0.558 Respiratory failure58 (2.0)6 (2.2)52 (1.9)1.12 (0.49, 2.58)0.786 Mechanical ventilation31 (1.0)1 (0.4)30 (1.0)0.37 (0.05, 2.70)0.326 Acute kidney injury380 (12.1)45 (14.8)335 (11.9)1.29 (0.91, 1.83)0.157 Shock44 (1.4)5 (1.9)39 (1.4)1.38 (0.53, 3.61)0.506 Bleeding10 (0.4)1 (0.4)9 (0.4)1.15 (0.14, 9.38)0.89830-day readmission ratea249 (8.0)24 (8.2)225 (8.0)1.03 (0.69, 1.54)0.89290-day readmission ratea412 (13.3)39 (13.6)373 (13.3)1.03 (0.76, 1.41)0.839Continuous data are presented as mean ± SE and performed as linear regression models; categorical variables are presented as unweighted counts (weighted percentage) and logistic regression modelsAMI acute myocardial infarction, CVA cerebrovascular accident, PSM propensity score matching; LOS length of stay, VTE venous thromboembolism, OR odds ratio, CI confidence intervalaExcluding patients with in-hospital mortalitybThousand US dollarscBeta was used for LOS and total hospital costsp-Values < 0.05 are shown in bold Table 1Demographic and clinical characteristics of frail older adult patients undergoing single-level lumbar fusion, after PSMCharacteristicsAll patients (n = 3135)Robotic surgeryp-ValueYes (n = 285)No (n = 2850)DemographyAge, years70.2 ± 0.170.0 ± 0.470.2 ± 0.10.685 60–64693 (22.3)63 (23.4)630 (22.2)0.999 65–69825 (26.4)75 (25.5)750 (26.5) 70–74825 (26.3)75 (26.8)750 (26.3) 75–79561 (18.0)51 (17.5)510 (18.0) 80–84165 (4.9)15 (5.0)150 (4.9) 85–8955 (1.6)5 (1.6)50 (1.7) 90 + 11 (0.4)1 (0.3)10 (0.4)Sex0.963 Male1231 (39.0)110 (38.8)1121 (39.0) Female1904 (61.0)175 (61.2)1729 (61.0)Insurance status/primary payer0.174 Medicare/Medicaid2514 (80.3)225 (80.1)2289 (80.3) Private including HMO507 (16.4)52 (18.2)455 (16.2) Self-pay/no-charge/other111 (3.3)6 (1.7)105 (3.5) Missing321Weekend admission0.230 No3109 (99.2)284 (99.7)2825 (99.1) Yes26 (0.8)1 (0.3)25 (0.9)Major comorbidities Hypertension2440 (78.4)227 (80.4)2213 (78.2)0.359 Diabetes980 (31.6)93 (32.4)887 (31.5)0.756 Obesity894 (28.0)80 (28.2)814 (28.0)0.943 Chronic pulmonary disease788 (25.1)65 (22.5)723 (25.3)0.261 Chronic kidney disease676 (22.1)69 (25.3)607 (21.8)0.254 Severe liver disease0 (0.0)0 (0.0)0 (0.0)– Rheumatic disease218 (7.1)17 (6.2)201 (7.2)0.507Hospital bed size0.424 Small607 (19.1)69 (22.4)538 (18.8) Medium801 (24.4)78 (26.6)723 (24.2) Large1727 (56.5)138 (51.0)1589 (57.0)Hospital location/teaching status0.088 Metropolitan non-teaching661 (19.5)57 (17.6)604 (19.6) Metropolitan teaching2370 (76.1)213 (73.3)2157 (76.4) Non-metropolitan hospital104 (4.4)15 (9.1)89 (4.0)CCI0.298 0985 (31.7)87 (29.5)898 (31.9) 1822 (25.9)66 (23.1)756 (26.1) 2532 (16.7)54 (19.3)478 (16.5) 3325 (10.2)37 (13.1)288 (9.9) 4+ 471 (15.5)41 (15.0)430 (15.6)Continuous variables are presented as mean ± SE; categorical variables are presented as unweighted counts (weighted percentage)CCI Charlson Comorbidity IndexaExcluding patients with in-hospital mortalitybThousand US dollarsp-Values < 0.05 are shown in bold Demographic and clinical characteristics of frail older adult patients undergoing single-level lumbar fusion, after PSM Continuous variables are presented as mean ± SE; categorical variables are presented as unweighted counts (weighted percentage) CCI Charlson Comorbidity Index aExcluding patients with in-hospital mortality bThousand US dollars p-Values < 0.05 are shown in bold Table 2Associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSMOutcomesAll patients (n = 3135)Robotic surgeryOR/betac (95% CI)p-ValueYes (n = 285)No (n = 2850)In-hospital mortality13 (0.4)1 (0.4)12 (0.4)1.03 (0.14, 7.79)0.978LOS (days)a5.6 ± 0.15.3 ± 0.35.6 ± 0.1−0.36 (−1.00, 0.27)0.263Total hospital chargesa,b155.6 ± 3.0203.7 ± 13.3151.2 ± 2.852.51 (26.41, 78.60) < 0.001Any complications820 (25.6)78 (26.4)742 (25.5)1.05 (0.79, 1.39)0.747 AMI16 (0.5)2 (0.6)14 (0.5)1.11 (0.25, 5.00)0.892 CVA131 (3.9)12 (4.3)119 (3.9)1.13 (0.53, 2.42)0.747 VTE51 (1.5)1 (0.4)50 (1.6)0.25 (0.03, 1.85)0.175 Sepsis52 (1.6)2 (0.8)50 (1.7)0.46 (0.11, 1.91)0.286 Infection273 (8.2)21 (7.3)252 (8.3)0.88 (0.56, 1.37)0.558 Respiratory failure58 (2.0)6 (2.2)52 (1.9)1.12 (0.49, 2.58)0.786 Mechanical ventilation31 (1.0)1 (0.4)30 (1.0)0.37 (0.05, 2.70)0.326 Acute kidney injury380 (12.1)45 (14.8)335 (11.9)1.29 (0.91, 1.83)0.157 Shock44 (1.4)5 (1.9)39 (1.4)1.38 (0.53, 3.61)0.506 Bleeding10 (0.4)1 (0.4)9 (0.4)1.15 (0.14, 9.38)0.89830-day readmission ratea249 (8.0)24 (8.2)225 (8.0)1.03 (0.69, 1.54)0.89290-day readmission ratea412 (13.3)39 (13.6)373 (13.3)1.03 (0.76, 1.41)0.839Continuous data are presented as mean ± SE and performed as linear regression models; categorical variables are presented as unweighted counts (weighted percentage) and logistic regression modelsAMI acute myocardial infarction, CVA cerebrovascular accident, PSM propensity score matching; LOS length of stay, VTE venous thromboembolism, OR odds ratio, CI confidence intervalaExcluding patients with in-hospital mortalitybThousand US dollarscBeta was used for LOS and total hospital costsp-Values < 0.05 are shown in bold Associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSM Continuous data are presented as mean ± SE and performed as linear regression models; categorical variables are presented as unweighted counts (weighted percentage) and logistic regression models AMI acute myocardial infarction, CVA cerebrovascular accident, PSM propensity score matching; LOS length of stay, VTE venous thromboembolism, OR odds ratio, CI confidence interval aExcluding patients with in-hospital mortality bThousand US dollars cBeta was used for LOS and total hospital costs p-Values < 0.05 are shown in bold After PSM, the baseline characteristics were well-balanced between the two comparison groups. The mean age was 70.0 years in the robotic group and 70.2 years in the conventional group (p = 0.685), and the percentage of female patients was 61.2% in the robotic group and 61.0% in the conventional group (p = 0.963). No significant differences were noted after PSM between the two groups in comorbidities, including diabetes (32.4% versus 31.5%, p = 0.756) and CKD (25.3% versus 21.8%, p = 0.254) (Table 1). Comparison of outcomes following robotic versus conventional fusion surgery in the matched population Table 2 presents the outcomes after robotic versus conventional fusion surgery in the matched population. Patients who underwent robotic surgery had significantly higher total hospital charges (mean 203,700 USD versus 151,200 USD; β = 52.51, 95% CI 26.41, 78.60). No significant differences were observed in in-hospital mortality (0.4% for both groups; OR 1.03, 95% CI 0.14, 7.79) or LOS (5.3 ± 0.3 versus 5.6 ± 0.1 days; β = −0.36, 95% CI −1.00, 0.27). Additionally, no significant differences were observed in complication rates, including those for myocardial infarction (0.6% versus 0.5%, p = 0.892), sepsis (0.8% versus 1.7%, p = 0.286), and AKI (14.8% versus 11.9%, p = 0.157) between the two groups (Table 2). Comparison of outcomes following robotic versus conventional fusion surgery in the matched population, stratified by age group and CCI Analyses stratified by age and CCI groups are presented in Table 3. Among patients younger than age 70 years, those who underwent robotic surgery incurred significantly higher total hospital charges (β = 45.93, 95% CI 16.66, 75.20). Similarly, in patients aged 70 years or older, total hospital charges were significantly higher in the robotic surgery group (β = 56.75, 95% CI 23.02, 90.47). Table 3Stratified associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSMSubgroupRobotic surgeryOR/betac (95% CI)p-ValueAge < 70 years In-hospital mortalityYes versus no3.07 (0.42, 22.52)0.270 LOSaYes versus no−0.50 (−1.22, 0.23)0.181 Total hospital chargesa,bYes vs. No45.93 (16.66, 75.20)0.002 Any complicationsYes versus no1.05 (0.72, 1.55)0.791 30-day readmission rateaYes versus no0.87 (0.44, 1.71)0.682 90-day readmission rateaYes versus no0.97 (0.59, 1.60)0.902Age ≥ 70 years In-hospital mortalityYes versus noNA– LOSaYes versus no−0.55 (−1.48, 0.38)0.245 Total hospital chargesa,bYes versus no56.75 (23.02, 90.47)0.001 Any complicationsYes versus no0.94 (0.66, 1.33)0.719 30-day readmission rateaYes versus no1.10 (0.67, 1.81)0.696 90-day readmission rateaYes versus no1.05 (0.72, 1.52)0.803CCI 0–1 In-hospital mortalityYes versus noNA– LOSaYes versus no−0.35 (−1.07, 0.37)0.341 Total hospital chargesa,bYes versus no68.85 (36.78, 100.91) < 0.001 Any complicationsYes versus no0.84 (0.54, 1.30)0.427 30-day readmission rateaYes versus no0.86 (0.46, 1.62)0.639 90-day readmission rateaYes versus no1.22 (0.80, 1.86)0.348CCI ≥ 2 In-hospital mortalityYes versus no1.74 (0.24, 12.68)0.584 LOSaYes versus no−0.81 (−1.74, 0.12)0.090 Total hospital chargesa,bYes versus no33.85 (6.07, 61.62)0.017 Any complicationsYes versus no1.09 (0.75, 1.59)0.650 30-day readmission rateaYes versus no1.07 (0.64, 1.78)0.812 90-day readmission rateaYes versus no0.80 (0.51, 1.26)0.339PSM propensity score matching, LOS length of stay, OR odd ratio, CI confidence interval, NA no event occurred in one groupaExcluding patients with in-hospital mortalitybThousand US dollarscBeta was used for LOS and total hospital costsp-Values < 0.05 are shown in bold Table 3Stratified associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSMSubgroupRobotic surgeryOR/betac (95% CI)p-ValueAge < 70 years In-hospital mortalityYes versus no3.07 (0.42, 22.52)0.270 LOSaYes versus no−0.50 (−1.22, 0.23)0.181 Total hospital chargesa,bYes vs. No45.93 (16.66, 75.20)0.002 Any complicationsYes versus no1.05 (0.72, 1.55)0.791 30-day readmission rateaYes versus no0.87 (0.44, 1.71)0.682 90-day readmission rateaYes versus no0.97 (0.59, 1.60)0.902Age ≥ 70 years In-hospital mortalityYes versus noNA– LOSaYes versus no−0.55 (−1.48, 0.38)0.245 Total hospital chargesa,bYes versus no56.75 (23.02, 90.47)0.001 Any complicationsYes versus no0.94 (0.66, 1.33)0.719 30-day readmission rateaYes versus no1.10 (0.67, 1.81)0.696 90-day readmission rateaYes versus no1.05 (0.72, 1.52)0.803CCI 0–1 In-hospital mortalityYes versus noNA– LOSaYes versus no−0.35 (−1.07, 0.37)0.341 Total hospital chargesa,bYes versus no68.85 (36.78, 100.91) < 0.001 Any complicationsYes versus no0.84 (0.54, 1.30)0.427 30-day readmission rateaYes versus no0.86 (0.46, 1.62)0.639 90-day readmission rateaYes versus no1.22 (0.80, 1.86)0.348CCI ≥ 2 In-hospital mortalityYes versus no1.74 (0.24, 12.68)0.584 LOSaYes versus no−0.81 (−1.74, 0.12)0.090 Total hospital chargesa,bYes versus no33.85 (6.07, 61.62)0.017 Any complicationsYes versus no1.09 (0.75, 1.59)0.650 30-day readmission rateaYes versus no1.07 (0.64, 1.78)0.812 90-day readmission rateaYes versus no0.80 (0.51, 1.26)0.339PSM propensity score matching, LOS length of stay, OR odd ratio, CI confidence interval, NA no event occurred in one groupaExcluding patients with in-hospital mortalitybThousand US dollarscBeta was used for LOS and total hospital costsp-Values < 0.05 are shown in bold Stratified associations between robotic surgery and short-term outcomes in frail older adult patients undergoing single-level lumbar fusion, after PSM PSM propensity score matching, LOS length of stay, OR odd ratio, CI confidence interval, NA no event occurred in one group aExcluding patients with in-hospital mortality bThousand US dollars cBeta was used for LOS and total hospital costs p-Values < 0.05 are shown in bold In the analysis stratified by CCI, patients with a CCI of 0 or 1 undergoing robotic surgery also had higher total hospital charges (β = 68.85, 95% CI 36.78, 100.91), and those with a CCI of 2 or more showed a similar trend (β = 33.85, 95% CI 6.07, 61.62). Despite these differences in hospital charges, no significant differences were found in in-hospital mortality or LOS between the robotic and conventional surgery groups across all subgroups (Table 3). The sensitivity analysis by study year, presented in Supplementary Table 3, evaluated the potential temporal learning curve effect in outcomes between robotic and conventional procedures. In the earliest year (2016), robotic surgery was associated with a higher risk of 30-day and 90-day readmission compared with conventional surgery, but this difference was not observed in later years. In addition, we analyzed the original study sample before PSM (Supplementary Table 4), and the associations were consistent with those in the matched cohort, supporting the robustness of our findings. Discussion After comparing robotic versus conventional single-level lumbar fusion surgery in a large sample of frail US older adults, findings of the present study showed that robotic surgery was associated with significantly higher hospital charges, with mean charges for robotic procedures exceeding mean conventional surgery by as much as 50,000 USD. However, no significant differences were found between the two surgery groups in in-hospital mortality, complications, or LOS. Both groups had a mean mortality rate of 0.4%, and the incidence of complications, including myocardial infarction, sepsis, and acute kidney injury, was similar. Additionally, hospital LOS was comparable for the robotic and conventional surgery groups. Subgroup analyses by age and CCI showed that the higher charges associated with robotic surgery were consistent across different subgroups, but no statistically significant differences were noted regarding short-term clinical outcomes. Accordingly, robotic-assisted fusion does not appear to provide clear cost-effective short-term clinical advantages for frail older adult patients. Frailty remains a critical factor in surgical decision-making, particularly among frail older adults, for whom the priority is to optimize outcomes while minimizing costs in older adults who are considered physically and mentally suitable for undergoing the stress of surgery. Frailty is reported to be a strong predictor of adverse clinical outcomes after spine surgery [19]. Nevertheless, although previous studies have compared robotic and conventional fusion procedures, few to date have specifically focused on frail populations, a particularly vulnerable group, which inspired conducting the present study. Other investigators focused on the influence of frailty on postoperative complications in older adult patients receiving single-level lumbar fusion surgery [20], concluding that frail patients should undergo surgery utilizing minimally invasive techniques to minimize risk of surgical complications and that future studies should explore implementing frailty in risk stratification assessments for patients undergoing spine surgery, and still other authors found that robotic-assisted surgery did not reduce overall postoperative complications [21]. Results of that study did not pertain to frail older adults but to those who appeared to be more suitable for fusion surgeries. Overall, between those collective previous study results and those of the present study, associations between preoperative frailty and outcomes of spine surgeries are still not conclusive and more studies are needed. In the present study, although total hospital charges were significantly higher for the robotic surgeries than conventional fusion procedures, the short-term outcomes did not demonstrate significant clinical benefits of the robotic procedure. Possible explanations may include that the focus with use of robotic-assisted procedures has been on improving precision over immediate clinical impact. In other words, robotic-assisted surgery is designed to enhance precision in hardware placement and reduce intraoperative variability, which may not translate into short-term clinical outcomes such as reducing in-hospital mortality, postoperative complications, or hospital LOS. Nevertheless, these benefits may become more apparent in the long term, such as in reduced reoperation rates or better fusion integrity, which need further evaluation in longitudinal studies. A comprehensive review and meta-analysis by Wang et al. (2024) noted a significant improvement in the intraoperative course of the robotic-assisted technique but found that it has not yet demonstrated significant advantages in terms of postoperative symptom relief and functional improvement [22]. Another fairly recent (2023) systematic review and meta-analysis [23] concluded that robot-assisted technique helps improve short-term clinical outcomes, reduces intraoperative blood loss, and shortens recovery time compared with the conventional freehand technique. Another possible explanation for the present study findings relative to robotic versus conventional fusion surgeries is the comparable complexity of single-level fusion. The inherent complexity of the procedure may not differ significantly between robotic and conventional approaches and the robotic technology may not provide a substantial advantage in this context. Further studies are warranted to investigate multilevel fusion. Learning curve issues may also have contributed to and may help to explain the present study results. The adoption of robotic-assisted surgery involves a learning curve for surgeons [24], which may diminish its effectiveness during the initial implementation period, as in situations such as the present report of early experience in a high-risk patient group. Variability in surgeon experience and technique may have diluted the potential benefits in the analyzed population. Despite the comparable short-term outcomes between robotic and conventional approaches, the higher hospital charges associated with robotic-assisted surgery raise questions regarding its cost-effectiveness in frail older adults. While the present study focused on immediate in-hospital outcomes, it is possible that potential long-term benefits, such as reduced revision rates or improved functional recovery, may justify the higher initial expenditure—an area that warrants future investigation. Furthermore, although robotic platforms offer technical precision, frailty itself may override these advantages, as frail patients inherently carry elevated baseline risks. Future research should therefore aim to identify specific subgroups of frail individuals—such as those with complex spinal pathology—who might benefit the most from robotic assistance, rather than applying a one-size-fits-all approach. These insights are critical for guiding surgical decision-making and optimizing value-based care in aging populations. Lastly, we have performed an additional sensitivity analysis stratified by study year to explore the potential impact of a temporal learning curve on postoperative outcomes. Notably, in 2016, robotic surgery was associated with a higher risk of 30-day and 90-day readmissions compared with conventional procedures. However, this difference was no longer observed in subsequent years, suggesting that early outcome disparities may have been related to limited surgeon familiarity during the initial adoption phase. As experience accumulated and proficiency improved, these differences appeared to diminish, supporting the presence of a learning curve effect. Strength and limitations Results of the present study are supported by its use of a large, nationally representative sample from the National Readmission Database, which assures generalizability of the findings to the broader population of frail older adults undergoing lumbar fusion surgery in the USA. PSM was used to balance baseline characteristics between the robotic and conventional surgery groups, reducing selection bias and allowing for a more accurate comparison of outcomes. However, several limitations remain. First, this study relied on administrative claims data from the NRD, which lack clinical granularity. Important patient-level variables—such as disease severity, lesion location, duration of symptoms, prior lumbar spine surgeries, socioeconomic status, and functional status—were not available and thus could not be adjusted for, potentially leading to residual confounding. Although we performed propensity score matching, unmeasured differences in patient selection (e.g., robotic surgery reserved for more complex cases) may still exist and influence outcomes. Second, the HFRS, while validated and suitable for large datasets, is derived from ICD codes and may not fully capture clinical frailty compared with physical assessments such as the Fried phenotype or Clinical Frailty Scale. The NRD’s reliance on ICD coding, including ICD-10-PCS codes for identifying robotic procedures, also means that potential coding errors or misclassification cannot be ruled out. Third, surgeon-level variables, such as the surgeon’s experience, annual robotic case volume, and details of the robotic platform used, were unavailable, which may have influenced surgical performance and limited our ability to evaluate institutional learning curves or technical expertise. No independent validation of robotic case identification was feasible within the NRD. Fourth, this study focused on short-term outcomes (in-hospital and 30-/90-day readmissions), which may not fully reflect the clinical benefit of robotic precision. Long-term outcomes—such as reoperations, fusion integrity, and functional recovery—remain unavailable in the NRD and are important for fully evaluating the benefits of robotic surgery, particularly in frail populations. Conclusions Among older adults with frailty, robotic-assisted single-level lumbar fusion surgery is associated with significantly higher total hospital charges but does not demonstrate clear short-term clinical advantages over conventional surgery. In-hospital mortality, complication rates, and length of stay were comparable between the two groups. Further long-term, prospective research is needed to evaluate potential long-term benefits that may justify the higher costs in this vulnerable population. Supplementary Information Supplementary Material 1. Supplementary Material 1.