Preoperative Mobile Health Data Improve Predictions of Recovery From Lumbar Spine Surgery.

BACKGROUND AND OBJECTIVES: Neurosurgeons and hospitals devote tremendous resources to improving recovery from lumbar spine surgery. Current efforts to predict surgical recovery rely on one-time patient report and health record information. However, longitudinal mobile health (mHealth) assessments integrating symptom dynamics from ecological momentary assessment (EMA) and wearable biometric data may capture important influences on recovery. Our objective was to evaluate whether a preoperative mHealth assessment integrating EMA with Fitbit monitoring improved predictions of spine surgery recovery. METHODS: Patients age 21-85 years undergoing lumbar surgery for degenerative disease between 2021 and 2023 were recruited. For up to 3 weeks preoperatively, participants completed EMAs up to 5 times daily asking about momentary pain, disability, depression, and catastrophizing. At the same time, they were passively monitored using Fitbit trackers. Study outcomes were good/excellent recovery on the Quality of Recovery-15 (QOR-15) and a clinically important change in Patient-Reported Outcomes Measurement Information System Pain Interference 1 month postoperatively. After feature engineering, several machine learning prediction models were tested. Prediction performance was measured using the c-statistic. RESULTS: A total of 133 participants were included, with a median (IQR) age of 62 (53, 68) years, and 56% were female. The median (IQR) number of preoperative EMAs completed was 78 (61, 95), and the median (IQR) number of days with usable Fitbit data was 17 (12, 21). 63 patients (48%) achieved a clinically meaningful improvement in Patient-Reported Outcomes Measurement Information System pain interference. Compared with traditional evaluations alone, mHealth evaluations led to a 34% improvement in predictions for pain interference (c = 0.82 vs c = 0.61). 49 patients (40%) had a good or excellent recovery based on the QOR-15. Including preoperative mHealth data led to a 30% improvement in predictions of QOR-15 (c = 0.70 vs c = 0.54). CONCLUSION: Multimodal mHealth evaluations improve predictions of lumbar surgery outcomes. These methods may be useful for informing patient selection and perioperative recovery strategies.

Predictors of Postoperative Segmental and Overall Lumbar Lordosis in Minimally Invasive Transforaminal Lumbar Interbody Fusion: A Consecutive Case Series.

STUDY DESIGN: Retrospective Case-Series. OBJECTIVES: Due to heterogeneity in previous studies, the effect of MI-TLIF on postoperative segmental lordosis (SL) and lumbar lordosis (LL) remains unclear. Therefore, we aim to identify radiographic factors associated with lordosis after surgery in a homogenous series of MI-TLIF patients. METHODS: A single-center retrospective review identified consecutive patients who underwent single-level MI-TLIF for grade 1 degenerative spondylolisthesis from 2015-2020. All surgeries underwent unilateral facetectomies and a contralateral facet release with expandable interbody cages. PROs included the ODI and NRS-BP for low-back pain. Radiographic measures included SL, disc height, percent spondylolisthesis, cage positioning, LL, PI-LL mismatch, sacral-slope, and pelvic-tilt. Surgeries were considered "lordosing" if the change in postoperative SL was ≥ +4° and "kyphosing" if ≤ -4°. Predictors of change in SL/LL were evaluated using Pearson's correlation and multivariable regression. RESULTS: A total of 73 patients with an average follow-up of 22.5 (range 12-61) months were included. Patients experienced significant improvements in ODI (29% ± 22% improvement, P < .001) and NRS-BP (3.3 ± 3 point improvement, P < .001). There was a significant increase in mean SL (Δ3.43° ± 4.37°, P < .001) while LL (Δ0.17° ± 6.98°, P > .05) remained stable. Thirty-eight (52%) patients experienced lordosing MI-TLIFs, compared to 4 (5%) kyphosing and 31 (43%) neutral MI-TLIFs. A lower preoperative SL and more anterior cage placement were associated with the greatest improvement in SL (ß = -.45° P = .001, ß = 15.06° P < .001, respectively). CONCLUSIONS: In our series, the majority of patients experienced lordosing or neutral MI-TLIFs (n = 69, 95%). Preoperative radiographic alignment and anterior cage placement were significantly associated with target SL following MI-TLIF.

Mapping the geographic migration of United States neurosurgeons across training and current practice regions: associations with academic productivity.

OBJECTIVE: Characterizing changes in the geographic distribution of neurosurgeons in the United States (US) may inform efforts to provide a more equitable distribution of neurosurgical care. Herein, the authors performed a comprehensive analysis of the geographic movement and distribution of the neurosurgical workforce. METHODS: A list containing all board-certified neurosurgeons practicing in the US in 2019 was obtained from the American Association of Neurological Surgeons membership database. Chi-square analysis and a post hoc comparison with Bonferroni correction were performed to assess differences in demographics and geographic movement throughout neurosurgeon careers. Three multinomial logistic regression models were performed to further evaluate relationships among training location, current practice location, neurosurgeon characteristics, and academic productivity. RESULTS: The study cohort included 4075 (3830 male, 245 female) neurosurgeons practicing in the US. Seven hundred eighty-one neurosurgeons practice in the Northeast, 810 in the Midwest, 1562 in the South, 906 in the West, and 16 in a US territory. States with the lowest density of neurosurgeons included Vermont and Rhode Island in the Northeast; Arkansas, Hawaii, and Wyoming in the West; North Dakota in the Midwest; and Delaware in the South. Overall, the effect size, as measured by Cramér's V statistic, between training stage and training region is relatively modest at 0.27 (1.0 is complete dependence); this finding was reflected in the similarly modest pseudo R2 values of the multinomial logit models, which ranged from 0.197 to 0.246. Multinomial logistic regression with L1 regularization revealed significant associations between current practice region and residency region, medical school region, age, academic status, sex, or race (p < 0.05). On subanalysis of the academic neurosurgeons, the region of residency training correlated with an advanced degree type in the overall neurosurgeon cohort, with more neurosurgeons than expected holding Doctor of Medicine and Doctor of Philosophy degrees in the West (p = 0.021). CONCLUSIONS: Female neurosurgeons were less likely to practice in the South, and neurosurgeons in the South and West had reduced odds of holding academic rather than private positions. The Northeast was the most likely region to contain neurosurgeons who had completed their training in the same locality, particularly among academic neurosurgeons who did their residency in the Northeast.

Feasibility and Acceptability of a Preoperative Multimodal Mobile Health Assessment in Spine Surgery Candidates.

BACKGROUND: Rapid growth in smartphone use has expanded opportunities to use mobile health (mHealth) technology to collect real-time patient-reported and objective biometric data. These data may have important implication for personalized treatments of degenerative spine disease. However, no large-scale study has examined the feasibility and acceptability of these methods in spine surgery patients. OBJECTIVE: To evaluate the feasibility and acceptability of a multimodal preoperative mHealth assessment in patients with degenerative spine disease. METHODS: Adults undergoing elective spine surgery were provided with Fitbit trackers and sent preoperative ecological momentary assessments (EMAs) assessing pain, disability, mood, and catastrophizing 5 times daily for 3 weeks. Objective adherence rates and a subjective acceptability survey were used to evaluate feasibility of these methods. RESULTS: The 77 included participants completed an average of 82 EMAs each, with an average completion rate of 86%. Younger age and chronic pulmonary disease were significantly associated with lower EMA adherence. Seventy-two (93%) participants completed Fitbit monitoring and wore the Fitbits for an average of 247 hours each. On average, participants wore the Fitbits for at least 12 hours per day for 15 days. Only worse mood scores were independently associated with lower Fitbit adherence. Most participants endorsed positive experiences with the study protocol, including 91% who said they would be willing to complete EMAs to improve their preoperative surgical guidance. CONCLUSION: Spine fusion candidates successfully completed a preoperative multimodal mHealth assessment with high acceptability. The intensive longitudinal data collected may provide new insights that improve patient selection and treatment guidance.

Clinically meaningful improvement in disabilities of arm, shoulder, and hand (DASH) following cervical spine surgery.

BACKGROUND CONTEXT: Patients with cervical spine disease suffer from upper limb disability. At present, no clinical benchmarks exist for clinically meaningful change in the upper limb function following cervical spine surgery. PURPOSE: Primary: to establish clinically meaningful metrics; the minimal clinically important difference (MCID) and substantial clinical benefit (SCB) of upper limb functional improvement in patients following cervical spine surgery. Secondary: to identify the prognostic factors of MCID and SCB of upper limb function following cervical spine surgery. STUDY DESIGN: Retrospective cohort study. PATIENT SAMPLE: Adult patients ≥18 years of age who underwent cervical spine surgery from 2012 to 2016. OUTCOME MEASURES: Patient-reported outcomes: Neck disability index (NDI) and Disabilities of Arm, Shoulder, and Hand (DASH). METHODS: MCID was defined as minimal improvement and SCB as substantial improvement in the DASH score at last follow-up. The anchor-based methods (ROC analyses) defined optimal MCID and SCB thresholds with area under curve (AUC) in discriminating improved vs. non-improved patients. The MCID was also calculated by distribution-based methods: half standard-deviation (0.5-SD) and standard error of the mean (SEM) method. A multivariable logistic regression evaluated the impact of baseline factors in achieving the MCID and SCB in DASH following cervical spine surgery. RESULTS: Between 2012 and 2016, 1,046 patients with average age of 57±11.3 years, 53% males, underwent cervical spine surgery. Using the ROC analysis, the threshold for MCID was -8 points with AUC of 0.73 (95% CI: 0.67-0.79) and the SCB was -18 points with AUC of 0.88 (95% confidence interval [CI]: 0.85-0.91). The MCID was -11 points by 0.5-SD and -12 points by SEM-method. On multivariable analysis, patients with myelopathy had lower odds of achieving MCID and SCB, whereas older patients and those with ≥6 months duration of symptoms had lower odds of achieving DASH MCID and SCB respectively. CONCLUSIONS: In patients undergoing cervical spine surgery, MCID of -8 points and SCB of -18 points in DASH improvement may be considered clinically significant. These metrics may enable evaluation of minimal and substantial improvement in the upper extremity function following cervical spine surgery.

Leveraging Artificial Intelligence and Synthetic Data Derivatives for Spine Surgery Research.

STUDY DESIGN: Retrospective cohort study. OBJECTIVES: Leveraging electronic health records (EHRs) for spine surgery research is impeded by concerns regarding patient privacy and data ownership. Synthetic data derivatives may help overcome these limitations. This study's objective was to validate the use of synthetic data for spine surgery research. METHODS: Data came from the EHR from 15 hospitals. Patients that underwent anterior cervical or posterior lumbar fusion (2010-2020) were included. Real data were obtained from the EHR. Synthetic data was generated to simulate the properties of the real data, without maintaining a one-to-one correspondence with real patients. Within each cohort, ability to predict 30-day readmissions and 30-day complications was evaluated using logistic regression and extreme gradient boosting machines (XGBoost). RESULTS: We identified 9,072 real and 9,088 synthetic cervical fusion patients. Descriptive characteristics were nearly identical between the 2 datasets. When predicting readmission, models built using real and synthetic data both had c-statistics of .69-.71 using logistic regression and XGBoost. Among 12,111 real and 12,126 synthetic lumbar fusion patients, descriptive characteristics were nearly the same for most variables. Using logistic regression and XGBoost to predict readmission, discrimination was similar with models built using real and synthetic data (c-statistics .66-.69). When predicting complications, models derived using real and synthetic data showed similar discrimination in both cohorts. Despite some differences, the most influential predictors were similar in the real and synthetic datasets. CONCLUSION: Synthetic data replicate most descriptive and predictive properties of real data, and therefore may expand EHR research in spine surgery.

Association Between Neighborhood-Level Socioeconomic Disadvantage and Patient-Reported Outcomes in Lumbar Spine Surgery.

BACKGROUND: Despite an increased understanding of the impact of socioeconomic status on neurosurgical outcomes, the impact of neighborhood-level social determinants on lumbar spine surgery patient-reported outcomes remains unknown. OBJECTIVE: To evaluate the impact of geographic social deprivation on physical and mental health of lumbar surgery patients. METHODS: A single-center retrospective cohort study analyzing patients undergoing lumbar surgery for degenerative disease from 2015 to 2018 was performed. Surgeries were categorized as decompression only or decompression with fusion. The area deprivation index was used to define social deprivation. Study outcomes included preoperative and change in Patient-Reported Outcomes Measurement (PROMIS) physical function (PF), pain interference (PI), depression, and anxiety (mean follow-up: 43.3 weeks). Multivariable imputation was performed for missing data. One-way analysis of variance and multivariable linear regression were used to evaluate the association between area deprivation index and PROMIS scores. RESULTS: In our cohort of 2010 patients, those with the greatest social deprivation had significantly worse mean preoperative PROMIS scores compared with the least-deprived cohort (mean difference [95% CI]-PF: -2.5 [-3.7 to -1.4]; PI: 3.0 [2.0-4.1]; depression: 5.5 [3.4-7.5]; anxiety: 6.0 [3.8-8.2], all P < .001), without significant differences in change in these domains at latest follow-up (PF: +0.5 [-1.2 to 2.2]; PI: -0.2 [-1.7 to 2.1]; depression: -2 [-4.0 to 0.1]; anxiety: -2.6 [-4.9 to 0.4], all P > .05). CONCLUSION: Lumbar spine surgery patients with greater social deprivation present with worse preoperative physical and mental health but experience comparable benefit from surgery than patients with less deprivation, emphasizing the need to further understand social and health factors that may affect both disease severity and access to care.

Resection of an Intradural Extramedullary Capillary Hemangioma in the Lumbar Spine.

BACKGROUND AND IMPORTANCE: Capillary hemangiomas are space-occupying lesions that rarely affect the central nervous system. When they present within the spinal canal, they can cause insidious symptoms and threaten neurological function. In this study, we present a case of an intradural extramedullary capillary hemangioma of the lumbar spine, discuss our management strategy, and review the current literature. For the first time for this diagnosis, we also provide an operative video. CLINICAL PRESENTATION: The patient is a previously healthy 40-year-old man who presented with complaints of progressive low back and leg pain, numbness, and intermittent subjective urinary incontinence. MRI revealed a discrete, homogenously enhancing intradural extramedullary lesion at L4. This lesion was resected by performing an L4 laminoplasty, which entails en bloc removal of the L4 lamina and then securing it back into place once the intradural resection and dural closure are completed. Histological analysis revealed a diagnosis of capillary hemangioma. The patient had full resolution of his symptoms postoperatively. DISCUSSION: Definitive management of spinal capillary hemangiomas involves gross total resection and can be accomplished with laminoplasty. Because these benign tumors can be adherent to adjacent structures, intraoperative neuromonitoring is helpful adjunct to preserve neurological function for a good outcome. CONCLUSION: Capillary hemangiomas rarely affect the spine but should be considered on the list of differential diagnoses of intradural lesions.

Outpatient versus inpatient lumbar decompression surgery: a matched noninferiority study investigating clinical and patient-reported outcomes.

OBJECTIVE: Spine surgery represents an ideal target for healthcare cost reduction efforts, with outpatient surgery resulting in significant cost savings. With an increased focus on value-based healthcare delivery, lumbar decompression surgery has been increasingly performed in the outpatient setting when appropriate. The aim of this study was to compare clinical and patient-reported outcomes following outpatient and inpatient lumbar decompression surgery. METHODS: The Quality Outcomes Database (QOD) was queried for patients undergoing elective one- or two-level lumbar decompression (laminectomy or laminotomy with or without discectomy) for degenerative spine disease. Patients were grouped as outpatient if they had a length of stay (LOS) < 24 hours and as inpatient if they stayed in the hospital ≥ 24 hours. Patients with ≥ 72-hour stay were excluded from the comparative analysis to increase baseline comparability between the two groups. To create two highly homogeneous groups, optimal matching was performed at a 1:1 ratio between the two groups on 38 baseline variables, including demographics, comorbidities, symptoms, patient-reported scores, indications, and operative details. Outcomes of interest were readmissions and reoperations at 30 days and 3 months after surgery, overall satisfaction, and decrease in Oswestry Disability Index (ODI), back pain, and leg pain at 3 months after surgery. Satisfaction was defined as a score of 1 or 2 in the North American Spine Society patient satisfaction index. Noninferiority of outpatient compared with inpatient surgery was defined as risk difference of < 1.5% at a one-sided 97.5% confidence interval. RESULTS: A total of 18,689 eligible one- and two-level decompression surgeries were identified. The matched study cohorts consisted of 5016 patients in each group. Nonroutine discharge was slightly less common in the outpatient group (0.6% vs 0.3%, p = 0.01). The 30-day readmission rates were 4.4% and 4.3% for the outpatient and inpatient groups, respectively, while the 30-day reoperation rates were 1.4% and 1.5%. The 3-month readmission rates were 6.3% for both groups, and the 3-month reoperation rates were 3.1% for the outpatient cases and 2.9% for the inpatient cases. Overall satisfaction at 3 months was 88.8% for the outpatient group and 88.4% for the inpatient group. Noninferiority of outpatient surgery was documented for readmissions, reoperations, and patient-reported satisfaction from surgery. CONCLUSIONS: Outpatient lumbar decompression surgery demonstrated slightly lower nonroutine discharge rates in comparison with inpatient surgery. Noninferiority in clinical outcomes at 30 days and 3 months after surgery was documented for outpatient compared with inpatient decompression surgery. Additionally, outpatient decompression surgery performed noninferiorly to inpatient surgery in achieving patient satisfaction from surgery.

Social risk factors predicting outcomes of cervical myelopathy surgery.

OBJECTIVE: Combinations of certain social risk factors of race, sex, education, socioeconomic status (SES), insurance, education, employment, and one's housing situation have been associated with poorer pain and disability outcomes after lumbar spine surgery. To date, an exploration of such factors in patients with cervical spine surgery has not been conducted. The objective of the current work was to 1) define the social risk phenotypes of individuals who have undergone cervical spine surgery for myelopathy and 2) analyze their predictive capacity toward disability, pain, quality of life, and patient satisfaction-based outcomes. METHODS: The Cervical Myelopathy Quality Outcomes Database was queried for the period from January 2016 to December 2018. Race/ethnicity, educational attainment, SES, insurance payer, and employment status were modeled into unique social phenotypes using latent class analyses. Proportions of social groups were analyzed for demonstrating a minimal clinically important difference (MCID) of 30% from baseline for disability, neck and arm pain, quality of life, and patient satisfaction at the 3-month and 1-year follow-ups. RESULTS: A total of 730 individuals who had undergone cervical myelopathy surgery were included in the final cohort. Latent class analysis identified 2 subgroups: 1) high risk (non-White race and ethnicity, lower educational attainment, not working, poor insurance, and predominantly lower SES), n = 268, 36.7% (class 1); and 2) low risk (White, employed with good insurance, and higher education and SES), n = 462, 63.3% (class 2). For both 3-month and 1-year outcomes, the high-risk group (class 1) had decreased odds (all p < 0.05) of attaining an MCID score in disability, neck/arm pain, and health-related quality of life. Being in the low-risk group (class 2) resulted in an increased odds of attaining an MCID score in disability, neck/arm pain, and health-related quality of life. Neither group had increased or decreased odds of being satisfied with surgery. CONCLUSIONS: Although 2 groups underwent similar surgical approaches, the social phenotype involving non-White race/ethnicity, poor insurance, lower SES, and poor employment did not meet MCIDs for a variety of outcome measures. This finding should prompt surgeons to proactively incorporate socially conscience care pathways within healthcare systems, as well as to optimize community-based resources to improve outcomes and personalize care for populations at social risk.

Classifying Patients Operated for Spondylolisthesis: A K-Means Clustering Analysis of Clinical Presentation Phenotypes.

BACKGROUND: Trials of lumbar spondylolisthesis are difficult to compare because of the heterogeneity in the populations studied. OBJECTIVE: To define patterns of clinical presentation. METHODS: This is a study of the prospective Quality Outcomes Database spondylolisthesis registry, including patients who underwent single-segment surgery for grade 1 degenerative lumbar spondylolisthesis. Twenty-four-month patient-reported outcomes (PROs) were collected. A k-means clustering analysis-an unsupervised machine learning algorithm-was used to identify clinical presentation phenotypes. RESULTS: Overall, 608 patients were identified, of which 507 (83.4%) had 24-mo follow-up. Clustering revealed 2 distinct cohorts. Cluster 1 (high disease burden) was younger, had higher body mass index (BMI) and American Society of Anesthesiologist (ASA) grades, and globally worse baseline PROs. Cluster 2 (intermediate disease burden) was older and had lower BMI and ASA grades, and intermediate baseline PROs. Baseline radiographic parameters were similar (P > .05). Both clusters improved clinically (P < .001 all 24-mo PROs). In multivariable adjusted analyses, mean 24-mo Oswestry Disability Index (ODI), Numeric Rating Scale Back Pain (NRS-BP), Numeric Rating Scale Leg Pain, and EuroQol-5D (EQ-5D) were markedly worse for the high-disease-burden cluster (adjusted-P < .001). However, the high-disease-burden cluster demonstrated greater 24-mo improvements for ODI, NRS-BP, and EQ-5D (adjusted-P < .05) and a higher proportion reaching ODI minimal clinically important difference (MCID) (adjusted-P = .001). High-disease-burden cluster had lower satisfaction (adjusted-P = .02). CONCLUSION: We define 2 distinct phenotypes-those with high vs intermediate disease burden-operated for lumbar spondylolisthesis. Those with high disease burden were less satisfied, had a lower quality of life, and more disability, more back pain, and more leg pain than those with intermediate disease burden, but had greater magnitudes of improvement in disability, back pain, quality of life, and more often reached ODI MCID.

Telemedicine in Neurosurgery: Standardizing the Spinal Physical Examination Using A Modified Delphi Method.

OBJECTIVE: The use of telemedicine has dramatically increased due to the coronavirus disease 2019 pandemic. Many neurosurgeons are now using telemedicine technologies for preoperative evaluations and routine outpatient visits. Our goal was to standardize the telemedicine motor neurologic examination, summarize the evidence surrounding clinical use of telehealth technologies, and discuss financial and legal considerations. METHODS: We identified a 12-member panel composed of spine surgeons, fellows, and senior residents at a single institution. We created an initial telehealth strength examination protocol based on published data and developed 10 agree/disagree statements summarizing the protocol. A blinded Delphi method was utilized to build consensus for each statement, defined as > 80% agreement and no significant disagreement using a 2-way binomial test (significance threshold of p < 0.05). Any statement that did not meet consensus was edited and iteratively resubmitted to the panel until consensus was achieved. In the final round, the panel was unblinded and the protocol was finalized. RESULTS: After the first round, 4/10 statements failed to meet consensus ( < 80% agreement, and p = 0.031, p = 0.031, p = 0.003, and p = 0.031 statistical disagreement, respectively). The disagreement pertained to grading of strength of the upper (3/10 statements) and lower extremities (1/10 statement). The amended statements clarified strength grading, achieved consensus ( > 80% agreement, p > 0.05 disagreement), and were used to create the final telehealth strength examination protocol. CONCLUSION: The resulting protocol was used in our clinic to standardize the telehealth strength examination. This protocol, as well as our summary of telehealth clinical practice, should aid neurosurgical clinics in integrating telemedicine modalities into their practice.

Comparison of cost and complication rates for profiling hospital performance in lumbar fusion for spondylolisthesis.

BACKGROUND CONTEXT: There is growing interest among payers in profiling hospital value and quality-of-care, including both the cost and safety of common surgeries, such as lumbar fusion. Nonetheless, there is sparse evidence describing the statistical reliability of such measures when applied to lumbar fusion for spondylolisthesis. PURPOSE: To evaluate the reliability of 90-day inpatient hospital costs, overall complications, and rates of serious complications for profiling hospital performance in lumbar fusion surgery for spondylolisthesis. STUDY DESIGN/SETTING: Data for this analysis came from State Inpatient Databases from nine states made available through the Healthcare Cost and Utilization Project. PATIENT SAMPLE: Patients undergoing elective lumbar spine fusion for spondylolisthesis from 2010 to 2017 in participating states. OUTCOME MEASURES: Statistical reliability, defined as the ability to distinguish true performance differences across hospitals relative to statistical noise. Reliability was assessed separately for 90-day inpatient costs (standardized across years to 2019 dollars), overall complications, and serious complication rates. METHODS: Statistical reliability was measured as the amount of variation between hospitals relative to the total amount of variation for each measure. Total variation includes both between-hospital variation ("signal") and within-hospital variation ("statistical noise"). Thus, reliability equals signal over (signal plus noise) and ranges from 0 to 1. To adjust for differences in patient-level risk and procedural characteristics, hierarchical linear and logistic regression models were created for the cost and complication outcomes. Random hospital intercepts were used to assess between-hospital variation. We evaluated the reliability of each measure by study year and examined the number of hospitals meeting different thresholds of reliability by year. RESULTS: We included a total of 66,571 elective lumbar fusion surgeries for spondylolisthesis performed at 244 hospitals during the study period. The mean 90-day hospital cost was $30,827 (2019 dollars). 12.0% of patients experienced a complication within 90 days of surgery, including 7.8% who had a serious complication. The median reliability of 90-day cost ranged from 0.97to 0.99 across study years, and there was a narrow distribution of reliability values. By comparison, the median reliability for the overall complication metric ranged from 0.22 to 0.44, and the reliability of the serious complication measure ranged from 0.30 to 0.49 across the study years. At least 96% of hospitals had high (> 0.7) reliability for cost in any year, whereas only 0-9% and 0-11% of hospitals reached this cutoff for the overall and serious complication rate in any year, respectively. By comparison, 10%-69% of hospitals per year achieved a more moderate threshold of 0.4 reliability for overall complications, compared to 21%-80% of hospitals who achieved this lower reliability threshold for serious complications. CONCLUSIONS: 90-day inpatient costs are highly reliable for assessing variation across hospitals, whereas overall and serious complications are only moderately reliable for profiling performance. These results support the viability of emerging bundled payment programs that assume true differences in costs of care exist across hospitals.