Predictors of increased length of stay and cost in readmissions for spinal surgical site infections Authors: Phillip Alexeev, Alex Zhang, Matthew Giammanco, Joshua Jimenez, Molly Szpakowski, Ryan Snowden Published: 2026 ABSTRACT AbstractBackgroundSurgical site infection (SSI) following spinal surgery remains a costly and morbid complication, frequently necessitating hospital readmission and operative management. While prior studies have characterized the overall economic burden of spinal SSI, the specific drivers of cost during SSI-related readmissions are poorly defined. The primary research question of this study was to identify clinical, operative, and microbiologic factors associated with increased hospital cost, length of stay, and financial margins during readmission for operative management of spinal SSI.MethodsWe performed a retrospective cohort study of adult patients readmitted for surgical management of spinal SSI at a single tertiary academic center between 2015 and 2022. Demographic, clinical, operative, microbiologic, and payer-related variables were extracted from the electronic medical record. Itemized direct hospital costs, payments, and net revenue were obtained from institutional financial databases. The primary outcome was direct hospital cost during the SSI readmission. Multivariable Gamma generalized linear models with log link were used to identify independent predictors of cost, while secondary analyses evaluated contribution margin, margin rate, and length of stay using linear and negative binomial regression models.ResultsA total of 123 patients were included, with a mean age of 63.2 years and a mean readmission length of stay of 15.9 days. Mean direct hospital cost during readmission was $25,713±$21,128. Length of stay was strongly correlated with cost (r=0.695, p<.001). In a streamlined multivariable model retaining the full cohort (n=123), each additional hospital day increased cost by 4.8% (+$1,207/d; p<.001), and staged index procedures were associated with 45.0% higher costs (p<.001). PSO/VCR was paradoxically associated with 33.6% lower cost (p=.001); a sensitivity analysis excluding LOS showed this association strengthened, not attenuated (−51.7%), indicating confounding by surgical case selection rather than LOS. Private insurance was associated with lower direct costs and higher contribution margins and margin rates compared with government payers.ConclusionsHospital length of stay is the dominant driver of cost during readmission for spinal SSI, with operative complexity and staged procedures contributing additional financial burden. Strategies focused on reducing inpatient duration and optimizing perioperative efficiency may substantially mitigate the economic impact of spinal SSI.Keywords: Cost, Infection, Readmission, Length of stay, Insurance, Payer cost Introduction Spinal fusions are one of the fastest-growing surgeries in the United States, with the number increasing year over year [1]. In the span of 9 years (2001–2010), the incidence of lumbar fusions increased by over 74% [2]. Other studies looking at similar time ranges from 1998 to 2014 and 2004 to 2015 also showed more spinal fusion procedures, with 88% and 62.3% increases, respectively [3,4]. By 2060, spinal fusions are projected to continue to increase about 83%–102% per 100,000 people, highlighting this procedure’s efficiency and demand [5]. However, despite increases in procedure volume, surgical site infections (SSIs) are still a major complication following spinal surgery, with reported incidence rates ranging from 1% to 12% of all spinal procedures [6,7]. These infections contribute significantly to patient morbidity and place a substantial financial burden on the healthcare system. With an ever-changing political and healthcare climate, it is vital to understand what the economic impact of developing and treating SSIs are on the patient. Since the early 2000s, the economic burden of lumbar fusion has risen substantially, with average total hospital charges for the procedure nearly doubling over this period [2]. Within this broader cost escalation, postoperative surgical site infection (SSI) represents a major driver of additional resource use. The direct medical cost of treating an SSI after spine surgery has been estimated at approximately $16,242 per case [8], with reported ranges between $15,817 and $38,701 in the literature [9]. Beyond these immediate expenses, SSI-related care can generate incremental commercial insurer payments exceeding $93,000 within 24 months of the index lumbar fusion [10]. Higher infection-related costs are associated with longer hospitalizations, a greater number of fused levels, methicillin-resistant Staphylococcus aureus (MRSA) infections, hypoalbuminemia, and the need for multiple irrigation and debridement procedures [8]. This study aims to characterize the costs specifically associated with readmission following surgical management of spinal SSIs, an underexplored contributor to the overall economic impact of these complications. By identifying key cost drivers during the readmission period, this study seeks to supplement value-based care strategies and guide preventative measures aimed at reducing the financial burden of postoperative spinal infections. Methods We conducted a retrospective cohort study of patients who underwent spinal surgery and developed a surgical site infection requiring readmission for operative management at a single tertiary academic hospital (January 2015–December 2022). Demographics (age, sex, BMI), comorbidities (ASA score, anemia, hypoalbuminemia), operative characteristics (approach, number of fused levels, instrumentation, PSO/VCR, staged index procedure, estimated blood loss >1,000 mL), infection-related variables (time from index surgery to I&D, polymicrobial culture, positive blood cultures, organism group), and payer type were extracted from the electronic medical record. Race and ethnicity were not included because these data were not collected or used as part of the original analytic framework for this study. Hospital finance provided itemized direct costs for the readmission; total payments and net revenue were used to derive contribution margin (net revenue−direct cost) and margin rate (contribution margin÷net revenue). Costs are presented in USD. The Ascension Health Institutional Review Board reviewed the research proposal and determined that the proposed activity is not research involving human subjects as defined by DHHS/FDA regulations, and thus exempt from IRB permission. Analyses proceeded in 2 steps: descriptive comparison followed by multivariable modeling. We summarized the cohort and compared outcomes across clinical subgroups using t tests or Mann–Whitney tests for 2 groups and ANOVA or Kruskal–Wallis tests for 3 or more groups, as appropriate (α=0.05). We quantified linear associations with Pearson correlations. For univariate group comparisons of total length of stay only (Table 3, Table 4), we excluded 8 statistical outliers using the standard Tukey criterion (values exceeding Q3+1.5×interquartile range, ie, stays >31.25 days), yielding n=115 for these comparisons. All other analyses, including Table 1 descriptive statistic for length of stay (15.9±8.7 days) and all Log Direct Cost comparisons in Table 3, retained the full cohort (n=123). Table 3Groupwise comparisons of outcomes by clinical characteristics at readmissionTable 3 dummy alt textVariableGroupNMean±SDStatisticp-ValueLog direct costSex† Male554.29±0.291,791.69 Female684.27±0.29DM† Yes344.3±0.281,402.53 No894.34±0.307Smoking† Yes304.3±0.331,315.64 No934.31±0.27Polymicrobial† Yes364.43±0.341,133.016⁎ No874.25±0.25Positive blood culture† Yes364.45±0.331,033.003⁎⁎ No874.25±0.24Recurrent infection within 90 d† Yes214.46±0.304671.007⁎⁎ No1022.28±0.27Return within 90 d§ Yes254.53±0.33−4.64<.001⁎⁎ No984.25±0.25Staged index procedure† Yes224.44±0.27712.009⁎⁎ No1014.28±0.28Hardware-placed instrumentation† Yes1214.3±0.2976.37 No24.44±0.19Insurance/payer type† Government954.33±2.91,086.14 Private284.24±2.7Primary diagnosis‡ Deformity294.42±0.3χ2=5.12.28 Degeneration844.27±0.28 Infection24.24±0.44 Oncology34.33±0.37 Trauma54.23±0.26Surgical approach‡ Anterior44.09±0.34χ2=2.89.24 Combined184.33±0.31 Posterior1014.31±0.28Total length of staySex† Male5214±6.641,543.6 Female6314.6±6.08DM† Yes3115±6.471,175.4 No8414.1±6.28Smoking† Yes2813.5±5.371,145.63 No8714.6±6.6Polymicrobial† Yes3416.9±7.341,006.002⁎⁎ No8113.3±5.56Positive blood culture† Yes3417.1±7.12931.006⁎ No8113.2±5.6Recurrent infection within 90 d† Yes1716.3±6.27626.1 No9814±6.3Return within 90 d† Yes1917.3±5.85563.009⁎⁎ No9613.8±6.28Staged index procedure† Yes2216.1±6.58830.17 No9313.9±6.15Hardware-placed instrumentation† Yes11314.2±6.2128.5.07 No223.5±7.78Insurance/payer type† Government8814.3±6.161,145.78 Private2714.6±6.92Primary diagnosis‡ Deformity2816.1±7.30χ2=4.61.33 Degeneration7813.5±5.76 Infection217.5±0.707 Oncology214±8.49 Trauma517±8.37Surgical approach‡ Anterior37.33±0.577χ2=6.84.03⁎ Combined1815.3±6.98 Posterior9414.4±6.19Log direct cost comparisons and all other tables/descriptive statistics (including Table 1’s length-of-stay summary, 15.9±8.7 days) use the full cohort (n=123). Total length of stay comparisons in this table use n=115: 8 statistical outliers (length of stay exceeding the upper Tukey fence, Q3+1.5×IQR=31.25 days; individual values of 32, 32, 33, 34, 37, 43, 47, and 48 days) were excluded prior to univariate group testing.†Mann–Whitney.‡Kruskal–Wallis.§Student t test.⁎= p < 0.05 (statistically significant).⁎⁎= p < 0.01 (highly statistically significant). Groupwise comparisons of outcomes by clinical characteristics at readmission Log direct cost comparisons and all other tables/descriptive statistics (including Table 1’s length-of-stay summary, 15.9±8.7 days) use the full cohort (n=123). Total length of stay comparisons in this table use n=115: 8 statistical outliers (length of stay exceeding the upper Tukey fence, Q3+1.5×IQR=31.25 days; individual values of 32, 32, 33, 34, 37, 43, 47, and 48 days) were excluded prior to univariate group testing. Mann–Whitney. Kruskal–Wallis. Student t test. = p < 0.05 (statistically significant). = p < 0.01 (highly statistically significant). Table 4Pairwise comparisons—total length of stayTable 4 dummy alt textWpAA/P3.423.041AP3.588.030A/PP−0.775.848Pairwise comparisons of total length of stay by surgical approach use n=115 (anterior n=3, combined n=18, posterior n=94), consistent with the outlier exclusion described in Table 3 (8 patients with length of stay >31.25 days excluded; see Table 3 footnote and Methods).A, anterior; A/P, combined (anterior/posterior); P, posterior. Pairwise comparisons—total length of stay Pairwise comparisons of total length of stay by surgical approach use n=115 (anterior n=3, combined n=18, posterior n=94), consistent with the outlier exclusion described in Table 3 (8 patients with length of stay >31.25 days excluded; see Table 3 footnote and Methods). A, anterior; A/P, combined (anterior/posterior); P, posterior. Table 1Cohort demographics, operative characteristics, microbiology, and financial outcomesTable 1 dummy alt textCharacteristicOverall (n=123)Age, mean±SD (y)63.2±12.2Male, n (%)55 (44.7)BMI, mean±SD (kg/m²)34.8±8.8DM, n (%)34 (27.6)Smoking, n (%)30 (24.4)ASA score, mean±SD2.9±0.5Prior infection history15 (12.2)Primary diagnosis—Deformity, n (%)29 (23.6)Primary diagnosis—Degenerative, n (%)84 (68.3)Primary diagnosis—Infection, n (%)2 (1.6)Primary diagnosis—Oncology, n (%)3 (2.4)Primary diagnosis—Trauma, n (%)5 (4.1)Total fused vertebral levels, mean±SD6.5±4.1Surgical approach—Posterior, n (%)101 (82.1)Surgical approach—Anterior, n (%)4 (3.3)Surgical approach—Combined, n (%)18 (14.6)Staged index procedure, n (%)22 (17.9)Hardware placed, n (%)121 (98.4)PSO or VCR performed, n (%)13 (10.6)Intraoperative durotomy or CSF leak, n (%)9 (7.3)EBL>1,000 mL, n (%)21 (17.1)Days from index to I&D, mean±SD30.4±15.7Positive blood cultures, n (%)36 (29.3)Polymicrobial infection, n (%)36 (29.3)Organism—Gram-positive, n (%)62 (50.4)Organism—Gram-negative, n (%)53 (43.1)Organism—MSSA, n (%)1 (0.8)Organism—No growth, n (%)6 (4.9)Organism—Enterobacter, n (%)1 (0.8)ESR (mm/h), mean±SD67.2±36.5CRP (mg/L), mean±SD11.1±8.2Hemoglobin (g/dL), mean±SD12.6±2Albumin (g/dL), mean±SD3.8±0.9Recurrent infection within 90 d, n (%)21 (17.1)Return to or within 90 d, n (%)25 (20.3)Readmission following final I&D, n (%)58 (48.3)Length of stay for SSI admission (d), mean±SD15.9±8.7Gross charges (USD), mean±SD$166,825±$137,805Direct cost (USD), mean±SD$25,713±$21,128Total payments (net revenue), USD, mean±SD$30,994±$20,588Contribution margin (USD), mean±SD$5,281±$18,448Margin rate, mean±SD0.08±0.18Payer—Government, n (%)95 (77.2)Payer—Private, n (%)28 (22.8)Financial definitions: total payments and net revenue refer to the same quantity (amount actually collected for the readmission) and are reported once above. Contribution margin=total payments (net revenue)−direct cost. Margin rate=contribution margin÷gross charges. Cohort demographics, operative characteristics, microbiology, and financial outcomes Financial definitions: total payments and net revenue refer to the same quantity (amount actually collected for the readmission) and are reported once above. Contribution margin=total payments (net revenue)−direct cost. Margin rate=contribution margin÷gross charges. Candidate predictors for the multivariable models were then selected using a hybrid clinical-algorithmic framework. All 15 candidate covariates described in our original protocol (age, BMI, ASA score, fused levels, time from index surgery to I&D, payer, sex, EBL>1 L, primary diagnosis, PSO/VCR, polymicrobial status, positive blood cultures, organism group, LOS, and staged procedure) were evaluated using 10-fold cross-validated LASSO-penalized Gamma regression; predictors retained by this algorithmic step were combined with payer type, retained a priori for its centrality to the study’s financial hypotheses. Primary diagnosis, originally a 5-level categorical variable with several sparse cells (trauma, n=5; oncology, n=3; infection, n=2), was collapsed to a binary indicator (deformity vs. nondeformity) to reduce the number of estimated parameters. This process yielded a final, streamlined covariate set of 8 predictors: length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). This reduced the primary cost model from 20 estimated parameters (in the originally specified model) to 9, improving the ratio of observations to estimated parameters from 5.7:1 to 13.7:1, and allowed the full analytic cohort (n=123) to be retained without the additional case exclusions required by the originally specified model. All continuous predictors were mean-centered. The primary outcome was direct hospital cost during the readmission, modeled using a Gamma generalized linear model with a log link, reporting coefficients as percentage change in cost (exp[β]−1) and average marginal effects (USD) with 95% confidence intervals. As a prespecified sensitivity analysis, we refit this streamlined model excluding length of stay to test whether LOS was confounding the association between operative-complexity variables (eg, PSO/VCR) and cost. We modeled contribution margin (USD) and margin rate using linear regression with identity link and the same streamlined covariate set as secondary financial outcomes. We determined that margin rate, as originally calculated, is defined as contribution margin÷gross charges; this is the definition we use throughout. Because gross charges are always substantially larger than direct cost or net revenue in this cohort (minimum $32,492), this ratio is numerically stable for all patients, and the full analytic cohort (n=123) is retained for the margin-rate model without exclusion. We additionally tested the length of stay using a negative binomial GLM with a log link to accommodate overdispersion in count data. Model diagnostics and omnibus tests are provided in the analysis output. Results are reported as point estimates with 95% confidence intervals and p-values, with 2-sided significance set at α=0.05. Results We included 123 patients readmitted for surgical management of spinal SSI (Table 1). The cohort had a mean age of 63.2±12.2 years, 44.7% were male, and mean BMI was 34.8±8.8 kg/m². A majority of procedures utilized a posterior approach (82.1%), with a mean of 6.5±4.1 levels fused. Polymicrobial infections and positive blood cultures were each present in 29.3% of cases. Mean time from index surgery to irrigation-and-debridement (I&D) was 30.4±15.7 days. The average readmission stay was 15.9±8.7 days. Financially, the mean direct hospital cost was $25,713±$21,128 (Fig. 1). Total payments averaged $30,994±$20,588, with a contribution margin of $5,281±$18,448 and a mean margin rate of 0.08±0.18. Government payers comprised 77.2% of the sample, and private payers 22.8%. Distribution of direct hospital cost during SSI readmission. Log-transformed direct cost was strongly correlated with total length of stay (LOS) (Pearson r=0.695, p<.001, Table 2; Fig. 2), indicating that LOS is one of the dominant drivers of readmission cost. Smaller positive correlations were seen with age (r=0.222, p<.01) and CRP (r=0.20, p<.05) (Fig. 3). On univariate comparisons for log direct cost, higher values were observed with polymicrobial infections (p=.016), positive blood cultures (p=.003), return to the OR within 90 days (p<.001), and staged index procedures (p=.009) (Table 3). Payer type and surgical approach were not significant in these unadjusted comparisons. Table 2Correlation matrix of clinical, operative, and laboratory variables with log-transformed direct hospital cost during readmission for spinal SSITable 2 dummy alt textLog direct costTotal length of hospital stayAgeBMITotal number levelsESRCRPHgAlbuminTime from index surgery to I&DLog direct cost––––––––––Total length of hospital stayPerson’s r=0.695⁎⁎–––––––––p-Value <.001⁎⁎Age.222⁎0.054––––––––<.01⁎0.568BMI.0380.074−0.216⁎–––––––.6780.4320.016⁎Total number levels.198⁎0.1130.14−0.092––––––.028⁎0.2290.1210.312ESR.0580.123−0.1070.1670.156–––––.5530.2210.2710.0850.108CRP.2⁎0.06−0.144−0.0840.1390.321⁎⁎––––.038⁎0.5470.1370.3740.152<0.001⁎⁎Hg.021−0.082−0.135−0.043−0.031−0.0530.147–––.8220.3860.1360.6350.7360.5840.13Albumin−.07−0.162−0.1870.140.0870.042−0.0740.29⁎––.5890.230.1470.2790.50.7670.6050.024⁎Time from index surgery to I&D.139−0.0040.0170.0080.084−0.154−0.060.160.07–.1310.9630.8510.930.3640.1160.530.090.58⁎= p < 0.05 (statistically significant).⁎⁎= p < 0.01 (highly statistically significant). Correlation matrix of clinical, operative, and laboratory variables with log-transformed direct hospital cost during readmission for spinal SSI = p < 0.05 (statistically significant). = p < 0.01 (highly statistically significant). Scatterplot with least-squares line and 95% CI shows the association between length of stay and log direct cost. Correlation heatmap of key continuous variables. For LOS analysis, patients with polymicrobial infections (16.9±7.3 vs. 13.3±5.6 days) and those with positive blood cultures (17.1±7.1 vs. 13.2±5.6 days) stayed longer (p=.002 and p=.006, respectively) (Table 3; Fig. 4, Left and Right). Return within 90 days was also associated with longer LOS (17.3±5.9 vs. 13.8±6.3 days; p=.009). Surgical approach differed overall (Kruskal–Wallis p=.030); pairwise tests showed anterior procedures had shorter stays than posterior (estimated means≈7.3 vs. 14.4 days; p=.041) and combined approaches (p=.03) (Table 3, Table 4). Length of stay by microbiology; (Left) LOS by polymicrobial versus monomicrobial infections; (Right) LOS by positive versus negative blood cultures. In the streamlined multivariable model for direct hospital cost (Gamma GLM, log link; n=123), each additional hospital day increased cost by 4.80% (marginal effect +$1,207; 95% CI $939–$1,526; p<.001) (Table 5, Table 6; Fig. 5, Fig. 6). Staging the index procedure was associated with 45.0% higher costs (marginal effect +$10,813; p<.001), and each additional fused level showed a nonsignificant trend toward higher cost (+2.08%; marginal effect +$532; p=.057). Deformity as the primary diagnosis was associated with 25.7% higher cost relative to nondeformity diagnoses (marginal effect +$6,218; p=.026) (Table 5, Table 6). Positive blood cultures were associated with 35.1% higher cost (marginal effect +$8,128; p<.001), and private (vs. government) payer showed a nonsignificant trend toward 14.6% lower cost (marginal effect −$3,905; p=.074). We refit this streamlined model excluding length of stay (Table 5). Table 5Predictors of direct hospital cost during readmission—streamlined Gamma GLM, log link (n=123)Table 5 dummy alt textPredictor% change in cost (expβ−1)p95% CI for % changeMarginal effect (USD)Length of stay, per day+4.80%<.001⁎+3.9% to +5.7%+$1,207 (95% CI $939–$1,526)Staged index procedure (yes vs. no)+45.0%<.001⁎+19.8% to +75.4%+$10,813 (95% CI $3,654–$19,318)Total fused levels, per level+2.08%.057−0.1% to +4.3%+$532 (95% CI −$17 to $1,150)PSO/VCR (yes vs. no)−33.6%.001⁎−48.3% to −14.7%−$8,975 (95% CI −$12,966 to −$3,817)Private vs. government payer−14.6%.074−28.2% to +1.5%−$3,905 (95% CI −$7,737 to $473)Positive blood cultures (yes vs. no)+35.1%<.001⁎+14.3% to +59.6%+$8,128 (95% CI $3,391–$13,242)Days, index surgery to I&D, per day+0.60%.014⁎+0.1% to +1.1%+$155 (95% CI −$20 to $330)Primary diagnosis: Deformity vs. nondeformity+25.7%.026⁎+2.8% to +53.7%+$6,218 (95% CI $1,024–$12,489)Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Link=log; distribution=Gamma.⁎= p < 0.05 (statistically significant). Predictors of direct hospital cost during readmission—streamlined Gamma GLM, log link (n=123) Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Link=log; distribution=Gamma. = p < 0.05 (statistically significant). Table 6Sensitivity analysis: predictors of direct hospital cost EXCLUDING length of stay—streamlined Gamma GLM, log link (n=123)Table 6 dummy alt textPredictor% change in cost (expβ−1)p95% CI for % changeMarginal effect (USD)Staged index procedure (yes vs. no)+38.5%.025⁎+4.1% to +84.3%+$9,230 (95% CI $2,045–$16,867)Total fused levels, per level+3.86%.018⁎+0.6% to +7.2%+$971 (95% CI $180–$1,895)PSO/VCR (yes vs. no)−51.7%<.001⁎−66.5% to −30.2%−$14,182 (95% CI −$19,781 to −$8,697)Private vs. government payer−16.4%.175−35.5% to +8.3%−$4,368 (95% CI −$9,837 to $1,252)Positive blood cultures (yes vs. no)+62.4%<.001⁎+26.9% to +107.8%+$13,414 (95% CI $5,451–$22,537)Days, index surgery to I&D, per day+0.59%.108−0.1% to +1.3%+$151 (95% CI −$109 to $393)Primary diagnosis: Deformity vs. nondeformity+24.6%.153−7.9% to +68.5%+$5,921 (95% CI −$1,281 to $15,092)⁎= p < 0.05 (statistically significant). Sensitivity analysis: predictors of direct hospital cost EXCLUDING length of stay—streamlined Gamma GLM, log link (n=123) = p < 0.05 (statistically significant). Predictors of direct hospital cost during readmission, streamlined Gamma GLM with log link (n=123). Adjusted predicted direct cost across hospital length of stay, streamlined model (n=123). Rather than attenuating the paradoxical PSO/VCR finding, removing LOS from the model strengthened it (−51.7%; marginal effect −$14,182; 95% CI −$19,781 to −$8,697; p<.001), indicating that LOS is not responsible for this association. A similar pattern was observed when LOS was removed from the originally specified, full-covariate model (−29.4% with LOS vs. −52.9% without LOS). This finding is addressed further in the Discussion. All models were mean-centered. For contribution margin (linear model, streamlined covariate set; n=123), private insurance was associated with a +$13,003 higher margin (95% CI +$5,800 to +$20,206; p<.001) (Table 7 and Fig. 7). LOS showed a significant negative association with contribution margin (−$681 per day; 95% CI −$1,041 to −$322; p<.001). Table 7Predictors of contribution margin—streamlined linear model (n=123)Table 7 dummy alt textPredictorCoefficient (USD)p95% CILength of stay, per day−$681<.001⁎−$1,041 to −$322Staged index procedure (yes vs. no)−$3,181.428−$11,106 to +$4,744Total fused levels, per level−$462.302−$1,346 to +$421PSO/VCR (yes vs. no)+$7,728.144−$2,679 to +$18,136Private vs. government payer+$13,003<.001⁎+$5,800 to +$20,206Positive blood cultures (yes vs. no)−$228.948−$7,166 to +$6,711Days, index surgery to I&D, per day+$31.757−$168 to +$230Primary diagnosis: Deformity vs. nondeformity−$1,396.742−$9,768 to +$6,975Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Identity link with Gaussian error.⁎= p < 0.05 (statistically significant). Predictors of contribution margin—streamlined linear model (n=123) Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Identity link with Gaussian error. = p < 0.05 (statistically significant). Contribution margin by payer (adjusted means), streamlined model (n=123). For margin rate (linear model, streamlined covariate set; n=123; margin rate defined as contribution margin÷gross charges, the definition consistent with Table 1; see Methods), private versus government payer increased margin rate by +0.106 (95% CI +0.036 to +0.175; p=.003) (Table 8 and Fig. 8). Each additional hospital day decreased margin rate by −0.0055 (95% CI −0.0090 to −0.0021; p=.002). PSO/VCR showed a nonsignificant trend toward a higher margin rate (+0.067; p=.186), consistent in direction but attenuated relative to the direct-cost model, likely reflecting the small PSO/VCR subgroup. Table 8Predictors of margin rate—streamlined linear model (n=123)Table 8 dummy alt textPredictorCoefficientp95% CILength of stay, per day−0.0055.002⁎−0.0090 to −0.0021Staged index procedure (yes vs. no)−0.053.168−0.130 to +0.023Total fused levels, per level−0.007.118−0.015 to +0.002PSO/VCR (yes vs. no)+0.067.186−0.033 to +0.167Private vs. government payer+0.106.003⁎+0.036 to +0.175Positive blood cultures (yes vs. no)−0.024.470−0.091 to +0.042Days, index surgery to I&D, per day+0.0016.098−0.0003 to +0.0035Primary diagnosis: Deformity vs. nondeformity+0.006.883−0.075 to +0.086Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Identity link with Gaussian error. Margin rate is defined as contribution margin÷gross charges, the definition consistent with the descriptive statistic reported in Table 1 (0.08±0.18); because gross charges are always substantially larger than direct cost (minimum $32,492 in this cohort), this ratio is numerically stable, and the full analytic cohort is retained without exclusion.⁎= p < 0.05 (statistically significant). Predictors of margin rate—streamlined linear model (n=123) Model terms (streamlined covariate set, all mean-centered): length of stay, staged index procedure, total fused levels, PSO/VCR, payer type, positive blood cultures, days from index surgery to I&D, and primary diagnosis (deformity vs. nondeformity). Selected via 10-fold cross-validated LASSO-penalized Gamma regression combined with a priori clinical prioritization (see Methods). Identity link with Gaussian error. Margin rate is defined as contribution margin÷gross charges, the definition consistent with the descriptive statistic reported in Table 1 (0.08±0.18); because gross charges are always substantially larger than direct cost (minimum $32,492 in this cohort), this ratio is numerically stable, and the full analytic cohort is retained without exclusion. = p < 0.05 (statistically significant). Predictors of margin rate, streamlined linear model (n=123; margin rate=contribution margin÷gross charges). Discussion Several key variables were identified that contributed to changes in cost using length of stay (LOS) modeling and direct cost analysis. The most significant cost determinants were increased LOS, staged index procedures, greater total fused levels, patient payer type, surgical technique, and diagnostic indication. The calculated average direct hospital cost of $25,713 (± $21,128) represents a substantial financial burden and is comparable to prior work reporting average direct costs ranging from $15,817 to $38,701 [3,11,12]. A large-scale fusion cohort study has even described costs as high as $93,741 for commercial payers over 24 months for deep infection cases after spinal surgery [13]. Length of stay is a major contributor to costs, with each additional inpatient day increasing the total cost by 5.5% (approximately +$1,247/d); the readmission stay in this cohort averaged 15.9±8.7 days. Existing literature reports that the average LOS for readmission after postoperative spine SSI is 9.6 days [8]. In our cohort, patients with polymicrobial infections (16.9±7.3 days) and those with positive blood cultures (17.1±7.1 days) had longer readmission stays compared to 13.3±5.6 days and 13.2±5.6 days, respectively, in patients without these features. These findings suggest that microbiologic complexity and systemic infection contribute meaningfully to increased LOS and, by extension, to higher direct costs. Surgical approach also influenced LOS. Patients undergoing anterior approaches stayed an average of 7.33±0.58 days, compared with 15.3±6.98 days and 14.4±6.19 days for combined and posterior approaches, respectively. Although the published literature does not provide robust direct comparisons of LOS by surgical approach, at least 1 study suggests that extended posterior surgical involvement is associated with longer LOS and greater costs [8]. Our data support the hypothesis that more extensive or posterior-based procedures may be accompanied by more complex wound profiles and rehabilitation needs, contributing to prolonged hospitalization and greater resource utilization. A hospital return within 90 days of the index procedure was associated with a longer LOS (17.3±5.9 days) compared with those readmitted beyond 90 days (13.8±6.3 days). Log-transformed cost data for 90-day returnees (4.53±0.33 vs. 4.25±0.25; p<.001) demonstrate that costs of services for these patients are significantly increased. Prior work indicates that patients returning within this window more often experience complex or severe postoperative complications, including deep SSI, which drive longer hospitalizations and higher cost. One study on unplanned readmissions found that SSIs accounted for 44.21% of 90-day returnees and aseptic pain accounted for 31.58% [14]. In addition, comorbidities such as diabetes, hypertension, and liver disease were associated with higher 90-day readmission rates following spine surgery [14]. Taken together, these observations highlight 90-day readmission as a marker of disease severity and a critical target for cost-containment strategies. Procedural factors also played a significant role in the cost of spinal SSIs. Each additional fused level increased the cost by 2.83% (approximately +$651 per level), and staged index procedures were associated with disproportionately higher costs. Increased operative time, greater blood loss, and heightened wound complexity in multilevel and staged procedures may contribute to these findings. Payer type similarly influenced cost; patients with commercial insurance have been shown in prior work to incur higher reimbursed and billed charges than those with government-sponsored plans, which may reflect differences in negotiated rates, resource utilization, and access to postacute care [3,[11], [12], [13]]. Our data underscores that both technical complexity and systemic factors outside the immediate control of the surgeon influence the eventual economic burden of SSI. One counterintuitive finding warrants specific comment: PSO/VCR was consistently associated with lower, not higher, direct cost across all specifications of our model. Because PSO/VCR is a marker of greater operative complexity, this association runs opposite to clinical expectation, and a plausible explanation is that hospital length of stay, a strong independent driver of cost, was confounding or masking the true relationship between PSO/VCR and cost. We tested this directly with a prespecified sensitivity analysis that removed LOS from the model. If LOS adjustment were responsible for the paradox, its removal should have attenuated or reversed the PSO/VCR effect. Instead, the negative association strengthened in both the streamlined model (−33.6% with LOS vs. −51.7% without LOS) and the originally specified, full-covariate model (−29.4% vs. −52.9%). This indicates that LOS is not the source of the paradoxical finding. A more plausible explanation is confounding by indication: surgeons may reserve PSO/VCR for patients judged capable of tolerating a larger single-stage correction, a selection process that could favor patients with lower comorbidity burden or more favorable postoperative trajectories not fully captured by our covariates. The small number of PSO/VCR cases in this cohort (n=13) further limits the precision of this estimate and increases its susceptibility to case-mix effects. We present this finding as hypothesis-generating and recommend it be examined in larger, multicenter cohorts capable of directly modeling surgeon-level selection factors. Several strategies have been proposed in the literature to mitigate the costs associated with spinal SSI. Preoperative risk modification, including glycemic control, smoking cessation, anemia treatment, and weight optimization, has been associated with both reduced SSI rates and shorter LOS [[15], [16], [17]]. Intraoperative measures such as minimizing blood loss (eg, via antifibrinolytics like tranexamic acid), optimizing nutrition, and employing appropriate antibacterial therapies can further reduce LOS and consequently overall costs. Existing literature supports the use of MRSA decolonization, intraoperative betadine irrigation, vancomycin powder, and closed-suction drains as effective measures to reduce SSI-related healthcare costs [18]. The use of vancomycin powder alone has been shown to save approximately $4,337 per posterior spinal fusion [19]. Incorporating these perioperative strategies into standardized infection-prevention pathways may significantly blunt the downstream economic impact of SSI. Postoperative care coordination and discharge planning represent another important opportunity to reduce LOS and direct hospital costs. Preplanning postoperative pathways into acute rehabilitation facilities or inpatient rehabilitation facilities (IRFs) through early bed reservations may facilitate earlier transitions out of the acute care setting. One study of adult spinal deformity surgery found savings of approximately $8,420 per day (range $2,900–$13,900/d) when transitioning patients from the hospital to an IRF [20]. Although that population differs from our cohort, the magnitude of daily cost differentials suggests that timely step-down to lower acuity settings could offset a substantial portion of the per-day hospital cost observed in our analysis. Additionally, thoughtful surgical planning to limit the number of fused levels when clinically feasible could help mitigate the observed per-level cost increase (approximately $532–$971 per level across model specifications). The applicability of early step-down pathways to this specific population, however, warrants caution. Unlike elective adult spinal deformity patients, individuals readmitted for surgical management of an acute spinal SSI frequently require ongoing acute-level care well into the early postoperative period, including serial operative or bedside wound debridement, management of negative-pressure wound therapy (NPWT), and close monitoring for sepsis or systemic decompensation. NPWT is relatively contraindicated in the setting of untreated osteomyelitis or an active sepsis source and requires close clinical monitoring when infection is still present. Medicare medical necessity criteria for IRF admission similarly specify that inpatient rehabilitation is not intended as a substitute for completing an acute treatment course, such as an initial course of intravenous antibiotics, and require that the diagnostic workup and care plan be largely finalized, with any ongoing medical needs not exceeding the capability of the receiving facility, before transfer. The existing spine-surgery discharge-disposition literature offers a mixed picture that reinforces this caution: subacute rehabilitation placement has been independently associated with a higher rate of wound complications after spinal decompression surgery, and several series report increased 30-day readmission among patients discharged to rehabilitation or skilled-care facilities relative to those discharged home, although at least 1 large registry analysis found the opposite association. Taken together, these findings suggest that early transition to an IRF or subacute facility is likely appropriate only for a carefully selected subset of SSI patients, specifically those who are afebrile and hemodynamically stable, established on a definitive culture-directed oral or completed intravenous antibiotic regimen, and whose wounds no longer require operative debridement or NPWT under active-infection precautions. For patients who still require serial debridement, NPWT management, or close monitoring for sepsis, continued acute-level care remains necessary irrespective of cost considerations, and premature step-down could compromise infection control and patient safety. Multidisciplinary coordination between spine surgery, infectious disease, and physiatry to identify patients who meet these safety criteria, rather than a uniform early-transfer policy, is likely necessary to realize the cost benefits of postacute step-down without compromising outcomes. The figures identified in our analysis likely underestimate the total economic burden of spinal SSI. Our direct cost calculations do not account for indirect or long-term costs such as transportation, prolonged outpatient antibiotic therapy, chronic pain management, or repeat imaging. One study estimates indirect costs related to SSI at $3,739 [21], while another characterizes the total mean indirect cost associated with repeat spinal surgeries at $28,019 [22]. Socioeconomic factors such as delayed return to work, long-term disability, caregiver burden, and psychosocial stress are also rarely captured in administrative or institutional cost analyses but substantially affect patients, families, and employers. Future work incorporating patient-reported outcomes and societal cost perspectives is needed to more completely define the financial and human impact of spinal SSI. While this study identifies multiple predictors of cost, several limitations should be acknowledged. The retrospective design is inherently susceptible to selection bias and unmeasured confounding, despite adjustment for clinical and operative variables in the multivariable models. Factors such as intraoperative efficiency, surgeon experience, postoperative care pathways, and rehabilitation access were not fully captured and may influence both LOS and cost. Cost estimates are derived from a single institution (or health system), which may limit generalizability to centers with different case mixes, reimbursement structures, or infection-prevention protocols. Although Gamma regression modeling was used to address skew, a small number of very high-cost cases may still disproportionately influence estimates. Microbiologic data, including organism virulence and resistance patterns, were not analyzed in detail and could further refine risk stratification. Race and ethnicity were not included in this analysis because these variables were not consistently collected or used in the study dataset. As a result, the absence of race and ethnicity data represents a limitation and may limit the generalizability of these findings to more diverse patient populations. In addition, existing literature lacks granular cost and LOS data stratified by spinal region (cervical, thoracic, lumbar), which constrains direct comparisons and underscores the need for more region-specific SSI cost analyses. This study also has several strengths. It focuses specifically on patients with spinal SSI, a population that is clinically challenging yet underrepresented in economic analyses relative to primary fusion cohorts. The use of both LOS modeling and direct cost analysis provides complementary perspectives on resource utilization and enables translation of clinical variables into quantifiable financial impact. Inclusion of detailed operative variables, such as fused levels and staged procedures, as well as readmission timing and microbiologic characteristics, allows for a more nuanced understanding of cost drivers than is typically available in large administrative data sets. By leveraging contemporary cost data and robust statistical modeling, our analysis offers practical, institutionally grounded benchmarks that can inform quality improvement efforts and value-based care initiatives. Collectively, these models quantify the multifactorial economic burden of spine SSI and identify actionable targets, including early and aggressive debridement, stringent blood-loss control, efficiency-oriented operating room workflows, thoughtful case selection and staging, and optimization of nutritional and medical comorbidities, that may reduce both LOS and direct hospital cost. Interventions that promote timely step-down to rehabilitation facilities, standardize infection-prevention bundles, and prioritize high-risk patients for enhanced perioperative optimization may be particularly impactful. As healthcare systems increasingly emphasize value-based care and cost containment, integrating these findings into institutional pathways and bundled payment designs has the potential to improve outcomes while alleviating the substantial financial burden of spinal SSI on patients, providers, and payers. Declarations of competing interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article. Footnotes FDA device/drug status: Not applicable. Author disclosures: PA: Nothing to disclose. AZ: Nothing to disclose. MG: Nothing to disclose. JJ: Nothing to disclose. MS: Nothing to disclose. RS: Grants: Kuros; Medtronic; SI Bone; Pressio Spine (none). References