2.1 Patients, demographic, treatment, and outcome data

After Institutional Review Board (IRB) approval was obtained, patients who underwent treatment for DCM between 2015 and 2021 were identified from an internal, web-based database at our tertiary referral center. The database captures all patients who underwent surgery with our institution's orthopaedic spine service within this timeframe. This study period reflects the time point at which the database was locked for analysis. Data were prospectively entered as part of routine clinical care and reviewed for completeness and consistency prior to inclusion. Inclusion criteria included patients aged ≥18 years undergoing surgery for DCM. Patients were excluded for diagnoses of isolated cervical radiculopathy, neuromuscular diseases, tumor, and traumatic spinal cord injury. Patients with concomitant myeloradiculopathy were included, provided that cervical myelopathy was present and represented the primary indication for operative intervention. No minimum postoperative follow-up duration was required for study inclusion. All patients meeting these inclusion and exclusion criteria were included in the study. Demographic and baseline data was collected, including sex, age, body mass index (BMI), race, ethnicity, and insurance type, and preoperative Nurick grade [Table 1].¹⁴,¹⁵ Severity of spinal cord compression on cervical spine MRI was assessed using the technique described by Kang et al., as follows: Grade 0, normal CSF visible around cord and no evidence of cord deformity; Grade I: cervical canal stenosis with obliteration of CSF space; Grade II: compression and deformity of spinal cord with no signal changes; Grade III: cervical stenosis with increased T2 signal intensity at the corresponding level [Fig. 1].¹⁶ Treatment data included number of levels and surgical approach/procedure, including anterior cervical discectomy and fusion (ACDF), cervical disc replacement (CDR), posterior spinal fusion (PSF), and laminoplasty. A hybrid treatment, involving a combination of ACDF and CDR at different levels, was also utilized. Outcomes included length of stay (LOS), postoperative Nurick grade at discharge, hospital discharge disposition (home vs skilled nursing vs other), and 30-day reoperation and readmission rates following index surgery. Patients initially evaluated in clinic were assigned to the “Elective group” and those initially evaluated in the ED were assigned to the “ED group.”

Close

Fig. 1 Sagittal (A) and axial (B) T2-weighted MRI demonstrating severe cord compression (Grade III) at C5-6 with evidence of myelomalacia.

Export to PPT

2.2 Sociodemographic indices

Sociodemographic status for each patient was estimated using SDI and ADI, which were obtained through aggregated Zip Code Tabulation Area (ZCTA).¹²,¹³ SDI is reported as percentile based on zip code data, ranging from 1 to 100. SDI accounts for levels of disadvantage by zip code based off seven demographic characteristics, including: percent living in poverty, percent with less than 12 years of education, percent single-parent household, percent living in rented housing unit, percent living in overcrowded housing unit, percent of households without a car, and percent non-employed adults under 65 years of age within the zip code. Higher percent SDI indicates higher degree of deprivation. ADI is a composite index based on United States Census data reported on a scale of 1 to 10. ADI evaluates neighborhoods based on income, education, employment, and housing quality, with higher ADI indicating higher level of disadvantage.¹⁷

2.3 Statistical analysis

Descriptive statistics were performed for all patient baseline, treatment, and outcome data. Univariate comparisons were performed between the Elective and ED groups. Numeric variables including age, BMI, cervical stenosis on MRI, SDI, ADI, and levels operated on, baseline Nurick grade, postoperative Nurick grade, and LOS were compared using a t-test or Wilcoxon rank-sum test. Categorical variables, including sex, race, ethnicity, insurance, surgical approach, discharge disposition, and reoperation and readmission rates were compared using a chi-square test or Fisher's exact test, as appropriate. Categorical variable Medicare/other Government insurance subgroup is comprised of Medicare with Military Insurance. “Other Government” portion consisting of Military insurance was minimal, 4/77 for ED and 2/145 in Elective. Furthermore, to assess the proportions of patients who underwent motion sparing surgery, CDR and laminoplasty were grouped together to compare against other surgical approaches. Univariate logistic regression models were conducted to analyze the associations between sociodemographic variables or surgery type and likelihood of ED presentation. For logistic regression models including surgery type, some procedures were performed infrequently; therefore, Firth penalized regression was used to reduce small-sample bias and produce finite estimates. No formal adjustment for multiple comparisons was performed, as univariate analyses were exploratory and results were interpreted with consideration to effect sizes and confidence intervals.

Lastly, multivariable logistic regression models were fit to assess the impact of ADI and SDI, separately, on patient presentation (ED or Elective), controlling for BMI, sex, race, and insurance status. These variables were included in the multivariable model due to clinical relevance and prior literature supporting their potential association with ED presentation.¹⁸,¹⁹ Two separate models were constructed due to potential collinearity between ADI and SDI. Statistical analysis was performed using SAS® Studio software (SAS Institute Inc., Cary, NC) and R version 4.2.1 with level of significance set at p = 0.05.²⁰ An a priori power calculation was not performed as the study used data from a database with a fixed cohort size; therefore, results should be interpreted based on effect sizes and 95% confidence intervals.

3.1 Patient demographics and preoperative characteristics

A total of 327 patients underwent surgery for DCM between 2015 and 2021. Of these, 227 patients were admitted through clinic (Elective group) while 100 were admitted through the ED (ED group). Mean age was similar between the elective and ED cohorts (61.3 years vs 61.9 years, respectively p = 0.663). The Elective group had a higher proportion of female (48.0 vs 30.0%, respectively p = 0.002) and White patients (72.7 vs 51.0%, respectively p < 0.001). A larger proportion of ED patients were male compared to Elective (70% vs 52%, p = 0.002). Overall, the distribution of race was significantly different between groups (p < 0.001). The ED group had a larger proportion of Black and Asian patients compared to the Elective group, which were 23% vs 7.5% and 8% vs 3.5%, respectively. Average BMI was significantly higher for patients admitted through clinic (30.6 vs 28.1, p = 0.003). There was no difference in the percentage of Non-Hispanic patients between the Elective and ED groups (90.7 vs 84%, respectively, p = 0.241). More patients in the ED cohort were uninsured or covered by Medicare/other Government insurance (78.0 vs 67.0%, respectively, p = 0.044). Patients in the ED cohort had higher median SDI (70 (51, 88) vs 63 (36, 82), p < 0.001) and ADI (9, (7, 9) vs 8 (6,9), p = 0.002) [Table 2].

ED patients were more likely to have Grade III stenosis on MRI as compared to patients from the Elective cohort (79.4.0% vs 42.9%, respectively, p < 0.001) [Table 3], yet baseline Nurick grade was similar between cohorts (2.2 vs 2.1, respectively, p = 0.34) [Table 2].

3.2 Treatment and outcomes results

The distribution of surgical approaches was statistically different between groups (p < 0.001), with ED patients more frequently managed using fusion and posterior approaches while elective patients more commonly underwent anterior procedures [Table 4]. A greater proportion of ED patients underwent PSF as compared to Elective patients (30% vs 15% respectively), but the same fraction of patients underwent motion-sparing procedures across the groups (33%). Interestingly, the only motion-sparing procedure ED patients received was laminoplasty.

Table 5 illustrates the results of the univariate logistic regression models, with a greater Odds Ratio indicating increased odds of being a patient admitted through the ED. A one unit increase in BMI was associated with a 6% decrease in the odds of being admitted to surgery through the ED compared to clinic. The odds of a Black patient being admitted to surgery through the ED was 4.38 times greater than the odds of a White patient being admitted to surgery through the ED, and an Asian patient was 3.24 times more likely. The odds of presenting in the ED for patients who had PSF surgery is 2.46 times greater than the odds for those who had ACDF surgery.

3.3 Surgical characteristics and outcomes of the ED and elective groups

There was no difference in the median number of levels operated on between the ED and Elective groups (3 (2, 4) vs 3 (2,4) levels, respectively p = 0.164). Elective patients had shorter median LOS (2 (2,4) vs 6.5 (4, 11.25), p < 0.001) and were more likely to be discharged home (82.3% vs 35.0% respectively, p < 0.001). There was a difference in 30-day reoperation rate (2.2% vs 10%, p = 0.003) but not in readmission rate (17.0% vs 11.9%, p = 0.213). Elective patients had greater functional improvement in Nurick grade at final follow-up (1.9 vs 1.1 improvement in Nurick grading, <0.001) [Table 7].

3.4 Multivariable logistic regression for ADI and SDI

After controlling for BMI, sex, race, and insurance status, a one unit increase (decile) in ADI score was associated with a 22% increase in the odds (95% CI: 7% to 39%) of a patient presenting through the ED. Controlling for the same variables, a one unit increase (percentile) in SDI score was associated with a 2% increase in the odds (95% CI: 1% to 3%) of a patient presenting through the ED [Table 6].

4.1 Demographics and clinical presentation

In our cohort, patients presenting through the ED were more likely to be male and non-white, which is consistent with prior work.¹⁸ This finding should be interpreted in the context of our institution's catchment area, and the overall cohort, which was predominantly male and included a substantial proportion of patients from racial and ethnic minority groups. Nevertheless, this association may reflect how demographic factors influence different pathways to DCM evaluation.

Prior studies have highlighted the association between socioeconomic disadvantage and delays in diagnosis and treatment of DCM. For example, Pope et al. found that delays in DCM diagnosis were associated with ethnicity and that disadvantaged populations presented with greater disease severity.¹¹ In our cohort, ED presentation was similarly associated with more advanced disease, as evidenced by a higher proportion of Grade III stenosis on MRI, reflecting myelomalacia. This greater disease severity may reflect delays in diagnosis related to limited access to care, lower health literacy, and historical distrust in the medical system, which disproportionately affect socially disadvantaged populations. Despite more radiographically advanced disease at presentation, Nurick scores were similar across presentation settings. Because Nurick grading is clinician-assessed and reflects functional status, this suggests a discrepancy between radiographic disease severity and clinical symptoms, with ED patients exhibiting more advanced structural disease than symptoms alone may indicate. Functional reserve and compensatory mechanisms may allow patients to maintain neurologic function despite progressive structural injury on imaging.²³ This radiographic-clinical mismatch may be particularly consequential in populations facing barriers to care, where reliance on symptom severity alone may further delay diagnostic evaluation. Given that the goal of surgical intervention is to stop myelopathy progression, early detection is critical. Advanced disease in ED patients suggests that they face barriers to timely evaluation and workup of DCM, which may be more pronounced when symptom severity does not correlate with disease progression. Future research efforts should aim to identify these barriers and implement interventions to mitigate their effects.

4.2 Management and outcomes

Our results are consistent with prior work showing that sociodemographic disparities are associated with worse outcomes among orthopedic conditions.²⁴ While delays in diagnosis may contribute one explanation for worse outcomes, our data demonstrate significant differences in procedure type distribution between the ED and elective cohorts, suggesting that disparities may also influence treatment pathways once patients enter the healthcare system.

Our finding that ED patients were more likely to undergo PSF is consistent with those of McClelland et al. In their study, McClelland et al. evaluated a nationwide inpatient sample from 2001 to 2010, and identified that non-private insurance and non-white race were independently associated with undergoing posterior cervical surgery, as opposed to an anterior approach, which has been associated with increased mortality in DCM treatment.²⁵ Park et al. similarly found an association between the anterior approach and patient employment status; however, there were no differences in outcomes between the anterior and posterior groups.²⁶ Their findings also showed that the posteriorly-treated group had worse myelopathy, more levels, and older age.

In our cohort, differences in surgical intervention type should also be interpreted in the context of more advanced disease in the ED group. PSF is indicated when reliable decompression and stabilization are required, particularly with advanced pathology and in the setting of unstable alignment.²⁷ Patients presenting through the ED may also have less opportunity for preoperative planning and optimization for surgery, in addition to greater disease severity, further favoring the use of PSF. Surgical decision-making may be influenced by surgeon preference and institutional practice patterns, which could affect procedure selection independently of disease severity. Thus, multi-center studies are needed to clarify the interactions of these factors and their implications for outcomes in disadvantaged populations.

Patients in the elective cohort experienced significantly greater improvement in Nurick grade, shorter LOS, and were more likely to discharge home. In contrast, the ED group demonstrated significantly worse post-operative improvement, despite similar baseline Nurick scores between groups, which was consistent with their greater disease severity at presentation.²,³,²⁸,²⁹ The significantly longer LOS in the ED group may reflect increased comorbidities and worse baseline health or differences in surgical complexity; although, our findings showed no difference in number of operated levels between groups. Our results also showed a lower 30-day reoperation rate in the elective group and no difference in 30-day readmission rate between the two cohorts. Postoperative reoperation reflects surgical or technical complications requiring operative treatment, whereas postoperative readmission can be influenced by a broad range of medical complications or social-support related factors. Accordingly, the higher rate of reoperation in the ED group is likely attributable to increased technical complexity and disease severity at the time of surgery. The absence of a difference in readmission rates may be in part influenced by the high proportion of ED patients discharging to skilled nursing facilities or other long-term rehabilitation programs, where postoperative issues may be managed without hospital readmission.

Discharge disposition reflects the combined influence of clinical recovery and social factors, including social support and resource availability. Prior work in traumatic injury populations has demonstrated that social support is key in recovery, with stronger social networks significantly influencing both mental and physical outcomes.³⁰ Discharge disposition may also serve as a proxy for access to healthcare services and patients’ confidence in managing postoperative recovery. In a national inpatient sample spanning 2016-2020, Kabangu et al. found that minority race and lower socioeconomic status were associated with unfavorable discharge following cervical spinal cord injury.³¹ Similarly, in our cohort, patients presenting through the ED were more likely to experience unfavorable discharge outcomes, suggesting that presentation setting may also capture underlying social vulnerability in addition to disease severity. These support the use of ADI and SDI as tools to help identify patients at increased risk for challenges with post-discharge care.

Overall, our findings are consistent with prior studies that have compared spine surgery patients on the basis of admission setting. In a study evaluating the impact of admission setting on perioperative outcomes in lumbar fusion, Kukreja et al. found that emergent admission was an independent risk factor for perioperative complications and longer LOS, and that the comorbidity and sociodemographic profiles of each group differed significantly.³² Extending these observations to DCM, our study identifies further differences in postoperative outcomes, demonstrating that ED presentation is associated with worse functional improvement in Nurick grade at final follow-up and greater 30-day reoperation rates.

Similar patterns have been observed in other spine populations. Zeoli et al. reported differences in surgical outcomes following metastatic spine surgery between patients who presented to the ED versus clinic.³³ In their cohort, ED patients were more likely to be older, have more comorbidities, and public insurance. Patients presenting through the ED had higher Bilsky scores but a similar rate of preoperative neurologic deficits compared with clinic patients. This finding reflects a discrepancy between structural disease severity and clinical symptoms, consistent with our data, and underscores how sociodemographic factors may influence the timing and setting of evaluation for DCM.

Taken together, our findings build upon prior work in demonstrating that patients with DCM who present through the ED often represent a population with greater disease severity and greater barriers to outpatient care. While referral bias may contribute to these differences, with patients presenting through the ED reflecting subsets with limited outpatient referral pathways and advanced disease, this bias represents a real-world disparity in access to care. Rather than a limitation, our study leverages presentation setting to contextualize these differences and to better understand the factors shaping DCM evaluation and treatment.

Our study differs from others in a few ways. First, our data was obtained from a single institution, allowing us to examine disparities in spine care observed in national level studies within our local community. By limiting analysis to patients within our institution's catchment area, we were able to compare outcomes among individuals living in a similar geographic location, with relatively comparable access to healthcare resources based on distance. Although this may limit generalizability, it reduces geographic variability and enables a more focused examination of disparities within a defined healthcare environment. From a policy and health-systems perspective, utilizing ADI and SDI to quantify social deprivation allowed our study to comprehensively evaluate multiple domains of socioeconomic disadvantage into a standardized score, rather than using individual categories such as race, insurance type, and income. ADI can be employed across populations and health systems, making it useful for comparisons in large-scale research and predictive modeling, as well as for identifying groups at higher risk for delayed presentation and care. Further, targeted interventions that incorporate socioeconomic risk stratification into outpatient referral pathways or preoperative assessment may help direct resources to patients who have difficulty accessing elective spine care.

4.3 Limitations

This study has a number of important limitations. First, this was a retrospective, single institution study, which may limit generalizability due to local practice patterns and access to care, and reflect surgeon preferences in evaluation and management of DCM. Institutional sociodemographic characteristics may also influence ED versus clinic presentation patterns. However, both cohorts were derived from the same institution and catchment area, and observed differences likely reflect disparities in access to outpatient care rather than regional variation in healthcare availability. Secondly, wide confidence intervals for some variables, particularly smaller subgroups, should be interpreted with caution due to limited statistical power. Due to the use of a pre-existing database without access to individual-level data, we were unable to calculate ROC curves, c-statistics, or Hosmer-Lemeshow tests to formally assess model fit. Similarly, we were unable to obtain data to construct Kaplan-Meier curves. Formal assessment of inter-observer reliability of MRI interpretation was not possible due to the nature of the database; however, prior studies have shown reliable inter-rater agreement using the Kang MRI grading system.³⁴,³⁵

Several system-level factors may have influenced our findings. First, regional healthcare infrastructure–including availability of outpatient spine services and proximity to tertiary referral centers–may play a role in both presentation setting and timing. The study period also overlapped with the COVID-19 pandemic, which likely resulted in disruptions to elective care, delays in outpatient evaluations, and shifts in healthcare utilization patterns. Consequently, these circumstances may have contributed to increased ED presentations and delayed diagnosis for some patients.

We also utilized SDI and ADI to quantify socioeconomic disadvantage. ADI and SDI encompass various dimensions of disparity using ZIP code and census-level data, but the geographic resolution of these measurements may not account for socioeconomic heterogeneity within geographic regions. SDI and ADI are area-level measures; however, they only apply to patients with a ZIP code of residence. Therefore, homeless patients may not have a value associated with them, potentially excluding a particularly vulnerable population from these analyses. Accordingly, these indices should be interpreted as population-level proxies rather than individual-level measures of deprivation. Despite these limitations, ADI and SDI are well-validated, standardized tools that provide a generalizable framework for comparing the impact of socioeconomic disadvantage on patient outcomes.

Granular data clinical comorbidity data, such as tobacco use and diabetes, were not available in the database used for this study. Future studies may also benefit from the inclusion of a formal comorbidity index to provide additional context for these findings. Furthermore, disease severity at presentation can be impacted by other factors that this study does not measure, including health-seeking behavior patterns, referral systems, and regional resource limitations that lead patients to rely on the ED. The impact of these variables is difficult to quantify but should be considered when interpreting findings. Given the retrospective nature of the study, patient-reported outcomes could not be obtained. Nurick grade has been validated as a reliable measure of functional outcomes in DCM, but future studies should seek to include PROMs to strengthen functional outcomes data. Finally, our study is limited by variable follow-up duration, as this study did not impose a minimum postoperative follow-up requirement. Because outcomes data were collected at heterogeneous time points, comparisons may be imprecise and limit the generalizability of our findings. Despite these limitations, our results are consistent with prior studies linking sociodemographic disparities to worse outcomes, and provide a framework for guiding future research aimed at mitigating these disparities.