Postoperative Neurological Complications Following Revision Spine Surgery: A State Inpatient Database Analysis

Database

We used the State Inpatient Database (SID) from the state of California, which has information about “present on admission” (POA). The POA data give information about diagnoses present when a patient is admitted to the hospital. These data are defined as conditions present at the time of inpatient admission orders and can be used to differentiate between preexisting conditions and conditions that developed during an inpatient admission. All conditions that develop during an outpatient encounter, emergency room visit, observation period, or outpatient/same-day surgery leading up to an inpatient hospitalization are included in POA data. The POA data allow for more precise medical diagnoses billing and coding. Data were extracted from the SID from 2008 to 2011 because POA reporting before 2008 was not mandatory, and data after 2011 were not available. The SID files encompassed both insured and uninsured patients. They contained clinical and nonclinical variables, including primary and secondary diagnosis, procedure, admission and discharge status, and patient demographics as defined by the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM).

Patients Included

The algorithm used for data extraction and all ICD-9-CM codes can be found in the Supplemental Material available online. Two samples were created for analysis (all spinal procedures and revision spinal procedures). Initial analysis on all spinal procedures was done to find the predictors for neurological deficit and the impact of revision procedures on the incidence of new neurological deficits. Subsequent subanalysis on revision spinal procedures was then done to find the effect of neurological deficit on both new-onset and exacerbation of preexisting morbidity. On the basis of previously published studies reporting the incidence of neurological deficit in revision spine surgeries,^7,8 we decided to conduct our analysis with all revision procedures grouped together as opposed to separated by individual procedure type. Within each sample, demographic data were gathered about all patients, regardless of whether they developed postoperative neurological deficits. The total number of patients who developed postoperative spinal deficits following revision procedures was then subdivided into causative procedures: revision anterior cervical discectomy and fusion, revision posterior cervical fusion, and revision thoracolumbar fusion. These data were then used to calculate the total incidence of postoperative neurological deficits following all spinal procedures and following revision spine procedures only.

In-hospital morbidity was also evaluated and subdivided by age, length of hospital stay, van Walraven patient comorbidity scores (VWS), race, gender, admission source, household income, and payment source. In general, morbidity is defined as any departure, subjective or objective, from a state of physiological or psychological well-being. In our case, morbidity is defined as development of cardiac, respiratory, gastrointestinal, renal, infection, wound, and/or stroke complications or length of stay greater than 14 days.

Statistical Analysis

Dataset construction and analysis according to our inclusion criteria were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC). We created 2 datasets for analysis: the first included all spinal surgical procedures, and the second was limited to revision procedures. After running a macro, a set of options in Microsoft Excel that can automate tasks, to calculate VWS (an Elixhauser comorbidity index quantifies disease burden in patients), we ran an initial bivariate analysis using χ² for categorical variables and t test for continuous variables. Continuous data were presented as means and standard deviations, and categorical data were presented as percentages. Final variables included in the multivariable models were based on initial statistical significance in the bivariate analysis with a .05 level of significance, clinical plausibility, and effect on model C statistics. After testing for multicollinearity, we ran several logistic models to calculate adjusted Wald odds ratios and 95% confidence intervals for predictors of neurological deficits, morbidity, and in-hospital mortality. The mortality model failed to produce any meaningful numbers because the number of patients who developed neurological deficits and died in the hospital was small; thus, we did not report the numbers. We opted to use California's SID starting in 2008 for the present on admission variable reporting to provide an improved accuracy given that postoperative complications scanning became mandatory in late 2007. Any cases fewer than 10 were replaced by an asterisk in accordance with Healthcare Cost and Utilization Project regulations to limit possible patient identification. Statistical significance was defined as a P value < .05.

Patient Characteristics and Bivariate Analysis of Predictors of Neurological Deficit in All Spinal Procedures

In our analysis of 130 868 spinal procedures, 69 622 patients (53.2%) were women. A total of 96 975 patients (74.10%) were white; 19 606 patients (14.98%) were Hispanic; 6676 patients (5.10%) were African American; 5238 patients (4.0%) were Asian or Pacific Islander; 67 patients (0.05%) were Native American; and 2306 patients (1.76%) were of another race. A total of 740 patients (0.57%) with a mean age of 59.7 years developed a neurological deficit; however, 130 128 patients (99.43%) with a mean age of 56.4 years did not develop a neurological deficit. There were 376 patients (0.29%) who died during the hospital stay. Of patients who developed neurological deficits, 16 patients (2.16%) died during the hospital stay.

In the bivariate analysis of all spinal procedures shown in Table 1, there were 7645 (5.84%) revision procedures. Among these patients, 67 patients (0.88%) developed a postoperative neurological deficit with an unadjusted odds ratio of 1.61 (95% CI, 1.25–2.07, P < .05). A total of 489 patients (0.44%) with a VWS less than 5 developed a neurological deficit. There were 213 patients (1.29%) with a VWS of 5 to 14 who developed a neurological deficit with an odds ratio of 2.99 (95% CI, 2.54–3.52 , P < .05) when compared with those with a VWS less than 5; whereas, 38 patients (1.7%) of patients with a VWS greater than 14 developed a neurological deficit with an odds ratio of 3.98 (95% CI, 2.85–5.55, P <.05) when compared with those with a VWS less than 5.

Multivariable Analysis Predicting Neurological Deficit in All Spinal Procedures

The multivariable analysis of all spinal procedures presented in Table 2 shows that revision surgery was associated with increased odds of developing a neurological deficit, with an odds ratio of 1.56 (1.207–2.004). When using a VWS less than 5 as a reference, a VWS of 5 to 14 was associated with increased odds of developing a neurological deficit, with an odds ratio of 2.66 (95% CI, 2.25–3.14, P < .05). A VWS greater than 14 was also associated with increased odds of developing a neurological deficit, with an odds ratio of 3.45 (95% CI, 2.46–4.84, P < .05). With private insurance as the reference group, patients insured through Medicare were at increased odds of developing neurological deficits, with an odds ratio of 1.384 (95% CI, 1.126–1.700, P < .05). Patients insured through Medicaid were also at increased odds of developing neurological deficits, with an odds ratio of 1.421 (95% CI, 1.025–1.969, P < .05).

Revision Spinal Procedures Breakdown

The breakdown of revision surgeries in Table 3 shows that revision spinal procedures accounted for 5.84% (7645) of all spinal procedures. Within revision surgeries, 2677 procedures (35.02%) were revision of the lumbar and lumbosacral spine anterior column done with a posterior technique. In addition, 1240 procedures (16.22%) were revision of the lumbar or lumbosacral spine posterior column done with a posterior technique, and 1181 procedures (15.45%) were revision of the cervical spine anterior column done with an anterior technique.

Patient Characteristics of Revision Procedures

Patient characteristics of revision spine surgeries in Table 4 show that 4209 patients (55.50%) were women. The average age of patients undergoing revision spine surgery was 56.36 years. Within all revision procedures, the incidence of neurological deficit was 0.88%. Revision of the dorsal and dorsolumbar spine, anterior column, done with an anterior technique had a neurological deficit incidence of 2.50%. Revision of dorsal and dorsolumbar spine, posterior column, done with a posterior technique had a neurological deficit incidence of 1.80%. There were 33 revision cases not otherwise specified in terms of spine segment and technique. The incidence of neurological deficit within this subgroup was 3.03%.

Impact of Neurological Deficit on Morbidity After Revision Procedures

Various patient characteristics were analyzed by bivariate analysis as predictors of morbidity following revision spinal procedures as shown in Table 4. Overall, 7645 patients were included in the analysis, with 6831 (89.35%) suffering no morbidity and 814 (10.65%) suffering morbidity. There were 67 patients who suffered from neurologic deficit following their revision procedures, with 39 of 67 (58.21%) suffering no morbidity and 28 of 67 (41.79%) suffering morbidity. The odds of suffering morbidity after sustaining neurological deficit was 6.20 (95% CI, 3.79–10.13, P < .05) when compared with those who had no neurological deficits following revision spinal procedure.

The morbid group had an average age of 58.81 years, whereas the nonmorbid group had an average age of 56.07 years. Average length of hospital stay was 13.27 days in the morbid group and 3.92 days in the nonmorbid group. Using those with a VWS of less than 5 as the reference group, the odds of morbidity was 4.09 (95% CI, 3.46–4.83, P < .05) in the group that scored 5 to 14 and 8.46 (95% CI, 5.78–12.38, P < .05) in the group that scored greater than 14. Using routine admission as the reference group, the odds of morbidity for emergency department admission was 4.20 (95% CI, 3.26–5.41, P < .05); for admission from another hospital, 6.21 (95% CI, 3.52–10.96, P < .05); and for admission from another health facility, 3.16 (95% CI, 1.94–5.15, P < .05). Using private insurance as the reference group, the odds of morbidity with Medicare was 1.67 (95% CI, 1.40–1.99, P < .05) and the odds of morbidity with Medicaid was 1.85 (95% CI, 1.33–2.56, P < .05).

Multivariable analysis of the same patient characteristics shown in Table 5 yielded slightly different results. The odds ratio for morbidity in those who sustained neurological deficit following revision procedure was 5.33 (95% CI, 3.15–9.04, P < .05). A VWS of 5 to 14 had 3.45 (95% CI, 2.90–4.11, P < .05) odds of morbidity, whereas a score greater than 14 had a 5.68 (95% CI, 3.80–8.49, P < .05) odds of morbidity. Emergency department admission had a 3.11 (95% CI, 2.37–4.09, P < .05) odds of morbidity, another hospital had a 3.76 (95% CI, 2.04–6.95 P < .05) odds of morbidity, and another health facility had a 2.46 (95% CI, 1.47–4.11, P < .05) odds of morbidity. With payer type, only Medicare had a significant odds of morbidity at 1.30 (95% CI, 1.09–1.56, P < .05).

Analysis of mortality was not included due to the very small number of patients who died following revision procedures. Thus, no significant data could be analyzed.