Predicting Clinical Outcomes Following Surgical Correction of Adult Spinal Deformity

Predicting Clinical Outcomes Following Surgical Correction of Adult Spinal Deformity Abstract BACKGROUND Deformity reconstruction surgery has been shown to improve quality of life (QOL) in cases of adult spinal deformity (ASD) but is associated with significant morbidity. OBJECTIVE To create a preoperative predictive nomogram to help risk-stratify patients and determine which would likely benefit from corrective surgery for ASD as measured by patient-reported health-related quality of life (HRQoL). METHODS All patients aged 25-yr and older with radiographic evidence of ASD and QOL data that underwent thoracolumbar fusion between 2008 and 2014 were identified. Demographic and clinical parameters were obtained. The EuroQol 5 dimensions questionnaire (EQ-5D) was used to measure HRQoL preoperatively and at 12-mo postoperative follow-up. Logistic regression of preoperative variables was used to create the prognostic nomogram. RESULTS Our sample included data from 191 patients. Fifty-one percent of patients experienced clinically relevant postoperative improvement in HRQoL. Seven variables were included in the final model: preoperative EQ-5D score, sex, preoperative diagnosis (degenerative, idiopathic, or iatrogenic), previous spinal surgical history, obesity, and a sex-by-obesity interaction term. Preoperative EQ-5D score independently predicted the outcome. Sex interacted with obesity: obese men were at disproportionately higher odds of improving than nonobese men, but obesity did not affect odds of the outcome among women. Model discrimination was good, with an optimism-adjusted c-statistic of 0.739. CONCLUSION The predictive nomogram that we developed using these data can improve preoperative risk counseling and patient selection for deformity correction surgery. Deformity, Nomogram, Outcomes, Scoliosis, Predictive modeling, Spinal Fusion, Quality of life ABBREVIATIONS ABBREVIATIONS ADS adult spinal deformity BMI body mass index CI confidence interval EMR electronic medical record EQ-5D EuroQol 5 dimensions questionnaire HRQoL health-related quality of life MCID minimum clinically important difference ODI Oswestry Disability Index OR odds ratios PRO patient-reported outcome QOL quality of life SRS Scoliosis Research Society Adult spinal deformity (ASD) refers to unnatural curvature of the spine. ASD can occur as a natural process of aging, as a secondary process to scoliosis or kyphosis, or can be a result of prior surgery.1-6 ASD may result in sagittal imbalance if the patient cannot compensate for the deformity with pelvic retroversion and/or knee flexion or through compensatory and reciprocal changes in the thoracic and cervical spine. This often leads to standing intolerance, pain, and disability.7 In select patients, deformity can also result in a reduction of the diameter of the spinal canal and neural foramina, resulting in debilitating spinal stenosis and radiculopathy.8,9 When conservative management fails, surgery to restore neutral sagittal alignment can improve quality of life (QOL) for ASD patients.2,5,10-13 The rate of surgical intervention for ASD in the United States has steadily increased over recent years.14 In a study of patients in the nationwide inpatient sample, the number of annual ASD surgeries increased from 9400 procedures in 2000 to 20 600 in 2010.15 Although QOL deficits are a key determinant in the decision to surgically treat ASD, few prior studies have identified patient level predictors of changes in QOL following surgery for ASD. As the number of patients undergoing surgery to treat ASD continues to increase, it is increasingly important to quantify the patient-level factors that are associated with surgical outcomes in ASD. Identifying these patient-level factors can improve surgical planning, further enrich patient counseling, and can potentially reduce surgery-associated complication and mortality. In the present study, we sought to better understand the preoperative predictors of patient-reported health-related quality-of-life (HRQoL) outcomes for ASD surgery. We built a predictive model for patients with ASD that underwent spinal fusion with improvements in QOL metrics as the primary outcome endpoint. We hypothesized that preoperative demographic and comorbid characteristics were associated with outcomes following surgical intervention for ASD. METHODS The Knowledge Program is a patient-reported outcome (PRO) assessment tool that is integrated with the electronic medical record (EMR) at our institution. It prospectively compiles self-assessment data taken at all outpatient visits. The PRO metrics used in this investigation have been validated to assess patient HRQoL after spinal surgery.16 Specifically, the EuroQol five dimensions questionnaire (version 3L-US; EQ-5D) is designed to measure and standardize health outcomes. Its scores are transformed into an index value known as the quality-adjusted life year, a number between 0 and 1, where 1 is equivalent to 1 yr in perfect health and 0 is equivalent to patient death.17 The minimum clinically important difference (MCID) for the EQ-5D at 1 yr selected was 0.1.18 Patient Population EMRs were used to identify adult patients with the radiographic evidence of thoracolumbar spinal deformity that underwent anterior, posterior, or circumferential lumbar and thoracolumbar fusion of greater than 3 vertebral levels between 2008 and 2014. Included in the sample were patients 25 yr of age or over with idiopathic, degenerative, or iatrogenic diagnoses of ASD. Any patients with spinal fracture, malignancy, or infection were excluded in order to eliminate the confounding effect of multiple surgical interventions. We excluded patients who were missing EQ-5D scores within 6 mo prior to surgery or between 9 and 15 mo following surgery. IRB approval at our institution was obtained prior to all data collection. As patient-care was not affected and there was no risk involved for patients, the requirement for patient consent was waived. Patients’ sex, race, date of birth, weight, height, medical comorbidities, medications, history of spinal trauma or surgical intervention, smoking status/history, length of fusion, type of fusion, operating surgeon, length of hospital stay, operative time, blood loss, in-hospital complications noted and patient complications following surgery were obtained. Preoperative sagittal parameters including pelvic incidence, pelvic tilt, and lumbar lordosis were recorded from relevant imaging studies. Statistical Methods Descriptive statistics were calculated for the entire sample and for the sample stratified by HRQoL group: improvement or no improvement. Group differences were examined using independent-samples T-tests or Mann–Whitney U-tests for continuous variables and Fisher's exact tests for categorical variables. For each patient, the preoperative EQ-5D score closest to the surgery date and the postoperative score closest to 365 d following the surgery date were selected for analysis. Patients excluded due to missing EQ-5D data but who otherwise met inclusion criteria were compared with patients included in the final sample to check for sampling bias. Independent-samples T-tests, Mann–Whitney U-tests, and Fisher's exact tests were used to determine group differences. Sixteen candidate predictors were considered for the model: age at procedure; sex; race; body mass index (BMI) at procedure; preoperative diagnosis; preoperative EQ-5D score; history of spinal trauma; history of spinal surgery; operative approach (anterior vs posterior); number of levels fused; and comorbid dyslipidemia, obesity, hypertension, diabetes mellitus 1 and 2, and tobacco use disorder. We used best subsets selection to determine which predictors would be included in the final logistic regression model. This technique uses a branch-and-bound algorithm; best models for each possible number of predictor variables were generated.19,20 Each of the best models were compared and the final model containing the best predictors of the outcome was selected based on the highest global chi-square statistic. Interactions between sex and other covariates were tested when calibrating the model and interaction terms were included as warranted if calibration improved. Linearity of each continuous predictor was assessed, and transformation or piecewise linear relationships were included if warranted. The discriminatory ability of the final model was evaluated using the concordance (c) statistic. Bootstrap resampling was used to validate the model.21,22 Five hundred alternate datasets of the same size as the original were generated by resampling the original sample with replacement. The final model was applied to each alternate dataset to calculate Harrell's optimism to adjust the c-statistic accordingly.23 Model calibration was examined by plotting observed probabilities against predicted probabilities from the final model, as well as by the Hosmer–Lemeshow Goodness-of-Fit Test,24 which aims to produce a P-value greater than .05 to indicate that the model fit is adequate. A nomogram was created based on the final logistic regression model. Sensitivity Analysis To evaluate the effect of varying MCID thresholds on classification of improvement or nonimprovement, sensitivity analysis was performed. The final model developed with an EQ-5D improvement cutoff of .1 was applied to cutoffs of improvement greater than zero, and greater than or equal to each .05, .15, and .2. RESULTS We identified 407 patients aged 25-yr and older with a diagnosis of ASD who underwent lumbar fusions of greater than three levels between 2008 and 2014. Two hundred sixteen patients were excluded due to missing EQ-5D data, leaving 191 patients in the final sample, all of whom had complete data on all variables of interest. Those excluded from the study due to incomplete EQ-5D data had a significantly lower proportion of anterior surgeries than those included in the sample. No other statistically significant differences were observed; see Table 1. Ninety-eight patients (51%) met the MCID for clinically relevant improvement in EQ-5D scores following surgery. Descriptive statistics for the entire sample and the sample stratified by outcome group can be found in Table 2. The average time between preoperative and postoperative EQ-5D assessment was 414.74 d (standard deviation [SD] = 56.75); patients who experienced clinically significant improvement following surgery completed their postoperative EQ-5D assessment on average 12 d later than those who did not improve (P = .04). Preoperative EQ-5D score differed between outcome groups, such that those who experienced postsurgical Eq-5D improvement had median preoperative EQ-5D scores 0.2 units lower than those who did not improve (P < .0001). TABLE 1. Characteristics of Those Included in the Sample and Those Excluded due to Missing Eq5D Mean (SD) or n(%) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) aFisher's exact tests and independent samples t-tests conducted on nonmissing data. Diagnosis, comorbidity information, and BMI missing at less than 5% for the excluded group; all other data complete. bSignificant group difference. View Large TABLE 1. Characteristics of Those Included in the Sample and Those Excluded due to Missing Eq5D Mean (SD) or n(%) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) aFisher's exact tests and independent samples t-tests conducted on nonmissing data. Diagnosis, comorbidity information, and BMI missing at less than 5% for the excluded group; all other data complete. bSignificant group difference. View Large TABLE 2. Descriptive Statistics and Comparisons Between Outcome Groups Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Mean (SD) or n (%). aMedian (IQR). bSignificant group difference. View Large TABLE 2. Descriptive Statistics and Comparisons Between Outcome Groups Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Mean (SD) or n (%). aMedian (IQR). bSignificant group difference. View Large Ultimately, too few patients had history of spinal trauma, tobacco use disorder, or diabetes mellitus 1, and these candidate predictors were eliminated from our model. Following best-subsets model selection, seven variables were included in the final model: preoperative EQ-5D score, sex, obesity, diagnosis (idiopathic or iatrogenic ASD both relative to degenerative ASD), age, surgical history, and a sex by obesity interaction term (global χ2 = 45.98) Model odds ratios (OR) are displayed in Table 3. TABLE 3. Model P-Values OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) * Female sex and obesity are the two variables interacting. View Large TABLE 3. Model P-Values OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) * Female sex and obesity are the two variables interacting. View Large Preoperative EQ-5D score independently predicted the outcome, such that every 0.1 unit lower (worse) preoperative EQ-5D yielded a 79% increase in odds of postoperative improvement (OR 1.79, 95% confidence interval [CI] 1.46-2.19; P < .0001). Sex interacted with obesity: among men, those who were obese were at disproportionately higher odds of improvement, when compared to nonobese men (OR 4.5, 95% CI 1.18-17.11). However, among women, obesity did not affect odds of improvement. To further illustrate, 56% of nonobese women improved, and a similar 50% of obese women improved. In contrast, 32% of nonobese men improved, but 63% of obese men improved. The optimism-adjusted c-statistic for the model was 0.739. Given 2 patients, 1 from each outcome group, there is a 73.9% chance that our model would correctly predict which of the 2 experienced improvement in HRQoL following surgery. In detail, the model applied to the bootstrapped samples yielded unadjusted c-statistics ranging from .718 to .782, with an average c-statistic of .764 (SD = .011, median .766, Q1 .758, Q2 .772). Model calibration was acceptable (Hosmer–Lemeshow test P = .33). Figure 1 shows a modest departure in the bias-corrected calibration curve when compared to the ideal line. The calibration curve displays that model-generated predicted probabilities of improvement in HRQoL between .45 and .6 are very slightly lower than actual probabilities of improvement. Additionally, model predictions above .65 are systematically higher than actual probabilities. Table 4 displays results from the sensitivity analysis. Model fit was acceptable for all but the cutoff of an MCID value of 0.2. Discriminative ability was similar among the models, with the cutoff of 0.1 providing the best raw and optimism-adjusted discrimination. FIGURE 1. View largeDownload slide Calibration plot for the model predicting EQ-5D improvement of .1 or greater at 1-yr postsurgical follow-up. Model prediction of improvement is plotted against actual outcome; perfect calibration is indicated by a 45° line. The dotted line (apparent) indicates calibration when the model is applied to the training dataset, and the solid line (bias-corrected) indicates calibration when the model is applied to the bootstrap validation datasets. FIGURE 1. View largeDownload slide Calibration plot for the model predicting EQ-5D improvement of .1 or greater at 1-yr postsurgical follow-up. Model prediction of improvement is plotted against actual outcome; perfect calibration is indicated by a 45° line. The dotted line (apparent) indicates calibration when the model is applied to the training dataset, and the solid line (bias-corrected) indicates calibration when the model is applied to the bootstrap validation datasets. TABLE 4. Model Performance Applied to Alternative MCID Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 aHosmer and Lemeshow test, P-value < .05 indicates unacceptable fit. View Large TABLE 4. Model Performance Applied to Alternative MCID Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 aHosmer and Lemeshow test, P-value < .05 indicates unacceptable fit. View Large The nomogram created from the model is presented in Figure 2. To use the nomogram, locate a patient's position on the scale associated with each predictor. The top axis displays prognostic points; connect the position on each variable axis with the number of points corresponding to that position. Continue for each variable to determine the patient's total points, and then use the same method with the bottom 2 axes to establish a patient's probability of improvement in EQ5D following surgery based on his or her number of points. For example, a typical spinal deformity patient with a preoperative EQ5D score of .3 (62 points), 61 yr of age (18 points), not obese and female (29 points), with a degenerative diagnosis (13 points) and a history of prior surgeries (7 points) would have 129 total points and a corresponding approximate 65% probability of improving in HRQoL following surgery. FIGURE 2. View largeDownload slide The prognostic nomogram. The nomogram can be used to predict the preoperative probability of postoperative improvement. Points (top line) are assigned for each corresponding predictor and totaled and correlated to a probability of improvement (bottom 2 lines). FIGURE 2. View largeDownload slide The prognostic nomogram. The nomogram can be used to predict the preoperative probability of postoperative improvement. Points (top line) are assigned for each corresponding predictor and totaled and correlated to a probability of improvement (bottom 2 lines). Of the 191 patients included in the study sample, 32 (16.75%) were missing data for pelvic incidence/lumbar lordosis (PI-LL) mismatch (32 were missing pelvic incidence and 27 of those also were missing lumbar lordosis). The final model was built on the sample of 191 patients with complete cases as previously described. To evaluate the influence of PI-LL on the outcome, first the full model and then the full model plus PI/LL were applied to the subgroup of patients with available radiographic data, and the c-statistics were compared. The inclusion of PI-LL did not improve the accuracy of the predictive model to a degree necessary to justify the reduction of the overall sample size (unadjusted c-statistic = .733 without PI-LL included and .736 with PI-LL included). There were no demographic or etiological differences between those with available and those with missing radiographic data. DISCUSSION In this study, we created a prognostic nomogram with clinical utility in predicting whether a patient would experience clinically significant improvement in health-related QOL following ASD surgery (Figure 2). Guidelines and tools to assist clinicians in optimizing PRO following surgical intervention, particularly for ASD, are limited. While predictors of outcomes for spinal deformity have been extensively documented, heterogeneity among study design and clinical endpoints imparts limited clinical utility to these findings.11,14,15,25-32 Thus, a simple tool that synthesizes multiple clinical characteristics and can predict clinical and QOL outcomes is integral to improving patient care. These tools provide objective data that allow clinicians to develop individualized and optimized clinical recommendations. The clinical utility of the predictive modeling has been well established in the field of spine care.33-39 Scheer and colleagues40 recently used a decision tree algorithm to predict the development of intraoperative and perioperative complications based off preoperative demographic, clinical, and radiographic characteristics in a multicenter retrospective review of 557 patients with ASD. The authors reported a model accuracy of 87% and an area under the receiver operative characteristic of 0.89. Publication of these models represents a growing trend in the spine care community to develop tools to aid with clinical decision making. In the context of this new trend, the selection of appropriate outcome endpoints is paramount to ensure clinical utility of such tools. PRO scales represent instruments that reconcile objective surgical outcomes with patient satisfaction, functional status, and QOL.41-43 Using PROs as the primary outcome endpoint, clinicians can approach decision making in the context of the maximum benefit of surgical treatment. The present study is the first study to develop a prognostic nomogram to predict postoperative outcomes following ASD surgery as measured by patient-reported QOL. It is imperative to recognize the differences in goals and interpretation when considering prediction models in contrast to the more commonly encountered associative regression analyses. Prediction models aim to achieve the highest discriminatory ability possible.20,22,23 That is, the best model is one that is able to classify patients into the correct outcome group with the highest degree of accuracy, regardless of the significance or magnitude of each predictor variable's individual relationship with the outcome. Associative regression analyses focus on group-level effects, estimating the group-level average effect of one independent variable on the dependent variable, while controlling for covariates as necessary. In contrast, prediction models operate on the individual level, assessing the probability of an outcome based on an individual's unique characteristics; the outcome is conditional on all the predictors.20 Therefore, the group-level effects of each model predictor on the outcome should be interpreted with caution and bear the caveat that the primary interpretation of the model should be its usefulness as a whole. The findings discussed below deserve further validation and exploration in future study. Predictors of Improvement Following Surgery for ASD The utility of preoperative HRQoL scores in predicting improvement following surgery for ASD is well documented.44-46 In a retrospective review of a multicenter database of 421 patients with ASD, Scheer et al44 found that increased preoperative pain severity was associated with larger improvements in pain severity postoperatively. In a similarly designed review of 365 patients, Bakhsheshian et al45 reported that patients with Oswestry Disability Index (ODI) scores greater than 40 (high disability) experienced similar 2-yr improvement in HRQoL measures to patients with scores less than 40 (low disability). Only high disability patients, however, achieved significant improvement Scoliosis Research Society (SRS) Mental scale scores and a significantly higher rate of reaching MCIDs in along that scale. The results of the present study indicate similar findings. Patients with lower baseline EQ-5D scores were more likely to experience a clinically significant improvement in scores following surgery. Importantly, the EQ-5D is a qualitative measure of health-related QOL over 1 yr and is not a direct measure of pain.18,47,48 Patients with lower index scores may have greater opportunity for improvement and may be more sensitive to postoperative benefits and restoration of function than those with more mild preoperative QOL deficits.44 While sex was significantly associated with increased odds of improvement in EQ-5D following surgery, we noted a significant interaction with obesity. This is the first study of HRQoL data among ASD patients to note an interaction of 2 demographic variables. However, several prior studies have quantified the associations between sex and postoperative outcomes and comorbidities, such as obesity, and postoperative outcomes without considering potential interactions of comorbidities and sex. For example, Worley et al29 found increased odds of morbidity (OR 1.18) but decreased odds of mortality (OR 0.30) among females undergoing ASD surgery. In a retrospective review of 241 ASD patients, Soroceanu et al26 found that obese patients (BMI ≥ 30) experienced significant improvement in ODI, SF-36, and SRS scores although improvement was statistically significantly less than that achieved in patients who were not obese. The interaction between sex and obesity noted in this current study could provide new insight on how these preoperative characteristics may be affecting outcomes in ASD surgery. Previous studies may not have adequately characterized this relationship, and as such further study is warranted to verify the relationship and fully understand this phenomenon and assist with patient optimization for surgery. Limitations and Future Directions There are several limitations to this study. First, this study is a retrospective review of a prospectively collected database. Therefore, the inherent limitations of a retrospective review apply to these data. Preoperative and postoperative radiographic parameters were not directly analyzed in this study, and therefore we cannot comment specifically on PRO measures as they relate to the magnitude of deformity and/or attainment of adequate alignment objectives after surgical correction. The primary goal of this study was to look holistically at patients undergoing ASD surgery and develop a nomogram based on preoperative factors (excluding radiographic factors) to help determine the likelihood of a favorable outcome. While this is the largest study of its kind to be performed at a single institution, the sample size is relatively small in comparison to similar models based on large multicenter databases. Our sample size was limited significantly due to incomplete outcomes data at 1-yr follow-up. The results of our post-hoc analysis revealed that no significant bias was introduced into our sample with this loss, lending confidence that this loss had no significant effect on our results; however, it is impossible to fully measure this effect and it would be irresponsible to assume that this loss of data has no effect at all on these findings. Larger, prospective, and longitudinal studies are warranted, namely those including analysis of preoperative and postoperative spinopelvic radiographic parameters. Specific investigation into the individual effects found in our study is desired. In particular, the interaction found when building our model may be unique to our institution and should be confirmed. Finally, the model would further benefit from external validation in a separate patient population. Overall, the bootstrap calibration curve was good, particularly considering the size of the sample. It must be emphasized that model predictions between .45 and .6 are very slightly pessimistic, and model predictions above .65 are optimistic. CONCLUSION ASD often leads to severe pain and disability. Surgical correction of deformity has been shown to significantly improve PRO and QOL. The present study developed a predictive nomogram to help predict improvement in HRQoL following surgery. This nomogram and the results of this study can help in patient education and patient selection for surgery in the management of ASD. Further prospective studies are needed to validate these findings. Disclosure The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Notes This work was presented in part at the Lumbar Spine Research Society Annual Meeting, April 6-7, 2017, Chicago, Illinois. The methods and results of this study were presented on March 17, 2018 at the AANS/CNS Joint Section on Disorders of the Spine and Peripheral Nerves Annual Meeting in Orlando, Florida. REFERENCES 1. Good CR , Auerbach JD , O’Leary PT , Schuler TC . Adult spine deformity . Curr Rev Musculoskelet Med . 2011 ; 4 ( 4 ): 159 – 167 . Google Scholar CrossRef Search ADS PubMed 2. Berven SH , Deviren V , Smith JA , Emami A , Hu SS , Bradford DS . Management of fixed sagittal plane deformity: results of the transpedicular wedge resection osteotomy . Spine . 2001 ; 26 ( 18 ): 2036 – 2043 . Google Scholar CrossRef Search ADS PubMed 3. Takemitsu Y , Atsuta Y , Kamo Y et al. Operative treatment of lumbar degenerative kyphosis . In: Kazuo Y , Keiro O , Yoshiharu T , eds. Lumbar Fusion and Stabilization . Japan : Springer ; 1993 : 150 – 159 . Available at: http://link.springer.com/chapter/10.1007/978-4-431-68234-9_17. Accessed April 11, 2015 . Google Scholar CrossRef Search ADS 4. Lee C-S , Lee C-K , Kim Y-T , Hong Y-M , Yoo J-H . Dynamic sagittal imbalance of the spine in degenerative flat back: significance of pelvic tilt in surgical treatment . Spine . 2001 ; 26 ( 18 ): 2029 – 2035 . Google Scholar CrossRef Search ADS PubMed 5. Farcy J-PC , Schwab FJ . Management of flatback and related kyphotic decompensation syndromes . Spine . 1997 ; 22 ( 20 ): 2452 – 2457 . Google Scholar CrossRef Search ADS PubMed 6. Panjabi MM . Clinical spinal instability and low back pain . J Electromyogr Kinesiol . 2003 ; 13 ( 4 ): 371 – 379 . Google Scholar CrossRef Search ADS PubMed 7. Jang J-S , Lee S-H , Min J-H , Maeng DH . Changes in sagittal alignment after restoration of lower lumbar lordosis in patients with degenerative flat back syndrome . J Neurosurg Spine . 2007 ; 7 ( 4 ): 387 – 392 . Google Scholar CrossRef Search ADS PubMed 8. Yabuki S , Fukumori N , Takegami M et al. Prevalence of lumbar spinal stenosis, using the diagnostic support tool, and correlated factors in Japan: a population-based study . J Orthop Sci . 2013 ; 18 ( 6 ): 893 – 900 . Google Scholar CrossRef Search ADS PubMed 9. Sinikallio S , Aalto T , Airaksinen O et al. Depression and associated factors in patients with lumbar spinal stenosis . Disabil Rehabil . 2006 ; 28 ( 7 ): 415 – 422 . Google Scholar CrossRef Search ADS PubMed 10. Takemitsu Y , Harada Y , Iwahara T , Miyamoto M , Miyatake Y . Lumbar degenerative kyphosis: clinical, radiological and epidemiological studies . Spine . 1988 ; 13 ( 11 ): 1317 – 1326 . Google Scholar CrossRef Search ADS PubMed 11. Booth KC , Bridwell KH , Lenke LG , Baldus CR , Blanke KM . Complications and predictive factors for the successful treatment of flatback deformity (fixed sagittal imbalance) . Spine . 1999 ; 24 ( 16 ): 1712 . Google Scholar CrossRef Search ADS PubMed 12. Uddin OM , Haque R , Sugrue PA et al. Cost minimization in treatment of adult degenerative scoliosis . J Neurosurg Spine . 2015 ; 23 ( 6 ): 1 – 9 . Google Scholar CrossRef Search ADS PubMed 13. Bess S , Boachie-Adjei O , Burton D et al. Pain and disability determine treatment modality for older patients with adult scoliosis, while deformity guides treatment for younger patients . Spine . 2009 ; 34 ( 20 ): 2186 – 2190 . Google Scholar CrossRef Search ADS PubMed 14. Turner JD , Walker CT , Mundis GM Jr , Kakarla UK . Health burden of adult spinal deformity compared with other chronic diseases . World Neurosurg . 2015 ; 84 ( 4 ): 876 – 877 . Google Scholar CrossRef Search ADS PubMed 15. McCarthy I , O’Brien M , Ames C et al. Incremental cost-effectiveness of adult spinal deformity surgery: observed quality-adjusted life years with surgery compared with predicted quality-adjusted life years without surgery . Neurosurg Focus . 2014 ; 36 ( 5 ): E3 . Google Scholar CrossRef Search ADS PubMed 16. Tharin S , Mayer E , Krishnaney A . Lumbar microdiscectomy and lumbar decompression improve functional outcomes and depression scores . Evid-Based Spine-Care J . 2012 ; 3 ( 4 ): 65 – 66 . Google Scholar PubMed 17. Shaw JW , Johnson JA , Coons SJ . US valuation of the EQ-5D Health States . Med Care . 2005 ; 43 ( 3 ): 203 – 220 . Google Scholar CrossRef Search ADS PubMed 18. Parker SL , Godil SS , Shau DN , Mendenhall SK , McGirt MJ . Assessment of the minimum clinically important difference in pain, disability, and quality of life after anterior cervical discectomy and fusion . J Neurosurg Spine . 2013 ; 18 ( 2 ): 154 – 160 . Google Scholar CrossRef Search ADS PubMed 19. Furnival GM , Wilson RW . Regressions by leaps and bounds . Technometrics . 1974 ; 16 ( 4 ): 499 – 511 . Google Scholar CrossRef Search ADS 20. Steyerberg EW . Clinical Prediction Models . New York, NY : Springer ; 2009 . Available at: http://link.springer.com/10.1007/978-0-387-77244-8 . Google Scholar CrossRef Search ADS 21. Efron B . Estimating the error rate of a prediction rule: improvement on cross-validation . J Am Statist Assoc . 1983 ; 78 ( 382 ): 316 – 331 . Google Scholar CrossRef Search ADS 22. Harrell F . Regression Modeling Strategies—With Applications to Linear Models, Logistic Regression, and Survival Analysis . New York, NY : Springer ; 2001 . 23. Harrell FE , Lee KL , Mark DB . Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors . Statist. Med . 1996 ; 15 ( 4 ): 361 – 387 . Google Scholar CrossRef Search ADS 24. Jr DWH , Lemeshow S . Applied Logistic Regression . Hoboken, New Jersey : John Wiley & Sons ; 2004 . 25. Verma K , Lonner B , Dean L , Vecchione D , Kean K . Predictors of postoperative infection in spinal deformity surgery—which curves are at greatest risk? Bull Hosp Jt Dis (2013) . 2013 ; 71 ( 4 ): 257 – 264 . Google Scholar PubMed 26. Soroceanu A , Burton DC , Diebo BG et al. Impact of obesity on complications, infection, and patient-reported outcomes in adult spinal deformity surgery . J Neurosurg Spine . 2015 : 1 – 9 , doi:10.3171/2015.3.SPINE14743 . 27. Pugely AJ , Martin CT , Gao Y , Mendoza-Lattes S . Causes and risk factors for 30-day unplanned readmissions after lumbar spine surgery . Spine . 2014 ; 39 ( 9 ): 761 – 768 . Google Scholar CrossRef Search ADS PubMed 28. Aoki Y , Nakajima A , Takahashi H et al. Influence of pelvic incidence-lumbar lordosis mismatch on surgical outcomes of short-segment transforaminal lumbar interbody fusion . BMC Musculoskelet Disord . 2015 ; 16 : 213 . doi:10.1186/s12891-015-0676-1 Google Scholar CrossRef Search ADS PubMed 29. Worley N , Marascalchi B , Jalai CM et al. Predictors of inpatient morbidity and mortality in adult spinal deformity surgery . Eur Spine J Off Publ Eur Spine Soc Eur Spinal Deform Soc Eur Sect Cerv Spine Res Soc . 2016 ; 25 ( 3 ): 819 – 827 . Google Scholar CrossRef Search ADS 30. Smith JS , Shaffrey CI , Glassman SD et al. Risk-benefit assessment of surgery for adult scoliosis: an analysis based on patient age . Spine . 2011 ; 36 ( 10 ): 817 – 824 . Google Scholar CrossRef Search ADS PubMed 31. Benz RJ , Ibrahim ZG , Afshar P , Garfin SR . Predicting complications in elderly patients undergoing lumbar decompression . Clin Orthop Relat Res . 2001 ; 384 : 116 – 121 . 32. Daubs MD , Lenke LG , Cheh G , Stobbs G , Bridwell KH . Adult spinal deformity surgery: complications and outcomes in patients over age 60 . Spine . 2007 ; 32 ( 20 ): 2238 – 2244 . Google Scholar CrossRef Search ADS PubMed 33. Lubelski D , Derakhshan A , Nowacki AS et al. Predicting C5 palsy via the use of preoperative anatomic measurements . Spine J . 2014 ; 14 ( 9 ): 1895 – 1901 . Google Scholar CrossRef Search ADS PubMed 34. Lubelski D , Thompson NR , Agrawal B et al. Prediction of quality of life improvements in patients with lumbar stenosis following use of membrane stabilizing agents . Clin Neurol Neurosurg . 2015 ; 139 : 234 – 240 . Google Scholar CrossRef Search ADS PubMed 35. Royuela A , Kovacs FM , Campillo C , Casamitjana M , Muriel A , Abraira V . Predicting outcomes of neuroreflexotherapy in patients with subacute or chronic neck or low back pain . Spine J . 2014 ; 14 ( 8 ): 1588 – 1600 . Google Scholar CrossRef Search ADS PubMed 36. Paulino Pereira NR , Janssen SJ , van Dijk E et al. Development of a prognostic survival algorithm for patients with metastatic spine disease . J Bone Joint Surg . 2016 ; 98 ( 21 ): 1767 – 1776 . Google Scholar CrossRef Search ADS PubMed 37. Osorio JA , Scheer JK , Ames CP . Predictive modeling of complications . Curr Rev Musculoskelet Med . 2016 ; 9 ( 3 ): 333 – 337 . Google Scholar CrossRef Search ADS PubMed 38. Azimi P , Benzel EC , Shahzadi S , Azhari S , Mohammadi HR . Use of artificial neural networks to predict surgical satisfaction in patients with lumbar spinal canal stenosis . J Neurosurg Spine . 2014 ; 20 ( 3 ): 300 – 305 . Google Scholar CrossRef Search ADS PubMed 39. Daubs MD , Hung M , Adams JR et al. Clinical predictors of psychological distress in patients presenting for evaluation of a spinal disorder . Spine J . 2014 ; 14 ( 9 ): 1978 – 1983 . Google Scholar CrossRef Search ADS PubMed 40. Scheer JK , Smith JS , Schwab F et al. Development of a preoperative predictive model for major complications following adult spinal deformity surgery . J Neurosurg Spine . 2017 : 26 ( 6 ): 736 – 743 . Google Scholar CrossRef Search ADS PubMed 41. Teles AR , Khoshhal KI , Falavigna A . Why and how should we measure outcomes in spine surgery? J Taibah Univ Med Sci . 2016 ; 11 ( 2 ): 91 – 97 . 42. Chapman JR , Norvell DC , Hermsmeyer JTB et al. Evaluating common outcomes for measuring treatment success for chronic low back pain . Spine . 2011 ; 36 ( 21 Suppl ): S54 – S68 . Google Scholar CrossRef Search ADS PubMed 43. Djurasovic M , Glassman SD , Dimar JR , Howard JM , Bratcher KR , Carreon LY . Does fusion status correlate with patient outcomes in lumbar spinal fusion? Spine . 2011 ; 36 ( 5 ): 404 – 409 . Google Scholar CrossRef Search ADS PubMed 44. Scheer JK , Smith JS , Clark AJ et al. Comprehensive study of back and leg pain improvements after adult spinal deformity surgery: analysis of 421 patients with 2-year follow-up and of the impact of the surgery on treatment satisfaction . J Neurosurg Spine . 2015 ; 22 ( 5 ): 540 – 553 . Google Scholar CrossRef Search ADS PubMed 45. Bakhsheshian J , Scheer JK , Gum JL et al. Comparison of structural disease burden to health-related quality of life scores in 264 adult spinal deformity patients with 2-year follow-up . Clin Spine Surg . 2017 ; 30 ( 2 ): E124 – E131 . Google Scholar CrossRef Search ADS PubMed 46. Smith JS , Shaffrey CI , Glassman SD et al. Clinical and radiographic parameters that distinguish between the best and worst outcomes of scoliosis surgery for adults . Eur Spine J . 2013 ; 22 ( 2 ): 402 – 410 . Google Scholar CrossRef Search ADS PubMed 47. Coretti S , Ruggeri M , McNamee P . The minimum clinically important difference for EQ-5D index: a critical review . Expert Rev Pharmacoecon Outcomes Res . 2014 ; 14 ( 2 ): 221 – 233 . Google Scholar CrossRef Search ADS PubMed 48. EuroQol Group . EuroQol—a new facility for the measurement of health-related quality of life . Health Policy . 1990 ; 16 ( 3 ): 199 – 208 . CrossRef Search ADS PubMed Acknowledgment We thank Nicolas Thompson, Department of Quantitative Health Sciences, Lerner Research Institute, Cleveland Clinic, for his support and guidance. Copyright © 2018 by the Congress of Neurological Surgeons This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Neurosurgery Oxford University Press

Predicting Clinical Outcomes Following Surgical Correction of Adult Spinal Deformity

Loading next page...
 
/lp/ou_press/predicting-clinical-outcomes-following-surgical-correction-of-adult-MMCz7yEUAP
Publisher
Congress of Neurological Surgeons
Copyright
Copyright © 2018 by the Congress of Neurological Surgeons
ISSN
0148-396X
eISSN
1524-4040
D.O.I.
10.1093/neuros/nyy190
Publisher site
See Article on Publisher Site

Abstract

Abstract BACKGROUND Deformity reconstruction surgery has been shown to improve quality of life (QOL) in cases of adult spinal deformity (ASD) but is associated with significant morbidity. OBJECTIVE To create a preoperative predictive nomogram to help risk-stratify patients and determine which would likely benefit from corrective surgery for ASD as measured by patient-reported health-related quality of life (HRQoL). METHODS All patients aged 25-yr and older with radiographic evidence of ASD and QOL data that underwent thoracolumbar fusion between 2008 and 2014 were identified. Demographic and clinical parameters were obtained. The EuroQol 5 dimensions questionnaire (EQ-5D) was used to measure HRQoL preoperatively and at 12-mo postoperative follow-up. Logistic regression of preoperative variables was used to create the prognostic nomogram. RESULTS Our sample included data from 191 patients. Fifty-one percent of patients experienced clinically relevant postoperative improvement in HRQoL. Seven variables were included in the final model: preoperative EQ-5D score, sex, preoperative diagnosis (degenerative, idiopathic, or iatrogenic), previous spinal surgical history, obesity, and a sex-by-obesity interaction term. Preoperative EQ-5D score independently predicted the outcome. Sex interacted with obesity: obese men were at disproportionately higher odds of improving than nonobese men, but obesity did not affect odds of the outcome among women. Model discrimination was good, with an optimism-adjusted c-statistic of 0.739. CONCLUSION The predictive nomogram that we developed using these data can improve preoperative risk counseling and patient selection for deformity correction surgery. Deformity, Nomogram, Outcomes, Scoliosis, Predictive modeling, Spinal Fusion, Quality of life ABBREVIATIONS ABBREVIATIONS ADS adult spinal deformity BMI body mass index CI confidence interval EMR electronic medical record EQ-5D EuroQol 5 dimensions questionnaire HRQoL health-related quality of life MCID minimum clinically important difference ODI Oswestry Disability Index OR odds ratios PRO patient-reported outcome QOL quality of life SRS Scoliosis Research Society Adult spinal deformity (ASD) refers to unnatural curvature of the spine. ASD can occur as a natural process of aging, as a secondary process to scoliosis or kyphosis, or can be a result of prior surgery.1-6 ASD may result in sagittal imbalance if the patient cannot compensate for the deformity with pelvic retroversion and/or knee flexion or through compensatory and reciprocal changes in the thoracic and cervical spine. This often leads to standing intolerance, pain, and disability.7 In select patients, deformity can also result in a reduction of the diameter of the spinal canal and neural foramina, resulting in debilitating spinal stenosis and radiculopathy.8,9 When conservative management fails, surgery to restore neutral sagittal alignment can improve quality of life (QOL) for ASD patients.2,5,10-13 The rate of surgical intervention for ASD in the United States has steadily increased over recent years.14 In a study of patients in the nationwide inpatient sample, the number of annual ASD surgeries increased from 9400 procedures in 2000 to 20 600 in 2010.15 Although QOL deficits are a key determinant in the decision to surgically treat ASD, few prior studies have identified patient level predictors of changes in QOL following surgery for ASD. As the number of patients undergoing surgery to treat ASD continues to increase, it is increasingly important to quantify the patient-level factors that are associated with surgical outcomes in ASD. Identifying these patient-level factors can improve surgical planning, further enrich patient counseling, and can potentially reduce surgery-associated complication and mortality. In the present study, we sought to better understand the preoperative predictors of patient-reported health-related quality-of-life (HRQoL) outcomes for ASD surgery. We built a predictive model for patients with ASD that underwent spinal fusion with improvements in QOL metrics as the primary outcome endpoint. We hypothesized that preoperative demographic and comorbid characteristics were associated with outcomes following surgical intervention for ASD. METHODS The Knowledge Program is a patient-reported outcome (PRO) assessment tool that is integrated with the electronic medical record (EMR) at our institution. It prospectively compiles self-assessment data taken at all outpatient visits. The PRO metrics used in this investigation have been validated to assess patient HRQoL after spinal surgery.16 Specifically, the EuroQol five dimensions questionnaire (version 3L-US; EQ-5D) is designed to measure and standardize health outcomes. Its scores are transformed into an index value known as the quality-adjusted life year, a number between 0 and 1, where 1 is equivalent to 1 yr in perfect health and 0 is equivalent to patient death.17 The minimum clinically important difference (MCID) for the EQ-5D at 1 yr selected was 0.1.18 Patient Population EMRs were used to identify adult patients with the radiographic evidence of thoracolumbar spinal deformity that underwent anterior, posterior, or circumferential lumbar and thoracolumbar fusion of greater than 3 vertebral levels between 2008 and 2014. Included in the sample were patients 25 yr of age or over with idiopathic, degenerative, or iatrogenic diagnoses of ASD. Any patients with spinal fracture, malignancy, or infection were excluded in order to eliminate the confounding effect of multiple surgical interventions. We excluded patients who were missing EQ-5D scores within 6 mo prior to surgery or between 9 and 15 mo following surgery. IRB approval at our institution was obtained prior to all data collection. As patient-care was not affected and there was no risk involved for patients, the requirement for patient consent was waived. Patients’ sex, race, date of birth, weight, height, medical comorbidities, medications, history of spinal trauma or surgical intervention, smoking status/history, length of fusion, type of fusion, operating surgeon, length of hospital stay, operative time, blood loss, in-hospital complications noted and patient complications following surgery were obtained. Preoperative sagittal parameters including pelvic incidence, pelvic tilt, and lumbar lordosis were recorded from relevant imaging studies. Statistical Methods Descriptive statistics were calculated for the entire sample and for the sample stratified by HRQoL group: improvement or no improvement. Group differences were examined using independent-samples T-tests or Mann–Whitney U-tests for continuous variables and Fisher's exact tests for categorical variables. For each patient, the preoperative EQ-5D score closest to the surgery date and the postoperative score closest to 365 d following the surgery date were selected for analysis. Patients excluded due to missing EQ-5D data but who otherwise met inclusion criteria were compared with patients included in the final sample to check for sampling bias. Independent-samples T-tests, Mann–Whitney U-tests, and Fisher's exact tests were used to determine group differences. Sixteen candidate predictors were considered for the model: age at procedure; sex; race; body mass index (BMI) at procedure; preoperative diagnosis; preoperative EQ-5D score; history of spinal trauma; history of spinal surgery; operative approach (anterior vs posterior); number of levels fused; and comorbid dyslipidemia, obesity, hypertension, diabetes mellitus 1 and 2, and tobacco use disorder. We used best subsets selection to determine which predictors would be included in the final logistic regression model. This technique uses a branch-and-bound algorithm; best models for each possible number of predictor variables were generated.19,20 Each of the best models were compared and the final model containing the best predictors of the outcome was selected based on the highest global chi-square statistic. Interactions between sex and other covariates were tested when calibrating the model and interaction terms were included as warranted if calibration improved. Linearity of each continuous predictor was assessed, and transformation or piecewise linear relationships were included if warranted. The discriminatory ability of the final model was evaluated using the concordance (c) statistic. Bootstrap resampling was used to validate the model.21,22 Five hundred alternate datasets of the same size as the original were generated by resampling the original sample with replacement. The final model was applied to each alternate dataset to calculate Harrell's optimism to adjust the c-statistic accordingly.23 Model calibration was examined by plotting observed probabilities against predicted probabilities from the final model, as well as by the Hosmer–Lemeshow Goodness-of-Fit Test,24 which aims to produce a P-value greater than .05 to indicate that the model fit is adequate. A nomogram was created based on the final logistic regression model. Sensitivity Analysis To evaluate the effect of varying MCID thresholds on classification of improvement or nonimprovement, sensitivity analysis was performed. The final model developed with an EQ-5D improvement cutoff of .1 was applied to cutoffs of improvement greater than zero, and greater than or equal to each .05, .15, and .2. RESULTS We identified 407 patients aged 25-yr and older with a diagnosis of ASD who underwent lumbar fusions of greater than three levels between 2008 and 2014. Two hundred sixteen patients were excluded due to missing EQ-5D data, leaving 191 patients in the final sample, all of whom had complete data on all variables of interest. Those excluded from the study due to incomplete EQ-5D data had a significantly lower proportion of anterior surgeries than those included in the sample. No other statistically significant differences were observed; see Table 1. Ninety-eight patients (51%) met the MCID for clinically relevant improvement in EQ-5D scores following surgery. Descriptive statistics for the entire sample and the sample stratified by outcome group can be found in Table 2. The average time between preoperative and postoperative EQ-5D assessment was 414.74 d (standard deviation [SD] = 56.75); patients who experienced clinically significant improvement following surgery completed their postoperative EQ-5D assessment on average 12 d later than those who did not improve (P = .04). Preoperative EQ-5D score differed between outcome groups, such that those who experienced postsurgical Eq-5D improvement had median preoperative EQ-5D scores 0.2 units lower than those who did not improve (P < .0001). TABLE 1. Characteristics of Those Included in the Sample and Those Excluded due to Missing Eq5D Mean (SD) or n(%) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) aFisher's exact tests and independent samples t-tests conducted on nonmissing data. Diagnosis, comorbidity information, and BMI missing at less than 5% for the excluded group; all other data complete. bSignificant group difference. View Large TABLE 1. Characteristics of Those Included in the Sample and Those Excluded due to Missing Eq5D Mean (SD) or n(%) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) Included Excluded n = 191 n = 216 Pa Age 64.3 (10.97) 63.79 (11.8) .652 Male 50 (26.18%) 60 (27.78%) .738 Female 141 (73.82%) 156 (72.22%) White 178 (93.19%) 206 (96.37) .393 Nonwhite 13 (6.81%) 10 (4.63) BMI 28.77 (6.45) 28.35 (6.065) .501 Diagnosis .111  Degenerative 118 (61.78%) 129 (62.62)  Idiopathic 36 (18.85%) 25 (12.14)  Iatrogenic 37 (19.37%) 52 (25.24) History of prior surgery .371  No 106 (49.07)  Yes 110 (50.93) Dyslipidemia .077  No 114 (59.69%) 145 (68.4)  Yes 77 (40.31%) 67 (31.6) Hypertension .148  No 63 (32.98%) 85 (40.09)  Yes 128 (67.02%) 127 (59.91) Obesity 1.0  No 122 (63.87%) 131 (64.22)  Yes 69 (36.13%) 73 (35.78) Diabetes II .889  No 162 (84.82%) 181 (85.38)  Yes 29 (15.18%) 31 (14.62) PI/LL mismatch 23.94 (19.72)(n = 159) 22.66 (17.57)(n = 133) .562 Estimated blood lossa 1400 (1500)(n = 176) 1200 (1700)(n = 168) .412 Length of staya 6 (3)(n = 189) 6 (3)(n = 205) .377 Fusion lengtha 8 (4) 7 (4) .612 Fusion type .0002b  Anterior 45 (23.56) 34 (15.74)  Posterior 143 (74.87) 169(78.24)  Anterior + posterior 3 (1.57) 13(6.02) aFisher's exact tests and independent samples t-tests conducted on nonmissing data. Diagnosis, comorbidity information, and BMI missing at less than 5% for the excluded group; all other data complete. bSignificant group difference. View Large TABLE 2. Descriptive Statistics and Comparisons Between Outcome Groups Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Mean (SD) or n (%). aMedian (IQR). bSignificant group difference. View Large TABLE 2. Descriptive Statistics and Comparisons Between Outcome Groups Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Total sample n = 191 No clinically relevant improvement in Eq5D n = 93 Clinically relevant improvement in Eq5D n = 98 P Age 64.3 (10.97) 64.12 (11.1) 64.47 (10.9) .826 Male 50 (26.18) 28 (30.11) 22 (22.45) .252 Female 141 (73.82) 65 (69.89) 76 (77.55) White 178 (93.19) 87 (93.55) 91 (92.86) 1.0 Nonwhite 13 (6.81) 6 (6.45) 7 (7.14) BMI 28.77 (6.45) 28.53 (6.33) 29.0 (6.59) .614 Dx  Degenerative 118 (61.78) 55 (59.14) 63 (64.29) ref  Idiopathic 36 (18.85) 15 (16.13) 21 (21.43) .184  Iatrogenic 37 (19.37) 23 (24.73) 14 (14.29) .06 History of prior surgery .468  No 103 (53.93) 53 (56.99) 50 (51.02)  Yes 88 (46.07 40 (43.01) 48 (48.98) Dyslipidemia .555  No 114 (59.69) 58 (62.37) 56 (57.14)  Yes 77 (40.31) 35 (37.63) 42 (42.86) Hypertension .759  No 63 (32.98) 31 (34.41) 31 (31.63)  Yes 128 (67.02) 61 (65.59) 67 (68.37) Obesity .654  No 122 (63.87) 61 (65.59) 61 (62.24)  Yes 69 (36.13) 32 (34.41) 37 (37.76) Diabetes II .546  No 162 (84.82) 77 (82.8) 85 (86.73)  Yes 29 (15.18) 16 (17.2) 13 (13.27) Baseline Eq5Da .397 (.358) .597 (.311) .397 (.138) < .0001b Preoperative assessmenta (days prior) 40 (65) 40 (70) 38.5 (53) .781 Postoperative assessment (days post) 362.53 (38.16) 356.7 (40.15) 368.1 (35.5) .039b Days between assessments 414.74 (56.75) 409.6 (54.9) 419.6 (58.32) .226 PI/LL mismatch 23.94 (19.72) 25.31 (21.15) 22.62 (18.28) .392 (n = 159) (n = 78) (n = 81) Estimated blood loss a 1400 (1500) 1325 (1425) 1445 (1600) .52 (n = 176) (n = 84) (n = 92) Length of stay a 6 (3) 6 (3) 6 (3) .532 (n = 189) (n = 92) (n = 97) Fusion lengtha 8 (4) 7 (4) 8 (4) .242 Fusion type .48  Anterior 45 (23.56) 19 (20.43) 26 (26.53)  Posterior 143 (74.87) 73(78.49) 70 (71.43)  Ant + post 3 (1.57) 1 (1.08) 2 (2.04) Mean (SD) or n (%). aMedian (IQR). bSignificant group difference. View Large Ultimately, too few patients had history of spinal trauma, tobacco use disorder, or diabetes mellitus 1, and these candidate predictors were eliminated from our model. Following best-subsets model selection, seven variables were included in the final model: preoperative EQ-5D score, sex, obesity, diagnosis (idiopathic or iatrogenic ASD both relative to degenerative ASD), age, surgical history, and a sex by obesity interaction term (global χ2 = 45.98) Model odds ratios (OR) are displayed in Table 3. TABLE 3. Model P-Values OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) * Female sex and obesity are the two variables interacting. View Large TABLE 3. Model P-Values OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) OR (95% Confidence Variable P interval) Baseline Eq5D <.0001 1.79 (1.46-2.19) Female*obese .022 Female sex .009 Obesity .028  Among males 4.50 (1.18-17.11)  Among females .73 (.328-1.62) Idiopathic diagnosis .026 3.59 (1.16-11.08) Degenerative diagnosis .221 1.78 (.71-4.5) Age .162 1.02 (.99-1.06) Surgical history .4 1.37 (.66-2.829) * Female sex and obesity are the two variables interacting. View Large Preoperative EQ-5D score independently predicted the outcome, such that every 0.1 unit lower (worse) preoperative EQ-5D yielded a 79% increase in odds of postoperative improvement (OR 1.79, 95% confidence interval [CI] 1.46-2.19; P < .0001). Sex interacted with obesity: among men, those who were obese were at disproportionately higher odds of improvement, when compared to nonobese men (OR 4.5, 95% CI 1.18-17.11). However, among women, obesity did not affect odds of improvement. To further illustrate, 56% of nonobese women improved, and a similar 50% of obese women improved. In contrast, 32% of nonobese men improved, but 63% of obese men improved. The optimism-adjusted c-statistic for the model was 0.739. Given 2 patients, 1 from each outcome group, there is a 73.9% chance that our model would correctly predict which of the 2 experienced improvement in HRQoL following surgery. In detail, the model applied to the bootstrapped samples yielded unadjusted c-statistics ranging from .718 to .782, with an average c-statistic of .764 (SD = .011, median .766, Q1 .758, Q2 .772). Model calibration was acceptable (Hosmer–Lemeshow test P = .33). Figure 1 shows a modest departure in the bias-corrected calibration curve when compared to the ideal line. The calibration curve displays that model-generated predicted probabilities of improvement in HRQoL between .45 and .6 are very slightly lower than actual probabilities of improvement. Additionally, model predictions above .65 are systematically higher than actual probabilities. Table 4 displays results from the sensitivity analysis. Model fit was acceptable for all but the cutoff of an MCID value of 0.2. Discriminative ability was similar among the models, with the cutoff of 0.1 providing the best raw and optimism-adjusted discrimination. FIGURE 1. View largeDownload slide Calibration plot for the model predicting EQ-5D improvement of .1 or greater at 1-yr postsurgical follow-up. Model prediction of improvement is plotted against actual outcome; perfect calibration is indicated by a 45° line. The dotted line (apparent) indicates calibration when the model is applied to the training dataset, and the solid line (bias-corrected) indicates calibration when the model is applied to the bootstrap validation datasets. FIGURE 1. View largeDownload slide Calibration plot for the model predicting EQ-5D improvement of .1 or greater at 1-yr postsurgical follow-up. Model prediction of improvement is plotted against actual outcome; perfect calibration is indicated by a 45° line. The dotted line (apparent) indicates calibration when the model is applied to the training dataset, and the solid line (bias-corrected) indicates calibration when the model is applied to the bootstrap validation datasets. TABLE 4. Model Performance Applied to Alternative MCID Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 aHosmer and Lemeshow test, P-value < .05 indicates unacceptable fit. View Large TABLE 4. Model Performance Applied to Alternative MCID Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 Eq-5D cutoff for Number classified Unadjusted Optimism-adjusted Goodness-of-fit improvement as improved c-statistic c-statistic testa 0 127 .7 .643 .293 .05 115 .736 .691 .277 .1 98 .777 .739 .33 .15 90 .755 .716 .244 .2 84 .765 .73 .003 aHosmer and Lemeshow test, P-value < .05 indicates unacceptable fit. View Large The nomogram created from the model is presented in Figure 2. To use the nomogram, locate a patient's position on the scale associated with each predictor. The top axis displays prognostic points; connect the position on each variable axis with the number of points corresponding to that position. Continue for each variable to determine the patient's total points, and then use the same method with the bottom 2 axes to establish a patient's probability of improvement in EQ5D following surgery based on his or her number of points. For example, a typical spinal deformity patient with a preoperative EQ5D score of .3 (62 points), 61 yr of age (18 points), not obese and female (29 points), with a degenerative diagnosis (13 points) and a history of prior surgeries (7 points) would have 129 total points and a corresponding approximate 65% probability of improving in HRQoL following surgery. FIGURE 2. View largeDownload slide The prognostic nomogram. The nomogram can be used to predict the preoperative probability of postoperative improvement. Points (top line) are assigned for each corresponding predictor and totaled and correlated to a probability of improvement (bottom 2 lines). FIGURE 2. View largeDownload slide The prognostic nomogram. The nomogram can be used to predict the preoperative probability of postoperative improvement. Points (top line) are assigned for each corresponding predictor and totaled and correlated to a probability of improvement (bottom 2 lines). Of the 191 patients included in the study sample, 32 (16.75%) were missing data for pelvic incidence/lumbar lordosis (PI-LL) mismatch (32 were missing pelvic incidence and 27 of those also were missing lumbar lordosis). The final model was built on the sample of 191 patients with complete cases as previously described. To evaluate the influence of PI-LL on the outcome, first the full model and then the full model plus PI/LL were applied to the subgroup of patients with available radiographic data, and the c-statistics were compared. The inclusion of PI-LL did not improve the accuracy of the predictive model to a degree necessary to justify the reduction of the overall sample size (unadjusted c-statistic = .733 without PI-LL included and .736 with PI-LL included). There were no demographic or etiological differences between those with available and those with missing radiographic data. DISCUSSION In this study, we created a prognostic nomogram with clinical utility in predicting whether a patient would experience clinically significant improvement in health-related QOL following ASD surgery (Figure 2). Guidelines and tools to assist clinicians in optimizing PRO following surgical intervention, particularly for ASD, are limited. While predictors of outcomes for spinal deformity have been extensively documented, heterogeneity among study design and clinical endpoints imparts limited clinical utility to these findings.11,14,15,25-32 Thus, a simple tool that synthesizes multiple clinical characteristics and can predict clinical and QOL outcomes is integral to improving patient care. These tools provide objective data that allow clinicians to develop individualized and optimized clinical recommendations. The clinical utility of the predictive modeling has been well established in the field of spine care.33-39 Scheer and colleagues40 recently used a decision tree algorithm to predict the development of intraoperative and perioperative complications based off preoperative demographic, clinical, and radiographic characteristics in a multicenter retrospective review of 557 patients with ASD. The authors reported a model accuracy of 87% and an area under the receiver operative characteristic of 0.89. Publication of these models represents a growing trend in the spine care community to develop tools to aid with clinical decision making. In the context of this new trend, the selection of appropriate outcome endpoints is paramount to ensure clinical utility of such tools. PRO scales represent instruments that reconcile objective surgical outcomes with patient satisfaction, functional status, and QOL.41-43 Using PROs as the primary outcome endpoint, clinicians can approach decision making in the context of the maximum benefit of surgical treatment. The present study is the first study to develop a prognostic nomogram to predict postoperative outcomes following ASD surgery as measured by patient-reported QOL. It is imperative to recognize the differences in goals and interpretation when considering prediction models in contrast to the more commonly encountered associative regression analyses. Prediction models aim to achieve the highest discriminatory ability possible.20,22,23 That is, the best model is one that is able to classify patients into the correct outcome group with the highest degree of accuracy, regardless of the significance or magnitude of each predictor variable's individual relationship with the outcome. Associative regression analyses focus on group-level effects, estimating the group-level average effect of one independent variable on the dependent variable, while controlling for covariates as necessary. In contrast, prediction models operate on the individual level, assessing the probability of an outcome based on an individual's unique characteristics; the outcome is conditional on all the predictors.20 Therefore, the group-level effects of each model predictor on the outcome should be interpreted with caution and bear the caveat that the primary interpretation of the model should be its usefulness as a whole. The findings discussed below deserve further validation and exploration in future study. Predictors of Improvement Following Surgery for ASD The utility of preoperative HRQoL scores in predicting improvement following surgery for ASD is well documented.44-46 In a retrospective review of a multicenter database of 421 patients with ASD, Scheer et al44 found that increased preoperative pain severity was associated with larger improvements in pain severity postoperatively. In a similarly designed review of 365 patients, Bakhsheshian et al45 reported that patients with Oswestry Disability Index (ODI) scores greater than 40 (high disability) experienced similar 2-yr improvement in HRQoL measures to patients with scores less than 40 (low disability). Only high disability patients, however, achieved significant improvement Scoliosis Research Society (SRS) Mental scale scores and a significantly higher rate of reaching MCIDs in along that scale. The results of the present study indicate similar findings. Patients with lower baseline EQ-5D scores were more likely to experience a clinically significant improvement in scores following surgery. Importantly, the EQ-5D is a qualitative measure of health-related QOL over 1 yr and is not a direct measure of pain.18,47,48 Patients with lower index scores may have greater opportunity for improvement and may be more sensitive to postoperative benefits and restoration of function than those with more mild preoperative QOL deficits.44 While sex was significantly associated with increased odds of improvement in EQ-5D following surgery, we noted a significant interaction with obesity. This is the first study of HRQoL data among ASD patients to note an interaction of 2 demographic variables. However, several prior studies have quantified the associations between sex and postoperative outcomes and comorbidities, such as obesity, and postoperative outcomes without considering potential interactions of comorbidities and sex. For example, Worley et al29 found increased odds of morbidity (OR 1.18) but decreased odds of mortality (OR 0.30) among females undergoing ASD surgery. In a retrospective review of 241 ASD patients, Soroceanu et al26 found that obese patients (BMI ≥ 30) experienced significant improvement in ODI, SF-36, and SRS scores although improvement was statistically significantly less than that achieved in patients who were not obese. The interaction between sex and obesity noted in this current study could provide new insight on how these preoperative characteristics may be affecting outcomes in ASD surgery. Previous studies may not have adequately characterized this relationship, and as such further study is warranted to verify the relationship and fully understand this phenomenon and assist with patient optimization for surgery. Limitations and Future Directions There are several limitations to this study. First, this study is a retrospective review of a prospectively collected database. Therefore, the inherent limitations of a retrospective review apply to these data. Preoperative and postoperative radiographic parameters were not directly analyzed in this study, and therefore we cannot comment specifically on PRO measures as they relate to the magnitude of deformity and/or attainment of adequate alignment objectives after surgical correction. The primary goal of this study was to look holistically at patients undergoing ASD surgery and develop a nomogram based on preoperative factors (excluding radiographic factors) to help determine the likelihood of a favorable outcome. While this is the largest study of its kind to be performed at a single institution, the sample size is relatively small in comparison to similar models based on large multicenter databases. Our sample size was limited significantly due to incomplete outcomes data at 1-yr follow-up. The results of our post-hoc analysis revealed that no significant bias was introduced into our sample with this loss, lending confidence that this loss had no significant effect on our results; however, it is impossible to fully measure this effect and it would be irresponsible to assume that this loss of data has no effect at all on these findings. Larger, prospective, and longitudinal studies are warranted, namely those including analysis of preoperative and postoperative spinopelvic radiographic parameters. Specific investigation into the individual effects found in our study is desired. In particular, the interaction found when building our model may be unique to our institution and should be confirmed. Finally, the model would further benefit from external validation in a separate patient population. Overall, the bootstrap calibration curve was good, particularly considering the size of the sample. It must be emphasized that model predictions between .45 and .6 are very slightly pessimistic, and model predictions above .65 are optimistic. CONCLUSION ASD often leads to severe pain and disability. Surgical correction of deformity has been shown to significantly improve PRO and QOL. The present study developed a predictive nomogram to help predict improvement in HRQoL following surgery. This nomogram and the results of this study can help in patient education and patient selection for surgery in the management of ASD. Further prospective studies are needed to validate these findings. Disclosure The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article. Notes This work was presented in part at the Lumbar Spine Research Society Annual Meeting, April 6-7, 2017, Chicago, Illinois. The methods and results of this study were presented on March 17, 2018 at the AANS/CNS Joint Section on Disorders of the Spine and Peripheral Nerves Annual Meeting in Orlando, Florida. REFERENCES 1. Good CR , Auerbach JD , O’Leary PT , Schuler TC . Adult spine deformity . Curr Rev Musculoskelet Med . 2011 ; 4 ( 4 ): 159 – 167 . Google Scholar CrossRef Search ADS PubMed 2. Berven SH , Deviren V , Smith JA , Emami A , Hu SS , Bradford DS . Management of fixed sagittal plane deformity: results of the transpedicular wedge resection osteotomy . Spine . 2001 ; 26 ( 18 ): 2036 – 2043 . Google Scholar CrossRef Search ADS PubMed 3. Takemitsu Y , Atsuta Y , Kamo Y et al. Operative treatment of lumbar degenerative kyphosis . In: Kazuo Y , Keiro O , Yoshiharu T , eds. Lumbar Fusion and Stabilization . Japan : Springer ; 1993 : 150 – 159 . Available at: http://link.springer.com/chapter/10.1007/978-4-431-68234-9_17. Accessed April 11, 2015 . Google Scholar CrossRef Search ADS 4. Lee C-S , Lee C-K , Kim Y-T , Hong Y-M , Yoo J-H . Dynamic sagittal imbalance of the spine in degenerative flat back: significance of pelvic tilt in surgical treatment . Spine . 2001 ; 26 ( 18 ): 2029 – 2035 . Google Scholar CrossRef Search ADS PubMed 5. Farcy J-PC , Schwab FJ . Management of flatback and related kyphotic decompensation syndromes . Spine . 1997 ; 22 ( 20 ): 2452 – 2457 . Google Scholar CrossRef Search ADS PubMed 6. Panjabi MM . Clinical spinal instability and low back pain . J Electromyogr Kinesiol . 2003 ; 13 ( 4 ): 371 – 379 . Google Scholar CrossRef Search ADS PubMed 7. Jang J-S , Lee S-H , Min J-H , Maeng DH . Changes in sagittal alignment after restoration of lower lumbar lordosis in patients with degenerative flat back syndrome . J Neurosurg Spine . 2007 ; 7 ( 4 ): 387 – 392 . Google Scholar CrossRef Search ADS PubMed 8. Yabuki S , Fukumori N , Takegami M et al. Prevalence of lumbar spinal stenosis, using the diagnostic support tool, and correlated factors in Japan: a population-based study . J Orthop Sci . 2013 ; 18 ( 6 ): 893 – 900 . Google Scholar CrossRef Search ADS PubMed 9. Sinikallio S , Aalto T , Airaksinen O et al. Depression and associated factors in patients with lumbar spinal stenosis . Disabil Rehabil . 2006 ; 28 ( 7 ): 415 – 422 . Google Scholar CrossRef Search ADS PubMed 10. Takemitsu Y , Harada Y , Iwahara T , Miyamoto M , Miyatake Y . Lumbar degenerative kyphosis: clinical, radiological and epidemiological studies . Spine . 1988 ; 13 ( 11 ): 1317 – 1326 . Google Scholar CrossRef Search ADS PubMed 11. Booth KC , Bridwell KH , Lenke LG , Baldus CR , Blanke KM . Complications and predictive factors for the successful treatment of flatback deformity (fixed sagittal imbalance) . Spine . 1999 ; 24 ( 16 ): 1712 . Google Scholar CrossRef Search ADS PubMed 12. Uddin OM , Haque R , Sugrue PA et al. Cost minimization in treatment of adult degenerative scoliosis . J Neurosurg Spine . 2015 ; 23 ( 6 ): 1 – 9 . Google Scholar CrossRef Search ADS PubMed 13. Bess S , Boachie-Adjei O , Burton D et al. Pain and disability determine treatment modality for older patients with adult scoliosis, while deformity guides treatment for younger patients . Spine . 2009 ; 34 ( 20 ): 2186 – 2190 . Google Scholar CrossRef Search ADS PubMed 14. Turner JD , Walker CT , Mundis GM Jr , Kakarla UK . Health burden of adult spinal deformity compared with other chronic diseases . World Neurosurg . 2015 ; 84 ( 4 ): 876 – 877 . Google Scholar CrossRef Search ADS PubMed 15. McCarthy I , O’Brien M , Ames C et al. Incremental cost-effectiveness of adult spinal deformity surgery: observed quality-adjusted life years with surgery compared with predicted quality-adjusted life years without surgery . Neurosurg Focus . 2014 ; 36 ( 5 ): E3 . Google Scholar CrossRef Search ADS PubMed 16. Tharin S , Mayer E , Krishnaney A . Lumbar microdiscectomy and lumbar decompression improve functional outcomes and depression scores . Evid-Based Spine-Care J . 2012 ; 3 ( 4 ): 65 – 66 . Google Scholar PubMed 17. Shaw JW , Johnson JA , Coons SJ . US valuation of the EQ-5D Health States . Med Care . 2005 ; 43 ( 3 ): 203 – 220 . Google Scholar CrossRef Search ADS PubMed 18. Parker SL , Godil SS , Shau DN , Mendenhall SK , McGirt MJ . Assessment of the minimum clinically important difference in pain, disability, and quality of life after anterior cervical discectomy and fusion . J Neurosurg Spine . 2013 ; 18 ( 2 ): 154 – 160 . Google Scholar CrossRef Search ADS PubMed 19. Furnival GM , Wilson RW . Regressions by leaps and bounds . Technometrics . 1974 ; 16 ( 4 ): 499 – 511 . Google Scholar CrossRef Search ADS 20. Steyerberg EW . Clinical Prediction Models . New York, NY : Springer ; 2009 . Available at: http://link.springer.com/10.1007/978-0-387-77244-8 . Google Scholar CrossRef Search ADS 21. Efron B . Estimating the error rate of a prediction rule: improvement on cross-validation . J Am Statist Assoc . 1983 ; 78 ( 382 ): 316 – 331 . Google Scholar CrossRef Search ADS 22. Harrell F . Regression Modeling Strategies—With Applications to Linear Models, Logistic Regression, and Survival Analysis . New York, NY : Springer ; 2001 . 23. Harrell FE , Lee KL , Mark DB . Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors . Statist. Med . 1996 ; 15 ( 4 ): 361 – 387 . Google Scholar CrossRef Search ADS 24. Jr DWH , Lemeshow S . Applied Logistic Regression . Hoboken, New Jersey : John Wiley & Sons ; 2004 . 25. Verma K , Lonner B , Dean L , Vecchione D , Kean K . Predictors of postoperative infection in spinal deformity surgery—which curves are at greatest risk? Bull Hosp Jt Dis (2013) . 2013 ; 71 ( 4 ): 257 – 264 . Google Scholar PubMed 26. Soroceanu A , Burton DC , Diebo BG et al. Impact of obesity on complications, infection, and patient-reported outcomes in adult spinal deformity surgery . J Neurosurg Spine . 2015 : 1 – 9 , doi:10.3171/2015.3.SPINE14743 . 27. Pugely AJ , Martin CT , Gao Y , Mendoza-Lattes S . Causes and risk factors for 30-day unplanned readmissions after lumbar spine surgery . Spine . 2014 ; 39 ( 9 ): 761 – 768 . Google Scholar CrossRef Search ADS PubMed 28. Aoki Y , Nakajima A , Takahashi H et al. Influence of pelvic incidence-lumbar lordosis mismatch on surgical outcomes of short-segment transforaminal lumbar interbody fusion . BMC Musculoskelet Disord . 2015 ; 16 : 213 . doi:10.1186/s12891-015-0676-1 Google Scholar CrossRef Search ADS PubMed 29. Worley N , Marascalchi B , Jalai CM et al. Predictors of inpatient morbidity and mortality in adult spinal deformity surgery . Eur Spine J Off Publ Eur Spine Soc Eur Spinal Deform Soc Eur Sect Cerv Spine Res Soc . 2016 ; 25 ( 3 ): 819 – 827 . Google Scholar CrossRef Search ADS 30. Smith JS , Shaffrey CI , Glassman SD et al. Risk-benefit assessment of surgery for adult scoliosis: an analysis based on patient age . Spine . 2011 ; 36 ( 10 ): 817 – 824 . Google Scholar CrossRef Search ADS PubMed 31. Benz RJ , Ibrahim ZG , Afshar P , Garfin SR . Predicting complications in elderly patients undergoing lumbar decompression . Clin Orthop Relat Res . 2001 ; 384 : 116 – 121 . 32. Daubs MD , Lenke LG , Cheh G , Stobbs G , Bridwell KH . Adult spinal deformity surgery: complications and outcomes in patients over age 60 . Spine . 2007 ; 32 ( 20 ): 2238 – 2244 . Google Scholar CrossRef Search ADS PubMed 33. Lubelski D , Derakhshan A , Nowacki AS et al. Predicting C5 palsy via the use of preoperative anatomic measurements . Spine J . 2014 ; 14 ( 9 ): 1895 – 1901 . Google Scholar CrossRef Search ADS PubMed 34. Lubelski D , Thompson NR , Agrawal B et al. Prediction of quality of life improvements in patients with lumbar stenosis following use of membrane stabilizing agents . Clin Neurol Neurosurg . 2015 ; 139 : 234 – 240 . Google Scholar CrossRef Search ADS PubMed 35. Royuela A , Kovacs FM , Campillo C , Casamitjana M , Muriel A , Abraira V . Predicting outcomes of neuroreflexotherapy in patients with subacute or chronic neck or low back pain . Spine J . 2014 ; 14 ( 8 ): 1588 – 1600 . Google Scholar CrossRef Search ADS PubMed 36. Paulino Pereira NR , Janssen SJ , van Dijk E et al. Development of a prognostic survival algorithm for patients with metastatic spine disease . J Bone Joint Surg . 2016 ; 98 ( 21 ): 1767 – 1776 . Google Scholar CrossRef Search ADS PubMed 37. Osorio JA , Scheer JK , Ames CP . Predictive modeling of complications . Curr Rev Musculoskelet Med . 2016 ; 9 ( 3 ): 333 – 337 . Google Scholar CrossRef Search ADS PubMed 38. Azimi P , Benzel EC , Shahzadi S , Azhari S , Mohammadi HR . Use of artificial neural networks to predict surgical satisfaction in patients with lumbar spinal canal stenosis . J Neurosurg Spine . 2014 ; 20 ( 3 ): 300 – 305 . Google Scholar CrossRef Search ADS PubMed 39. Daubs MD , Hung M , Adams JR et al. Clinical predictors of psychological distress in patients presenting for evaluation of a spinal disorder . Spine J . 2014 ; 14 ( 9 ): 1978 – 1983 . Google Scholar CrossRef Search ADS PubMed 40. Scheer JK , Smith JS , Schwab F et al. Development of a preoperative predictive model for major complications following adult spinal deformity surgery . J Neurosurg Spine . 2017 : 26 ( 6 ): 736 – 743 . Google Scholar CrossRef Search ADS PubMed 41. Teles AR , Khoshhal KI , Falavigna A . Why and how should we measure outcomes in spine surgery? J Taibah Univ Med Sci . 2016 ; 11 ( 2 ): 91 – 97 . 42. Chapman JR , Norvell DC , Hermsmeyer JTB et al. Evaluating common outcomes for measuring treatment success for chronic low back pain . Spine . 2011 ; 36 ( 21 Suppl ): S54 – S68 . Google Scholar CrossRef Search ADS PubMed 43. Djurasovic M , Glassman SD , Dimar JR , Howard JM , Bratcher KR , Carreon LY . Does fusion status correlate with patient outcomes in lumbar spinal fusion? Spine . 2011 ; 36 ( 5 ): 404 – 409 . Google Scholar CrossRef Search ADS PubMed 44. Scheer JK , Smith JS , Clark AJ et al. Comprehensive study of back and leg pain improvements after adult spinal deformity surgery: analysis of 421 patients with 2-year follow-up and of the impact of the surgery on treatment satisfaction . J Neurosurg Spine . 2015 ; 22 ( 5 ): 540 – 553 . Google Scholar CrossRef Search ADS PubMed 45. Bakhsheshian J , Scheer JK , Gum JL et al. Comparison of structural disease burden to health-related quality of life scores in 264 adult spinal deformity patients with 2-year follow-up . Clin Spine Surg . 2017 ; 30 ( 2 ): E124 – E131 . Google Scholar CrossRef Search ADS PubMed 46. Smith JS , Shaffrey CI , Glassman SD et al. Clinical and radiographic parameters that distinguish between the best and worst outcomes of scoliosis surgery for adults . Eur Spine J . 2013 ; 22 ( 2 ): 402 – 410 . Google Scholar CrossRef Search ADS PubMed 47. Coretti S , Ruggeri M , McNamee P . The minimum clinically important difference for EQ-5D index: a critical review . Expert Rev Pharmacoecon Outcomes Res . 2014 ; 14 ( 2 ): 221 – 233 . Google Scholar CrossRef Search ADS PubMed 48. EuroQol Group . EuroQol—a new facility for the measurement of health-related quality of life . Health Policy . 1990 ; 16 ( 3 ): 199 – 208 . CrossRef Search ADS PubMed Acknowledgment We thank Nicolas Thompson, Department of Quantitative Health Sciences, Lerner Research Institute, Cleveland Clinic, for his support and guidance. Copyright © 2018 by the Congress of Neurological Surgeons This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

NeurosurgeryOxford University Press

Published: Jun 5, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off