Abstract
BACKGROUND
Recent studies in surgical and non-surgical specialities have suggested that patients admitted on the weekend may have worse outcomes. In particular, patients with stroke and acute cardiovascular events have shown worse outcomes with weekend treatment. It is unclear whether this extends to patients with spinal cord injury (SCI). This study was designed to evaluate factors for readmission after index hospitalization for spinal cord injury.
METHODS
This cohort was constructed from the State Inpatient Databases of California, New York, and Florida. 14,396 patients with SCI were identified. The primary outcome measure evaluated was 30-day readmission. Secondary measures include in-hospital complications.
Univariate and multivariate analysis were utilized to evaluate covariates. χ2, Fisher’s exact, and linear, logistic, and modified Poisson regression methods were utilized for statistical analysis. Propensity score methods were used with matched pairs analysis performed by the McNemar’s test.
RESULTS
Weekend admission was not associated with increased 30- day readmission rates in multivariate analysis. Race and discharge to a facility (RR 1.60 (1.43–1.79)) or home with home care (RR 1.23 (1.07–1.42)), were statistically significant risk factors for readmission. Payor status did not affect rates of readmission. In propensity score matched pairs analysis, weekend admission was not associated with increased odds of 30-day readmission (OR 1.04 (0.89–1.21)). Patients admitted to high volume centers had significantly lower risk of readmission when compared with patients admitted to low volume centers.
CONCLUSION
Our results suggest that the weekend effect, described previously in other patient populations, may not play as important a role in patients with SCI.
Keywords: spinal cord injury, trauma, weekend effect, spine surgery, insurance, healthcare disparities, race
Introduction
Spinal cord injury (SCI) is a significant public health problem afflicting approximately 12,400 patients in the United States annually.1 The incidence of new SCIs is estimated at nearly 40 cases per million persons annually,2 and the average age of affected individuals is less than 40 years.3 It is estimated that approximately 259,000 Americans are currently living with these injuries.4 Recovery from a complete SCI is exceedingly rare, leaving most patients with significant permanent disability. Treatment of these injuries is multimodal, and is often complicated by other significant associated comorbidities and the necessity for long-term and expensive disease management. Given the individual and societal costs associated with SCI, there is interest in identifying specific practice patterns and demographic factors that influence patient outcomes. A key driver of costs related to the index hospitalization includes perioperative complications and readmissions.5 Thus, both health care providers and insurers are interested in identifying of risk factors for these events.
Previous studies on a variety of different acute medical conditions have raised concerns that patients admitted to hospital on a weekend have worse outcomes than those admitted during the week.6–13 A recent study found that nurse staffing ratios, rather than day of the week, impacted mortality after stroke.14 These studies raise interesting questions concerning care of patients with SCI, as it is unclear what factors may influence outcome after neurologic injury, including the timing of the event and related care in the hospital.
Population-based databases allow construction of large cohorts of patients to address research questions and generate hypotheses. This study utilized census-level state inpatient databases (SID) to identify patient factors associated with readmission to the hospital, and to study weekend treatment and its effect on in-hospital complications and readmission to the hospital. Based on analysis of our institutional data (data not published), we hypothesize that weekend care for SCI patients is not associated with increased 30-day readmissions or increased hospital complications.
Methods
Study design and patient population
This was a retrospective cohort study using the Agency for Healthcare Research & Quality Healthcare Cost and Utilization Project SID for California (CA), Florida (FL), and New York (NY) from 2006 to 2010 to identify adults patients suffering traumatic spinal injuries identified by ICD-9-CM diagnosis codes (see Supplemental Table 1). Patients were required to have a traumatic injury (ICD-9-CM diagnosis codes 800.x–959.x excluding 905.x–909.x) as the principle diagnosis or an E-code for traumatic mechanism of injury (see Supplemental Table 1) to be included in the index study population. Index hospitalization complications were identified using ICD-9-CM diagnosis codes (Supplemental Table 1). Unique patient-level and encrypted patient-day identifiers were used to track sequential visits over time within the same state.15 Patients under age 18 years, missing identifiers, with hospital stay over 365 days, and those with an admission for spinal injury in 2005 were excluded from analysis. Patients residing in a different state than the admitting hospital were excluded, as those patients may not have equal risk for readmission.
Index admissions for spinal injury between 2006 and 2010 were identified. The SIDs from 2005 provided a baseline period to identify comorbidities and to decrease the likelihood that patients with chronic SCI would be included. The SIDs from 2011 provided minimum one-year follow-up for all patients with index admission in 2010. ICD-9-CM procedure and diagnosis codes were utilized to identify patients undergoing spine surgery, surgical approaches, and complications during the index hospitalization (see Supplemental Table 1).
The final cohort of patients with admission to hospital for an acute SCI from 2006–2010 included 14,396 patients. Given approximately 12,400 new SCI per year in the United States,1 one would expect approximately 15,000 new SCIs in a 5 year cohort of patients from California, New York, and Florida based on the 2010 Census data.16 For readmission analyses, patients who died during index admission were excluded. Using the index admission for SCI and any hospitalization during the previous year, pre-existing comorbidities were identified according to two different methods: the algorithm defined by Elixhauser and colleagues17 and the Charlson Comorbidity Index.18 Missing information for race is likely not missing at random in administrative databases,19 so was included as a separate category in analysis. Native Americans, Pacific Islanders, and “Other” race were included together as “Other.” Since spinal cord injury care is concentrated at high volume centers, we categorized treating hospital by annual volume of spinal cord injury.20
We utilized the Trauma Mortality Prediction Model (TMPM) to model the risk of death based on presenting injuries as scored using ICD-9-CM diagnosis codes.21, 22 The TMPM was augmented with age, sex, and injury mechanism to model probability of death.
Propensity Score Modeling
We used nonparsimonious multivariable logistic regression to create the propensity score model with weekend admission as the dependent variable.23, 24 The predicted probabilities of weekend admission were used to identify two populations of patients, one with weekday admissions and one with weekend admissions, with the same probability of weekend admission and balanced covariates. After propensity score matching on the predicted probability for weekend admission, comparison of the matched pairs can be used to determine whether individuals admitted with SCI on a weekend had higher or lower risk of 30-day readmission than patients with SCI admitted during the week with similar mix of injury severity, as modeled by the TMPM, and medical comorbidities. Because we studied readmission, patients who died during initial hospital admission were excluded before calculating the propensity score. Patients admitted on the weekend were matched 1:1 with patients admitted during the week using a greedy matching algorithm with caliper width of 0.2 standard deviations of the logit of the propensity score.25 Standardized differences between matched covariates were calculated to assess balance between matched pairs.26
Statistical Analysis
Chi-square and Student’s t-tests were used for univariate comparisons. Variables with p <0.10 in univariate testing were included in the initial multivariate models. Because of the commonality of the expected outcome, odds ratios likely provide an inflated value of the measure of association. Thus, a modified Poisson regression was utilized to calculate the relative risk (RR) for readmission.27 Univariate analysis of propensity score-matched pairs was performed with the McNemar’s test. 2-sided p values of <0.05 were considered significant. All statistical analysis was performed using SAS Enterprise 6.1 (SAS Institute, Cary, NC).
Results
Cohort Analysis
14,396 unique patients with SCI were admitted to hospitals in the three states from 2006–2010 were identified. 362 (10.44%) of patients admitted on the weekend and 902 (8.26%) of patients admitted on a weekday died during index hospitalization. After removing patients who died, 13,129 patients remained for readmission analysis. 414 of 3107 (13.32%) patients admitted on the weekend, and 1423 of 10022 (14.20%) patients admitted on a weekday were readmitted within 30 days of discharge. Univariate analysis of weekday versus weekend admission is reported in Table 1. Age, race, payer, anatomic site of SCI, and many comorbidities differed between patients admitted on a weekday and those admitted on the weekend (Table 1). Patients admitted on the weekend were more likely to be admitted to a high volume center. Univariate analysis of patient characteristics associated with 30-day readmission is reported in Table 2. Older age, sex, race, and most comorbidities were significantly associated with 30-day readmission. Though the difference in means of the TMPM predictions was not statistically different between readmitted and non-readmitted patients, it is a validated measure of injury severity,21, 22 and therefore we included it in multivariate analysis. Patients with Medicare had higher risk of readmission, while private insurance showed a protective effect (Table 2). Patient residence and weekend admission were not associated with 30 day readmission (Table 2). Each additional day in the hospital during the index SCI admission was associated with increased risk of readmission in univariate analysis.
Table 1. Univariate analysis of patient demographics, comorbidities, and anatomic location of spinal cord injury differ by presentation day.
| Characteristic | Weekday Admission 10022 (76.33%) |
Weekend Admission 3107 (23.67%) |
p Value |
|---|---|---|---|
| N (%) | N (%) | ||
| Age | <0.0001 | ||
| Mean ± St.Dev. | 52.40 ± 20.61 | 49.51±20.13 | |
| Sex | 0.2693 | ||
| Male | 6870 (68.82) | 2164 (69.87) | |
| Race | 0.0021 | ||
| White | 5928 (62.24) | 1936 (64.62) | |
| Black | 1458 (15.31) | 422 (14.09) | |
| Hispanic | 1360 (14.28) | 418 (13.95) | |
| Other | 779 (7.77) | 220 (7.08) | |
| Missing | 497 (4.96) | 111 (3.57) | |
| TMPM | 0.0171 | ||
| Mean ± St.Dev. | 9.37±11.88 | 8.81±11.62 | |
| Elixhauser Comorbidities | |||
| Congestive Heart Failure | 573 (5.72) | 134 (4.31) | 0.0024 |
| Valvular Disease | 341 (3.40) | 99 (3.19) | 0.5586 |
| Peripheral Vascular Disease | 378 (3.77) | 85 (2.74) | 0.0062 |
| Hypertension | 3855 (38.57) | 1028 (33.09) | <0.0001 |
| Chronic Pulmonary Disease | 1213 (12.10) | 348 (11.20) | 0.1743 |
| Pulmonary Vascular Disease | 236 (2.35) | 55 (1.77) | 0.0531 |
| Paralysis | 1858 (18.54) | 473 (15.22) | <0.0001 |
| Neurological Disease | 828 (8.26) | 244 (7.85) | 0.4674 |
| Diabetes | 1688 (16.84) | 413 (13.29) | <0.0001 |
| Renal Disease | 501 (5.00) | 104 (3.35) | <0.0001 |
| Liver Disease | 231 (2.30) | 78 (2.51) | 0.5091 |
| Peptic Ulcer Disease | <11 | <11 | 0.7600 |
| Rheumatological Disease | 294 (2.93) | 75 (2.41) | 0.1257 |
| AIDS | 21 (0.21) | <11 | 0.8679 |
| Lymphoma | 43 (0.43) | <11 | 0.0305 |
| Coagulopathy | 549 (5.48) | 155 (4.99) | 0.2902 |
| Metastatic Cancer | 93 (0.93) | 19 (0.61) | 0.0938 |
| Solid Tumor | 104 (1.04) | 25 (0.80) | 0.2498 |
| Hypothyroidism | 595 (5.94) | 155 (4.99) | 0.0466 |
| Obesity | 632 (6.31) | 157 (5.05) | 0.0102 |
| Weight Loss | 666 (6.65) | 226 (7.27) | 0.2238 |
| Electrolyte Disorder | 2523 (25.17) | 759 (24.43) | 0.4015 |
| Anemia, Blood Loss | 148 (1.48) | 27 (0.87) | 0.0099 |
| Anemia, Deficiency | 1688 (16.84) | 462 (14.87) | 0.0094 |
| Alcohol Abuse | 1341 (13.38) | 554 (17.83) | 0.4676 |
| Drug Abuse | 815 (8.13) | 291 (9.37) | 0.0305 |
| Psychoses | 602 (6.01) | 197 (6.34) | 0.4966 |
| Depression | 1062 (10.60) | 309 (9.95) | 0.2996 |
| Spinal Cord Injury | |||
| Cervical | 5833 (58.20) | 1906 (61.35) | 0.0019 |
| Thoracic | 2557 (25.51) | 712 (22.92) | 0.0034 |
| Lumbosacral | 2202 (21.97) | 646 (20.79) | 0.1633 |
| Weekend Admission | 2693 (23.85) | 414 (22.54) | 0.2198 |
| State | <0.0001 | ||
| California | 4727 (47.17) | 1335 (42.97) | |
| Florida | 2793 (27.87) | 1099 (35.37) | |
| New York | 2502 (24.97) | 673 (21.66) | |
| No. transferred from another hospital | 3729 (37.21) | 261 (8.40) | <0.0001 |
| Patient Residence | 0.0601 | ||
| Large Metro Area (>1 million) | 6869 (68.55) | 2062 (66.39) | |
| Small Metro Area (<1 million) | 2568 (25.63) | 840 (27.04) | |
| Micropolitan Area or Other | 584 (5.83) | 204 (6.57) | |
| Expected Payer | <0.0001 | ||
| Medicare | 3054 (30.48) | 766 (24.66) | |
| Medicaid | 1874 (18.70) | 509 (16.39) | |
| Private Insurance | 3390 (33.83) | 1279 (41.18) | |
| Other | 1702 (16.99) | 552 (17.77) | |
| Hospital, Annual Number of SCI Admissions | <0.0001 | ||
| <10 | 4096 (40.87) | 1117 (35.95) | |
| 10–19 | 2137 (21.32) | 767 (24.69) | |
| 20–29 | 1229 (12.26) | 430 (13.84) | |
| >30 | 2560 (25.54) | 793 (25.52) |
All numeric data (age) were evaluated with t-tests. All categorical data (all other variables above) were evaluated with χ2 tests.
Trauma Mortality Prediction Model (TMPM) score is modeled as described in Glance et al, 2009.
Table 2. Association of demographic and presenting characteristics of 13,129 patients surviving admission with 30 day readmission.
| Characteristic | No 30-day readmission 11292 (86.01%) |
Readmitted within 30 days 1837 (13.99%) |
p Value |
|---|---|---|---|
| N (%) | N (%) | ||
| Age, y | <0.0001 | ||
| Mean ± St.Dev. | 50.85 ± 20.38 | 57.07 ± 20.67 | |
| Sex | 0.1344 | ||
| Male | 7794 (69.32) | 1240 (67.57) | |
| Race | 0.0189 | ||
| White | 6750 (62.77) | 1114 (63.01) | |
| Black | 1583 (14.72) | 297 (16.80) | |
| Hispanic | 1549 (14.41) | 229 (12.95) | |
| Other | 871 (7.71) | 128 (6.97) | |
| Missing | 539 (4.77) | 69 (3.76) | |
| Elixhauser Comorbidities | |||
| Congestive Heart Failure | 514 (4.55) | 193 (10.51) | <0.0001 |
| Valvular Disease | 337 (2.98) | 103 (5.61) | <0.0001 |
| Peripheral Vascular Disease | 345 (3.06) | 118 (6.42) | < 0.0001 |
| Hypertension | 3997 (35.40) | 886 (48.23) | <0.0001 |
| Chronic Pulmonary Disease | 1262 (11.18) | 299 (16.28) | <0.0001 |
| Pulmonary Vascular Disease | 232 (2.05) | 59 (3.21) | 0.0018 |
| Paralysis | 1932 (17.11) | 399 (21.72) | <0.0001 |
| Neurological Disease | 842 (7.46) | 230 (12.52) | <0.0001 |
| Diabetes | 1666 (14.75) | 435 (23.68) | <0.0001 |
| Renal Disease | 435 (3.85) | 170 (9.25) | <0.0001 |
| Liver Disease | 239 (2.12) | 70 (3.81) | <0.0001 |
| Peptic Ulcer Disease | <11 | <11 | 0.9821 |
| Rheumatological Disease | 297 (2.63) | 72 (3.92) | 0.0019 |
| AIDS | 21 (0.19) | < 11 | 0.0928 |
| Lymphoma | 34 (0.30) | 14 (0.76) | 0.0024 |
| Coagulopathy | 559 (4.95) | 145 (7.89) | <0.0001 |
| Metastatic Cancer | 78 (0.69) | 34 (1.85) | <0.0001 |
| Solid Tumor | 105 (0.93) | 24 (1.31) | 0.1291 |
| Hypothyroidism | 603 (5.34) | 147 (8.00) | <0.0001 |
| Obesity | 646 (5.72) | 143 (7.78) | 0.0006 |
| Weight Loss | 683 (6.05) | 209 (11.38) | <0.0001 |
| Electrolyte Disorder | 2639 (23.37) | 643 (35.00) | <0.0001 |
| Anemia, Blood Loss | 132 (1.17) | 43 (2.34) | <0.0001 |
| Anemia, Deficiency | 1706 (15.11) | 444 (24.17) | <0.0001 |
| Alcohol Abuse | 1640 (14.52) | 255 (13.88) | 0.4676 |
| Drug Abuse | 939 (8.32) | 167 (9.09) | 0.2672 |
| Psychoses | 632 (5.60) | 167 (9.09) | <0.0001 |
| Depression | 1109 (9.82) | 262 (14.26) | <0.0001 |
| Spinal Cord Injury | |||
| Cervical | 6679 (59.15) | 1060 (57.70) | 0.2429 |
| Thoracic | 2745 (24.31) | 524 (28.52) | <0.0001 |
| Lumbosacral | 2470 (21.87) | 378 (20.58) | 0.2110 |
| Weekend Admission | 2693 (23.85) | 414 (22.54) | 0.2198 |
| State | 0.20 | ||
| California | 5227 (46.29) | 835 (45.45) | |
| Florida | 3316 (29.37) | 576 (31.36) | |
| New York | 2749 (24.34) | 426 (23.19) | |
| No. transferred from another hospital | 3462 (30.66) | 528 (28.74) | 0.0977 |
| Patient Residence | 0.9962 | ||
| Large Metro Area (>1 million) | 7681 (68.03) | 1250 (68.05) | |
| Small Metro Area (<1 million) | 2932 (25.97) | 476 (25.91) | |
| Micropolitan Area or Other | 677 (6.00) | 111 (6.04) | |
| Expected Payer | <0.0001 | ||
| Medicare | 3060 (27.11) | 760 (41.37) | |
| Medicaid | 2036 (18.04) | 347 (18.89) | |
| Private Insurance | 4142 (36.69) | 527 (28.69) | |
| Other | 2051 (18.17) | 203 (11.05) | |
| Hospital, Annual Number of SCI Admissions | <0.0001 | ||
| <10 | 4372 (38.72) | 841 (45.78) | |
| 10–19 | 2549 (22.57) | 355 (19.32) | |
| 20–29 | 1430 (12.66) | 229 (12.47) | |
| >30 | 2941 (26.04) | 412 (22.43) | |
| In-Hospital Complications | |||
| Pneumonia | 1520 (13.46) | 404 (21.99) | <0.0001 |
| Decubitus Ulcer | 1158 (10.26) | 321 (17.47) | <0.0001 |
| Respiratory Failure | 1177 (10.42) | 328 (17.86) | <0.0001 |
| Acute Renal Failure | 396 (3.51) | 150 (8.17) | <0.0001 |
| Central Venous Catheter Infection | 40 (0.35) | 23 (1.25) | <0.0001 |
| Sepsis | 713 (6.31) | 224 (12.19) | <0.0001 |
| Clostridium difficile | 254 (2.25) | 82 (4.46) | <0.0001 |
| Urinary Tract Infection | 2442 (21.63) | 509 (27.71) | <0.0001 |
All numeric data (age) were evaluated with t-tests. All categorical data (all other variables above) were evaluated with χ2 tests. Trauma Mortality Prediction Model (TMPM) score is modeled as described in Glance et al, 2009.
In a multivariable Poisson regression model (Table 3), black race, older age, Medicare primary payer, and discharge location other than home were associated with significantly increased risk of 30-day readmission. Patients treated at hospitals with increased volume of SCI admissions had lower risk of readmission. We studied both Charlson comorbidity categories and Elixhauser comorbidities. In comparison with the Charlson categories, inclusion of the individual Elixhauser comorbidities in the Poisson model resulted in better model fit (−2 log (Likelihood Ratio) difference 1365.63, χ2 with 10 degrees of freedom, p<0.0001). Several Elixhauser comorbidities were associated with increased risk of readmission, including peripheral vascular disease, neurological disorder, renal failure, lymphoma, metastatic cancer, weight loss, psychosis, depression, and diabetes (Table 3). Patients treated at hospitals with a high annual volume of patients with SCI had the lowest risk of readmission (Table 3). 10402 of 13,129 (79.23%) patients with spine injuries in this cohort were treated in hospitals in the top quartile of SCI admissions (data not shown).
Table 3. Multivariable regression of parameters associated with risk of 30-day readmission.
| Characteristic | Multivariable Regression | |
|---|---|---|
| Estimated Relative Risk | 95% CI | |
| Age (10 year increase) | 1.06 | 1.03–1.09 |
|
| ||
| Race | ||
|
| ||
| White | Reference | |
|
| ||
| African-American | 1.17 | 1.04–1.32 |
|
| ||
| Hispanic | 1.02 | 0.90–1.17 |
|
| ||
| Other | 0.94 | 0.80–1.11 |
|
| ||
| Missing | 0.89 | 0.71–1.11 |
|
| ||
| Expected Payer | ||
|
| ||
| Medicare | Reference | |
|
| ||
| Medicaid | 1.05 | 0.90–1.22 |
|
| ||
| Private Insurance | 0.87 | 0.77–1.00 |
|
| ||
| Other | 0.80 | 0.67–0.95 |
|
| ||
| Discharge Location | ||
|
| ||
| Home | Reference | |
|
| ||
| Home with Home Care | 1.23 | 1.07–1.42 |
|
| ||
| Other | 1.60 | 1.43–1.79 |
|
| ||
| Hospital, Annual Number of SCI Admissions | ||
|
| ||
| <10 | 1.13 | 1.01–1.27 |
|
| ||
| 10–19 | 1.00 | 0.88–1.14 |
|
| ||
| 20–29 | 1.14 | 0.98–1.32 |
|
| ||
| >30 | Reference | |
|
| ||
| Comorbidities | ||
|
| ||
| Peripheral Vascular Disease | 1.27 | 1.08–1.50 |
|
| ||
| Neurological Disorder | 1.17 | 1.03–1.33 |
|
| ||
| Renal Failure | 1.28 | 1.10–1.50 |
|
| ||
| Lymphoma | 1.67 | 1.05–2.67 |
|
| ||
| Metastatic Cancer | 1.62 | 1.18–2.21 |
|
| ||
| Weight Loss | 1.21 | 1.06–1.38 |
|
| ||
| Psychoses | 1.33 | 1.15–1.53 |
|
| ||
| Depression | 1.21 | 1.07–1.36 |
|
| ||
| Diabetes | 1.23 | 1.11–1.37 |
|
| ||
| Injuries and in-hospital complications | ||
|
| ||
| Thoracic SCI | 1.15 | 1.04–1.26 |
|
| ||
| Spinal Fusion | ||
|
| ||
| No surgery | Reference | |
|
| ||
| 1 level | 1.00 | 0.75–1.33 |
|
| ||
| 2–3 levels | 0.80 | 0.71–0.89 |
|
| ||
| 4–8 levels | 0.83 | 0.73–0.94 |
|
| ||
| >9 levels | 0.89 | 0.60–1.32 |
|
| ||
| Pneumonia | 1.25 | 1.11–1.40 |
|
| ||
| Decubitus Ulcer | 1.27 | 1.13–1.42 |
|
| ||
| Respiratory Failure | 1.27 | 1.12–1.43 |
|
| ||
| Acute Renal Failure | 1.25 | 1.07–1.46 |
|
| ||
| Central Venous Catheter Infection | 1.78 | 1.27–2.49 |
In multivariable regression, several variables did not achieve significance, including multiple Elixhauser comorbidities, weekend admission, wound dehiscence, osteomyelitis, prolonged ventilation, and surgical complication.
Pneumonia, decubitus ulcer, respiratory failure, acute renal failure, and central venous catheter infections during the index SCI hospitalization were associated with increased risk of readmission (Table 3). Admission on the weekend was not significant in either univariate (Table 2) or multivariate analysis. Initial SCI hospitalization length of stay was significantly longer in patients readmitted within 30 days compared to those not readmitted (data not shown). Length of stay was collinear with TMPM, so was removed for multivariate regression. Thoracic injury was associated with higher risk of readmission (Table 3).
Propensity-Score Matched Pairs Analysis
To further evaluate the effect of weekend admission on readmission and in-hospital complications, we constructed propensity-score matched pairs as described above. We included TMPM, age, sex, race, patient residence, hospital SCI volume, comorbidities, and location of SCI in the model to create the propensity score, with weekend admission as the dependent variable. The TMPM predicts the probability of mortality given traumatic injuries,21 so was used to balance severity of injury between groups. Supplemental Table 2 reports the frequencies of variables included in the logistic regression model to create the propensity score for the matched-pairs cohort. We were able to match 3099 of 3107 (99.7%) patients admitted on the weekend to a control patient with equal likelihood for weekend admission, but who was admitted on a weekday. We analyzed the adequacy of the match by comparing standardized differences26 and probability distributions of the propensity score (data not shown).
We performed analyses of 30-day readmission and other outcomes (Table 4) using the propensity score matched pairs. Numbers of discordant pairs, where one of the pair was affected and the other was not, are presented in Table 4. The ratio of discordant pairs allows for calculation of the odds ratio. Weekend admission was not associated with readmission within 30 days or discharge location. Patients admitted on the weekend had lower odds of index hospitalization medical complications, including deep venous thrombosis, decubitus ulcer, urinary tract infection, and sepsis compared to weekday admission (Table 4). Length of stay was shorter in patients admitted on the weekend (16.54 (15.65–17.44)) versus those admitted on a weekday (26.0 (24.78–27.33)).
Table 4. Weekend treatment does not increase odds of readmission within 30 days, but is associated with lower in-hospital medical complications when analyzed using propensity score matched pairs.
| Outcome | N, Weekend patient affected with Weekday patient Unaffected | N, Weekend patient unaffected with Weekday patient affected | Odds Ratio | OR (95 % CI) | Test Statistic | P value | |
|---|---|---|---|---|---|---|---|
| Lower Limit | Upper Limit | ||||||
| Primary Outcome of Interest | |||||||
|
| |||||||
| Readmitted within 30 days | 352 | 339 | 1.04 | 0.89 | 1.21 | 0.245 | 0.6209 |
|
| |||||||
| Secondary Outcomes (in-hospital medical and surgical complications) | |||||||
|
| |||||||
| Discharge Home | 716 | 725 | 0.99 | 0.89 | 1.10 | 0.056 | 0.8126 |
|
| |||||||
| DVT | 71 | 141 | 0.50 | 0.38 | 0.67 | 23.113 | <0.0001 |
|
| |||||||
| Decubitus Ulcer | 189 | 325 | 0.58 | 0.49 | 0.70 | 35.984 | <0.0001 |
|
| |||||||
| UTI | 340 | 566 | 0.60 | 0.53 | 0.69 | 56.375 | <0.0001 |
|
| |||||||
| Wound Infection | 59 | 91 | 0.65 | 0.47 | 0.90 | 6.827 | 0.0090 |
|
| |||||||
| C. difficile | 55 | 75 | 0.73 | 0.52 | 1.04 | 3.077 | 0.0794 |
|
| |||||||
| Central Line Infection | <11 | 13 | 0.62 | 0.26 | 1.48 | 1.190 | 0.2752 |
|
| |||||||
| Pneumonia | 340 | 393 | 0.87 | 0.75 | 1.00 | 3.832 | 0.0503 |
|
| |||||||
| Renal Failure | 95 | 117 | 0.81 | 0.62 | 1.06 | 2.283 | 0.1308 |
|
| |||||||
| Sepsis | 167 | 192 | 0.87 | 0.71 | 1.07 | 1.741 | 0.1870 |
|
| |||||||
| Fusion | 677 | 813 | 0.83 | 0.75 | 0.92 | 12.413 | 0.0004 |
|
| |||||||
| Surgical Complication | 74 | 94 | 0.79 | 0.58 | 1.07 | 2.381 | 0.1228 |
The discordant pairs from the propensity score matching are presented.
Discussion
In this study, we analyzed a retrospective cohort of SCI patients admitted to hospitals in California, New York, and Florida over a 5-year period. We found no evidence of a “weekend” effect with regards to readmission or index hospital complications in multivariate analysis of patients with SCI. Patients readmitted within 30 days were older and more likely to have medical comorbidities (Table 2). Multivariate analysis showed increased risk of readmission for African-Americans, patients discharged home with home care or to another facility (e.g., long-term care, skilled nursing facility), and those admitted to lower SCI volume hospitals (Table 3). Patients admitted on a weekend with acute SCI were less likely to have perioperative medical complications and did not have increased risk of readmission within 30 days (Table 4).
Previous authors have studied the effect of weekend admission on patient outcomes.6–12 Several of these studies have shown worse outcomes with weekend admission.6–8, 10, 11, 13 A more recent study suggested that nurse staffing ratios, rather than the weekend itself, may be the key factor affecting mortality.14 A multitude of factors, including the availability of diagnostic studies, may also be at play.28 A population study of spinal oncology patients requiring surgical intervention did not show a deleterious effect of weekend admission, but did report a higher likelihood of delay in surgical treatment if a patient was admitted on the weekend.29 However, a different population-based study using the Nationwide Inpatient Sample of patients undergoing surgical intervention in cervical spine trauma did report longer length of stay or hospital costs for patients admitted on the weekend compared to weekday.30 Use of the SIDs of California, Florida, and New York offers important advantages. Patients are trackable over time with an encrypted identifier, and the time to readmission can be calculated.15 Although the Nationwide Inpatient Sample allows a large sample of inpatient admission across the United States to be studied, the study data are cross-sectional by definition and do not allow analysis at the patient-level since no patient identifiers are available. The SID billing data allows construction of longitudinal cohorts for study of readmission risks and longer-term complications.
Factors that do not impact billing, but may impact patient care, show lower accuracy in administrative databases.31 The focus on billable coding presents an inherent bias, though the bias is likely not affected by the timing of a trauma admission. The incentive for accurate billing does not change with time of admission.
In our analysis, weekend admission was not associated with readmission at 30 days. However, consistent with our institutional cohort (data not published), weekend admission was not associated with length of hospital stay, readmission within 30 days, and was associated with lower in-hospital medical complications. Bray et al previously reported that negative health outcomes may be more related to staffing levels than with the particular day of the week during which one receives care.14 Higher volume trauma centers may not suffer the same staffing shortfalls outside of normal work hours, and may be equipped to provide similar care no matter when patients arrive. Thus, admission of patients to primarily large Trauma Centers may nullify any possible weekend effect.
Limitations
This dataset is derived from administrative billing data provided by California, New York, and Florida. There is inherent bias in this data, as described above, in that coding affecting remuneration is more likely to be accurate. Additionally, some factors may be missing in non-random fashion, limiting the applicability of this analysis. Additionally, due to using 3 states, there is chance of readmission to a neighboring state, though we minimized this by restricting the cohort to patients residing in the same state as the admitting hospital.
Significance
This study used a large retrospective cohort of patients from the HCUP State Inpatient Databases to determine whether outcomes in patients with SCI are impacted by weekend admission. Despite the inherent limitations this report represents the largest cohort of SCI patients and evaluation of readmission risk factors.
Conclusion
In this study, we showed in two separate analyses that a cohort of patients with SCI injury had similar risk of readmission independent of day of admission. African-American patients have increased risk of 30-day readmission. Hospitalization in high SCI volume centers was associated with lower risk of readmission compared with low volume centers.
Supplementary Material
Acknowledgments
We would like to thank Alan Cook, MD FACS, who graciously provided the SAS macro for calculation of the Trauma Mortality Prediction Model. We would also like to thank Dongsheng Yang at the Cleveland Clinic Foundation for his gracious help with calculating standardized differences.
The Center for Administrative Data Research is supported in part by the Washington University Institute of Clinical and Translational Sciences grant UL1 TR000448 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH), Grant Number R24 HS19455 through the Agency for Healthcare Research and Quality (AHRQ), and Grant Number KM1CA156708 through the National Cancer Institute (NCI) at the National Institutes of Health (NIH).
Footnotes
Disclosures:
Wilson Ray – Consulting - Depuy Synthes, Ulrich. SAB – Acera Surgical, Harvest technology, Stock – LDR Holding, Grant Support – NINDS/NIH, DARPA, Missouri Spinal Cord Injury Foundation; Ian Dorward – Consulting – Depuy Synthes; Margaret Olsen – Consulting – Sanofi Pasteur, Merck, Pfizer, Grant Support – Sanofi Pasteur, Cubist, Pfizer
References
- 1.Devivo MJ. Epidemiology of traumatic spinal cord injury: trends and future implications. Spinal cord. 2012 May;50(5):365–72. doi: 10.1038/sc.2011.178. [DOI] [PubMed] [Google Scholar]
- 2.Oppenheim JS, Spitzer DE, Winfree CJ. Spinal cord bypass surgery using peripheral nerve transfers: review of translational studies and a case report on its use following complete spinal cord injury in a human. Experimental article. Neurosurg Focus. 2009 Feb;26(2):E6. doi: 10.3171/FOC.2009.26.2.E6. Epub 2009/05/14. eng. [DOI] [PubMed] [Google Scholar]
- 3.Nobunaga AI, Go BK, Karunas RB. Recent demographic and injury trends in people served by the Model Spinal Cord Injury Care Systems. Arch Phys Med Rehabil. 1999 Nov;80(11):1372–82. doi: 10.1016/s0003-9993(99)90247-2. Epub 1999/11/24. eng. [DOI] [PubMed] [Google Scholar]
- 4.Center NSCIS. National Spinal Cord Injury Statistical Center: Spinal Cord Injury Facts and Figures at a Glance. Birmingham, Alabama: Apr, 2009. [cited 2010 June 6]; Available from: www.nscisc.uab.edu/public_content/facts_figures_2009.aspx. [Google Scholar]
- 5.Friedman B, Basu J. The rate and cost of hospital readmissions for preventable conditions. Medical care research and review : MCRR. 2004 Jun;61(2):225–40. doi: 10.1177/1077558704263799. [DOI] [PubMed] [Google Scholar]
- 6.Aylin P, Yunus A, Bottle A, Majeed A, Bell D. Weekend mortality for emergency admissions. A large, multicentre study. Quality & safety in health care. 2010 Jun;19(3):213–7. doi: 10.1136/qshc.2008.028639. [DOI] [PubMed] [Google Scholar]
- 7.Bell CM, Redelmeier DA. Mortality among patients admitted to hospitals on weekends as compared with weekdays. The New England journal of medicine. 2001 Aug 30;345(9):663–8. doi: 10.1056/NEJMsa003376. [DOI] [PubMed] [Google Scholar]
- 8.Cram P, Hillis SL, Barnett M, Rosenthal GE. Effects of weekend admission and hospital teaching status on in-hospital mortality. The American journal of medicine. 2004 Aug 1;117(3):151–7. doi: 10.1016/j.amjmed.2004.02.035. [DOI] [PubMed] [Google Scholar]
- 9.Handel AE, Patel SV, Skingsley A, Bramley K, Sobieski R, Ramagopalan SV. Weekend admissions as an independent predictor of mortality: an analysis of Scottish hospital admissions. BMJ open. 2012;2(6) doi: 10.1136/bmjopen-2012-001789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Kostis WJ, Demissie K, Marcella SW, Shao YH, Wilson AC, Moreyra AE, et al. Weekend versus weekday admission and mortality from myocardial infarction. The New England journal of medicine. 2007 Mar 15;356(11):1099–109. doi: 10.1056/NEJMoa063355. [DOI] [PubMed] [Google Scholar]
- 11.Mohammed MA, Sidhu KS, Rudge G, Stevens AJ. Weekend admission to hospital has a higher risk of death in the elective setting than in the emergency setting: a retrospective database study of national health service hospitals in England. BMC health services research. 2012;12:87. doi: 10.1186/1472-6963-12-87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Saposnik G, Baibergenova A, Bayer N, Hachinski V. Weekends: a dangerous time for having a stroke? Stroke; a journal of cerebral circulation. 2007 Apr;38(4):1211–5. doi: 10.1161/01.STR.0000259622.78616.ea. [DOI] [PubMed] [Google Scholar]
- 13.Thomas CJ, Smith RP, Uzoigwe CE, Braybrooke JR. The weekend effect: short-term mortality following admission with a hip fracture. The bone & joint journal. 2014 Mar;96-B(3):373–8. doi: 10.1302/0301-620X.96B3.33118. [DOI] [PubMed] [Google Scholar]
- 14.Bray BD, Ayis S, Campbell J, Cloud GC, James M, Hoffman A, et al. Associations between stroke mortality and weekend working by stroke specialist physicians and registered nurses: prospective multicentre cohort study. PLoS medicine. 2014 Aug;11(8):e1001705. doi: 10.1371/journal.pmed.1001705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Steiner C, Elixhauser A, Schnaier J. The healthcare cost and utilization project: an overview. Effective clinical practice : ECP. 2002 May-Jun;5(3):143–51. [PubMed] [Google Scholar]
- 16.Mackun PJ, Wilson S, Fischetti TR, Goworowska J. Population distribution and change: 2000 to 2010. US Department of Commerce; 2011. Contract No.: C2010BR-01. [Google Scholar]
- 17.Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Medical care. 1998 Jan;36(1):8–27. doi: 10.1097/00005650-199801000-00004. [DOI] [PubMed] [Google Scholar]
- 18.Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of chronic diseases. 1987;40(5):373–83. doi: 10.1016/0021-9681(87)90171-8. [DOI] [PubMed] [Google Scholar]
- 19.Henry AJ, Hevelone ND, Lipsitz S, Nguyen LL. Comparative methods for handling missing data in large databases. Journal of vascular surgery. 2013 Nov;58(5):1353–9.e6. doi: 10.1016/j.jvs.2013.05.008. [DOI] [PubMed] [Google Scholar]
- 20.Macias CA, Rosengart MR, Puyana JC, Linde-Zwirble WT, Smith W, Peitzman AB, et al. The effects of trauma center care, admission volume, and surgical volume on paralysis after traumatic spinal cord injury. Annals of surgery. 2009 Jan;249(1):10–7. doi: 10.1097/SLA.0b013e31818a1505. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Glance LG, Osler TM, Mukamel DB, Meredith W, Wagner J, Dick AW. TMPM-ICD9: a trauma mortality prediction model based on ICD-9-CM codes. Annals of surgery. 2009 Jun;249(6):1032–9. doi: 10.1097/SLA.0b013e3181a38f28. [DOI] [PubMed] [Google Scholar]
- 22.Cook A, Weddle J, Baker S, Hosmer D, Glance L, Friedman L, et al. A comparison of the Injury Severity Score and the Trauma Mortality Prediction Model. The journal of trauma and acute care surgery. 2014 Jan;76(1):47–52. doi: 10.1097/TA.0b013e3182ab0d5d. discussion -3. [DOI] [PubMed] [Google Scholar]
- 23.D’Agostino RB., Jr Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Statistics in medicine. 1998 Oct 15;17(19):2265–81. doi: 10.1002/(sici)1097-0258(19981015)17:19<2265::aid-sim918>3.0.co;2-b. [DOI] [PubMed] [Google Scholar]
- 24.Rubin DB. Estimating causal effects from large data sets using propensity scores. Annals of internal medicine. 1997 Oct 15;127(8 Pt 2):757–63. doi: 10.7326/0003-4819-127-8_part_2-199710151-00064. [DOI] [PubMed] [Google Scholar]
- 25.Austin PC. Optimal caliper widths for propensity-score matching when estimating differences in means and differences in proportions in observational studies. Pharm Stat. 2011 Mar-Apr;10(2):150–61. doi: 10.1002/pst.433. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Statistics in medicine. 2009 Nov 10;28(25):3083–107. doi: 10.1002/sim.3697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zou G. A modified poisson regression approach to prospective studies with binary data. American journal of epidemiology. 2004 Apr 1;159(7):702–6. doi: 10.1093/aje/kwh090. [DOI] [PubMed] [Google Scholar]
- 28.Becker DJ. Do hospitals provide lower quality care on weekends? Health services research. 2007 Aug;42(4):1589–612. doi: 10.1111/j.1475-6773.2006.00663.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Dasenbrock HH, Pradilla G, Witham TF, Gokaslan ZL, Bydon A. The impact of weekend hospital admission on the timing of intervention and outcomes after surgery for spinal metastases. Neurosurgery. 2012 Mar;70(3):586–93. doi: 10.1227/NEU.0b013e318232d1ee. [DOI] [PubMed] [Google Scholar]
- 30.Nandyala SV, Marquez-Lara A, Fineberg SJ, Schmitt DR, Singh K. Comparison of perioperative outcomes and cost of spinal fusion for cervical trauma: weekday versus weekend admissions. Spine. 2013 Dec 1;38(25):2178–83. doi: 10.1097/BRS.0000000000000020. [DOI] [PubMed] [Google Scholar]
- 31.Campbell SE, Campbell MK, Grimshaw JM, Walker AE. A systematic review of discharge coding accuracy. Journal of public health medicine. 2001 Sep;23(3):205–11. doi: 10.1093/pubmed/23.3.205. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.