On any given night, ∼550,000 individuals are experiencing homelessness in the United States,1 and ∼7.5% of these individuals are Veterans.2 Persons who are homeless have a higher disease burden compared with their housed counterparts3 (ie, “nonhomeless individuals” or “housed individuals”). This includes higher rates of tuberculosis, human immunodeficiency virus, hepatitis B, hepatitis C, uncontrolled diabetes mellitus, uncontrolled hypertension, mental illness, and substance use disorder when compared with the population at large.4–8 In addition to the increased prevalence of these chronic conditions, homeless individuals face psychosocial barriers in obtaining appropriate health care (ie, comorbid behavioral conditions, cultural, familial, socioeconomic, structural factors including but not limited to access to transport and access to care) and frequently lack social support.9 As such, homeless patients undergoing surgery are potentially at higher risk for readmission as they lack both the physical and social supports necessary to facilitate postdischarge recovery and may also be vulnerable to poor or inadequate care coordination.10
Only a few studies have examined hospital readmissions among homeless persons, and none have focused specifically on surgical patients. Saab et al11 compared 30-day hospital readmission rates between 400 homeless-experienced patients and a control group of low-income, housed patients in a matched cohort study in Ontario, Canada. Homeless patients had a 22.2% readmission rate compared with a rate of only 7% for the controls.11 Doran et al12 found homeless patients’ readmission rate was 50% within a single hospital in a mid-sized northeastern city in the summer months, which is well over the national average readmission rate of 17.5% in 2013 among all patients.13,14 These and other studies suggest homeless patients are at increased risk of readmission; however, it is not clear whether this is related to the health care system factors, patient factors, social contextual factors, or some combination of these 3 factors. Homeless populations may be particularly vulnerable to readmission following surgery as they may not have access to sanitary conditions necessary for proper wound care, may have the inability to properly store medications, and may have difficulty getting to follow-up appointments.
To determine the extent to which homelessness affects rates of postoperative readmissions, we conducted a large cohort study in the Veterans Health Administration (VHA), an integrated health care system serving >5 million individuals annually. The main objective of this study was to examine predictors of surgical readmissions among homeless Veterans. We hypothesized that homeless Veterans were at higher risk for readmission than housed Veterans, and that this risk was partly explained by an increased risk of postdischarge wound–related complications.
A retrospective cohort study design was used to examine our objective. The study protocol was reviewed and approved by the VHA Central Institutional Review Board, which granted a waiver of informed consent.
Cases were included if they had an admission for a qualifying general, vascular, or orthopedic surgery as assessed by the Veterans Affairs Surgical Quality Improvement Program (VASQIP) between October 1, 2007 and September 30, 2014.15 These 3 specialties account for the majority (70%) of surgeries performed during this time interval. Inclusion criteria included inpatient length of stay from 2 to 30 days and discharged alive. The study population has been previously described by Morris et al.16
For this study, we used VASQIP data to identify surgical procedures; the data provide detailed, nurse-abstracted, preoperative, and intraoperative surgical characteristics (discussed in the Variables section). VASQIP data represent a sample of surgical procedures from 134 VHA hospitals performing surgery.15,17 Surgical Quality Nurses at each facility are trained to perform standardized chart abstraction on eligible procedures. These are defined based on the probability of postoperative mortality and morbidity using clinically derived and validated mortality and morbidity calculators published and updated annually by the VHA National Surgery Office.15 Among eligible procedures, VASQIP samples 70%–80% for chart abstraction. Because of the sampling methodology, these procedures represent a higher proportion of the more complicated surgeries performed within the VHA.
In addition to surgical characteristics, patients’ sex, race, marital status, homeless status, health care system factors (ie, length of stay in the index hospitalization and arrangement of postdischarge care setting), location of patient (rural vs. urban), and mental health diagnosis history were collected from the Veteran Affairs (VA) Corporate Data Warehouse (CDW).18 Thirty-day postdischarge readmission was also captured from the CDW.
The independent variable in this cohort study was homeless status around the time of the surgical procedure. A Veteran is considered homeless by the VHA if they lack an adequate night time residence. This is the definition of homelessness in The McKinney Homeless Assistance Act as amended by S. 896 The Homeless Emergency Assistance and Rapid Transition to Housing Act of 2009.19 For our study, we identified patients as homeless by the presence of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes or clinic stop codes (also referred to as decision support system outpatient identifiers) related to homelessness in the 30 days before hospital admission through 30 days posthospital discharge.20 The ICD-9-CM codes used included: V60.0 (lack of housing), V60.1 (inadequate housing), V60.89 (other specified housing or economic circumstances), and V60.9 (unspecified housing or economic circumstances). Clinic stops are defined by the VHA to help identify what recourses their patients at specific facilities require to better facilitate resource allocation. The clinic stops used were: 522 [Department of Housing and Urban Development VA-shared housing (HUD-VASH)], 528 (telephone/homeless mentally ill), 529 (health care for homeless Veterans), 530 (telephone/HUD-VASH), and 590 (community outreach to homeless Veterans by staff). In addition, such codes do not assure literal homelessness on the day of surgery, but they do indicate that the patient was seeking VA care possibly for their homelessness or other medical reasons within a time frame close in time to the day of surgery, and align well with prior studies of homelessness with VA data.21,22
While VASQIP captures a substantial number of nurse-abstracted preoperative risk factors and postoperative outcomes, the CDW was used to identify additional clinical and mental health comorbidities. The clinical comorbidities we investigated are listed in Table 1. We identified any of the following mental health ICD-9-CM codes in the 6 months before a hospital admission for surgery as comorbidities: (depression: 296.2, 296.3, 296.82, 298.0, 300.4, 301.12, 309.0, 309.1, 309.28, 311.0; anxiety: 300.0, 300.2. 309.20. 309.21. 309.24; post traumatic stress disorder: 309.81; bipolar disorder: 296.1, 296.4, 296.5-8; and psychosis or schizophrenia: 295, 297.1, 297.3, 298.8, 298.9, 301.22). Behavioral health measures included indicators for tobacco dependence (305.1, V15.82, or use of smoking cessation services), recent alcohol abuse/dependence, defined as >2 drinks per day in the 2 weeks before surgery (291, 303, 305.0), substance use disorders (including alcohol and all recreational drugs ICD-9-CM codes as demonstrated in Derrington et al)23,24 in the 2 years before readmission. Psychosocial dimensions of a patient were defined as factors that influenced a patient’s psychological/psychiatric well-being and/or social well-being.
We also investigated additional perioperative risk factors traditionally identified by VASQIP including Work Relative Value System [relative value unit (RVU)], operation time, and length of stay. VASQIP determines RVUs based on the surgical complexity of current procedural terminology-coded procedures in terms of intensity of effort and complexity of operation. The highest possible work RVU, based on all available current procedural terminology code fields for each patient, was used.
The primary outcome, 30-day unplanned readmission, was identified from the CDW. Unplanned readmissions were those remaining after excluding Centers for Medicare & Medicaid Services-defined planned procedures according to the current algorithm25 which considers organ transplants, maintenance chemotherapy, rehabilitation, and other planned inpatient procedures not related to an acute complication of care to be planned readmissions. Predischarge complications were identified from VASQIP postoperative complications identified while the patient was still admitted to the hospital. We did not use VASQIP postoperative results to assess postdischarge complications and reasons for readmission, as VASQIP looks at data within 30 days of surgery, which would include the patient’s hospital course. Secondary outcomes of interest included the occurrence of a VASQIP-identified postdischarge surgical complication or a postdischarge emergency department (ED) visit per clinic stop code (eg, CL 130).
Discharge destination was categorized as discharge to the community, nursing home, domiciliary/boarding home, or other (irregular, other placement, VA-Paid Home, Home-Basic Primary Care, Hospice, Medicare Home Health, and Other Agency Home Health/Placement). Region and surgery timing (season) were calculated from available variables within VASQIP (ie, facility identifier and operation date). We also identified the teaching status of hospitals in which the surgeries occurred by manually identifying hospitals that had associations with residency programs.
Bivariate frequencies were used to compare demographic, preoperative, operative, and postoperative characteristics between homeless patients and housed patients. χ2 tests were used to assess relationships among categorical variables and homeless status, while t tests were used to compare continuous variables by homeless status. Differences between predictors of readmission in homeless and housed individuals were examined using a stratified analysis. Two-way interactions for differences between predictors of readmission in homeless and housed individuals were examined for all covariates. Multivariate logistic regression with generalized estimating equations to account for clustering within facilities was used to estimate the final adjusted model of predictors of unplanned readmissions among individuals stratified by homelessness. All variables assessed in bivariate analyses were included in the final adjusted model. The correlation matrix for this analysis was autoregressive. All analyses were conducted using SAS version 9.4 and an α of 0.05 was statistically significant.
Characteristics of Homeless Veterans Undergoing Surgery
Our sample included 237,441 surgeries and 199,879 unique patients: 42.8% orthopedic, 39.2% general, and 18.1% vascular as previously described by Morris et al.16 A total of 5068 (2.1%) operations were performed on 4309 homeless patients, with 10.1% classified as emergent procedures compared with 8.9% in the housed group (P<0.01, Table 1). Homeless individuals were younger (56±1.1 vs. 64±11 y, P<0.01), had fewer medical comorbidities (eg, chronic obstructive pulmonary disease: 13.2% vs. 15.0%; hypertension: 57.1% vs. 70.8%; P<0.01), but more psychiatric comorbidities (57.3% vs. 25.9% with at least one, P<0.01), tobacco dependence (62.8% vs. 32.3%, P<0.01), substance use disorder (36.9% vs. 8.5%, P <0.001), and recent alcohol abuse (13.4% vs. 7.9%, P<0.01). Surgeries performed on homeless patients were less complex (RVU: 17.6±0.3 vs. 19.6±7.4, P<0.01) with shorter operation times (2.4 vs. 2.5 h, P<0.01, Table 2) and longer postoperative lengths of stay (8.1. 6.8 d, P<0.01, Table 2).
Differences in Postdischarge Health Care Utilization
The median time to readmission was 8 days postdischarge, regardless of homelessness status. In unadjusted analyses, homeless individuals were more likely to experience an unplanned readmission within 30 days postdischarge [13.3% vs. 9.3%; odds ratio (OR), 1.43; 95% confidence interval (CI), 1.30–1.56; P<0.001] and to return to the ED within 30 days of discharge as compared with their housed counterparts (23.7% vs. 14.2%; OR, 1.87; CI, 1.75–2.00, P<0.01, Table 2). In contrast, homeless individuals experienced fewer predischarge complications (5.2% vs. 5.9%, P=0.05, Table 2) and were less likely to experience a postdischarge complication (5.4% vs. 6.3%; OR, 0.86; 95% CI, 0.76–0.97, P=0.02, Table 2), despite having higher readmission rates. In addition, Homeless patients were more likely to be readmitted in the summer months [July to September (15.7%, P<0.01)], while their housed counterparts’ readmission rate minimally varied throughout the year (9.1%–9.6%, P=0.03, Table 3). Homeless patients were more likely to be readmitted if they had evidence of recent alcohol abuse (16.5%, P=0.01, Table 3).
The most notable predictor for readmission was discharge destination in our unadjusted analysis (Table 3). In contrast to housed patients, homeless patients were more likely to be readmitted when discharged to the community rather than a domiciliary or nursing home (14.1% vs. 11.0% and 8.8%, respectively, P<0.01, Fig. 1). Compared with homeless patients discharged to the community, homeless patients discharged to nursing homes were older (69±11.3 vs. 63±11.1 y), more likely to have orthopedic surgeries, and more likely to be partially dependent. The opposite was seen in their housed counterparts, with the lowest readmission rates among patients discharged to the community as compared with a domiciliary or nursing home (9.0% vs. 13.6% and 11.9%, respectively, P<0.01, Fig. 1).
Wound complication was a common readmission-related code among all patients. Homeless and housed patients discharged to the community had similar rates of predischarge wound complications (1.6% vs. 1.5%, P=0.59); however, homeless-experienced patients had slightly higher rates of postdischarge wound complication–related readmissions compared with their housed counter parts (3.7% vs. 2.8%; OR, 1.35; P<0.01). Over the course of the year wound complication rates did not significantly vary in either population. Regardless of housing status, 2 of 3 patients who developed a wound complication were readmitted.
In the final adjusted model, homeless-experienced patients were more likely to experience an unplanned readmission even after adjusting for demographics, preoperative comorbidities, and operative characteristics (Table 4). While many demographics and operative characteristics associated with unplanned readmission in the overall population were also associated with readmission at a similar magnitude for homeless patients, there were some differences. Among homeless patients, recent alcohol abuse was more strongly associated with readmission than across the entire study population (OR, 1.45; CI, 1.15–1.84; P<0.01; OR, 1.05; CI 0.99–1.10; P=0.09, respectively, Table 4). However, evidence of a alcohol/substance use disorder was not associated with increased readmission risk in either cohort (OR, 0.88; CI, 0.72–10.7; P=0.19 for homeless, OR, 0.89; CI, 0.85–0.93; P<0.0001 for housed, Table 4). In addition, readmissions among homeless patients were significantly lower for blacks as compared with whites (OR, 0.65; CI, 0.53–0.80; P<0.01, Table 4). How sick the patient was at baseline, represented by American Society Anesthesiologists (ASA) classification, was associated with increased readmission rates within the entire population; however, higher ASA classifications were associated with even greater risks for readmission among homeless patients (OR, 1.86; CI, 1.30–2.68; P<0.01 vs. OR, 1.38; CI, 1.31–1.46; P<0.01, Table 4). Readmissions were significantly higher for homeless patients discharged in July to September (OR, 1.54; CI, 1.16–2.05; P<0.01, Table 4) as compared with patients discharged in January to March. Interestingly, while patients discharged to a nursing home in the overall population were 15% more likely to experience a readmission (OR, 1.15; CI, 1.08–1.23; P<0.01, Table 4), homeless patients discharged to a nursing home were less likely to experience a postdischarge readmission as compared with patients discharged to the community (OR, 0.57; 95% CI, 0.44–0.74; P<0.001, Table 4).
Our study found that homeless patients have an increased risk of postoperative readmission despite having lower measured postoperative complication rates. Our findings are consistent with prior reports that homeless patients are at greater risk for hospital readmission than their housed counterparts;11,12 however, the magnitude of the risk was lower in our study.11,12 This lower risk of readmission seen in our cohort may be influenced by the study population (medical vs. surgical), and length of follow-up (30 vs. 90-d). Prior work by Kertesz and colleagues did not distinguish what proportion of their population was surgical versus medical. Another difference relates to the timing of the readmission.22 Doran et al12 described a single-site study in an urban center and found a 30-day readmission rate of 50% in a population made up of mostly medical patients (92% medical and 8% surgical). Their study was single-site and started in the summer, as part of a joint hospital and community-wide effort to screen all patients for homelessness. Kertesz et al22 also conducted a single-center study in Boston among nonmaternity medical or surgical patients who had been discharged to respite housing programs with access to outpatient clinics; they found a 90-day readmission rate of 21%. We believe our 30-day readmissions rate of 13% captures readmissions related to patients’ index surgical hospitalization. This belief is supported by prior studies including Merkow et al26 who found after surgery readmissions within 30 days of index operation were mostly associated with postdischarge complications related to the procedure and not with exacerbation of preadmission complications. Therefore, the differences in magnitude in readmission rates compared with prior studies may in part be explained by our study setting given that VA patients have access to significantly more outreach programs and have fewer barriers to health care than homeless patients at large.27
The strongest predictors for readmission among homeless surgical patients were discharge destination, recent alcohol abuse, season of discharge, race, and ASA classification. Discharge to locations other than the community was associated with lower risk for readmission, highlighting the importance of discharging patients to supportive environments where they will receive ongoing services to stabilize their medical and caregiver needs. Among homeless patients, those discharged to nursing homes were older and sicker compared with those discharged to the community, yet their readmission rates were lower than those discharged to the community. The opposite was true for their housed counterparts. This supports prior literature’s findings that the lack of having a place to live perpetuates hospitalizations.
Homeless patients may require more health care system resources to make up for a lack of social and tangible support. Alcohol abuse within 2 weeks of an operation was significantly associated with readmissions in homeless patients, but not in their housed counter parts. It is unknown if the same is true for recent drug abuse as the data are not available. Interestingly, this significant correlation between recent alcohol abuse was not observed in patients with long-term substance/alcohol use disorder(s). This suggests recent alcohol abuse before surgery significantly increases a homeless patient’s risk for readmission, yet engaging in long-term high risk behaviors (ie, alcohol/drug use disorders) does not increase a homeless patient’s risk to the same extent. Therefore, patients with recent alcohol abuse should be considered for rehabilitation either preoperatively or at hospital discharge.
Homeless readmission rates were the highest in the summer months, which would lead one to speculate that more homeless patients were discharged to the community in the summer, but the ratio of patients discharged to each type of location remained constant throughout the year (results not reported). Region of the country was not associated with readmission rates regardless of time of year. In addition, wound complication rates did not vary with the seasons in either population, in contrast to prior literature.28
Wound complication was a commonly associated with readmission, regardless of housing status; however, homeless patients discharged to the community had higher rates of wound complications requiring readmission compared to their housed counterparts. This difference arose despite similar predischarge complication rates between the 2 groups. Possibly homeless patients were unable to maintain wound hygiene or lacked the self-care skills needed to avoid wound-related complications. However, once any patient developed a wound complication regardless of homeless status, ∼2 of 3 of them were readmitted. Therefore, it is critical to manage wounds appropriately on an outpatient basis, and address self-care deficits associated with homelessness proactively, as once a patient develops a wound complication their risk for readmission increases dramatically.
Our study has several strengths. This is one of the first studies investigating readmissions among homeless and housed surgical patients. The VA population includes a relatively high proportion of homeless individuals. This allowed us adequate power to examine small differences in predictors of readmission between housed and homeless individuals. In addition, the VA health care system provides continuity of care, which allows us to follow patients over time and across EDs and inpatient facilities. Lastly, we were able to take advantage of the rich clinical and demographic data available through VASQIP, a chart abstracted dataset, and the VA’s CDW of data extracted from the comprehensive VA electronic medical record.
While the VA is an excellent resource for scientific research, our results may have limited generalizability in the private sector as women are under-represented and the age distribution reflects military recruitment trends rather than population growth trends. VA patients also tend to have worse health and lower socioeconomic status than non-VA Veterans or civilians.
Although homelessness status was defined by ICD-9-CM codes and clinic stop codes in attempts to minimize having an ascertainment bias, it is possible we underestimated the number of patients who were homeless. This study found significant differences in readmission rates depending on the patient’s discharge location and if they had recent alcohol abuse; however, we were unable to identify whether patients with substance use disorders or mental health disorders were discharged to locations with appropriate services. With VASQIP we are limited in only identifying resources available to our patient within the health care system and as a result are unable to determine if the patient had any social supports outside of the hospital system. Furthermore, we did not capture care that occurred outside the VA, and it is possible surgical patients, homeless and housed, had non-VA ED visits and admissions. Despite it being well documented that homeless patients have more comorbities, we found homeless VA patients have fewer documented comorbities compared with their housed counterparts, which may in part be due to their underutilization primary care services and under diagnosis.29
In summary, within the VA system, the 30-day hospital readmission rate among surgical patients who were homeless was higher than among their housed counterparts. Lack of housing significantly impacts a homeless patient’s risk for readmission. Discharge to the community, homelessness status, race, recent alcohol abuse, and elevated ASA classification were the most significant risk factors associated with readmission for homeless patients. A holistic evaluation of homeless surgical patients, to include psychosocial support and housing needs assessment may provide opportunities to improve patient care.
1. Henry M, Wat RT, Rosenthal L, et al. Abt Associates. The 2017 Annual Homeless Assessment Report (AHAR) to Congress. The US Department of Housing and Urban Development. 2017.
2. Henry M, Watt R, Rosenthal L, et al. Abt Associates. The 2016 Annual Homeless Assessment Report (AHAR) to Congress (Part 1). United States Department of Housing and Urban Development, Washingon, DC. 2016. Available at: www.hudexchange.info/resources/documents/2016-AHAR-Part-1.pdf
. Accessed November 5, 2017.
3. Schanzer B, Dominguez B, Shrout PE, et al. Homelessness, health status, and health care use. Am J Public Health. 2007;97:464–469.
4. D’Amore J, Hwang O, Chiang W, et al. The epidemiology of the homeless population and its impact on an urban emergency department. Acad Emerg Med. 2001;8:1051–1055.
5. Desai RA, Rosenheck RA, Agnello V. Prevalence of hepatitis C virus infection in a sample of homeless Veterans. Soc Psychiatry Psychiatr Epidemiol. 2003;38:396–401.
6. Eyrich-Garg KM, Cacciola JS, Carise D, et al. Individual characteristics of the literally homeless, marginally housed, and impoverished in a US substance abuse treatment-seeking sample. Soc Psychiatry Psychiatr Epidemiol. 2008;43:831–842.
7. Haddad M, Wilson T, Ijaz K, et al. Tuberculosis and homelessness in the United States, 1994-2003. JAMA. 2005;293:2762–2766.
8. Lee T, Hanlon J, Ben-David J, et al. Risk factors for cardiovascular disease in homeless adults. Circulation. 2005;111:2629–2635.
9. Bao Y, McGuire TG, Chan YF, et al. Value-based payment in implementing evidence-based care: the Mental Health Integration Program in Washington state. Am J Manag Care. 2017;23:48–53.
10. Hwang SW, Kirst MJ, Chiu S, et al. Multidimensional Social Support and the Health of Homeless Individuals. J Urban Health. 2009;86:791–803.
11. Saab D, Nisenbaum R, Dhalla I, et al. Hospital readmissions in a community-based sample of homeless adults: a matched-cohort study. J Gen Intern Med. 2016;31:1011–1018.
12. Doran KM, Ragins KT, Iacomacci AL, et al. The revolving hospital door: hospital readmissions among patients who are homeless. Med Care. 2013;51:767–773.
13. US Department of Health and Human Services. New HHS data shows major strides made in patient safety, leading to improved care and savings. 2014. Available at: http://innovation.cms.gov/Files/reports/patient-safety-results.pdf
. Accessed November 5, 2017.
14. Centers for Medicare and Medicaid Services. Better care, smarter spending, healthier people: improving our health care delivery system. 2015. Available at: www.cms.gov/Newsroom/MediaReleaseDatabase/Fact-sheets/2015-Fact-sheets-items/2015-01-26.html
. Accessed November 5, 2017.
15. Khuri SF, Daley J, Henderson W, et al. The Department of Veterans Affairs’ NSQIP: the first national, validated, outcome-based, risk-adjusted, and peer-controlled program for the measurement and enhancement of the quality of surgical care. National VA Surgical Quality Improvement Program. Ann Surg. 1998;228:491–507.
16. Morris MS, Graham LA, Richman JS, et al. Postoperative 30-day readmission: time to focus on what happens outside the hospital. Ann Surg. 2016;264:621–631.
17. Khuri SF, Daley J, Henderson WG. The comparative assessment and improvement of quality of surgical care in the Department of Veterans Affairs. Arch Surg. 2002;137:20–27.
18. Copeland LA, Graham LA, Richman JS, et al. A study to reduce readmissions after surgery in the Veterans Health Administration: design and methodology. BMC Health Serv Res. 2017;17:198.
19. The McKinney-Vento Homeless Assistance Act As amended by S. 896 The Homeless Emergency Assistance and Rapid Transition to Housing (HEARTH) Act of 2009. 2009. Available at: www.hudexchange.info/resources/documents/HomelessAssistanceActAmendedbyHEARTH.pdf
. Accessed November 5, 2017.
20. Peterson R, Gundlapalli AV, Metraux S, et al. Identifying homelessness among Veterans using VA administrative data: opportunities to expand detection criteria. PLoS One. 2015;10:e0132664.
21. Tsai J, Rosenheck RA. Risk factors for homelessness among US veterans. Epidemiol Rev. 2015;37:177–195.
22. Kertesz SG, Holt CL, Steward JL, et al. Comparing homeless persons’ care experiences in tailored versus nontailored primary care programs. Am J Public Health. 2013;103(suppl 2):S331–S339.
23. Derrington TM, Bernstein J, Belanoff C, et al. Refining measurement of substance use disorders among women of child-bearing age using hospital records: the development of the Explicit-Mention Substance Abuse Need for Treatment in Women (EMSANT-W) Algorithm. Matern Child Health J. 2015;19:2168–2178.
24. Elixhauser A, Steiner CA, Palmer L. Clinical Classifications Software (CCS) US Agency for Healthcare Research and Quality. 2012. Available at: www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp
. Accessed November 5, 2017.
25. Horwitz LI, Grady JN, Cohen DB, et al. Centers for Medicare & Medicaid services planned readmission algorithm—version 2.1. 2013.
26. Merkow RP, Ju MH, Chung JW, et al. Underlying reasons associated with hospital readmission following surgery in the United States. JAMA. 2015;313:483–495.
27. Watkins KE, Smith B, Akincigil A, et al. The quality of medication treatment for mental disorders in the Department of Veterans Affairs and in Private-Sector Plans. Psychiatr Serv. 2016;67:391–396.
28. Sagi HC, Cooper S, Donahue S, et al. Seasonal variations in posttraumatic wound infections after open extremity fractures. J Trauma Acute Care Surg. 2015;79:1073–1078.
29. Kushel MB, Perry S, Bangsberg D, et al. Emergency department use among the homeless and marginally housed: results from a community-based study. Am J Public Health. 2002;92:778–784.