Weaver, Kristin J. MD, PhD*; Neal, Dan PhD*; Hoh, Daniel J. MD*; Mocco, J MD, MS*; Barker, Fred G. II MD‡; Hoh, Brian L. MD*
The “July phenomenon” is a belief that there are worse patient outcomes in teaching hospitals during the month of July, a month in which recent medical graduates start residency and current residents advance to the next training year with increasing responsibilities. The literature has been inconsistent regarding whether a July phenomenon exists.1-13 The “July phenomenon” is just one of several resident training issues, including resident duty hours and fatigue, that have been raised as possible contributors to medical errors and poor patient outcomes.14-19
Our hypothesis is that there is no “July phenomenon” in neurosurgical training hospitals. We tested our hypothesis by analyzing the Nationwide Inpatient Sample (NIS) database, the largest all-payer inpatient care database in the United States, for differences in mortality and complications in patients with neurosurgical diagnoses in the month of July compared with other months at neurosurgical teaching hospitals and then performed the same analysis in nonteaching hospitals to serve as an experimental control. We chose the neurosurgical diagnoses to be included in the analysis based on the following criteria: (1) diagnoses affecting hospital inpatients because neurosurgical residents are most involved and most likely to contribute to complications and medical errors in the inpatient setting; (2) common disorders at neurosurgical training institutions so there would be adequate numbers of patients in respective cohorts; (3) disorders with reasonable likelihood for complications or medical errors so there would be a sufficient event rate; and (4) disorders that are representative of a broad spectrum of neurosurgical patients treated by neurosurgery residents and not subspecialty specific. The 4 neurosurgical diagnoses that we included in the analysis are nontraumatic central nervous system (CNS) hemorrhage, CNS trauma, CNS tumors, and hydrocephalus.
PATIENTS AND METHODS
We obtained the NIS database from the Agency for Healthcare Quality and Research's Healthcare Cost and Utilization Project (Rockville, Maryland). The NIS contains data approximating a 20% stratified sample of hospitals in the United States. For more information regarding the NIS database, visit www.hcup-us.ahrq.gov/nisoverview.jsp.
Hospitalizations for nontraumatic hemorrhage, CNS trauma, CNS tumors, and hydrocephalus from 1998 to 2008 were collected from the NIS core database by International Classification of Diseases, 9th Revision codes (nontraumatic hemorrhage 430, 431, 423.0-.9; CNS Trauma 800.0-.9, 192.0-.9, 803.0-.9, 804.0-.9, 805.0-.9, 806.0-.9, 851.0-.9; 852.0-.9, 853.0-.9, 854.0-.9, 852.0-.9; CNS Tumors 191.0-.9, 192.0-.9, 198.3, 225.0-.9, 237.0, 237.1, 237.1, 237.5, 239.6; Hydrocephalus 331.3, 331.4, 742.3, 741). For the years 2007 to 2008, we combined these data with corresponding data from the NIS hospital database. For other years, hospital information is included in the main NIS yearly data files.
The analysis was adjusted for the following patient-specific factors, which are coded in the NIS database: sex, age, race, median income level in patient's postal (zip) code (low income, low to middle income, middle to upper income, and upper income), and payer (Medicare, Medicaid, private insurance, self-pay, or no charge). The models also accounted for hospital-level factors: hospital region (Northeast, Midwest, South, and West), bed size (small, medium, or large), hospital location (rural or urban location), annual hospital volume of patients, and primary payer (Medicare, Medicaid, private insurance, self-pay, no charge, other).
The 2 clinical outcomes of mortality and complications were evaluated for each diagnosis in teaching hospitals in July compared with other months. The complications evaluated were bleeding, intraoperative or postoperative (998.11); foreign body inadvertently left in wound (998.4); medical care or postoperative nervous center complication (997.00); hematoma intraoperative or postoperative (998.12); therapeutic misadventure (999.9); medicine poisoning, wrong substance given or taken in error (977.9); accidental puncture or laceration (998.2); postoperative respiratory complication (997.3); ventricular shunt complication or infection (996.75, 996.63); and therapeutic misadventure, surgical treatment (998.9).
We used the SAS statistical software package version 9.1 (SAS Institute, Cary, North Carolina) to calculate all descriptive statistics and perform all analyses. For each combination of hospital type and diagnosis category, we used a generalized linear mixed model (SAS PROC GLIMMIX) to determine whether mortality or complications rates differed in July for that hospital type and diagnosis category. In all analyses, either mortality (yes or no) or complications of any kind (yes or no) was our outcome variable, so we modeled both responses as binary variables and used a logit link function. Our primary predictor variable was month of hospitalization (July or other), and we included as covariates sex, hospital case volume (ie, the number of cases with the diagnosis of interest treated at the hospital during the year of the observation), hospital size (small, medium, or large), hospital location (rural or nonrural), hospital region (North, South, East, or West), patient's primary payer (Medicare, Medicaid, private insurance, self-insured, no charge, or other), patient age, patient race, and year of hospitalization. We included hospital as a random factor in the analysis to account for the fact that NIS data contain multiple observations on individual hospitals, and we retained the default convergence criteria set by the SAS system when fitting the model. To assess model fit, we evaluated the generalized χ2/df fit statistic generated by PROC GLIMMIX, which was near 1 for all models.
A search of the NIS database for the years 1998 to 2008 yielded a total of 858 222 total admissions: 129 927 nontraumatic hemorrhage, 90 133 CNS trauma, 204 150 CNS tumor, and 71 687 hydrocephalus in teaching hospitals; and 91 179 nontraumatic hemorrhage, 93 757 CNS trauma, 138 328 CNS tumor, and 36 287 hydrocephalus in nonteaching hospitals. The patient demographic and hospital characteristics of admissions for July vs all other months in teaching hospitals are shown in Table 1. The patient demographic and hospital characteristics of admissions for July vs all other months in nonteaching hospitals are shown in Table 2. With only very small deviations, the 2 groups do not differ significantly.
TABLE 1-b Patient De...Image Tools
TABLE 2-b Patient De...Image Tools
In a generalized linear mixed-model analysis, the probability of dying in a teaching hospital was not significantly different in July compared with other months for nontraumatic hemorrhage (P = .071), CNS trauma (P = .485), CNS tumor (P = .578), and hydrocephalus (P = .151) (Table 3). The probability of dying in a nonteaching hospital showed similar findings in each diagnosis: nontraumatic hemorrhage (P = .168), CNS trauma (P = .568), CNS tumor (P = .544), and hydrocephalus (P = .581) (Table 4).
In a generalized linear mixed-model analysis, the probability of any complication in the month of July vs any other month in a teaching hospital was not statistically different for any of the 4 diagnoses (nontraumatic hemorrhage [P = .529], CNS trauma [P = .378], CNS tumor [P = .461], hydrocephalus [P = .441]) (Table 3). Similarly, the probability of any complication in the month of July vs any other month in a nonteaching hospital was not statistically different for nontraumatic hemorrhage (P = .104), CNS trauma (P = .012), CNS tumor (P = .886), and hydrocephalus (P = .104) (Table 4).
The “July phenomenon” has been studied across different medical specialties including general surgery, obstetrics and gynecology, and internal medicine.1-13 The evidence of or against the presence of a “July phenomenon” has been inconsistent. The majority of studies of the “July phenomenon” have found no difference in patient outcomes in July compared with other months. Specifically, Buchwald et al4 evaluated common medical and surgical indications for admission to a teaching hospital. For neck and back problems, they found no increase in the length of stay or total charges for July admissions compared with May. Banco et al20 specifically looked at the incidence of perioperative spinal infection rates at different times during the academic year. They found that there was no association between infection rates in the month of July or other times of the year when residents or fellows were new on the service. Finkielman et al6 found that there was no increase in mortality, length of stay, or discharge disposition in the intensive care unit of an academic medical center from 1994 to 2002.
Several studies, on the other hand, have suggested a “July phenomenon.” Blumberg3 found an increase in the surgical mortality rate for July medium-risk nonelective and nontrauma high-risk emergency admissions from 1984 to 1985. Another group found a slight increase in overall surgical mortality during the years 1991 to 1997 at teaching hospitals in Ohio.2 Walling and Veremakis13 found that first-year residents during early months of their training compared with later in the year made more prescribing errors. Last, Inaba et al21 demonstrated an increased risk of errors resulting in preventable or potentially preventable complications at the beginning of the academic year (July and August) compared with the end of the academic year (May and June) at an academic level 1 trauma center.
To our knowledge, there are only 2 neurosurgical studies of the “July phenomenon” in the literature. Smith et al12 used the NIS to determine whether more adverse endpoints occur during the months of July and August in pediatric brain tumor or shunt surgery. This group looked at mortality rate, discharge disposition, neurological complications, transfusion rates, hospital charges, length of stay, and operative hematoma; they found no evidence of worse outcomes or inefficient care in pediatric neurosurgery at teaching hospitals during July and August. Some of the data from Kestle et al7 on the outcomes related to shunt surgery suggested a “July phenomenon,” although the effect was small. Given the limited data on the “July phenomenon” in the field of neurosurgery, this study was used to further address this question.
In this retrospective, cohort study of greater than 850 000 admissions, we found that there was no increase in the mortality or complication rate for the month of July in teaching hospitals compared with all other months for the diagnoses nontraumatic hemorrhage, CNS trauma, CNS tumor, and hydrocephalus.
In this study, there are several limitations. First, this study is a retrospective study, which has the potential for selection bias. However, a study of mortality and complications during different times of the year cannot be prospective in nature. This type of study is also prone to confounding; however, with only very small deviations, the 2 groups do not differ significantly in the demographic characteristics. Interestingly, in all analyses, approximately 30% of observations were excluded because they were missing information on at least 1 of the covariates. Investigation showed that most (∼93%) of the excluded observations were missing information on patient race. Further investigation showed that the 2 analysis groups (July vs other months) were not imbalanced on the missing data. The proportion of observations missing race information that occurred in July often exactly coincided with the proportion of observations that occurred in July in the total dataset. Hence, we believe that our analysis groups are free of any bias attributable to missing data. Details of the analysis of missing observations for mortality and complications are shown in Tables 5 and 6, respectively.
Another limitation is the risks inherent to coding error. Given the size of the NIS database, there is the potential for inconsistent, inaccurate, and/or incomplete coding variability or error in patient characteristics, diagnoses, and hospital characteristics.
This is indeed a limitation of the NIS database, but we have attempted to account for this as much as possible, and the large numbers of patients in each cohort (July vs other months) should mitigate isolated cases of coding inaccuracy and still prove our hypothesis. Moreover, the list of complications and medical errors included in the analysis is a good representative sample of the types of complications and medical errors that would be committed by a neurosurgery resident.
Another limitation to this study is the criteria used to define the patient population. Specifically, our data are limited to in-hospital observations as collected in the NIS database. Therefore, mortality and complications occurring after discharge would not be captured by the NIS database. We believe that this should not significantly affect the findings of our study, however, because complications and medical errors committed by neurosurgery residents in July vs other months are most likely going to be in-hospital and not after discharge.
Moreover, although there was no increase in mortality or complication rate, this study did not account for “near misses.” These events are not recorded in the NIS database and therefore not included in our study. It is reasonable to assume that at the start of an academic year, the number of errors made by residents would be higher compared with later months. It is possible that these mistakes result in events that do not affect the mortality and complication rates and therefore are not able to be monitored. Alternatively, the lack of a “July phenomenon” may be attributable to the increased supervision by faculty and more senior residents during the start of the academic year. This has been an argument as to why the ACGME resident duty hour reform has not translated into any reduction in medical errors or mortality, as discussed by Pape and Pfeifer.22 Nevertheless, even with the limitations, our study adds to the very limited field of knowledge on the “July phenomenon” in the field of neurosurgery.
No “July phenomenon” was found for neurosurgical mortality or complications in patients with nontraumatic hemorrhage, CNS trauma, CNS tumor, or hydrocephalus.
Dr Mocco is a consultant to Lazarus Effect Inc, Edge Therapeutics, and NFocus and an investor/stockholder in Blockade Therapeutics. The other authors have no personal financial or institutional interest in any of the drugs, materials, or devices described in this article.
1. Aylin P, Majeed FA. The killing season—fact or fiction. BMJ. 1994;309(6970):1690.
2. Barry WA, Rosenthal GE. Is there a July phenomenon? The effect of July admissions on intensive care morality and LOS in teaching hospitals. J Gen Intern Med. 2003;18(8):639–645.
3. Blumberg MS. Measuring surgical quality in Maryland: a model. Health Aff (Millwood). 1988;7(1):62–78.
4. Buchwald D, Komaroff AL, Cook EF, Epstein AM. Indirect costs for medical education. Is there a July phenomenon? Arch Intern Med. 1989;149(4):765–768.
5. Claridge JA, Schulman AM, Sawyer RG, Ghezel-Ayagh A, Young JS. The ‘July phenomenon’ and the care of the severely injured patient: fact or fiction? Surgery. 2001;130(2):346–353.
6. Finkielman JD, Moreles J, Peters SG, et al.. Mortality rate and length of stay of patients admitted to the intensive care unit in July. Crit Care Med. 2004;32(5):1161–1164.
7. Kestle JR, Cochrane DD, Drake JM. Shunt insertion in the summer: is it safe? J Neurosurg. 2006;105(3 suppl):165–168.
8. Myles TD. Is there an obstetric July phenomenon? Obstet Gynecol. 2003;102(5 pt 1):1080–1084.
9. Rich EC, Gifford G, Luxenberg M, Dowd B. The relationship of house staff experience to the cost and quality of inpatient care. JAMA. 1990;263(7):953–957.
10. Rich ED, Hillson SD, Dowd B, Morris N. Specialty differences in the “July Phenomenon” for Twin Cities teaching hospitals. Med Care. 1993;31(1):73–83.
11. Shulkin DJ. The July phenomenon revisited: are hospital complications associated with new house staff? Am J Med Qual. 1995;10(1):14–17.
12. Smith ER, Bulter WE, Barker FG. Is there a “July phenomenon” in pediatric neurosurgery at teaching hospitals? J Neurosurg. 2006;105(3 suppl):169–176.
13. Walling HW, Veremakis C. Ordering errors by first-year residents: evidence of learning from mistakes. Mo Med. 2004;101(2):128–131.
14. Barger LK, Cade BE, Ayas NT, et al.. Extended work shifts and the risk of motor vehicle crashes among interns. N Engl J Med. 2005;352(2):125–134.
15. Davydov L, Caliendo G, Mehl B, Smith LG. Investigation of correlation between house-staff work hours and prescribing errors. Am J Health Syst Pharm. 2004;61(11):1130–1134.
16. Friedman WA. Resident duty hours in American neurosurgery. Neurosurgery. 2004;54(4):925–933.
17. Landrigan CP, Rothschild JM, Cronin HW, et al.. Effect of reducing interns' work hours on serious medical errors in intensive care units. N Eng J Med. 2004;351(18):1838–1848.
18. Lockley SW, Cronin JW, Evans EE, Cade BE, Lee CJ, Landrigan CP. Effect of reducing interns' weekly work hours on sleep and attentional failures. N Engl J Med. 2004;351(18):1829–1837.
19. Zhan C, Miller MR. Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization. JAMA. 2003;290(14):1868–1874.
20. Banco SP, Vaccar AR, Blam O, et al.. Spine infections: variations in incidence during the academic year. Spine. 2002;27(9):962–965.
21. Inaba K, Recinos G, Teixeira PG, et al.. Complications and death at the start of the new academic year: is there a July phenomenon? J Trauma. 2010;68(1):19–22.
22. Pape HC, Pfeifer R. Restricted duty hours for surgeons and impact on residents quality of life, education, and patient care: a literature review. Patient Saf Surg. 2009;3(1):3.
The authors present an analysis of the Nationwide Inpatient Sample (NIS) database to compare mortality and in-hospital complications between July and other months in broad categories of neurosurgical admissions to see whether there is support for the concept of a “July phenomenon.” Using generalized linear mixed-model analysis, they found no difference in mortality or complications in July compared with other months. Their models adjusted for factors such as demographics, geographic location, and blunt measures of socioeconomic status including zip code and payer status. They used elegant statistical techniques within the confines of a large database in an attempt to answer an intriguing concept that has bearing on the delivery of quality care. The idea of a July phenomenon certainly on the surface seems like a credible concept given the influx of inexperienced clinicians to the patient-care team and promotion of more junior residents to more senior levels with evolving responsibilities. However, one might anticipate, particularly for mortality and major complications, that increased oversight from more senior residents and faculty likely offsets these effects during this annual period of transition. We imagine that the July phenomenon could be a bigger issue for the more nuanced components of surgical care delivery, such as near misses, process adherence, and efficiency. These nuances would not be detectable using a large database such as the NIS. Furthermore, the ICD-9 codes used to categorize complications in this study may not encompass all major complications. Neurosurgical training, like many surgical specialties, is renowned for chain-of-command supervision and multiple levels of oversight. Could this training model contribute to the lack of confirmation of a phenomenon that intuitively seems likely to occur? Furthermore, it is our belief that a July phenomenon, if it does exist, may actually occur in March-April, when oversight is relaxed because of attendings' growing confidence of trainees' skills and at a time of the year when residents' growing confidence in patient care and/or increasing emotional factors may, in fact, lead to errors in judgment, as illuminated in Groopman's book How Doctors Think.1 It would be interesting to evaluate these effects, not only in neurosurgery, but in other surgical and medical specialties as well.
Judith M. Wong
A. John Popp
1. Groopman JE. How Doctors Think. Boston, MA: Houghton Mifflin; 2007. Cited Here...