Advertisement

Sharps-related injuries in health care workers: a case-crossover study

      Injuries caused by sharp medical devices are common among health care workers in the United States. The best available data suggest that between 400,000 and 800,000 such injuries occur in hospitals each year (
      National Institute of Occupational Safety and Health, Department of Health and Human Services
      ,
      • Henry K
      • Campbell S
      Needlestick/sharps injuries and HIV exposure among health care workers. National estimates based on a survey of U.S. hospitals.
      , ). These injuries are a source of concern because of their potential to transmit various infectious agents, including hepatitis B virus (
      • Shapiro C
      Occupational risk of infection with hepatitis B and hepatitis C virus.
      ), hepatitis C virus (
      • Lanphear B
      • Linemann C
      • Cannon C
      • DeRonde M
      • Pendy L
      • Kerley L
      Hepatitis C virus infection in health care workers risk of exposure and infection.
      ), and human immunodeficiency virus (HIV) (
      Centers for Disease Control and Prevention
      Case-control study of HIV seroconversion in health-care workers after percutaneous exposure to HIV-infected blood—France, United Kingdom, and United States, January 1988-August 1994.
      ,
      • Cardo D.M
      • Culver D.H
      • Ciesielski C.A
      • et al.
      A case-control study of HIV seroconversion in health care workers after percutaneous exposure. Centers for Disease Control and Prevention Needlestick Surveillance Group.
      ). The evaluation and treatment of these injuries and subsequent illnesses impose a heavy societal burden in terms of economic cost (
      • Jagger J
      • Bentley M
      • Juillet E
      Direct cost of follow-up for percutaneous and mucocutaneous exposures to at-risk body fluids data from two hospitals.
      ), worker anxiety and distress (
      • David H.T
      • David Y.M
      Living with needlestick injuries.
      , ), and future morbidity (
      National Institute of Occupational Safety and Health, Department of Health and Human Services
      , ,
      • Shapiro C
      Occupational risk of infection with hepatitis B and hepatitis C virus.
      ,
      • Lanphear B
      • Linemann C
      • Cannon C
      • DeRonde M
      • Pendy L
      • Kerley L
      Hepatitis C virus infection in health care workers risk of exposure and infection.
      , ,
      • Mitsui T
      • Iwano K
      • Masuko K
      • et al.
      Hepatitis C virus infection in medical personnel after needlestick accident.
      ,
      • Petrosillo N
      • Puro V
      • Jagger J
      • Ippolito G
      The risks of occupational exposure and infection by human immunodeficiency virus, hepatitis B virus, and hepatitis C virus in the dialysis setting. Italian Multicenter Study on Nosocomial and Occupational Risk of Infections in Dialysis.
      ).
      The current hospital environment presents employees with various challenges, including worker fatigue, rushing, and distraction (
      • Sexton J.B
      • Thomas E.J
      • Helmreich R.L
      Error, stress, and teamwork in medicine and aviation cross sectional surveys.
      ,
      • Chisholm C
      • Collinson E
      • Nelson D
      • Cordell W
      Emergency department workplace interruptions are emergency physicians “interrupt-driven” and “multi-tasking”?.
      ,
      • Gander P.H
      • Merry A
      • Millar M.M
      • Weller J
      Hours of work and fatigue-related error a survey of New Zealand anaesthetists.
      ). We postulated that such brief, transient factors might increase the risk of sharps-related injuries in health care workers. We performed a case-crossover study—a design in which each “case” serves as his or her own control—to identify and quantify transient factors that might increase the risk of such injuries (
      • Maclure M
      • Mittleman M.A
      Should we use a case-crossover design?.
      ,
      • Maclure M
      The case-crossover design a method for studying transient effects on the risk of acute events.
      ). This design has been used to investigate risk factors for injuries and accidents in other settings (
      • Sorock G
      • Lombardi D
      • Hauser R
      • Eisen E
      • Herrick R
      • Mittleman M
      A case-crossover study of occupational traumatic hand injury methods and initial findings.
      ,
      • Reidelmeier D
      • Tibshirani R
      Association between cellular-telephone calls and motor vehicle collisions.
      ,
      • Valent F
      • Brusaferro S
      • Barbone F
      A case-crossover study of sleep and childhood injury.
      ,
      • Petridou E
      • Mittleman M.A
      • Trohanis D
      • et al.
      Transient exposures and the risk of childhood injury a case-crossover study in Greece.
      ,
      • Sorock G
      • Lombardi D
      • Gabel C
      • et al.
      Case-crossover studies of occupational trauma methodological caveats.
      ).

      Methods

      Subjects

      Subjects were health care workers employed at the University of Maryland Medical Center in Baltimore, Maryland, or at the Beth Israel Deaconess Medical Center in Boston, Massachusetts. Employees who reported a sharps-related injury sustained while handling a contaminated medical device were considered eligible for the study, and were invited to participate by occupational health services staff at the time of postexposure risk assessment. Participating subjects were interviewed by telephone using a standardized questionnaire. Employees were considered ineligible for participation if they had been injured by a clean or unused medical device, or if they had suffered a skin or mucus membrane splash with blood or body fluids without skin puncture or laceration. All subjects provided written informed consent, and the study was approved by the institutional review boards of the University of Maryland Medical Center and Beth Israel Deaconess Medical Center.

      Study design

      The case-crossover study design is characterized by self-matching, in which the case and control data are obtained from the same subject (
      • Maclure M
      • Mittleman M.A
      Should we use a case-crossover design?.
      ,
      • Sorock G
      • Lombardi D
      • Gabel C
      • et al.
      Case-crossover studies of occupational trauma methodological caveats.
      ). The use of such an approach permits assessment of the relation between brief, transient exposures (such as fatigue or rushing) and an acute event (in this case, the occurrence of a sharps-related injury). This design adjusts for many of the differences between cases and controls that might confound the results of a more traditional case-control study.
      Between February 2000 and October 2001, 139 subjects who reported sharps-related injuries to employee health services were recruited at the two medical centers. The median time from assessment in employee health services to interview was 3 days, and 90% (n = 123) of subjects were interviewed within a week of injury.

      Exposure assessment

      In all subjects, we assessed rushing, distraction, anger, fatigue, performance of a task in an emergency situation, and teaching. In subjects who were not scrubbed in an operating room or procedure suite at the time of injury, we also assessed repeated attempts to perform procedures, staffing shortages, and the presence of an uncooperative patient. Subjects who were scrubbed in an operating room or procedure suite at the time of injury were asked about the presence of a bloody operative field, excess noise in the operating room or procedure suite, and the performance of highly complex procedures. Trainees were asked whether or not they were being taught at the time of the injury. Subjects were asked whether these factors had been present on the day of the injury, and if so, whether they had been present at the time of the injury.
      Subjects were asked to estimate their average time at risk of sharps-related injuries. Those who were not scrubbed at the time of injury were considered to have a noncontinuous risk of sharps-related injury, and time at risk was estimated as the product of the number of procedures performed in the last month and average procedure duration. Among those who were scrubbed at the time of injury, risk of injury while scrubbed was considered to be continuous, and time at risk was estimated as the number of hours spent scrubbed in an operating room or procedure suite within the past week. The usual frequencies of exposures were estimated as the proportion of time (while at risk of sharps-related injuries) that subjects experienced a given exposure. Example questions are provided in the Appendix.

      Reliability of measurements

      From December 2000 to October 2001, we assessed the reliability (reproducibility) of subjects’ estimates of time at risk, exposure frequency, and presence of exposures at the time of injury in a subset of 40 study participants. At the termination of the original interview, subjects were asked whether they would be willing to participate in a second interview, in exchange for additional reimbursement. They were re-interviewed between 2 and 5 days after their initial interview, using an abbreviated version of the original questionnaire.

      Statistical analysis

      Usual frequency estimates were used to estimate the subject-specific person-time exposed and unexposed to each transient risk factor during the past week (among those at continuous risk) or past month (among those at noncontinuous risk). Data were analyzed using standard methods for case-crossover data (
      • Mittleman M.A
      • Maclure M
      • Robins J.M
      Control sampling strategies for case-crossover studies an assessment of relative efficiency.
      ). We estimated the incidence rate ratio as a measure of relative risk using the Mantel-Haenszel estimators for person-time data (
      • Rothman K
      • Greenland S
      Applications of stratified analysis methods.
      ); its variance was computed using standard methods (
      • Greenland S
      • Robins J
      Estimation of a common effect parameter from sparse follow-up data.
      ).
      We calculated intraclass correlation coefficients for subjects’ estimates of time at risk and exposure frequency using a random-effects model for repeated measurements by a single observer (
      • Fleiss J
      ). Reproducibility of subjects’ assessment of the risk status of the source patient, and of exposure at the time of injury was assessed using the κ statistic (
      • Cohen J
      A coefficient of agreement for nominal scales.
      ) and classified (
      • Landis J
      • Koch G
      An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers.
      ). We tested for an interaction between exposure and whether (or not) the source patient had a known infection with HIV or hepatitis virus (
      • Rothman K
      • Greenland S
      Introduction to stratified analysis.
      ).
      All statistical analyses were performed using SAS, version 8.01 (SAS Institute, Cary, North Carolina), except for reliability analyses, which were performed using SPSS, version 10.0 (SPSS Inc, Chicago, Illinois).

      Results

      The majority of subjects were women, and most injuries were incurred by nurses or trainees, including medical students, nursing students, and residents (Table 1). Most injuries occurred with hollow-bore devices. Definite or suspected exposure to HIV, hepatitis B virus, or hepatitis C virus was reported by approximately half of the subjects. A minority of subjects incurred their injuries in a continuous-risk environment, such as an operating room or procedure suite.
      Table 1Characteristics of the Sample (n = 139)
      CharacteristicNumber (%), Mean ± SD, or Median (Range)
      Percentages may not add up to 100 due to rounding.
      Age (years)34 ± 10
      Days until interview3 (0–15)
      Female sex100 (72)
      Job description
      Nurse63 (45)
      Medical or surgical resident36 (26)
      Surgical or laboratory technician9 (7)
      Staff physician or surgeon7 (5)
      Student9 (7)
      Other15 (11)
      Device causing injury
      Hollow-bore needle91 (66)
      Suture needle24 (17)
      Scalpel, blade, or scissors10 (7)
      Stylette or trocar6 (4)
      Other8 (5)
      Continuous risk
      Defined as scrubbed in an operating room or procedure suite at the time of injury.
      32 (23)
      Known exposure to HIV or hepatitis C virus24 (17)
      Suspected exposure to HIV or hepatitis C virus38 (27)
      HIV = human immunodeficiency virus.
      * Percentages may not add up to 100 due to rounding.
      Defined as scrubbed in an operating room or procedure suite at the time of injury.

      Risk factors for sharps-related injuries

      There was substantial variability in the prevalence of exposures of interest during the hazard period (i.e., the period during which the injury occurred), and in the average frequency of exposure while performing similar tasks in the control time period (Figure). An increased risk of sharps-related injury was associated with rushing, anger, distraction, and multiple passes. A trend toward increased risk was seen when subjects were fatigued, working with an uncooperative patient, or working as part of a team that was short staffed, and among surgeons working in a noisy operating room environment (Table 2). Among trainees scrubbed in an operating room or procedure suite, the risk of injury appeared to decrease in association with being taught, and a trend toward decreased risk of injury was seen with emergency procedures. No change in the risk of sharps-related injuries was found in association with teaching, performing highly complex operative procedures, or working in a bloody operative field.
      Figure thumbnail GR1
      FigureFrequency of exposure during hazard and control periods. Black bars represent the proportion of study subjects reporting an exposure during the hazard period (i.e., at the time of injury). White bars represent the average frequency of exposure during the control period. Incidence rate ratios for exposures of interest were calculated using stratified analytic methods, so that ratios cannot be calculated directly from the .
      Table 2Estimates of the Effect of Workplace and Individual Level Exposures on the Risk of Sharps-Related Injuries
      ExposureNumber (%) Exposed at the Time of InjuryIncidence Rate Ratio (95% Confidence Interval)P Value
      All subjects (n = 139)
      Rushing67 (48)3.3 (2.5–5.0)<0.001
      Fatigued39 (28)1.5 (0.9–2.3)0.11
      Distracted49 (35)8.9 (5.1–16)<0.001
      Angry17 (12)5.5 (2.7–11.1)<0.001
      Emergency situation or procedure9 (6)0.5 (0.15–1.5)0.19
      Teaching11 (8)0.9 (0.3–2.3)0.77
      Noncontinuous risk (n = 106)
      Continuous risk was defined as scrubbed in an operating room or procedure suite at the time of injury. All other subjects were considered to be at noncontinuous risk of shaprs-related injury.
      Uncooperative patient19 (18)1.8 (1.0–3.3)0.07
      Short-staffed
      Question introduced in later versions of questionnaire (n = 75 subjects).
      24 (32)1.6 (0.9–2.8)0.11
      Multiple passes
      Defined as more than one attempt to perform the procedure. Question posed only to subjects (n = 59) performing phlebotomy, arterial blood gas measurement, intravenous catheter insertion, central venous catheter placement, arthrocentesis, and lumbar puncture at the time of injury.
      24 (41)3.1 (1.8–5.5)<0.001
      Continuous risk (n = 33)
      Continuous risk was defined as scrubbed in an operating room or procedure suite at the time of injury. All other subjects were considered to be at noncontinuous risk of shaprs-related injury.
      Being taught
      Question posed only to trainees (n = 19).
      10 (53)0.1 (0.02–0.7)0.02
      Bloody operative field4 (12)1.5 (0.4–5.0)0.53
      Highly complex procedure9 (26)0.9 (0.3–2.4)0.87
      Noisy work environment5 (15)2.1 (0.7–6.6)0.21
      * Continuous risk was defined as scrubbed in an operating room or procedure suite at the time of injury. All other subjects were considered to be at noncontinuous risk of shaprs-related injury.
      Question introduced in later versions of questionnaire (n = 75 subjects).
      Defined as more than one attempt to perform the procedure. Question posed only to subjects (n = 59) performing phlebotomy, arterial blood gas measurement, intravenous catheter insertion, central venous catheter placement, arthrocentesis, and lumbar puncture at the time of injury.
      § Question posed only to trainees (n = 19).
      Stratified analyses were performed to evaluate whether exposure effects were modified by the source patient’s infection status. The risk of injury with multiple passes was higher when source patients were known or suspected to be infected with HIV or hepatitis C virus than when such infection was not suspected (relative risk = 6.1 [95% confidence interval {CI}: 2.6 to 14.4] vs. 1.3 [95% CI: 0.4 to 3.6], P for interaction = 0.02). Infection status did not affect the risk associated with other exposures.

      Reproducibility of measures

      The reproducibility of estimated time at risk was excellent (Table 3). Subjects’ assessments of whether the exposure was “high risk” (defined as exposure to a source patient known or suspected to be infected with HIV or hepatitis C virus) was also highly reproducible. Intraclass correlation coefficients for the usual frequency of exposures while at risk of sharps injury were good or excellent for all exposures. Reproducibility of the presence of exposures at the time of injury ranged from substantial (in the case of anger), to almost perfect (in the case of rushing).
      Table 3Reproducibility of Responses to Needlestick Questionnaire among 40 Subjects Re-interviewed after 2 to 5 Days
      Intraclass Correlation Coefficient (95% Confidence Interval)κ (95% Confidence Interval)
      Time at risk0.96 (0.93–0.98)
      High-risk exposure
      Defined as exposure to a source patient known or suspected to be infected with human immunodeficiency virus or hepatitis C virus.
      0.88 (0.80–0.96)
      ExposuresUsual frequencyExposed when injured
      Rushing0.88 (0.79–0.93)0.88 (0.80–0.96)
      Fatigued0.79 (0.63–0.88)0.84 (0.75–0.94)
      Distracted0.69 (0.49–0.82)0.89 (0.82–0.97)
      Angry0.98 (0.97–0.99)0.75 (0.62–0.89)
      Emergency situation or procedure0.99 (0.97–0.99)0.89 (0.79–1.0)
      Teaching0.97 (0.93–0.98)0.88 (0.75–1.0)
      * Defined as exposure to a source patient known or suspected to be infected with human immunodeficiency virus or hepatitis C virus.

      Discussion

      We used a case-crossover study design to identify and evaluate the effects of common workplace factors on the risk of sharps-related injuries in health care workers. Among subjects participating in the study, distraction, anger, and rushing were associated with the largest increase in the risk of sharps-related injuries. Free text recorded by interviewers suggests that the source of distraction was frequently an interaction with a coworker while performing a procedure. Distractions were often trivial in nature, an observation supported by the finding that emergency situations did not increase the risk of sharps-related injuries. Anger was most often associated with conflict with another employee or with a patient. This suggests that worker education on the importance of not disturbing colleagues working with sharp devices, and on the importance of “cooling off” after an interpersonal conflict before attempting a procedure, could help to prevent future sharps-related injuries.
      The association between rushing and injury is more problematic, as rushing may be difficult to avoid in the current health care environment. However, if rushing is associated with sharps-related injuries, critical evaluation of the current pace of work and staffing ratios in health care institutions could lead to strategies that reduce such injuries.
      Multiple attempts to complete a procedure were also associated with an increased risk of injury. Repeated attempts at procedure completion have also been associated with patient injury (
      • Johnson E.M
      • Saltzman D.A
      • Suh G
      • Dahms R.A
      • Leonard A.S
      Complications and risks of central venous catheter placement in children.
      ,
      • Mansfield P
      • Hohn D
      • Fornage B
      • Gregurich M
      • Ota D
      Complications and failures of subclavian-vein catheterization.
      ). In contrast with previous studies, we failed to find a significant increase in injury risk among surgeons exposed to bloody operative fields (
      • Tokars J.I
      • Bell D.M
      • Culver D.H
      • et al.
      Percutaneous injuries during surgical procedures.
      ,
      • Panlilio A.L
      • Foy D.R
      • Edwards J.R
      • et al.
      Blood contacts during surgical procedures.
      ), although our study may have lacked statistical power to demonstrate such a relation, given the relatively small number of surgeons.
      As with any study that involves retrospective data collection from human subjects, our results are potentially vulnerable to recall bias; however, this limitation would apply equally to case-control or cohort studies of sharps-related injuries. Relative to these designs, the case-crossover study design may have the advantage of internal consistency in exposure assessment, in that the same person ascertains the presence of the exposure in both the hazard and control periods. Nonetheless, underreporting of sharps-related injuries could be a source of selection bias in our study, particularly if certain factors associated with injuries made subjects more likely to report them (
      • Maclure M
      • Mittleman M.A
      Should we use a case-crossover design?.
      ).
      Some exposures that we evaluated, such as being short staffed, might be less susceptible to biased reporting, as they would be more “objective” than such abstract exposures as anger or distraction. Although we cannot compare subjects’ assessments of such exposures with a gold standard, these assessments were extremely reliable when obtained again at a later date. This reliability suggests that subject estimates were based on clear perceptions of time at risk, exposure frequency, and the presence of exposure at injury.
      Our findings that distraction, anger, and rushing increase the risk of sharps-related injuries are robust and have substantial face validity. Interventions to minimize these factors should be explored by those who wish to prevent sharps-related injuries in the health care workplace.

      Acknowledgements

      The authors wish to acknowledge the expert assistance of Lydia Napper and Cindy Aiello in the completion of this project.

      APPENDIX.

      Sample questions

      A. Noncontinuous risk, exposure of interest: fatigue
      1. At any time on the day of your injury, did you feel fatigued or tired?
      Yes No
      If yes, then ask:
      2. Did you feel fatigued or tired at the time of your injury?
      Yes No
      Always ask:
      3. You said you performed (task) (#) times in the last month. How many times in the last month did you perform (task) while fatigued or tired, up to and including your injury?
      __ Times
      B. Continuous risk, exposure of interest: bloody operative field
      1. At any time on the day of your injury, was the operative field obscured by blood?
      Yes No
      If yes, then ask:
      2. Was the operative field obscured by blood at the time of injury?
      Yes No Not applicable
      Always ask:
      3. What percentage of the time, while you scrubbed in the operating room or procedure suite in the past week, was the operative field obscured by blood?
      __%

      References

        • National Institute of Occupational Safety and Health, Department of Health and Human Services
        Preventing Needlestick Injuries In Healthcare Settings. National Institute of Occupational Safety and Health, Department of Health and Human Services, Washington, DC1999 (Report No. 2000-108)
        • Henry K
        • Campbell S
        Needlestick/sharps injuries and HIV exposure among health care workers. National estimates based on a survey of U.S. hospitals.
        Minn Med. 1995; 78: 41-44
        • EpiNet
        (Accessed January 28, 2003)
        • Shapiro C
        Occupational risk of infection with hepatitis B and hepatitis C virus.
        Surg Clin North Am. 1995; 78: 1047-1056
        • Lanphear B
        • Linemann C
        • Cannon C
        • DeRonde M
        • Pendy L
        • Kerley L
        Hepatitis C virus infection in health care workers.
        Infect Control Hosp Epidemiol. 1994; 15: 745-750
        • Centers for Disease Control and Prevention
        Case-control study of HIV seroconversion in health-care workers after percutaneous exposure to HIV-infected blood—France, United Kingdom, and United States, January 1988-August 1994.
        MMWR Morb Mortal Wkly Rep. 1996; 44: 929-933
        • Cardo D.M
        • Culver D.H
        • Ciesielski C.A
        • et al.
        A case-control study of HIV seroconversion in health care workers after percutaneous exposure. Centers for Disease Control and Prevention Needlestick Surveillance Group.
        N Engl J Med. 1997; 337: 1485-1490
        • Jagger J
        • Bentley M
        • Juillet E
        Direct cost of follow-up for percutaneous and mucocutaneous exposures to at-risk body fluids.
        Adv Exposure Prev. 1998; 3 (Available at: http://www.med.Virginia.EDU/medcntr/centers/epinet. Accessed January 28, 2003): 1-3
        • David H.T
        • David Y.M
        Living with needlestick injuries.
        J Can Dent Assoc. 1997; 63: 283-286
        • Rosenstock L
        (Accessed January 28, 2003.)
        • National Surveillance System for Health Care Workers
        (Accessed January 28, 2003)
        • Mitsui T
        • Iwano K
        • Masuko K
        • et al.
        Hepatitis C virus infection in medical personnel after needlestick accident.
        Hepatology. 1992; 16: 1109-1114
        • Petrosillo N
        • Puro V
        • Jagger J
        • Ippolito G
        The risks of occupational exposure and infection by human immunodeficiency virus, hepatitis B virus, and hepatitis C virus in the dialysis setting. Italian Multicenter Study on Nosocomial and Occupational Risk of Infections in Dialysis.
        Am J Infect Control. 1995; 23: 278-285
        • Sexton J.B
        • Thomas E.J
        • Helmreich R.L
        Error, stress, and teamwork in medicine and aviation.
        BMJ. 2000; 320: 745-749
        • Chisholm C
        • Collinson E
        • Nelson D
        • Cordell W
        Emergency department workplace interruptions.
        Acad Emerg Med. 2000; 7: 1239-1243
        • Gander P.H
        • Merry A
        • Millar M.M
        • Weller J
        Hours of work and fatigue-related error.
        Anaesth Intensive Care. 2000; 28: 178-183
        • Maclure M
        • Mittleman M.A
        Should we use a case-crossover design?.
        Annu Rev Public Health. 2000; 21: 193-221
        • Maclure M
        The case-crossover design.
        Am J Epidemiol. 1991; 133: 144-153
        • Sorock G
        • Lombardi D
        • Hauser R
        • Eisen E
        • Herrick R
        • Mittleman M
        A case-crossover study of occupational traumatic hand injury.
        Am J Ind Med. 2001; 39: 171-179
        • Reidelmeier D
        • Tibshirani R
        Association between cellular-telephone calls and motor vehicle collisions.
        N Engl J Med. 1997; 336: 453-458
        • Valent F
        • Brusaferro S
        • Barbone F
        A case-crossover study of sleep and childhood injury.
        Pediatrics. 2001; 107 (Available at: http://www.pediatrics.org/cgi/content/full/107/2/e23. Accessed January 28, 2003): e23
        • Petridou E
        • Mittleman M.A
        • Trohanis D
        • et al.
        Transient exposures and the risk of childhood injury.
        Epidemiology. 1998; 9: 622-625
        • Sorock G
        • Lombardi D
        • Gabel C
        • et al.
        Case-crossover studies of occupational trauma.
        Inj Prev. 2001; 7: 38-42
        • Mittleman M.A
        • Maclure M
        • Robins J.M
        Control sampling strategies for case-crossover studies.
        Am J Epidemiol. 1995; 142: 91-98
        • Rothman K
        • Greenland S
        Applications of stratified analysis methods.
        in: Rothman K Greenland S Modern Epidemiology. 2nd ed. Lippincott-Raven, Philadelphia1998: 281-300
        • Greenland S
        • Robins J
        Estimation of a common effect parameter from sparse follow-up data.
        Biometrics. 1985; 41: 55-68
        • Fleiss J
        The Design and Analysis of Clinical Experiments. Wiley, New York1986
        • Cohen J
        A coefficient of agreement for nominal scales.
        Educ Psychol Meas. 1960; 20: 37-46
        • Landis J
        • Koch G
        An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers.
        Biometrics. 1977; 33: 363-374
        • Rothman K
        • Greenland S
        Introduction to stratified analysis.
        in: Rothman K Greenland S Modern Epidemiology. 2nd ed. Lippincott-Raven, Philadelphia1998: 253-279
        • Johnson E.M
        • Saltzman D.A
        • Suh G
        • Dahms R.A
        • Leonard A.S
        Complications and risks of central venous catheter placement in children.
        Surgery. 1998; 124: 911-916
        • Mansfield P
        • Hohn D
        • Fornage B
        • Gregurich M
        • Ota D
        Complications and failures of subclavian-vein catheterization.
        N Engl J Med. 1994; 331: 1735-1738
        • Tokars J.I
        • Bell D.M
        • Culver D.H
        • et al.
        Percutaneous injuries during surgical procedures.
        JAMA. 1992; 267: 2899-2904
        • Panlilio A.L
        • Foy D.R
        • Edwards J.R
        • et al.
        Blood contacts during surgical procedures.
        JAMA. 1991; 265: 1533-1537