Potential bias in studies of accidental needle sticks

Abstract

There are many types of bias in clinical and epidemiologic studies that may distort the results. In his classic paper, Sackett1 cataloged 35 biases in analytic research. Others have suggested useful subclassifications of these biases.2 The most common type of bias is confounding of the association between exposure or intervention and the outcome by external factors. Equally critical biases in comparative studies, particularly those using retrospectively collected data, are selection, recall, and nonresponse bias. Selection bias is often created by erroneous sampling and selection by design or self-selection of study participants. Nonresponse bias, a form of selection bias, may exist in studies in which the response rate is low and the exposure and/or outcome among respondents is not representative of that in the study population. Recall bias may affect the study results when the ability of recalling past experiences and exposures in the groups of study participants is unequal. Case-control studies often suffer from recall bias because all data on risk factors are collected retrospectively. For example, cases of a disease tend to recall history of exposures and other illnesses much better than the control subjects. We chose 2 recently published studies of accidental needle sticks (ANSs) in allergy practices as examples for a discussion of how potential selection, recall, and nonresponse bias can affect the results of retrospective, survey-based studies. In response to continued concern about the risk of bloodborne diseases from ANSs, the National Institute of Occupational Safety and Health released needle stick guidelines,3 which advised employers to use safer devices, particularly needles with safety features (described hereafter as safety needles). Subsequently, the US Congress passed a bill that became the Needlestick Safety and Prevention Act of 2000.4 It authorized the Occupational Safety and Health Administration (OSHA) to revise its Bloodborne Pathogens Standard, redefine the engineering controls, and require employers to use this newer technology and maintain a log of the number of needle sticks and similar cutaneous injuries.5,6 Enforcement of this law began July 17, 2001. Considerable variation exists in the previously reported ANS incidence (between 14 and 839 ANSs per 1,000 population per year) among health care workers, and recent, reliable estimates for all health care facilities in the United States are not known.7 Although claims of safety needle use resulting in significant declines in the incidence of ANSs have been documented,8,9 the data were of questionable quality, the studies were restricted to 1 or a few selected hospitals (some outside the United States), and no randomized controlled trials were conducted. In particular, the tracking and recording of injury events by respondents have been problematic. For example, in a study of phlebotomy procedures using safety needles vs conventional needles in 6 teaching hospitals, 46% of all ANSs had not been reported.10 To determine the incidence of ANSs in allergy practices, Kanter and Siegel11 conducted a study of 400 selected practices in the United States, mostly in California. They collected data on ANSs by conducting a 2-year retrospective survey. The time frame of their survey and the details of the type (mail, telephone, or e-mail) and methods of their survey were not provided. The survey response rate was 30%. Kanter and Siegel found an incidence of 6.4 ANSs per million injections, which was much lower than the ANS incidence in general medical practices. (The source of data on general medical practices was not described.) In a subset of their data, the ANS incidence using safety needles was reported to be much higher than that using conventional needles. However, they acknowledged that these data were unreliable in establishing whether safety needles were actually less safe than traditional needles and were only an “initial step in what should be an ongoing process of evaluation.”11 Even though this study found 45 cases of ANSs, they mention that there have been no known reported cases of disease transmission in allergy practices and so argue that there is “no real risk to nursing personnel in the allergy setting.”11 In 2004, Wolf et al12 conducted a similar e-mail survey of 250 allergy practices. Seventy practices (28%) responded to the survey, and 29% and 59% of the respondents had switched to using safety needles for intradermal testing and allergy injections, respectively. The ANS incidence in the use of safety needles for allergy injections was higher than that for conventional needle use (39.6 vs 17.8 ANSs per million Affiliations: * Department of Medicine, Division of Allergy & Inflammation, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts; † Department of Health Policy & Management, University of Kansas School of Medicine, Kansas City, Kansas. Disclosures: Authors have nothing to disclose. Received for publication December 16, 2007; Received in revised form February 26, 2008; Accepted for publication March 3, 2008.