Effectiveness of and compliance to preventive measures against the occupational transmission of human immunodeficiency virus.

To prevent the occupational transmission of human immunodeficiency virus (HIV) and other blood-borne infections in health care settings, guidelines have been issued as universal precautions and body substance isolation. Patient testing has also been advocated. The literature on the compliance to and effectiveness of these measures was reviewed and analyzed to establish the state of knowledge and make appropriate recommendations to improve guidelines. It showed that workers' compliance to recommended measures is relatively poor. The effectiveness and cost-effectiveness of universal precautions and body substance isolation remain to be demonstrated. Testing patients for HIV infection and other bloodborne pathogens does not appear to be a more appropriate solution. Focus should be placed on preventing parenteral exposures and applying risk assessment methods to identify health care settings and procedures at higher risk. These measures would allow safer medical devices to be targeted and would ensure that financial resources would be available to implement appropriate preventive measures.

. Of these, 63 are do cumented seroconv ersions , meaning that the health care worker had a negative, postexposure HIV antibody test result and subsequently tested po siti ve for HIV. The other I I I cases are con sidered to be probabl e occupational infe ction s, as seroc onvers ion was not documented and no other risk fa ctors could be identifi ed. The risk of occupatio nal HIV infection fo llowing parenteral ex posure to contaminated blood is estimated to be 0.3% (13). The risk of transmission resulting from exposure to mucous membranes or nonintact skin le sions is con sidered to be much low er , around 0.09% ( 14).
Oth er bloodborn e pathogens are also at the origin of occupational infections, for example, the hepatitis B virus, for which the risk of transmi ssion fo llowing pe rcutaneou s exposure is es timated to be up to 100 times greater than th at of HIV . Ca ses of hep-  (15)(16)(17)(18)(19)(20) and the risk of tran smi ssion following a parenteral exposure to contaminated blood, although not pre ci sely qu antified, could be as high as 10% (21).
Faced with these very serious infections, measures have been proposed to avoid transmission to health care workers, namely, universal prec autions and bod y subs ta nce isol ation. These two sys te ms are based On the prin ciple that all patients are con sid ered to be infecti ou s. This notion of universality cons titutes a maj or change in the app roach to preventing the transmi ssion of bloodborne pathogens in health ca re setting s that, until recently, was more selective, relying on the pat ient' s pr esumed or establi shed diagnosis. There is resistance to thi s change am on g health care workers, and so me ev en suggest that the presence of infec tions should be acti vel y sou ght o ut in all pati ents since kno wledge that a patient is infec ted would allow health care workers to protect themsel ves better (22,23). For example, in certain cases, a therapeutic decision could be changed when a pat ient is known to be in fected so that alternat ives that arc sa fer for health ca re personnel could be presc ribed.
Thi s article examines the uni ver sal and se lec tive approaches by reviewing the state of knowledg e concerning worker compliance, and the eff ectivenes s of both of these appro aches in regard to the occupation al transmission of HIV in the workplace.

Universal precautions
Universal precaution s were proposed by the Cent ers for Disease Control and Prevention to prevent bloodborne infe ctions in health care sett ings (24 , 25) . The level of exposure to body fluids constitutes the main determinant of occupational risk of infection . Added to this concept is the principle of universalit y of risk , all patient s being assumed to be infected .
Guidelines were issued regarding work practic es, per sonal protective equ ipment (barriers), cleaning practices, disinfection and decontamination , and the man ipulation and disposal of needl es and sharp obje cts. These measures aim to prevent exposure to spec ific body fluid s such as blood, semen, vaginal secretions, and any body fluid visibly contaminated with blood, as well as cerebrospinal flu id, synovial flu id, pleural fluid, peritoneal fluid, pericardial fluid , and amniotic fluid . Saliva is excluded except in the context of dent al and oral procedures. In practical term s, the guidelines recomm end the use of individu al prot ective equipment, such as gloves , masks, protective eyewear, apro ns, and gowns, to avoid exp osure by direct contact or by splashes. To prevent inj uries cause d by needles or other sharp objects, needle recapp ing is proscribed and sharp objects mu st be disposed of in puncture-resistant contai ners .

Compliance to universal precautions
The degree of compliance by health care workers to some or all uni versal precautio n measures was stud ied in differe nt clinical setti ngs. In a study of 806 observations made of 24 health care workers in emergency roo ms, dental clinics , and plastic surgery clinics, 3 1.3 % were compliant, 40.7% were parti ally compliant, and 28% were noncom pliant (26) . Use of punctu re-resistant containers for the disposal of sharp objects was satisfactory, but needl e recapping was still very prevalent and com pliance to hand washing very low. Another study combining two methods, questionn aire and observation, found that universal precauti ons were followed in only 19% of the 1441 proce dures observed (27). Compliance according to profe ssion was 27% for nurses, 5% for laboratory technic ians, and 3% for physi cians. Compliance to procedures of safe pipettin g (97%) and the handl ing of needl es (72%) was goo d but dropped for hand was hing (28 %) and the wearing of gloves (35 %). In 8 1% of the cases , physicians did not manipulate needles in compliance with recommendations. Discomfort, interfere nce with procedures , forgetfulness, and lack of safety materials were the reasons most ofte n listed for noncompl iance to universa l prec autio ns.
Most studies eva luated only so me of the meas ures proposed by univ ersal precautions, particularly barrier methods such as use of gloves . According to their 394 find ings, the rate of glov e use and the use of other barrier methods varie d according to several par ameters, mainly the profession , the clie ntele, and the type of procedure. Physicians (surgeo ns, specialists, generalists, resid ent s, and medical stude nts) used gloves much less than nurses, in a prop ortion varying from 10 to 40 % for the form er and 40 to 75 % for the latter (28)(29)(30)(31)(32)(33). Swed ish denti sts with priv ate practi ces used gloves during only 54% of their procedures, while those in public practice did so in 77 % of the cases (34) . Paramedics only rarely complied to the systematic use of barrier methods, the percentage of utili zation varying from 8 to 37% (29 , 35) .
As to the age of clientele, those treating childre n had a lower utilization rate (29, 32) than those treating patients 15--44 years of age, who had the highest rate (29) .
As to the type of pro cedure carr ied out, the rate of glove use amon g nurses varied from 70% for phlebotomy to 92% for arter ial punctures (36) . Th e use of barr ier methods was 17% for major procedu res, 56 % for minor ones, 19% when the patient was bleeding, and 44 % when the patient was not bleeding (29). The lack of easy access to prote ctive methods in emergenc y situations could explain these surprising results, as sugges ted by a study showi ng that supplying protective equipment in readil y available kits in strateg ic locations of emergency roo ms had increased the use of universal preca utions fro m 16 to 62% in a Florida hosp ital (37) .
Compli ance to the adequ ate handling and dispo sal of needles is genera lly higher than with barrier methods. A study showe d that the manipul ation of needles by emergency room resi dents was adequate in 97% of the cases at an Americ an hospital (37). However, despi te rec ommendations about needle recapping, it was observe d that, in four Michi gan hospitals, 8 to 51 % of the needles found in containers had been recapped before disposal (38 ). The situation seems eve n worse for paramedics, who recapped needles in 78% of the cases (35), and for physicians who did so in a proportion varyi ng from 50 to 60 % (3 1).
Most of these studies were conducted with a questionnaire administe red to worke rs, but one study using two methods, questionn aire and direct observation, showe d that workers tend ed to overesti mate their compliance to univer sal precautions (27). Nonetheless , the tendencies drawn fro m all the aforementio ned studies are similar and show that co mplia nce to recomm endation s relating to needle manipul ation is higher than that relating to the use of barrier methods , that nurses are more compliant than physici ans, and that acce ss to, and the ease of use of, protective equip ment well integrated into regular task perfo rmance tended to increase compliance. Furthermore, the notion of universality is still not well integrated since the potent ial risk presented by the pat ient is taken into con sideration by health care work ers in their decision to use preventi ve measures.
Effe ctiveness and cost of universal precauti ons Effectiveness studies. With respect to the effectiveness of universal precautions, a study targeted a cohort of health care workers participating in the United States National Institutes of Health (NIH ) surveillance program of occupational expo sure s to HIV (39). Workers had to answer a que stionn aire on cutaneous exposures sustained during the previous year. The analyses compared expo sure s before and after the implementation of universal precautions. The average annual number of exposures per worker dropped from 35.8 to 18.1 for blood (P::; O.OOI) and from 77.8 to 40.8 for other body fluid s (P ::; O.OOl).
Another study gathered data on physician s' exposure s in hospital medical care unit s before and after the implementation of univer sal precautions. Everyday they had to answer a questionnaire on cutaneous and parenteral exposures sustained during the day (40). A global reduction in exp osures of 5.07 to 2.66 per patient-month per physician, associated with an increase in aborted exposures from 3.41 to 5.90, was found . As for accidental needle sticks, the rate fell from 0.39 to 0.15, but this difference was not statistically significant (P =0.123). The methodological differen ces between these two studies and the absence of a reference group make it difficult to draw conclusions on the effectiveness of universal precautions.
Other authors have attempted to evaluate the effectiveness of certain components of universal precautions. For rigid containers, there was no reduction in the total number of needlestick s nor in the practice of recapping in association with their implement ation (4 1--44). Two of the studies (4 1,42) showed however a reduction in needlesticks in relation to needle disposal , and one (45) report ed a global reduction of all needles ticks, as well as those related to recapping among nurse s.
Othe r studies have dealt with the effectivenes s of using some barrier methods , such as gowns and surgical sheets (46--48). It was concluded that the effectiv eness in preventing blood contact varied accord ing to the type of material, its imperm eability, and wear and tear.
Even if barrier methods such as gloves succeeded in reducing the number of cutaneous exposures (39,40), their role in preventing the transmission of bloodborne pathogens among health care workers is not clear since needlesticks and cut s, which are not prevented by the use of gloves, pre sent the greatest risk for the transmi ssion of bloodb orne infections.
Recent studies have been condu cted with the purpose of evaluating new medical devices designed for preventin g accidental needle sticks, such as selfresheathing needles, new connectin g devices for intravenous lines or blunt suture needles. The results are promising and would allow immediate causes of exposures to be eliminated (ie, unnecessary needles or ill-co nceived devices) (49,50). Such an approach Scand J Work Environ Health 1994, vol 20, no 6 is complementary to the guidelines of the Centers for Disease Control and Prevention on universal precautions.
Cost and cost-effectiveness studies . An American stud y looked at the annual costs of implementing universal precaution s in a 900-bed university hospital (5 1). The conclusion was an annual increa se in costs of USD 350 900 . However neither the effectiveness nor the benefits of universal precautions were evaluated. In a Canadian stud y the cost-effectiveness ratio of universal precautions in a 450-bed hospital was found to be CAD 128 862 000 per case of occupational HIV infection prevented , according to the authors' best-case estimate (52) . It was concluded that universal precautions were not cost-effective and that priority should be put on the prevention of parenteral exposures, notably through a better design of materials and techniques used.
Despite the scarcity of studies on costs, universal precautions seem to constitute an expensive solution with to which prevent the occupational transmi ssion of HIV. However direct and indirect costs of exp osure to body fluids were ignored in the studies. These costs could be substantial, for example , those related to (i) reporting exposure, (ii) a 6-to 12-month postexpo sure follow-up of workers possibly admini stering immunoglobulin or zido vudine , (iii) the loss of productivity associated with these visits, and (iv) the postexposure stress of worker s awaiting the ir HIV antibody test results. Otherwise, preventing other bloodborne infections (hepatitis B and C) was not taken into account although it constituted another important parameter in the cost-effectiveness ratio equation . Indeed Stock et al found that universal precautions present a more interesting cost-effectiveness ratio when the risk of hepatiti s B is considered , although vaccination is certainly the most effective preventive means for this infection (52).

Body substance isolation
The expressi on body substance isolation describes a system of prevention of nosocomial infecti ons, including the transmission of occupational infections (53). Like universal precautions, this system is applied to all patient s, regardless of their diagnosis. A major difference is that body substance isolation targets all body fluids and moist body substances. It include s six components (54), the following four of which are aimed at bloodbome infections: (i) gloves must be worn for all expe cted contact with a body substance, (ii) other barriers such as masks, glasses, gowns, and aprons must be worn when contamination of clothing or skin or splashe s are anticipated, (iii) reusable items must be securely contained to prevent leakage, (iv) needles and other sharp objects must be disposed of in puncture-resistant containers; needles should not be recapped. There are few other difference s with universal precautions, for example, 395 Scand J Work Env iron Health 1994, vol 20, no 6 with body substance isolation , the health care worker is ca lled upon to judge the pert inen ce of using barrier method s such as wearin g gloves during phlebotomy, a practice which is optional if the worker is experienced and does not deem it necessary (55).
Some have found that body substance isolation is easier to implement than universal precautions (56), but few studies have evaluated its application and effectiveness, particularly concerning bloodborne infections. In one stud y, an increa se of 60 to 80% in adequate glove use was observed after body substance isolation was implemented, but adequate hand washing was not prevalent (30 to 45%) (57). In a survey, despite adequate training and knowledge of the princ iples of body substance isolation, nearly half of the respondents frequ ently using needles recapped them occasionally or often . It can be concluded that, as for universal precautions, worker compliance to body substance isolation is not optimal (58).

Selective measures
As an alternative to universal approaches, identifying patients infected with HIV has been propo sed so that supplemental pre ventive measure s can be taken with these patients and their speci mens (59-6 I). This proposition is in line with label ing infectious specimens to advise the worker to take appropriate pre ventive measures.

Antibody testing f or human immunodeficiency virus
HIV-infected person s can be identified with a simple serological test. Current tests for HIV antibodies, including two immun oenzymatic tests (ELISA) follo wed by a more spec ific confirmation test, are very reliable (62). The high sensi tivity and speci ficity of the test sequence allow virtual identification of all individuals (except those who are in the postinfection period) with a very low proportion of false positive results. The importance of quality control assuring the high specificity of confirmation tests must be emphasized because a lowering of specificity could actually compromise the predictive value of a positive result in populations in which the prevalence of infe ction is low.
Despite the good perform ance of the test, there is a delay for obtaining the result that could limit its widespread use, especiall y in emergency situations. The arrival on the market of rapid and effective tests with sensitivity and specificity approaching those of conventional techniques (63,64) co uld be a solution to this problem. The utiliz ation of these tests as diagnostic tools is not currently authorized in all countries.
Another aspect that cannot go unmentioned is that all persons undergoing an HIV antibody test must freely give their informed consent and receive appropriate counseling. The only e xcepti on to this rule of informed consent is in the cas e of blood, semen, and organ donati ons , for which HIV antibody testing is systematic. Upon nearing the end of the first decade of the HIV epidemic, howe ver, what some call " HIV exceptionalism" could be challenged and therefore lead to less stringent rules on informed consent (65).
Eff ectiveness and cost of a testing p rogram fo r patients Effectiveness studies. The effectiveness of pati ent testing in preventing occupati onal HIV infection has not been widel y studied. Some authors have evaluated the asso ciation between knowing the serological status of the patient and the rate of accidental exposure to blood among operating room personnel. In one study, no association was found between the perceived or known status of a patient and the rate of exposures (66). A similar tendency was observed in anoth er study (67).
Apart from the identification of infected patients and the labelin g of their specimens, other means of reducing the risk of HIV infection have been proposed, for example, therapeutic substitution (proposing alternat ive treatments to the more conventional ones) or reference to a more-experienced professional. To our knowledge, these "preventive methods" have not been studied from the point of view of effec tiveness .
Cost-effectiveness studies. The cost-effectiveness ratio of a testing program aimed at patients mainly varies as a function of the prevalence and inciden ce of infection and of the preventive measures taken after testing. According to an American study, the cost of a mandatory testing program for all patients in a 350bed publi c hospital would surpass the benefits, and such a program would constitute a waste of resources (68) .
In 1989 in the Province of Quebec, it was estimated that a testing program for patients admitted to acute care hospitals would cost more than 18 million Canadian dollars (around 13 million US dollars) per year for 800 000 admissions (69). Thi s calculation reflects only a part of the real costs of such an operation since it only took into account direct costs relat ed to the testin g and coun seling of hospitalized patient s, excludin g all those in ambulatory care or out-of-hospital settings. Other direct cos ts, such as those related to the infrastructur e necessary to answe r the incremental demand for tests, those caused by the repetiti on of tests, or indir ect expend iture s such as travelin g cos ts or worktime lost by the individuals being tested, were not includ ed in the calculation.
With this ava ilable ev idence, systematica lly testing all patient s appears to be an extremely expensive solution to the risk of occupational HIV transmission in health care sett ings.
An alternate so lution to universal testin g could be selective testing of certain populations, for exampl e, of ind ividual s in clin ical settings in which a high preval ence of HlV is expected. Thu s one author found that testin g clients 15 to 54 years of age (12% of the patients admitted) upon adm ission to hospitals where the annual rate of AIDS diagnosis is I for 1000 discharges permitted the identification of twothird s of infected person s (70) . We would therefore expect that a testin g program directed toward a clientele selected according to cert ain criteria would allow HlV -infected patients to be detected more effectiv ely and would cost less than if all patients were tested. One stud y looked at this question in regard to various prevalence scenarios by estimating the cost per patient tested and the cost per occ upational infection prevent ed (7 1). The authors found that the additio nal cos ts related to testing hospitalized patient s in the United State s would amount to USD 4 billion, or USD 23 000 per infected person dete cted if the seroprevalence rate had been I%. Thi s cost would have been brought down to USD 14000 per patient if the prevalence had been 10%, and it would have increased to USD 2 17 000 ifit had been 0.05%. The add itional costs for every HIV occ upational infection prevented would be USD 751 000 (ra nge USD 186 000 to USD 133 trillion). The auth or concluded that identifying infected patients with the goal of preventing occupational HlV infecti ons would not be j ustified. In the same way, the Cent ers for Disease Control and Preventi on do not recomm end substituting the use of uni versal pr ecaut ions with a testing program for patien ts (72) .
Finally, an HIV antibody testing program for patients would be difficult to justify without inclu ding one for hepatitis Band C (73)(74). Morbidity associated with hep atitis B, where I % of adult cases develop a fulminant form and 10% a chronic form , with a risk of developing cirrhosis or liver cancer, would ju stify testing for hepatiti s B as much as for Hl V antibodies , at least in the absence of vacci natio n for all health care professionals exposed to body fluids. Furtherm ore , a significant chronic lon g-term morbidity associ ated with hepatitis C virus infection has been shown (75) .

Discussion
HIV tran smission in health care setti ngs is still a matt er of conc ern for health care work ers. For some of them , the risk of infecti on is relatively high, notably in areas where the prevalence of HIV infection is elevat ed. For others, althoug h the actual risk is low, it is percei ved as being high, which can sometim es induce undesirable beh avior , such as denying access to care for infected patient s (76 , 77 ). Th is worr y, whether justified or not, has divided health care workers between advocates of universal and selective preventive approaches. Con sequ entl y, seve n Scand J Work Environ Health 1994, vol 20, no 6 years after the Centers for Disease Control and Preventio n's first recommendations on HIV preventi ve measures in health ca re settings, the situation is still confu sed. Anecdotal reports suggest that prevention practices in health care settings are often mixed , universal prec auti ons havin g been partially implemented, whereas so me professionals test patient s who " look infec ted." Otherwi se the literature shows that universal precautions are not very co ncrete, as regards innumerable work situations with in health care settin gs. Thi s situ ation has brought on major differences in their interpretation and appli cation, as shown in a study in which, although 66.4 % of Can adian hospit als stated having implemented universal prec aut ions, only 5.6% had applied all the minimal measure s requ ired by the Cent ers for Disease Cont rol and Prevention (78) . Moreover, the effectiveness of uni versal precautions in preventing bloodborne infections has not been demonstrated. Ce rtain meth odological weaknesses of evaluation studies, like the absen ce of a contro l group and poor external validity, have prevented such a demonstration.
Many health care workers do not compl y with universa l precauti ons in that they interfere with their work, are expensive, and are not very effect ive (29,79,80). On one hand , some fail to comply sys tematic ally bec ause they feel that they are at low risk and prefer to apply recommended mea sur es only when they suspect a patient of being infected. On the other hand , some do not trust universal precaution s and consider that te sting is a more effec tive way to pre vent HlV transm ission. In fact, testin g can be somew hat reassur ing since it allows the hazard to be identified, necessary and perhap s extraordinary measure s to be taken, and immediate knowledge of what to do in case of exposure. Testing also fit s in with the tradit ion of infe ction pre vent ion based on diagnosis.
However, as shown in the literature, HIV antibody testing of patients is not a more cost-effect ive solution than universal precautions, and it surely does not make sens e without concurrent testing for other bloodborne infections. Moreover it is impo ssible to test every patient each time the y consult, and the magni tude of occ upatio nal risk does not ju stify such expenses . while the epidemic is spreading and services to infected people are insufficient. Furth ermore , such a program would co nsiderably redu ce ava ilable resources for protec tive equipment, and therefore rende r obsol ete the efforts allocated to testin g.
Clearly, universality of preventive measures should be maintain ed. It is however imperative to improve the effectiveness of these measur es, particularly in the pre vention of parenteral expos ures, by developing safer devices and by makin g prevention measures "user friendl y." It is also imperative to improve the cost-effectivene ss of the prevention program by targeting health care settings where risk of infection is higher.
The prerequisites for risk assessment are the knowledge of local epidemiology of the bloodbome infections included in the prevention program and knowledge of the extent of exposure to body fluids and its associated factors. As for the first prerequisite, information on infection rates in the general population is usually available through surveillance systems or prevalence studies monitoring the epidemic. As for the second, however, the epidemiology of blood and body fluid exposures is not well known mainly because of the wide variability of the information gathered by hospitals and the small size of the individual data bases not allowing in-depth analyses. Therefore there is a need to develop a standardized system of surveillance of exposures on a wide enough scale to allow the factors associated with these exposures to be identified. This step would permit more effective strategies and devices for preventing parenteral exposures to be developed, and, consequently, it should improve compliance.
Such a prevention program, based on risk assessment, is deemed necessary in most countries in the context of the financial restrictions imposed upon health care systems worldwide. Other elements, like local legislation and the already implemented preventive measures, must be accounted for in the planning process, but the magnitude of the risk of bloodborne infections for health care workers seems to us to be the main driving force in the implementation of preventive measures.