Current Review

The Changing Profile of Bacterial Endocarditis as Seen at an Australian Provincial Centre†

Allen Cheng, MB BS, FRACP,1 Eugene Athan, MB BS, FRACP,2 Alan Appelbe, MB BS, FRACP2 andMalcolm McDonald, MB BS, FRACP, FRCPA2

1The Alfred Hospital, Prahran and 2Geelong Hospital, Geelong, Victoria, Australia

Background: The clinical profile of endocarditis has changed over the past four decades withstudies showing trends towards increasing age, more nosocomial and prosthetic valve infectionand increasing rates of Staphylococcus aureus infection. However, these studies have beenbiased by referral patterns.Methods: We reviewed data collected at three hospitals in the Barwon–South-West region inVictoria, Australia. All cases identified between 1994 and 1999 were reviewed according to theDuke criteria.Results: During this period, 58 patients were diagnosed as having endocarditis. The incidencerate during this time was 3.0 per 100 000 with a rise in the rate of admissions from 0.15 to 0.26 per1000 from 1995 to 1999. Sixteen (28%) were nosocomial with the majority from line-relatedsepsis. No intravenous drug users were identified. Eighteen (31%) involved endovascularprosthetic material. S. aureus was the causative pathogen in 23 (40%), with ‘viridans’ streptococcicontributing 12 (21%) and other organisms accounting for 12 (21%). Attributable mortality in thisseries was 17%.Conclusions: We have seen a rise in the rate of endocarditis during this time. The proportions ofendocarditis due to S. aureus and ‘viridans’ streptococci, as well as rates of nosocomial andprosthetic valve infection, are consistent with more recent series at referral hospitals and districthospitals, representing a change since studies performed in the 1960s and 1970s. Our findingsconfirm a trend towards a clinical profile seen at referral centres and reinforce the emergingimportance of S. aureus, nosocomial bacteraemia and prosthetic valve endocarditis. (Heart, Lungand Circulation 2002; 11: 26–31)

Key words: Australia, cross-infection, endocarditis, epidemiology, rural population.

ver the last four decades, the epidemiologyand clinical profile of bacterial endocarditishave changed considerably. However, despite

improvements in diagnostic techniques and treatment,the diagnosis often remains elusive and the disease still

carries a significant mortality. The ‘classic’ picture ofendocarditis described in textbooks1,2 still predominatesin developing countries and indigenous communities.3,4

This is characterised by vegetations on native heartvalves damaged by previous rheumatic heart disease or,less commonly, on congenital heart abnormalities. Thesecommunity-acquired infections tend to occur in childrenand younger adults, and the most important pathogensare alpha-haemolytic (‘viridans’) streptococci, with con-tributions from staphylococci and enterococci.

Correspondence: Allen Cheng, Menzies School of Health Research, PO Box 41096, Casuarina, Northern Territory 0811, Australia. Email: allencheng@hotmail.com†This work was presented in part at the Australian Society for Infectious Diseases Annual Meeting, Melbourne, April, 2001.

O

Heart, Lung and Circulation 2002; 11 A. Cheng et al. 27Changing profile of bacterial endocarditis

As long ago as 1967,5 it was noted that endocarditiswas more frequently being seen in older patients withnon-rheumatic heart disease with an increasing propor-tion of cases due to Staphylococcus aureus. Subsequentstudies at large metropolitan centres confirmed thesetrends and also noted a rise in nosocomial infection,6

infections of prosthetic devices and endocarditis relatedto injecting drug use.7 For want of a better word, thismay be termed the ‘metropolitan’ picture of endocarditis.

It is likely that most studies done in tertiary referralcentres are biased by referral patterns and their conclu-sions may not be widely applicable, especially in pro-vincial and rural centres. There have been fewepidemiological studies or clinical reviews of bacterialendocarditis relating to smaller community settings lessaffected by referral bias. Based on his analysis of thesource of referral of cases to a Sydney teaching hospital,Dwyer8 has suggested that the community pattern mayreflect the ‘classic’ picture. However, there have not beenany Australian reviews since the introduction of theDuke criteria in 1994,9 and none in the provincial context;thus, it is not clear if the profile of the disease wouldreflect more of the ‘classic’ or ‘metropolitan’ picture.

The aim of the present study was to use prospectivelycollected data from a provincial infectious diseases serv-ice, together with material from chart review, to delin-eate the clinical picture of endocarditis seen in Geelongand the Barwon–South-West region. The finding of a‘classic’ picture, a ‘metropolitan picture’, or some othermay be of value in formulating diagnostic decision pathsand initial empirical treatment in similar settings else-where in Australia.

MethodsThe provincial city of Geelong is the second largest cityin Victoria, Australia, with a population of approxi-mately 180 000. In August 1994 the Geelong InfectiousDisease Service (GIDS) was established which nowcovers the region’s 450-bed public hospital, GeelongHospital, and two private hospitals, St John of GodHealthcare and Geelong Private Hospital. It also includesseveral public and private outpatient clinics. They collec-tively serve the Barwon–South-West region with a popu-lation of close to 330 000, encompassing Geelong and thesurrounding rural communities. At present, there aretwo infectious diseases physicians in the service.

The region is also served by an experienced and well-equipped cardiology practice with facilities for trans-oesophageal echocardiography and coronary angio-graphy. In 1998, a cardiothoracic surgical unit wasestablished, but prior to that time, patients requiringsurgery were referred to tertiary centres in Melbourne.

The infectious diseases and cardiology services, togetherwith other physicians in the Division of Medicine andsubsequently the cardiothoracic surgeons, have had along-standing and successful arrangement whereby anypatient with a provisional diagnosis of endocarditis wasautomatically referred to both services immediately. Thiswas to ensure that no one with known or suspectedendocarditis escaped the notice of the key referralservices.

Patients were identified from a database prospectivelycollated by the infectious disease service specifically setup to study the profile of several diseases in a provincialsetting, including endocarditis. The database includeddemographic data, date and source of referral, site ofinfection and risk factors including the presence of pros-thetic material, nosocomial infection, the source of infec-tion, the infecting organism and details of treatmentincluding antibiotic therapy, the need for surgery and theoutcome. Other information, including clinical details,diagnostic investigations and additional details ofmanagement and outcome, was obtained through retro-spective chart review.

The diagnostic criteria for each patient identified ashaving endocarditis in the database were reviewed by anindependent clinician. Cases fell into three categories:those satisfying the clinical criteria for definite infectiveendocarditis using the Duke criteria (Tables 1,2); thosewhere all the criteria were not fulfilled (‘possible endo-carditis’) but were nevertheless regarded as having endo-carditis and treated as such; and those initially identifiedas having endocarditis but, on later reflection, regardedas doubtful (‘rejected endocarditis’) and excluded fromanalysis in the study series. Following the publication ofthe modified Duke criteria10 the data were re-analysed todetermine their effect on the classification of the patients.

The annual ratio of cases to hospital admissions wascalculated from 1995, being the first full year of theGeelong Infectious Diseases Service. Comparison of rateswere calculated using Intercooled Stata for Windows 6.0(Stata Corporation, College Station, TX, USA). For thepurpose of population-based calculations, patients wereexcluded if they lived outside the Barwon–South-Westreferral area as defined by the postcode given at the timeof admission.

ResultsDuring the period 1994–99, 62 patients were identifiedas having bacterial endocarditis. Following chartreview, four were excluded as it was felt that the diag-nosis of endocarditis was doubtful and all fell into the‘rejected’ category. Of the remaining 58, the median agewas 63 years with a range of 20–94. They were 41

28 A. Cheng et al. Heart, Lung and Circulation 2002; 11Changing profile of bacterial endocarditis

(70.6%) males and 17 (29.3%) females. Of these, 44(75.9%) were classified as definite on Duke criteria, and14 (24.1%) as possible endocarditis. Twenty-six peoplehad involvement of the mitral valve only, 24 hadinvolvement of the aortic valve only. More than onevalve was infected in three and other cardiac structureswere involved in five.

The proposed modifications to the original Duke crite-ria would not have resulted in the inclusion or exclusionof any of the other patients in this series. However,

because of the elimination of suggestive echocardio-graphic findings as minor criteria, three patients wouldhave been reclassified as ‘possible endocarditis’ ratherthan ‘definite endocarditis’.

IncidenceThe annual number of cases of endocarditis referred tothe Geelong Infectious Diseases Service rose from five in1995 to 15 in 1999. The proportion with prosthetic valveendocarditis and nosocomial infection did not change

Table 1. Summary of Duke criteria for diagnosis of infective endocarditis9 with proposed modifications†10

Major criteria(i) Positive blood culture for infective endocarditis

Typical organism for endocarditis from two separate blood cultures‘Viridans’ streptococci, Streptococcus bovis, HACEK group, S. aureus, enterococciPersistently positive blood cultures†Single positive culture for Coxiella burnetii or IgG Ab titre > 1:800

(ii) Evidence of endocardial involvementOscillating intracardiac mass on valve, in path of regurgitant jet or on implanted materialAbscessNew partial dehiscence of prosthetic valve

New valvular regurgitation (not only change in pre-existing murmur)Minor criteria

(i) Predisposing heart lesion or intravenous drug use(ii) Fever > 38.0°C

(iii) Vascular phenomena including major arterial emboli, mycotic aneurysm, intracranial haemorrhage, Janeway lesions

(iv) Immunological phenomena including glomerulonephritis, Osler’s nodes, Roth’s spots(v) Microbiological evidence; positive blood culture not meeting major criterion or serological evidence of organism

consistent with endocarditis(vi) Echocardiogram consistent with endocarditis but not meeting major criterion (†eliminated in proposed

modification)

HACEK, Haemophilus parainfluenzae, Haemophilus aphrophilus, Actinobacillus actinomycetemcomitans, Cardio-bacterium hominis, Eikenella corrodens and Kingella kingae.

Table 2. Summary of ‘definite’, ‘possible’ and ‘rejected’ endocarditis9 with proposed modifications†10

Definite(i) Pathological criteria

(ii) Clinical criteriaTwo major criteriaOne major and three minor criteriaFive minor criteria

Possible infective endocarditis(i) Findings consistent with infective endocarditis that fall short of ‘definite’ but not ‘rejected’

(ii) †One major criterion and one minor criterion; or three minor criteriaRejected

(i) Firm alternative diagnosis for manifestations of endocarditis(ii) Resolution of manifestations of endocarditis with antibiotic therapy for 4 days or less

(iii) No pathological evidence of infective endocarditis at surgery or autopsy after antibiotic therapy for 4 days or less

Heart, Lung and Circulation 2002; 11 A. Cheng et al. 29Changing profile of bacterial endocarditis

significantly during the study period. Similarly, therewere no obvious changes in the proportions of Staphylo-coccus aureus infection, although cases of endocarditisdue to methicillin-resistant S. aureus (MRSA) were notencountered until late 1998.

Forty-seven out of 58 cases were treated at GeelongHospital and the rate of admissions for endocarditis rosefrom 0.15 cases per 1000 admissions in 1995 to 0.26 casesper 1000 in 1998 (P = 0.21).11 The mid-interval populationof the Barwon–South-West region during this studyperiod was estimated at 327 210 in 1997.12 Four patientswere excluded from the incidence calculation as theylived outside the Barwon–South-West region and fivecases in 1994 were excluded due to incomplete data forthat year, leaving 49 cases within the region. Thus, theincidence of endocarditis over this period was thus esti-mated at 3.0 per 100 000 per annum for the period1994–99.

EchocardiographyEchocardiographic imaging was performed in 56 out of58 cases. Of these, findings satisfied major criteria in44 (78.6%), minor criteria in eight (14.3%) and were nega-tive in four (7.1%). In the 43 patients that had tran-soesophageal echocardiograms performed, findingssatisfied major criteria in 36 (83.7%), satisfied minorcriteria in five (11.6%) and was negative in two (4.7%).

Two of the five patients with transoesophageal find-ings that fell short of major criteria and the one withsuggestive transthoracic findings would have beenreclassified as ‘possible’ endocarditis rather than ‘defi-nite’ endocarditis when applying the proposed modifica-tions to the Duke criteria.

Causative OrganismsCausative agents could be classed into three groups.S. aureus was the pathogen in 23 (39.7%) cases, of whom

four had MRSA (Table 3). All cases of MRSA endocardi-tis were acquired nosocomially and were seen afterNovember 1998. Twelve patients (20.7%) had alpha-haemolytic (‘viridans’) streptococci; other causativeorganisms included four Enterococcus spp., five coagu-lase-negative staphylococci, one Streptococcus pneumoniae,one Streptococcus bovis and one Actinobacillus actinomyce-temcomitans. In 11 (19.0%) cases, blood cultures werenegative.

Source of InfectionA source of infection was identified in 18 (31.0%)patients: 10 cases were due to infected intravasculardevices (four peripheral cannulae, three central venouscatheters, two renal dialysis catheters, one Hickman’sline). Sixteen (27.6%) were classed as having nosocomialinfection including the 10 infected intravascular device-related cases, plus three with early postvalve replace-ment (within 2 months of surgery), and three with post-operative infection of unknown source. There were noinjecting drug users identified in this series.

Of the 10 patients with infected intravascular devicesas a source of endocarditis, four had methicillin-sensitiveS. aureus and four MRSA. Ten of the 23 S. aureus infec-tions were nosocomial.

Infections of Prosthetic Material and Predisposing Cardiac LesionsThere were 18 (31.0%) infections of endovascular pros-thetic material; 13 prosthetic aortic valves, two mitralvalves, one involving both aortic and mitral valves, oneinvolving both aortic and tricuspid valves and one infec-tion involving a prosthetic conduit. Of the seven patientswith pacing wires, six were felt to be infected and, ofthese, five had concurrent valvular infections. Of theendovascular infections, the majority were late; onlythree occurring within 2 months of valve replacement.

Table 3. Summary of causative organisms, type of valve infection and source of infection

Pathogen Native-valve Prosthetic-valve Nosocomial Community-acquired

Total

Staphylococcus aureus 19 4 10 13 23‘Viridans’ streptococci 9 3 1 11 12Other 12

Enterococcus spp. 1 3 1 3 4Coagulase-negative Staphylococcus 3 2 1 4 5Streptococcus pneumoniae 1 – – 1 1Streptococcus bovis 1 – 1 1 Actinobacillus sp. 1 – 1 1

Culture negative 6 5 3 8 11Total 40 18 16 42 58

30 A. Cheng et al. Heart, Lung and Circulation 2002; 11Changing profile of bacterial endocarditis

One-third of the 18 cases of prosthetic valve endo-carditis were staphylococcal with four due to S. aureusand two due to coagulase-negative staphylococci. Therewere three enterococcal infections, three due to alpha-haemolytic streptococci and one S. bovis, with theremaining five culture negative. Staphylococci were alsothe causative organism in four of the six pacemaker wireinfections, with three due to S. aureus and one coagulase-negative staphylococci; two were culture negative.

Twelve (20.6%) patients had documented predispos-ing cardiac abnormalities prior to the diagnosis of endo-carditis; seven had mitral regurgitation, two had mitralstenosis, one had aortic stenosis, one had complex con-genital heart disease and one had hypertrophic obstruc-tive cardiomyopathy.

Management and OutcomeTreatment with intravenous antibiotics was started in allcases and, in most instances, continued according toestablished guidelines.13

Eighteen patients (31.0%) required surgery, withsimilar rates in native valve (32%) and prosthetic valve(28%) endocarditis. Thirty-one (53.4%) patients were suc-cessfully treated the ‘Hospital in the Home’ programmefor completion of their intravenous therapy.

The four deaths (22%) in patients with prostheticvalve endocarditis occurred in patients in whom medicalmanagement was the only option. In the six patients withinfected pacing wires, three had successful removal ofthe wire, and two patients died while undergoingmedical therapy. Overall, 12 patients died, but two ofthese deaths were felt to be unrelated to the infection(metastatic parotid carcinoma; multiple myeloma) in thisseries of 58, giving a attributable case fatality rate of 17%.

DiscussionThe population-based incidence of endocarditis seenhere, estimated at 3.0 per 100 000 people, is similar to apreviously reported estimate for New South Wales in1980 (3.8 per 100 000 people).14 Population-based surveysin other countries have reported higher rates, from 4.2per 100 000 in Olmsted County, Minnesota USA15 to 6.2per 100 000 in Goteburg, Denmark16 to 11.6 per 100 000 inDelaware County, Pennsylvania USA.17 We cannot dis-count the possibility that some cases of endocarditis mayhave been referred directly to large metropolitan hospi-tals, leading to under-reporting in this series.

It is likely that the increase in the number of casesseen at the Geelong hospitals represents a true rise in theincidence of endocarditis, as the total population of theBarwon–South-West region is relatively stable, increas-ing by only 1.3% over 1991–97.12 Although there was a

rise in the proportion of cases of endocarditis per totaladmissions, rising from 0.17 to 0.26 per 1000 admissions,due to the relatively small number of cases annually,we were unable to demonstrate statistical significance.This proportion represents the lower end of previoushospital-based incidences (0.16–5.4 per 1000 admissions)8

probably reflecting referral patterns.The proportion of endocarditis due to S. aureus was

almost 40% with ‘viridans’ streptococci contributing 20%,consistent with more recent studies since the 1980s atreferral hospitals,8 at district hospitals15,18 and popula-tion-based series.16 This is in contrast to series publishedprior to 1970 which noted alpha-haemolytic streptococciin up to 60% of cases7 and S. aureus at between 8% and14% of cases.19 In Sydney, Dwyer and coworkers8 sug-gested that the rise in the prevalence of injecting druguse and nosocomial infection may account for thisincrease. Certainly, the rate of nosocomial infection inGeelong is higher than the 14–16% reported in otherseries.8,18

We are unable to explain the lack of intravenous drugusers in this series. It is possible that injecting practices orattendance patterns at hospitals outside the region mayhave led to this finding. The age distribution of our cases,with only four patients under the age of 30, differsmarkedly from the age distribution of injecting drugusers.19 We feel that this makes it unlikely that thisrisk factor was not sought during clinical assessment.The rate of culture-negative endocarditis reported hereat 20% is high when compared with other series(9–14%):8,9,16,18 there are many possible reasons, includ-ing lower suspicion of endocarditis, the administration ofantibiotics prior to the taking of blood cultures andvariations in laboratory procedures.

Almost one-third of cases involved prosthetic valves, ofwhich only three cases were early-onset infections seensoon after valve replacement. This is higher than the pro-portions reported at some referral centres (22–26%)8,15 andin community studies (14–24%).15,16,18 Clearly, prostheticvalve endocarditis is not just a metropolitan disease, andthe incidence may well rise even further with increasingavailability of cardiac surgery in provincial regions.

There has been a general decline the overall case-fatality ratio in the modern era from 31 to 35% in the1970s20–22 to the present rates. However, it has improvedlittle since then and the mortality of 17% seen here isconsistent with other series in the late 1980s and early1990s8,16 A significant proportion of patients used theoutpatient antibiotic therapy program, reflecting theincreasing acceptability and safety of this managementoption both in Australia23,24 and overseas.25 In this regard,provincial centres should be able to follow the trend nowwidely established in many industrialised countries.

Heart, Lung and Circulation 2002; 11 A. Cheng et al. 31Changing profile of bacterial endocarditis

Echocardiography was performed in almost allpatients, with findings indicative of endocarditis in 93%;however, the major Duke criteria were fulfilled in only76%. In the original series by Durack et al. only 51% ofoverall cases fulfilled major criteria and 23% fulfilledminor criteria9 reflecting reliance on transthoracicechocardiography at that time. Without the ‘gold stand-ard’ of pathological diagnosis, it is difficult to assess theeffect of the proposed modifications to the Duke crite-ria.10 Our experience shows that there may be a smallnumber of patients in which transoesophgeal echocardi-ography is not conclusive. Some of these patients arelogically classified as ‘possible’ endocarditis, reflectingthe high sensitivity of the test. In other provincial centres,however, transoesophgeal echocardiography may notalways be available and, in this context, it may be worthretaining minor echocardiographic findings.

It is difficult to compare this series with previousreviews due to differences in the definition of endo-carditis, the introduction of transoesphageal echocardi-ography and local variations in the referring populationsuch as patterns of intravenous drug use and prevalenceof patients with prosthetic valves. Given these funda-mental differences, we did not consider it appropriate toperform statistical tests.

Previous studies of the nature of the ‘referral filter’, inthe 1970s and 1980s8,15 have indicated that the epidemiol-ogy of endocarditis in the community may reflect morethe ‘classic’ picture. Our findings indicate that the profileof infective endocarditis in the provincial setting may betrending towards that of referral centres in large metro-politan areas, characterised by an increasing incidence ofnosocomial infection and endocarditis involving pros-thetic valves, together with increasingly pathogenic andantibiotic-resistant organisms. Diagnostic pathways andempirical treatment options will need to be adjustedaccordingly.

AcknowledgementsWe thank Dr Mark Veitch for denominator data for theBarwon–South-West region and the medical records staffat Geelong Hospital, St John of God Hospital andGeelong Private Hospital, Victoria, Australia.

References1. Levinson M. Infective endocarditis. In: Goldman L, Bennett JC

(Eds), Cecil Textbook of Medicine. WB Saunders, Philadelphia,2000, pp. 1631–40.

2. Kaye D. Infective endocarditis. In: Braunwald E, Fauci AS,Isselbacher KJ et al. (Eds), Harrison’s Principles of Internal Medi-cine. McGraw-Hill, New York, 1998, pp. 785–791.

3. Dhawan A, Grover A, Marwaha RK et al. Infective endocarditisin children: profile in a developing country. Ann. Trop. Paediatr.1993; 13: 189–94.

4. Coard KC, Prabhakar P, Bansal AS. Infective endocarditis at theUniversity Hospital of the West Indies. A postmortem evalua-tion. West Indian Med. J. 1989; 38: 217–21.

5. Anonymous. Bacterial endocarditis: a changing pattern. Lancet1967; 1: 605–6.

6. Finland M, Barnes MW. Changing aetiology of bacterial endo-carditis in the antibacterial era. Ann. Intern. Med. 1970; 72: 341–8.

7. Cherubin CE, Neu HC. Infective endocarditis at the Presby-terian Hospital in New York City from 1938–1967. Am. J. Med.1938; 51: 83–96.

8. Dwyer DE, Chen SC, Wright EJ, Crimmins D, Collignon PJ,Sorrell TC. Hospital practices influence the pattern of infectiveendocarditis. Med. J. Aust. 1994; 160: 716–18.

9. Durack DT, Lukes AS, Bright DK. New criteria for diagnosis ofinfective endocarditis: utilization of specific echocardiographicfindings. Duke Endocarditis Service. Am. J. Med. 1994; 96:200–9.

10. Li JS, Sexton DJ, Mick N et al. Proposed modifications to theDuke criteria for the diagnosis of infective endocarditis. Clin.Infect. Dis. 2000; 30: 633–8.

11. Barwon Health. Annual Reports 1995–99. Barwon Health,Geelong, 1999.

12. Estimated Resident Population Local Government Areas. Popu-lation, Victoria (catalogue no. 3234.2). Australian Bureau ofStatistics, Canberra, 1997.

13. TG Writing Group. Therapeutic Guidelines − Antibiotics, 1998/1999, 10th edn. Therapeutic Guidelines, Melbourne, 1998.

14. Hickey AJ, MacMahon SW, Wilcken DE. Mitral valve prolapseand bacterial endocarditis: when is antibiotic prophylaxis neces-sary? Am. Heart J. 1985; 109: 431–5.

15. Steckelberg JM, Melton LJD, Ilstrup DM, Rouse MS, WilsonWR. Influence of referral bias on the apparent clinical spectrumof infective endocarditis. Am. J. Med. 1990; 88: 582–8.

16. Hogevik H, Olaison L, Andersson R, Lindberg J, Alestig K.Epidemiologic aspects of infective endocarditis in an urbanpopulation. A 5-year prospective study. Medicine (Baltimore)1995; 74: 324–39.

17. Berlin JA, Abrutyn E, Strom BL et al. Incidence of infectiveendocarditis in the Delaware Valley, 1988–1990. Am. J. Cardiol.1995; 76: 933–6.

18. Watanakunakorn C, Burkert T. Infective endocarditis at a largecommunity teaching hospital 1980–1990. A review of 210 epi-sodes. Medicine (Baltimore) 1993; 72: 90–102.

19. Gerostamoulos J, Staikos V, Drummer OH. Heroin relateddeaths in Victoria: a review of cases for 1997 and 1998. DrugAlcohol Depend. 2001; 61: 123–7.

20. Bailey IK, Richards JG. Infective endocarditis in a Sydney teach-ing hospital – 1962–1971. ANZ J. Med. 1975; 5: 413–20.

21. Cacoub P, Leprince P, Nataf P et al. Pacemaker infective endo-carditis. Am. J. Cardiol. 1998; 82: 480–4.

22. McGivern D, Ispahani P, Banks D. Factors influencing mortalityfrom infective endocarditis in two district general hospitals.Postgrad. Med. J. 1987; 63: 345–9.

23. Grayson ML, Silvers J, Turnidge J. Home intravenous antibiotictherapy. A safe and effective alternative to inpatient care. Med. J.Aust. 1995; 162: 249–53.

24. Grayson ML. Hospital-in-the-home care: is it worth the hassle?Med. J. Aust. 1998; 168: 262–3.

25. Tice AD, Marsh PK, Craven PC, McEniry DW. Home intrave-nous antibiotic therapy. Am. J. Med. 1993; 94: 114–15.