Those who work in any area of public health are exposed to large amounts of data and processed information. Some acquaintance with statistical methods can make that exposure more profitable. The objectives of this paper are to describe a suitable framework for governing scientific investigations which involve applied statistics, and to set down guidelines for research workers preparing material for publication. Final sections suggest further reading and give a brief description of some of the statistical software available to assist analysis.

Bacteriological quality of potable water supplying Morogoro municipality and its outskirts (population 135000 people) was assessed by the determination of the most probable number (MPN) of faecal coliforms, Escherichia coli, faecal streptococci and Clostridium perfringens for a period of 6 months. River water and chlorinated pipeborne water were found contaminated with microorganisms in the order of 3·8 x 101 to 4·95 x 103; 3·2 x 101 to 4·5 x 103; 101 to 6·4 x 101 and 1·2 x 101 to 2·5 x 102/100 ml of water respectively. Injured coliforms in treated water averaged 87 %. Pollution of river water by organic matter was much above acceptable standard. These findings indicate that there is a need for further treatment of water before consumption in order to avoid potential health hazards.

An outbreak of cryptosporidiosis resulted in 516 cases in Wiltshire and Oxfordshire. The outbreak caused widespread interest and led to an official inquiry. The majority of cases were in children; 8% of cases were admitted to hospital and the median duration of illness was 3 weeks. The geographical distribution of cases matched the distribution of water supplies from three treatment works and eryptosporidium oocysts were found at these works and in the treated water. Attack rates in electoral wards supplied by the three treatment works were significantly higher than in other wards. The cause of the outbreak appeared to be the failure of normal treatment to remove oocysts. Measures at the treatment works reduced the number of oocysts detected in treated water, after which the outbreak came to an end. The conclusion of the investigations was that cryptosporidiosis is a risk of conventionally treated public water supplies.

In August 1988 an increase was noted in the number of cases of cryptosporidiosis identified by the microbiology laboratory at Doncaster Royal Infirmary. By 31 October, 67 cases had been reported. Preliminary investigations implicated the use of one of two swimming pools at a local sports centre and oocysts were identified in the pool water. Inspection of the pool revealed significant plumbing defects which had allowed ingress of sewage from the main sewer into the circulating pool water. Epidemiological investigation confirmed an association between head immersion and illness. The pools were closed when oocysts were identified in the water and extensive cleaning and repair work was undertaken. The pool water was retested for cryptosporidial oocysts and found to be negative before the pool re-opened.

An outbreak of cryptosporidiosis occurred in the Isle of Thanet during December 1990 and January 1991. A total of 47 cases ranging in age from 2 months to 85 years were identified in residents from the Margate, Broadstairs and Ramsgate areas, with dates of onset of illness from 3 December to 14 January. A case-control study demonstrated a strong statistical association between illness and the consumption of unboiled tap water from a particular source, with evidence of a dose–response relationship. Although no cryptosporidial oocysts were identified in samples of untreated or treated water taken during the investigation, the results were consistent with the view that the source of infection was treated river water which was used to supplement borehole water.

Environmental contamination has been shown to be an important aspect of the epidemiology of salmonellosis in calves. Markets and transport vehicles are important links in the calf marketing chain and these were investigated to determine the level of salmonella contamination.

Salmonellas were isolated from 7 of the 14 markets surveyed, with 31 of 838 samples (3.7%) being positive. Nine different salmonella serotypes, of which the commonest was Salmonella typhimurium, were isolated. Four different phage types of S. typhimurium were detected, the commonest being DT204C.

Salmonellas were isolated from 22 of the 107 vehicles (20.6%) examined before washing and from 4 of the 62 vehicles (6.5%) examined after cleaning. Twelve different salmonella serotypes were isolated, of which the most frequent was S. typhimurium. The commonest of the six different S. typhimurium phage types was DT204C.

These results indicate that improved cleaning and disinfection routines both for vehicles and markets are necessary to control salmonellosis in calves.

The diagnosis of maternal infection in early pregnancy depends on tests which are sensitive to recent infection, such as specific IgM. Two types of test are considered: those where the response persists for a period following infection and then declines, such as IgM. and those whose response increases with time since infection, such as IgG-avidity. However, individuals vary in their response to infection, and it may not always be possible to determine whether an infection occurred during pregnancy or before it. Mathematical methods are developed to evaluate the performance of these tests, and are applied to the diagnosis of toxoplasmosis in pregnancy. It is shown that, based on existing information, tests of recent infection are unlikely to be both sensitive and predictive. More data on these tests are required, before they can be reliably used to determine whether infection has occurred during pregnancy or before it.

The population structure and transmission of Escherichia coli in a small group of individuals isolated for 26 weeks on an Antarctic base were studied by multilocus electrophoresis of eight enzymes and plasmid analysis. Two hundred and sixty-nine strains were isolated. They were grouped into 60 allozyme types (ETs). Half of these ETs were only isolated once; others were repeatedly isolated from single subjects. Eleven were found in more than one subject and the pattern of the occurrence of some of them was considered to provide evidence of their spread from subject to subject.

Genotypic and phenotypic analysis of 42 strains of Streptococcus uberis isolated from mammary secretions of 17 cows collected at different periods of the lactation cycle and from episodes of clinical mastitis were performed. Seventeen restriction endonuclease fingerprint (REF) patterns and 12 bacteriocin-like inhibitory substance (BLIS) fingerprints were observed. REF identified and differentiated closely related strains of S. uberis isolated from mammary secretions collected from the same cow at different periods of the lactation cycle and from episodes of clinical mastitis. BLIS fingerprinting of S. uberis complemented REF results. REF and BLIS fingerprinting provided evidence concerning persistence of infection in the same quarter or different quarters of the mammary gland over different periods of the lactation cycle, and occurrence of infection with similar and dissimilar strains of S. uberis. Biochemical profiles could not identify closely related strains nor did they complement REF results. Antibiotic resistance patterns alone were of little value in differentiating closely related strains, but were identical with isolates having same REF pattern. None of the S. uberis strains was found to carry plasmids. REF and BLIS fingerprinting can be utilized effectively to differentiate closely related and unrelated strains of S. uberis isolated from bovine mammary secretions.

This study included 502 staphylococcus strains; Staphylococcus saprophyticus (297 strains) S. cohnii (47), S. xylosus (10), S. epidermidis (67) and S. aureus (81). Mitomycin C induction was performed on 100 isolates of S. saprophyticus and all induced strains were reacted with each other. Twenty-six strains proved to be lysogenic. Phages were propagated and titrated. With 12 of the phages there were three frequent associations, named lytic groups A, B and C, which included 75% of all typable strains. Typability of the system was 45% and reproducibility was between 94.2% and 100%. Phages did not lyse S. aureus and S. epidermidis strains, but they lysed S. saprophyticus and only rare strains of other novobiocin resistant species. Effective S. saprophyticus typing serves ecological purposes and tracing the origin of urinary strains from the skin or mucous membranes. Phage typing in association with plasmid profiling previously described, are anticipated as complementary methods with strong discriminatory power for differentiating among S. saprophyticus strains.

Characterization of 169 strains of Salmonella typhi of phage types C1, C4, D1 and D9 isolated in 1975–88 was carried out by rDNA gene restriction pattern analysis. Twenty-four isolates had been recovered during four large waterbone outbreaks in the last 20 years in Sicily; 145 strains, isolated from apparently sporadic cases of infection in Southern Italy in the same period of time, were also examined.

Application of rRNA–DNA hybridization technique after digestion of chromosomal DNA with Cla I showed the identity of patterns of the epidemic strains of phage types C1 and D1, confirming attribution of the outbreaks to single bacterial clones. Patterns of the two available strains of lysotype D9 were slightly different, whilst the 12 epidemic strains of phage type C4 could be assigned to two distinct patterns scarcely related to each other and, consequently, to two different clones. A considerable heterogeneity was detected among all apparently sporadic isolates of the four phage types under study.

This fingerprinting method appears a reliable tool to complement phage typing in characterizing isolates of S. typhi. In particular, epidemiological features of spread of this salmonella serovar in areas, where simultaneous circulation of indigenous and imported strains occurs, can be elucidated.

The indirect haemagglutination (IHA) test using sensitized turkey erythrocytes and the indirect immunofluorescence assay (IgM-IFA) was confirmed to be sensitive in the detection of a recent or current Pseudomonas pseudomallei infection in 19 culture-confirmed Singapore melioidosis patients. All were found to have antibody titres from 4 to 32 768 in the IHA test and 10 to 320 in the IgM-IFA test. When these tests were employed on sera from 16 immigrant Thai construction workers who died of sudden unexplained death syndrome (SUDS) and 73 healthy Thai fellow workers, 93·8% and 68·8% of SUDS cases had IHA titre of ≥ 4 and IgM-IFA titre of ≥ 10 respectively, in contrast to 39·7% and 12·3% found among healthy Thai workers. These data indicate that at the time of death, most of the SUDS patients had an active infection with P. pseudomallei, possibly resulting from reactivation of a latent infection. The aetiological role of P. pseudomallei as the major cause of SUDS is discussed.

In an attempt to explain the recent resurgence of homosexually-acquired gonorrhoea in the Lothian region of Scotland the number of infections and pattern of infection (urethral, rectal and pharyngeal) of all gonococcal isolates from homosexual men attending the Department of Genitourinary Medicine at Edinburgh Royal Infirmary between 1985 and 1990 were analysed. Serovar typing data were available from infections acquired between January 1986 and December 1990. A correlation between one serovar, Bacejk/Brpyust, and the overall pattern of gonorrhoea was observed. The number of infections caused by minor serovars also correlated with rates of gonococcal infection. The number of minor serovars isolated, which may represent strains from other geographical locations, is related to the total incidence of gonorrhoea. It is possible that the incidence of Bacejk/Brpyust may be determined by the size of the infected pool of gonorrhoea. The most likely explanation for the recent increase in gonorrhoea is a change in sexual behaviour and/or an influx of strains from other geographical areas.

In a setting where potable water is contaminated with Legionella pneumophila serogroup 1, we performed two case control studies. The first case control study consisted of 17 cases of nosocomial Legionnaires' disease (LD) and 33 control (the patients who were admitted to the ward where the case was admitted immediately before and after the case) subjects. Cases had a higher mortality rate 65% vs 12% (P < 0.004); were more likely to have received assisted ventilation (P < 0.00001); to have nasogastric tubes (P < 0.0004) and to be receiving corticosteroids or other immunosuppressive therapy (P < 0.0001). Based on the results of this study, sterile water was used to flush nasogastric tubes and to dilute nasogastric feeds. Only 3 cases of nosocomial LD occurred during the next year compared with 12 the previous year (P < 0.0001). Nine cases subsequently occurred and formed the basis for the second case-control study. Eighteen control subjects were those patients admitted to the same unit where the case developed LD, immediately before and after the case. The mortality rate for the cases was 89% vs 6% for controls (P < 0.00003). The only other significant difference was that cases were more likely to be receiving corticosteroids or other immunosuppressive therapy 89% vs 39% ( < 0.01). We hypothesized that microaspiration of contaminated potable water by immunocompromised patients was a risk factor for nosocomial Legionnaires' disease. From 17 March 1989 onwards such patients were given only sterile potable water. Only two cases of nosocomial LD occurred from June 1989 to September 1990 and both occurred on units where the sterile water policy was not in effect. We conclude that aspiration of contaminated potable water is a possible route for acquisition of nosocomial LD in our hospital and that provision of sterile potable water to high risk patients (those who are receiving corticosteroids or other immunosuppressive drugs; organ transplant recipients or hospitalized in an intensive care unit) should be mandatory.

Twenty-three pre- and post-treatment isolates of Helicobacter pylori from the antral mucosa of eight patients with dyspepsia and gastritis were compared using 1-D SDS PAGE of proteins. The protein patterns were highly reproducible and were used as the basis for two numerical analyses. The first, based on the total protein patterns, showed that a number of the strains did not cluster with their respective patient set. This was thought to be due to differences in both mobility and intensity of proteins in the major band region. The second analysis, based on partial patterns, excluding the major band region (51–68 kDa), divided the clinical isolates into clearly defined groups corresponding to the patient sets. Although there was a degree of heterogeneity with respect to protein pattern between the pre- and post-treatment isolates of some patients, there was nonetheless clear evidence that each patient was harbouring strains of only a single type. These results suggested that patients were not being reinfected with a different strain but that there was recrudescence of the pre-treatment strain. Protein ‘fingerprints’ provided a precise and reproducible means of strain differentiation, and revealed that in each patient the same strain persisted after drug therapy even though there was marked patient-to-patient strain variation.

The colonization resistance (CR) of the digestive tract was determined in 10 healthy volunteers by oral contamination with a neomycin resistant Escherichia coli (NR-E. coli) strain and measurement of the faecal concentration of this strain during 14 days after the contamination. This ‘gold standard’ was compared with another parameter of CR; the determination of the mean number of different biotypes of Enterobacteriaceae isolated from four faecal samples per volunteer. Both measures are significantly correlated (P < 0.01). The NR-E. coli strain could be cultured from faecal samples of 4/10 volunteers as long as 300 days after contamination. Serum antibody titres against endogenous E. coli strains and the NR-E. coli strain used for experimental oral contamination were measured by an indirect immunofluorescence (IIF) assay. The assay was read by a video camera connected to an image processing system. The 95% confidence limits of antibody titres (log2) against endogenous E. coli strains ranged between < 3 and 7.1 for IgA, between < 3 and 8.7 for IgG and between < 3 and 7.4 for IgM. Antibody titres against the NR-E. coli 4 strain were within this (normal) range. The serum antibody titres against the NR-E. coli strain increased slowly after oral contamination, especially IgG and IgM. Little increase in IgA titres could be observed. An increase of serum antibody titres did not correlate with the elimination of the oral contaminant from the intestines. Therefore, we conclude that the CR is not IgG nor IgM antibody mediated.

Relapse of Clostridium, difficile-associated diarrhoea occurs in 15–20% of patients; however, whether relapse is due to an endogenous source of the organism or reinfection from the environment remains unclear. Restriction enzyme analysis (REA) of chromosomal DNA was used to type multiple isolates from ten patients who had experienced apparent relapses. More than half the relapses were due to infection with a new strain of C. difficile. The remaining patients were infected with the same strain, but whether this strain was acquired from the environment or from endogenous sources could not be determined. Relapses with a different strain of C. difficile could occur if an individual harboured more than one strain in their gastrointestinal tract. To investigate this possibility ten other patients were assessed for carriage of multiple strains. Ten colonies from a primary culture plate from each patient were typed by REA and tested for their ability to produce cytotoxin. All isolates from the same patient were identical by both methods, indicating that multiple carriage of strains may be a rare event.

Although vaccination is widely practiced against infection by Erysipelothrix rhusiopathiae in pigs and turkeys, the protective antigen(s) involved have not been fully characterized or purified to homogeneity. Antigens of E. rhusiopathiae strain T28, serotype 2b, and of FRANKFURT XI, serotype N, in culture supernatant and in extracts made with hot acid, 10 mM NaOH, ultrasound or EDTA were compared by SDS-PAGE and immunoblotting and in a mouse protection test. EDTA and 10 mM NaOH yielded highly protective extracts; culture supernatant was less protective and ultrasonic or hot acid extracts stimulated little or no protection in mice. Protective antisera from swine, horses and mice recognized prominent bands of molecular mass (m.m.) of 66–64 and 40–39 kDa in EDTA and 10 mM NaOH extracts. Mice immunized with preparations of the 66–64 kDa band purified by preparative electrophoresis were protected. Both antigens were trypsin sensitive, contained no detectable polysaccharide, and showed a marked tendency to aggregate in the absence of SDS.

A matched and serologically confirmed case-control study was carried out to investigate the source of an outbreak of acute hepatitis involving 290000 cases in the suburbs of Shanghai, in January 1988. A total of 132 patients with acute hepatitis from six different hospitals were chosen as cases and the same number of control patients without hepatitis were matched for gender, age, admission date and area of residence. Serum specimens from both case and control patients were detected for specific anti-hepatitis A (HA) IgM antibody and a questionnaire was used to investigate probable risk factors related to the outbreak. The positive rate of anti-HA IgM was 98·48% in the case group and only 0·76% in the control, indicating that the infection was caused by HA virus. The results revealed that the source and mode of transmission were due to the consumption of contaminated and inadequately cooked clams (Anadara subcrenata lischke). There was a highly positive dose-response relationship between the odds ratio of contracting HA and the quantity or frequency of clam consumption. The odds ratios of acquiring HA from clams were up to 62·4–63·4 by both group stratification and multiple unconditional logistic regression analyses.

Cats and dogs being treated at two veterinary clinics were investigated for gastrointestinal carriage of Clostridium difficile using selective solid and en-richment media. Thirty-two (39.5%) of 81 stool samples yielded C. difficile. There were significant differences in isolation rates between clinics, 61.0% of animals being positive at one clinic compared to 17.5% at the other (Chi-square, P < 0.005). Of 29 animals receiving antibiotics, 15 (52.0%) harboured C. difficile while 11 (23.9%) of 46 animals not receiving antibiotics were positive (Chi-square, P < 0.01). There was no difference in carriage rate between cats (38.1%) and dogs (40.0%). The environment at both veterinary clinics was surveyed for the presence of C. difficile. Fifteen of 20 sites at one clinic were positive compared to 6 of 14 sites at the other clinic. Both cytotoxigenic and noncytotoxigenic isolates of C. difficile were recovered from animals and environmental sites. These findings suggest that household pets may be a potentially significant reservoir of infection with C. difficile.